diff --git a/llvm/lib/Target/X86/CMakeLists.txt b/llvm/lib/Target/X86/CMakeLists.txt
--- a/llvm/lib/Target/X86/CMakeLists.txt
+++ b/llvm/lib/Target/X86/CMakeLists.txt
@@ -46,6 +46,7 @@
   X86FastISel.cpp
   X86FixupBWInsts.cpp
   X86FixupLEAs.cpp
+  X86FixupInstTuning.cpp
   X86AvoidStoreForwardingBlocks.cpp
   X86DynAllocaExpander.cpp
   X86FixupSetCC.cpp
diff --git a/llvm/lib/Target/X86/X86.h b/llvm/lib/Target/X86/X86.h
--- a/llvm/lib/Target/X86/X86.h
+++ b/llvm/lib/Target/X86/X86.h
@@ -63,6 +63,10 @@
 /// instructions, in order to eliminate execution delays in some processors.
 FunctionPass *createX86FixupLEAs();
 
+/// Return as pass that replaces equivilent slower instructions with faster
+/// ones.
+FunctionPass *createX86FixupInstTuning();
+
 /// Return a pass that removes redundant LEA instructions and redundant address
 /// recalculations.
 FunctionPass *createX86OptimizeLEAs();
@@ -167,6 +171,7 @@
 void initializeFPSPass(PassRegistry &);
 void initializeFixupBWInstPassPass(PassRegistry &);
 void initializeFixupLEAPassPass(PassRegistry &);
+void initializeX86FixupInstTuningPassPass(PassRegistry &);
 void initializeWinEHStatePassPass(PassRegistry &);
 void initializeX86AvoidSFBPassPass(PassRegistry &);
 void initializeX86AvoidTrailingCallPassPass(PassRegistry &);
diff --git a/llvm/lib/Target/X86/X86FixupInstTuning.cpp b/llvm/lib/Target/X86/X86FixupInstTuning.cpp
new file mode 100644
--- /dev/null
+++ b/llvm/lib/Target/X86/X86FixupInstTuning.cpp
@@ -0,0 +1,146 @@
+//===-- X86FixupInstTunings.cpp - replace instructions -----------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file does a tuning pass replacing slower machine instructions
+// with faster ones. We do this here, as opposed to during normal ISel, as
+// attempting to get the "right" instruction can break patterns. This pass
+// is not meant search for special cases where an instruction can be transformed
+// to another, it is only meant to do transformations where the old instruction
+// is always replacable with the new instructions. For example:
+//
+//      `vpermq ymm` -> `vshufd ymm`
+//          -- BAD, not always valid (lane cross/non-repeated mask)
+//
+//      `vpermilps ymm` -> `vshufd ymm`
+//          -- GOOD, always replaceable
+//
+//===----------------------------------------------------------------------===//
+
+#include "X86.h"
+#include "X86InstrInfo.h"
+#include "X86Subtarget.h"
+#include "llvm/ADT/Statistic.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineRegisterInfo.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "x86-fixup-inst-tuning"
+
+STATISTIC(NumInstChanges, "Number of instructions changes");
+
+namespace {
+class X86FixupInstTuningPass : public MachineFunctionPass {
+public:
+  static char ID;
+
+  X86FixupInstTuningPass() : MachineFunctionPass(ID) {}
+
+  StringRef getPassName() const override { return "X86 Fixup Inst Tuning"; }
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+  bool processInstruction(MachineFunction &MF, MachineBasicBlock &MBB,
+                          MachineBasicBlock::iterator &I);
+
+  // This pass runs after regalloc and doesn't support VReg operands.
+  MachineFunctionProperties getRequiredProperties() const override {
+    return MachineFunctionProperties().set(
+        MachineFunctionProperties::Property::NoVRegs);
+  }
+
+private:
+  const X86InstrInfo *TII = nullptr;
+  const X86Subtarget *ST = nullptr;
+};
+} // end anonymous namespace
+
+char X86FixupInstTuningPass::ID = 0;
+
+INITIALIZE_PASS(X86FixupInstTuningPass, DEBUG_TYPE, DEBUG_TYPE, false, false)
+
+FunctionPass *llvm::createX86FixupInstTuning() {
+  return new X86FixupInstTuningPass();
+}
+
+bool X86FixupInstTuningPass::processInstruction(
+    MachineFunction &MF, MachineBasicBlock &MBB,
+    MachineBasicBlock::iterator &I) {
+  MachineInstr &MI = *I;
+  unsigned Opc = MI.getOpcode();
+  unsigned NumOperands = MI.getDesc().getNumOperands();
+
+  // `vpermilps r, i` -> `vshufps r, r, i`
+  // `vshufps` is always as fast or faster than `vpermilps` and takes 1 less
+  // byte of code size.
+  auto ProcessVPERMILPSri = [&](unsigned NewOpc) -> bool {
+    unsigned MaskImm = MI.getOperand(NumOperands - 1).getImm();
+    MI.removeOperand(NumOperands - 1);
+    MI.addOperand(MI.getOperand(1));
+    MI.setDesc(TII->get(NewOpc));
+    MI.addOperand(MachineOperand::CreateImm(MaskImm));
+    return true;
+  };
+
+  // `vpermilps m, i` -> `vpshufd m, i` iff no domain delay penalty on shuffles.
+  // `vpshufd` is always as fast or faster than `vpermilps` and takes 1 less
+  // byte of code size.
+  auto ProcessVPERMILPSmi = [&](unsigned NewOpc) -> bool {
+    // TODO: Might be work adding bypass delay if -Os/-Oz is enabled as
+    // `vpshufd` saves a byte of code size.
+    if (!ST->hasNoDomainDelayShuffle())
+      return false;
+    MI.setDesc(TII->get(NewOpc));
+    return true;
+  };
+
+  // TODO: Add masked predicate execution variants.
+  switch (Opc) {
+  case X86::VPERMILPSri:
+    return ProcessVPERMILPSri(X86::VSHUFPSrri);
+  case X86::VPERMILPSYri:
+    return ProcessVPERMILPSri(X86::VSHUFPSYrri);
+  case X86::VPERMILPSZ128ri:
+    return ProcessVPERMILPSri(X86::VSHUFPSZ128rri);
+  case X86::VPERMILPSZ256ri:
+    return ProcessVPERMILPSri(X86::VSHUFPSZ256rri);
+  case X86::VPERMILPSZri:
+    return ProcessVPERMILPSri(X86::VSHUFPSZrri);
+  case X86::VPERMILPSmi:
+    return ProcessVPERMILPSmi(X86::VPSHUFDmi);
+  case X86::VPERMILPSYmi:
+    // TODO: See if there is a more generic way we can test if the replacement
+    // instruction is supported.
+    return ST->hasAVX2() ? ProcessVPERMILPSmi(X86::VPSHUFDYmi) : false;
+  case X86::VPERMILPSZ128mi:
+    return ProcessVPERMILPSmi(X86::VPSHUFDZ128mi);
+  case X86::VPERMILPSZ256mi:
+    return ProcessVPERMILPSmi(X86::VPSHUFDZ256mi);
+  case X86::VPERMILPSZmi:
+    return ProcessVPERMILPSmi(X86::VPSHUFDZmi);
+  default:
+    return false;
+  }
+}
+
+bool X86FixupInstTuningPass::runOnMachineFunction(MachineFunction &MF) {
+  LLVM_DEBUG(dbgs() << "Start X86FixupInstTuning\n";);
+  bool Changed = false;
+  ST = &MF.getSubtarget<X86Subtarget>();
+  TII = ST->getInstrInfo();
+  for (MachineBasicBlock &MBB : MF) {
+    for (MachineBasicBlock::iterator I = MBB.begin(); I != MBB.end(); ++I) {
+      if (processInstruction(MF, MBB, I)) {
+        ++NumInstChanges;
+        Changed = true;
+      }
+    }
+  }
+  LLVM_DEBUG(dbgs() << "End X86FixupInstTuning\n";);
+  return Changed;
+}
diff --git a/llvm/lib/Target/X86/X86TargetMachine.cpp b/llvm/lib/Target/X86/X86TargetMachine.cpp
--- a/llvm/lib/Target/X86/X86TargetMachine.cpp
+++ b/llvm/lib/Target/X86/X86TargetMachine.cpp
@@ -571,6 +571,7 @@
     addPass(createX86FixupBWInsts());
     addPass(createX86PadShortFunctions());
     addPass(createX86FixupLEAs());
+    addPass(createX86FixupInstTuning());
   }
   addPass(createX86EvexToVexInsts());
   addPass(createX86DiscriminateMemOpsPass());
diff --git a/llvm/test/CodeGen/X86/2012-01-12-extract-sv.ll b/llvm/test/CodeGen/X86/2012-01-12-extract-sv.ll
--- a/llvm/test/CodeGen/X86/2012-01-12-extract-sv.ll
+++ b/llvm/test/CodeGen/X86/2012-01-12-extract-sv.ll
@@ -6,7 +6,7 @@
 ; AVX1-LABEL: endless_loop:
 ; AVX1:       # %bb.0: # %entry
 ; AVX1-NEXT:    vmovaps (%eax), %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-NEXT:    vxorps %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
diff --git a/llvm/test/CodeGen/X86/SwizzleShuff.ll b/llvm/test/CodeGen/X86/SwizzleShuff.ll
--- a/llvm/test/CodeGen/X86/SwizzleShuff.ll
+++ b/llvm/test/CodeGen/X86/SwizzleShuff.ll
@@ -21,8 +21,8 @@
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps (%rdi), %xmm0
 ; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1],mem[1,2]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,2,2]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,2]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,3,2,2]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,0,2]
 ; CHECK-NEXT:    vxorps %xmm0, %xmm1, %xmm0
 ; CHECK-NEXT:    retq
   %A = load <4 x i32>, ptr %pA
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -1942,7 +1942,7 @@
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[1,3],ymm0[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 16(%rdx), %xmm1, %xmm1
 ; AVX-NEXT:    vpaddb (%rdx), %xmm0, %xmm0
@@ -4242,7 +4242,7 @@
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm2
 ; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[1,3],ymm2[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX-NEXT:    vpaddb 16(%rdx), %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -1562,7 +1562,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],mem[1,3],ymm0[4,4],mem[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 16(%rsi), %xmm1, %xmm1
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
@@ -3439,7 +3439,7 @@
 ; AVX-NEXT:    vmovaps 48(%rdi), %xmm0
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = mem[0,1,0,1]
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[1,3],ymm1[4,4],ymm0[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = mem[0,1,0,1]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX-NEXT:    vpaddb 16(%rsi), %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
@@ -1275,7 +1275,7 @@
 define <4 x float> @test_mm_permute_ps(<4 x float> %a0) nounwind {
 ; CHECK-LABEL: test_mm_permute_ps:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %res = shufflevector <4 x float> %a0, <4 x float> %a0, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %res
@@ -1284,7 +1284,7 @@
 define <4 x float> @test2_mm_permute_ps(<4 x float> %a0) nounwind {
 ; CHECK-LABEL: test2_mm_permute_ps:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %res = shufflevector <4 x float> %a0, <4 x float> %a0, <4 x i32> <i32 2, i32 1, i32 2, i32 3>
   ret <4 x float> %res
@@ -1293,7 +1293,7 @@
 define <8 x float> @test_mm256_permute_ps(<8 x float> %a0) nounwind {
 ; CHECK-LABEL: test_mm256_permute_ps:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %res = shufflevector <8 x float> %a0, <8 x float> %a0, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %res
@@ -1934,7 +1934,7 @@
 ; X86-LABEL: test_mm256_set1_epi32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; X86-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; X86-NEXT:    retl
 ;
@@ -2002,13 +2002,13 @@
 ; X86-LABEL: test_mm256_set1_ps:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; X86-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: test_mm256_set1_ps:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; X64-NEXT:    retq
   %res0 = insertelement <8 x float> undef, float %a0, i32 0
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll b/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
--- a/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-x86-upgrade.ll
@@ -879,13 +879,13 @@
 define <4 x float> @test_x86_avx_vpermil_ps(<4 x float> %a0) {
 ; AVX-LABEL: test_x86_avx_vpermil_ps:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps $7, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x07]
+; AVX-NEXT:    vshufps $7, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x07]
 ; AVX-NEXT:    # xmm0 = xmm0[3,1,0,0]
 ; AVX-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
 ;
 ; AVX512VL-LABEL: test_x86_avx_vpermil_ps:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps $7, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x07]
+; AVX512VL-NEXT:    vshufps $7, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x07]
 ; AVX512VL-NEXT:    # xmm0 = xmm0[3,1,0,0]
 ; AVX512VL-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx.vpermil.ps(<4 x float> %a0, i8 7) ; <<4 x float>> [#uses=1]
@@ -897,13 +897,13 @@
 define <8 x float> @test_x86_avx_vpermil_ps_256(<8 x float> %a0) {
 ; AVX-LABEL: test_x86_avx_vpermil_ps_256:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps $7, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x04,0xc0,0x07]
+; AVX-NEXT:    vshufps $7, %ymm0, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0xc6,0xc0,0x07]
 ; AVX-NEXT:    # ymm0 = ymm0[3,1,0,0,7,5,4,4]
 ; AVX-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
 ;
 ; AVX512VL-LABEL: test_x86_avx_vpermil_ps_256:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps $7, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x04,0xc0,0x07]
+; AVX512VL-NEXT:    vshufps $7, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0xc6,0xc0,0x07]
 ; AVX512VL-NEXT:    # ymm0 = ymm0[3,1,0,0,7,5,4,4]
 ; AVX512VL-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <8 x float> @llvm.x86.avx.vpermil.ps.256(<8 x float> %a0, i8 7) ; <<8 x float>> [#uses=1]
diff --git a/llvm/test/CodeGen/X86/avx-splat.ll b/llvm/test/CodeGen/X86/avx-splat.ll
--- a/llvm/test/CodeGen/X86/avx-splat.ll
+++ b/llvm/test/CodeGen/X86/avx-splat.ll
@@ -145,7 +145,7 @@
 define <8 x float> @funcG(<8 x float> %a) nounwind uwtable readnone ssp {
 ; CHECK-LABEL: funcG:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; CHECK-NEXT:    ret{{[l|q]}}
 entry:
@@ -157,7 +157,7 @@
 ; CHECK-LABEL: funcH:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,1,1,5,5,5,5]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1,1,1,5,5,5,5]
 ; CHECK-NEXT:    ret{{[l|q]}}
 entry:
   %shuffle = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
diff --git a/llvm/test/CodeGen/X86/avx-vbroadcast.ll b/llvm/test/CodeGen/X86/avx-vbroadcast.ll
--- a/llvm/test/CodeGen/X86/avx-vbroadcast.ll
+++ b/llvm/test/CodeGen/X86/avx-vbroadcast.ll
@@ -624,12 +624,12 @@
 define <4 x i32> @H(<4 x i32> %a) {
 ; X86-LABEL: H:
 ; X86:       ## %bb.0: ## %entry
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: H:
 ; X64:       ## %bb.0: ## %entry
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X64-NEXT:    retq
 entry:
   %x = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
@@ -982,7 +982,7 @@
 ; X64:       ## %bb.0: ## %bb
 ; X64-NEXT:    movdq2q %xmm0, %mm0
 ; X64-NEXT:    movq2dq %mm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; X64-NEXT:    retq
 bb:
   %tmp1 = bitcast x86_mmx %tmp to i64
diff --git a/llvm/test/CodeGen/X86/avx-vinsertf128.ll b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
--- a/llvm/test/CodeGen/X86/avx-vinsertf128.ll
+++ b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
@@ -29,7 +29,7 @@
 ; CHECK-NEXT:    vminpd %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vminsd %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vcvtsd2ss %xmm0, %xmm0, %xmm0
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; CHECK-NEXT:    vmovups %xmm0, (%rax)
 ; CHECK-NEXT:    retq
 allocas:
diff --git a/llvm/test/CodeGen/X86/avx-vperm2x128.ll b/llvm/test/CodeGen/X86/avx-vperm2x128.ll
--- a/llvm/test/CodeGen/X86/avx-vperm2x128.ll
+++ b/llvm/test/CodeGen/X86/avx-vperm2x128.ll
@@ -337,7 +337,7 @@
 ; ALL-LABEL: shuffle_v8f32_uu67ucuf:
 ; ALL:       # %bb.0: # %entry
 ; ALL-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
 ; ALL-NEXT:    retq
 entry:
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 6, i32 7, i32 undef, i32 12, i32 undef, i32 15>
diff --git a/llvm/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
@@ -2015,7 +2015,7 @@
 define <4 x i64> @test_mm256_shuffle_epi32(<4 x i64> %a0) {
 ; CHECK-LABEL: test_mm256_shuffle_epi32:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,3,0,0,7,7,4,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3,0,0,7,7,4,4]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <4 x i64> %a0 to <8 x i32>
   %shuf = shufflevector <8 x i32> %arg0, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 0, i32 0, i32 7, i32 7, i32 4, i32 4>
diff --git a/llvm/test/CodeGen/X86/avx512-cvt.ll b/llvm/test/CodeGen/X86/avx512-cvt.ll
--- a/llvm/test/CodeGen/X86/avx512-cvt.ll
+++ b/llvm/test/CodeGen/X86/avx512-cvt.ll
@@ -211,7 +211,7 @@
 define <4 x i64> @f32to4sl(<4 x float> %a) {
 ; NODQ-LABEL: f32to4sl:
 ; NODQ:       # %bb.0:
-; NODQ-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; NODQ-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; NODQ-NEXT:    vcvttss2si %xmm1, %rax
 ; NODQ-NEXT:    vmovq %rax, %xmm1
 ; NODQ-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1707,7 +1707,7 @@
 ; NOVL:       # %bb.0:
 ; NOVL-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; NOVL-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; NOVL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; NOVL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; NOVL-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; NOVL-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
@@ -987,7 +987,7 @@
 define <16 x float> @test_mm512_permute_ps(<16 x float> %a0) {
 ; CHECK-LABEL: test_mm512_permute_ps:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[2,0,0,0,6,4,4,4,10,8,8,8,14,12,12,12]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[2,0,0,0,6,4,4,4,10,8,8,8,14,12,12,12]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %res = shufflevector <16 x float> %a0, <16 x float> undef, <16 x i32> <i32 2, i32 0, i32 0, i32 0, i32 6, i32 4, i32 4, i32 4, i32 10, i32 8, i32 8, i32 8, i32 14, i32 12, i32 12, i32 12>
   ret <16 x float> %res
@@ -1128,7 +1128,7 @@
 define <8 x i64> @test_mm512_shuffle_epi32(<8 x i64> %a0) {
 ; CHECK-LABEL: test_mm512_shuffle_epi32:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[1,0,0,0,5,4,4,4,9,8,8,8,13,12,12,12]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[1,0,0,0,5,4,4,4,9,8,8,8,13,12,12,12]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <8 x i64> %a0 to <16 x i32>
   %res0 = shufflevector <16 x i32> %arg0, <16 x i32> undef, <16 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4, i32 9, i32 8, i32 8, i32 8, i32 13, i32 12, i32 12, i32 12>
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
@@ -1022,7 +1022,7 @@
 define <16 x float>@test_int_x86_avx512_vpermil_ps_512(<16 x float> %x0, <16 x float> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_vpermil_ps_512:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpermilps $22, %zmm0, %zmm0 ## encoding: [0x62,0xf3,0x7d,0x48,0x04,0xc0,0x16]
+; CHECK-NEXT:    vshufps $22, %zmm0, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0xc6,0xc0,0x16]
 ; CHECK-NEXT:    ## zmm0 = zmm0[2,1,1,0,6,5,5,4,10,9,9,8,14,13,13,12]
 ; CHECK-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
   %res = call <16 x float> @llvm.x86.avx512.mask.vpermil.ps.512(<16 x float> %x0, i32 22, <16 x float> %x2, i16 -1)
@@ -1072,7 +1072,7 @@
 define <16 x i32>@test_int_x86_avx512_pshuf_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_pshuf_d_512:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpermilps $3, %zmm0, %zmm0 ## encoding: [0x62,0xf3,0x7d,0x48,0x04,0xc0,0x03]
+; CHECK-NEXT:    vshufps $3, %zmm0, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0xc6,0xc0,0x03]
 ; CHECK-NEXT:    ## zmm0 = zmm0[3,0,0,0,7,4,4,4,11,8,8,8,15,12,12,12]
 ; CHECK-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.pshuf.d.512(<16 x i32> %x0, i32 3, <16 x i32> %x2, i16 -1)
diff --git a/llvm/test/CodeGen/X86/avx512-shuffles/in_lane_permute.ll b/llvm/test/CodeGen/X86/avx512-shuffles/in_lane_permute.ll
--- a/llvm/test/CodeGen/X86/avx512-shuffles/in_lane_permute.ll
+++ b/llvm/test/CodeGen/X86/avx512-shuffles/in_lane_permute.ll
@@ -6,7 +6,7 @@
 define <4 x float> @test_4xfloat_perm_mask0(<4 x float> %vec) {
 ; CHECK-LABEL: test_4xfloat_perm_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,1]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,3,1]
 ; CHECK-NEXT:    retq
   %res = shufflevector <4 x float> %vec, <4 x float> undef, <4 x i32> <i32 2, i32 1, i32 3, i32 1>
   ret <4 x float> %res
@@ -92,7 +92,7 @@
 define <4 x float> @test_4xfloat_perm_mask3(<4 x float> %vec) {
 ; CHECK-LABEL: test_4xfloat_perm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,2]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2,3,2]
 ; CHECK-NEXT:    retq
   %res = shufflevector <4 x float> %vec, <4 x float> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 2>
   ret <4 x float> %res
@@ -342,7 +342,7 @@
 define <8 x float> @test_8xfloat_perm_imm_mask3(<8 x float> %vec) {
 ; CHECK-LABEL: test_8xfloat_perm_imm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,1,0,6,6,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,1,0,6,6,5,4]
 ; CHECK-NEXT:    retq
   %res = shufflevector <8 x float> %vec, <8 x float> undef, <8 x i32> <i32 2, i32 2, i32 1, i32 0, i32 6, i32 6, i32 5, i32 4>
   ret <8 x float> %res
@@ -835,7 +835,7 @@
 define <16 x float> @test_16xfloat_perm_imm_mask3(<16 x float> %vec) {
 ; CHECK-LABEL: test_16xfloat_perm_imm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[1,1,0,2,5,5,4,6,9,9,8,10,13,13,12,14]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[1,1,0,2,5,5,4,6,9,9,8,10,13,13,12,14]
 ; CHECK-NEXT:    retq
   %res = shufflevector <16 x float> %vec, <16 x float> undef, <16 x i32> <i32 1, i32 1, i32 0, i32 2, i32 5, i32 5, i32 4, i32 6, i32 9, i32 9, i32 8, i32 10, i32 13, i32 13, i32 12, i32 14>
   ret <16 x float> %res
diff --git a/llvm/test/CodeGen/X86/avx512-shuffles/shuffle.ll b/llvm/test/CodeGen/X86/avx512-shuffles/shuffle.ll
--- a/llvm/test/CodeGen/X86/avx512-shuffles/shuffle.ll
+++ b/llvm/test/CodeGen/X86/avx512-shuffles/shuffle.ll
@@ -2091,7 +2091,7 @@
 define <4 x i32> @test_4xi32_perm_mask0(<4 x i32> %vec) {
 ; CHECK-LABEL: test_4xi32_perm_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,3,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,3,0]
 ; CHECK-NEXT:    retq
   %res = shufflevector <4 x i32> %vec, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 3, i32 0>
   ret <4 x i32> %res
@@ -2171,7 +2171,7 @@
 define <4 x i32> @test_4xi32_perm_mask3(<4 x i32> %vec) {
 ; CHECK-LABEL: test_4xi32_perm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,0,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,0,3]
 ; CHECK-NEXT:    retq
   %res = shufflevector <4 x i32> %vec, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 0, i32 3>
   ret <4 x i32> %res
@@ -2325,7 +2325,7 @@
 define <8 x i32> @test_8xi32_perm_mask0(<8 x i32> %vec) {
 ; CHECK-LABEL: test_8xi32_perm_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,1,0,6,7,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,1,0,6,7,5,4]
 ; CHECK-NEXT:    retq
   %res = shufflevector <8 x i32> %vec, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 1, i32 0, i32 6, i32 7, i32 5, i32 4>
   ret <8 x i32> %res
@@ -2405,7 +2405,7 @@
 define <8 x i32> @test_8xi32_perm_mask3(<8 x i32> %vec) {
 ; CHECK-LABEL: test_8xi32_perm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,1,0,5,7,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3,1,0,5,7,5,4]
 ; CHECK-NEXT:    retq
   %res = shufflevector <8 x i32> %vec, <8 x i32> undef, <8 x i32> <i32 1, i32 3, i32 1, i32 0, i32 5, i32 7, i32 5, i32 4>
   ret <8 x i32> %res
@@ -2559,7 +2559,7 @@
 define <16 x i32> @test_16xi32_perm_mask0(<16 x i32> %vec) {
 ; CHECK-LABEL: test_16xi32_perm_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,1,3,0,7,5,7,4,11,9,11,8,15,13,15,12]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,1,3,0,7,5,7,4,11,9,11,8,15,13,15,12]
 ; CHECK-NEXT:    retq
   %res = shufflevector <16 x i32> %vec, <16 x i32> undef, <16 x i32> <i32 3, i32 1, i32 3, i32 0, i32 7, i32 5, i32 7, i32 4, i32 11, i32 9, i32 11, i32 8, i32 15, i32 13, i32 15, i32 12>
   ret <16 x i32> %res
@@ -2639,7 +2639,7 @@
 define <16 x i32> @test_16xi32_perm_mask3(<16 x i32> %vec) {
 ; CHECK-LABEL: test_16xi32_perm_mask3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,2,0,3,7,6,4,7,11,10,8,11,15,14,12,15]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,2,0,3,7,6,4,7,11,10,8,11,15,14,12,15]
 ; CHECK-NEXT:    retq
   %res = shufflevector <16 x i32> %vec, <16 x i32> undef, <16 x i32> <i32 3, i32 2, i32 0, i32 3, i32 7, i32 6, i32 4, i32 7, i32 11, i32 10, i32 8, i32 11, i32 15, i32 14, i32 12, i32 15>
   ret <16 x i32> %res
diff --git a/llvm/test/CodeGen/X86/avx512-trunc.ll b/llvm/test/CodeGen/X86/avx512-trunc.ll
--- a/llvm/test/CodeGen/X86/avx512-trunc.ll
+++ b/llvm/test/CodeGen/X86/avx512-trunc.ll
@@ -270,7 +270,7 @@
 define <2 x i32> @trunc_qd_128(<2 x i64> %i) #0 {
 ; ALL-LABEL: trunc_qd_128:
 ; ALL:       ## %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; ALL-NEXT:    retq
   %x = trunc <2 x i64> %i to <2 x i32>
   ret <2 x i32> %x
@@ -279,7 +279,7 @@
 define void @trunc_qd_128_mem(<2 x i64> %i, ptr %res) #0 {
 ; KNL-LABEL: trunc_qd_128_mem:
 ; KNL:       ## %bb.0:
-; KNL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; KNL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; KNL-NEXT:    vmovlps %xmm0, (%rdi)
 ; KNL-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
--- a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
+++ b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll
@@ -1211,7 +1211,7 @@
 ; AVX512-LABEL: test46:
 ; AVX512:       ## %bb.0:
 ; AVX512-NEXT:    vcmpeqps %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc2,0xc1,0x00]
-; AVX512-NEXT:    vpermilps $212, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xd4]
+; AVX512-NEXT:    vshufps $212, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xd4]
 ; AVX512-NEXT:    ## xmm0 = xmm0[0,1,1,3]
 ; AVX512-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0x54,0x05,A,A,A,A]
 ; AVX512-NEXT:    ## fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
diff --git a/llvm/test/CodeGen/X86/avx512fp16-mov.ll b/llvm/test/CodeGen/X86/avx512fp16-mov.ll
--- a/llvm/test/CodeGen/X86/avx512fp16-mov.ll
+++ b/llvm/test/CodeGen/X86/avx512fp16-mov.ll
@@ -1334,7 +1334,7 @@
 define half @extract_f16_6(<8 x half> %x) {
 ; CHECK-LABEL: extract_f16_6:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; CHECK-NEXT:    ret{{[l|q]}}
    %res = extractelement <8 x half> %x, i32 6
    ret half %res
@@ -1606,14 +1606,14 @@
 define void @extract_store_f16_6(<8 x half> %x, ptr %y) {
 ; X64-LABEL: extract_store_f16_6:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; X64-NEXT:    vmovsh %xmm0, (%rdi)
 ; X64-NEXT:    retq
 ;
 ; X86-LABEL: extract_store_f16_6:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; X86-NEXT:    vmovsh %xmm0, (%eax)
 ; X86-NEXT:    retl
    %res = extractelement <8 x half> %x, i32 6
diff --git a/llvm/test/CodeGen/X86/avx512fp16-mscatter.ll b/llvm/test/CodeGen/X86/avx512fp16-mscatter.ll
--- a/llvm/test/CodeGen/X86/avx512fp16-mscatter.ll
+++ b/llvm/test/CodeGen/X86/avx512fp16-mscatter.ll
@@ -31,7 +31,7 @@
 ; CHECK-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; CHECK-NEXT:    vpextrq $1, %xmm4, %rax
 ; CHECK-NEXT:    vmovsh %xmm3, (%rax)
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[3,3,3,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; CHECK-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
 ; CHECK-NEXT:    vmovq %xmm0, %rax
 ; CHECK-NEXT:    vmovsh %xmm3, (%rax)
@@ -58,7 +58,7 @@
 ; CHECK-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; CHECK-NEXT:    vpextrq $1, %xmm3, %rax
 ; CHECK-NEXT:    vmovsh %xmm1, (%rax)
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; CHECK-NEXT:    vextracti32x4 $3, %zmm2, %xmm2
 ; CHECK-NEXT:    vmovq %xmm2, %rax
 ; CHECK-NEXT:    vmovsh %xmm1, (%rax)
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
@@ -891,7 +891,7 @@
 define <8 x float>@test_int_x86_avx512_vpermil_ps_256(<8 x float> %x0, <8 x float> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_vpermil_ps_256:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps $22, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x04,0xc0,0x16]
+; CHECK-NEXT:    vshufps $22, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0xc6,0xc0,0x16]
 ; CHECK-NEXT:    # ymm0 = ymm0[2,1,1,0,6,5,5,4]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 -1)
@@ -943,7 +943,7 @@
 define <4 x float>@test_int_x86_avx512_vpermil_ps_128(<4 x float> %x0, <4 x float> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_vpermil_ps_128:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps $22, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x16]
+; CHECK-NEXT:    vshufps $22, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x16]
 ; CHECK-NEXT:    # xmm0 = xmm0[2,1,1,0]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 -1)
@@ -1967,7 +1967,7 @@
 define <4 x i32>@test_int_x86_avx512_pshuf_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_pshuf_d_128:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps $3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x03]
+; CHECK-NEXT:    vshufps $3, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x03]
 ; CHECK-NEXT:    # xmm0 = xmm0[3,0,0,0]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.pshuf.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
@@ -2019,7 +2019,7 @@
 define <8 x i32>@test_int_x86_avx512_pshuf_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_pshuf_d_256:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps $3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x04,0xc0,0x03]
+; CHECK-NEXT:    vshufps $3, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0xc6,0xc0,0x03]
 ; CHECK-NEXT:    # ymm0 = ymm0[3,0,0,0,7,4,4,4]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.pshuf.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
diff --git a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
--- a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
+++ b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
@@ -352,7 +352,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9241421688590303745,9241421688590303745]
@@ -617,7 +617,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [9241421688590303745,9241421688590303745,9241421688590303745,9241421688590303745]
 ; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
@@ -627,7 +627,7 @@
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,5,5]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,7,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
 ; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm3, %xmm3
diff --git a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
--- a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
+++ b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
@@ -451,7 +451,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9241421688590303745,9241421688590303745]
@@ -806,7 +806,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [9241421688590303745,9241421688590303745,9241421688590303745,9241421688590303745]
 ; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
@@ -821,7 +821,7 @@
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,5,5]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,7,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
 ; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm3, %xmm3
diff --git a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
--- a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
+++ b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
@@ -246,7 +246,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9241421688590303745,9241421688590303745]
diff --git a/llvm/test/CodeGen/X86/buildvec-extract.ll b/llvm/test/CodeGen/X86/buildvec-extract.ll
--- a/llvm/test/CodeGen/X86/buildvec-extract.ll
+++ b/llvm/test/CodeGen/X86/buildvec-extract.ll
@@ -83,7 +83,7 @@
 ;
 ; AVX-LABEL: extract1_i32_zext_insert0_i64_zero:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
 ; AVX-NEXT:    retq
@@ -128,7 +128,7 @@
 ;
 ; AVX-LABEL: extract2_i32_zext_insert0_i64_zero:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
 ; AVX-NEXT:    retq
@@ -187,7 +187,7 @@
 ;
 ; AVX-LABEL: extract0_i32_zext_insert1_i64_undef:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/combine-and.ll b/llvm/test/CodeGen/X86/combine-and.ll
--- a/llvm/test/CodeGen/X86/combine-and.ll
+++ b/llvm/test/CodeGen/X86/combine-and.ll
@@ -624,7 +624,7 @@
 ; AVX1-LABEL: neg_scalar_broadcast_v8i64:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,3]
@@ -790,7 +790,7 @@
 ;
 ; AVX1-LABEL: casted_neg_scalar_broadcast_v2i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vandnps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/combine-concatvectors.ll b/llvm/test/CodeGen/X86/combine-concatvectors.ll
--- a/llvm/test/CodeGen/X86/combine-concatvectors.ll
+++ b/llvm/test/CodeGen/X86/combine-concatvectors.ll
@@ -91,7 +91,7 @@
 ; AVX1-NEXT:    vmovaps (%rdi), %ymm0
 ; AVX1-NEXT:    vmovaps (%rsi), %ymm1
 ; AVX1-NEXT:    vmovaps (%rdx), %ymm2
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4],ymm2[5,6],ymm0[7]
diff --git a/llvm/test/CodeGen/X86/copy-low-subvec-elt-to-high-subvec-elt.ll b/llvm/test/CodeGen/X86/copy-low-subvec-elt-to-high-subvec-elt.ll
--- a/llvm/test/CodeGen/X86/copy-low-subvec-elt-to-high-subvec-elt.ll
+++ b/llvm/test/CodeGen/X86/copy-low-subvec-elt-to-high-subvec-elt.ll
@@ -47,7 +47,7 @@
 define <4 x float> @vec128_eltty_float_source_subvec_0_target_subvec_mask_1_unary(<4 x float> %x) nounwind {
 ; CHECK-LABEL: vec128_eltty_float_source_subvec_0_target_subvec_mask_1_unary:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; CHECK-NEXT:    retq
   %r = shufflevector <4 x float> %x, <4 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 0>
   ret <4 x float> %r
@@ -65,7 +65,7 @@
 define <4 x i32> @vec128_eltty_i32_source_subvec_0_target_subvec_mask_1_unary(<4 x i32> %x) nounwind {
 ; CHECK-LABEL: vec128_eltty_i32_source_subvec_0_target_subvec_mask_1_unary:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; CHECK-NEXT:    retq
   %r = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 0>
   ret <4 x i32> %r
@@ -322,7 +322,7 @@
 define <4 x i64> @vec256_eltty_i64_source_subvec_1_target_subvec_mask_2_binary(<4 x i64> %x, <4 x i64> %y) nounwind {
 ; CHECK-LABEL: vec256_eltty_i64_source_subvec_1_target_subvec_mask_2_binary:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
 ; CHECK-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; CHECK-NEXT:    retq
   %r = shufflevector <4 x i64> %x, <4 x i64> %y, <4 x i32> <i32 0, i32 1, i32 2, i32 6>
@@ -448,7 +448,7 @@
 ;
 ; CHECK-FAST-LABEL: vec256_eltty_float_source_subvec_1_target_subvec_mask_2_binary:
 ; CHECK-FAST:       # %bb.0:
-; CHECK-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
+; CHECK-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
 ; CHECK-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
 ; CHECK-FAST-NEXT:    retq
   %r = shufflevector <8 x float> %x, <8 x float> %y, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 12>
@@ -577,7 +577,7 @@
 ;
 ; CHECK-FAST-LABEL: vec256_eltty_i32_source_subvec_1_target_subvec_mask_2_binary:
 ; CHECK-FAST:       # %bb.0:
-; CHECK-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
+; CHECK-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
 ; CHECK-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
 ; CHECK-FAST-NEXT:    retq
   %r = shufflevector <8 x i32> %x, <8 x i32> %y, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 12>
@@ -598,7 +598,7 @@
 ; CHECK-SLOW-LABEL: vec256_eltty_i32_source_subvec_1_target_subvec_mask_3_binary:
 ; CHECK-SLOW:       # %bb.0:
 ; CHECK-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; CHECK-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
+; CHECK-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
 ; CHECK-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
 ; CHECK-SLOW-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/extract-concat.ll b/llvm/test/CodeGen/X86/extract-concat.ll
--- a/llvm/test/CodeGen/X86/extract-concat.ll
+++ b/llvm/test/CodeGen/X86/extract-concat.ll
@@ -68,9 +68,9 @@
 ;
 ; AVX1-LABEL: catcat:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm4
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    vmovddup {{.*#+}} ymm2 = ymm0[0,0,2,2]
diff --git a/llvm/test/CodeGen/X86/extract-store.ll b/llvm/test/CodeGen/X86/extract-store.ll
--- a/llvm/test/CodeGen/X86/extract-store.ll
+++ b/llvm/test/CodeGen/X86/extract-store.ll
@@ -326,7 +326,7 @@
 ; AVX-X86-LABEL: extract_i64_1:
 ; AVX-X86:       # %bb.0:
 ; AVX-X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX-X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; AVX-X86-NEXT:    vmovlps %xmm0, (%eax)
 ; AVX-X86-NEXT:    retl
 ;
diff --git a/llvm/test/CodeGen/X86/fdiv-combine-vec.ll b/llvm/test/CodeGen/X86/fdiv-combine-vec.ll
--- a/llvm/test/CodeGen/X86/fdiv-combine-vec.ll
+++ b/llvm/test/CodeGen/X86/fdiv-combine-vec.ll
@@ -61,7 +61,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vdivss %xmm1, %xmm2, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %vy = insertelement <4 x float> undef, float %y, i32 0
@@ -84,7 +84,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vdivss %xmm1, %xmm2, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX-NEXT:    vmulps %ymm1, %ymm0, %ymm0
 ; AVX-NEXT:    retq
@@ -115,7 +115,7 @@
 ; AVX-NEXT:    vsubss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm2, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm2, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %vy = insertelement <4 x float> undef, float %y, i32 0
@@ -146,7 +146,7 @@
 ; AVX-NEXT:    vsubss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm2, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm2, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX-NEXT:    vmulps %ymm1, %ymm0, %ymm0
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/fmaddsub-combine.ll b/llvm/test/CodeGen/X86/fmaddsub-combine.ll
--- a/llvm/test/CodeGen/X86/fmaddsub-combine.ll
+++ b/llvm/test/CodeGen/X86/fmaddsub-combine.ll
@@ -490,8 +490,8 @@
 ; NOFMA-NEXT:    vsubss %xmm5, %xmm4, %xmm4
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[2,3]
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm3[0],xmm1[3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm2, %xmm0, %xmm0
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
 ; NOFMA-NEXT:    retq
@@ -578,8 +578,8 @@
 ; NOFMA-NEXT:    vsubss %xmm9, %xmm8, %xmm8
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[2,3]
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm3[0],xmm1[3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm2, %xmm0, %xmm0
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
 ; NOFMA-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm4[1,1,3,3]
@@ -587,8 +587,8 @@
 ; NOFMA-NEXT:    vsubss %xmm2, %xmm1, %xmm1
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[2,3]
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm7[0],xmm1[3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm2 = xmm4[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm3 = xmm5[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm3 = xmm5[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm3, %xmm2, %xmm2
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[0]
 ; NOFMA-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
@@ -719,12 +719,12 @@
 ; NOFMA-NEXT:    vsubss %xmm15, %xmm14, %xmm14
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[2,3]
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0],xmm2[3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm3, %xmm0, %xmm0
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm2 = xmm6[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm3 = xmm7[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm2 = xmm6[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm3 = xmm7[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm3, %xmm2, %xmm2
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm2 = xmm8[0,1,2],xmm2[0]
 ; NOFMA-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
@@ -732,16 +732,16 @@
 ; NOFMA-NEXT:    vsubss %xmm4, %xmm3, %xmm3
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm3 = xmm9[0],xmm3[0],xmm9[2,3]
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1],xmm10[0],xmm3[3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm4 = xmm5[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm4 = xmm5[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm4, %xmm1, %xmm1
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[0]
 ; NOFMA-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm11[1,1,3,3]
 ; NOFMA-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm13[1,1,3,3]
 ; NOFMA-NEXT:    vsubss %xmm4, %xmm3, %xmm3
 ; NOFMA-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,0],xmm12[0,0]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm4 = xmm11[3,3,3,3]
-; NOFMA-NEXT:    vpermilps {{.*#+}} xmm5 = xmm13[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm4 = xmm11[3,3,3,3]
+; NOFMA-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[3,3,3,3]
 ; NOFMA-NEXT:    vsubss %xmm5, %xmm4, %xmm4
 ; NOFMA-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
 ; NOFMA-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
diff --git a/llvm/test/CodeGen/X86/haddsub-2.ll b/llvm/test/CodeGen/X86/haddsub-2.ll
--- a/llvm/test/CodeGen/X86/haddsub-2.ll
+++ b/llvm/test/CodeGen/X86/haddsub-2.ll
@@ -907,12 +907,12 @@
 ; AVX-LABEL: not_a_hsub_2:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vsubss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX-NEXT:    vsubss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vsubss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
diff --git a/llvm/test/CodeGen/X86/haddsub-4.ll b/llvm/test/CodeGen/X86/haddsub-4.ll
--- a/llvm/test/CodeGen/X86/haddsub-4.ll
+++ b/llvm/test/CodeGen/X86/haddsub-4.ll
@@ -66,13 +66,13 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: hadd_reverse_v8f32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    retq
   %lhs = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 7, i32 5, i32 15, i32 13, i32 3, i32 1, i32 11, i32 9>
@@ -96,17 +96,17 @@
 ; AVX1-LABEL: hadd_reverse2_v8f32:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: hadd_reverse2_v8f32:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,2,1,0,7,6,5,4]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,3,0,1]
 ; AVX2-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
@@ -133,13 +133,13 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vhaddps %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: hadd_reverse3_v8f32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vhaddps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    retq
   %shuf0 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 2, i32 8, i32 10, i32 4, i32 6, i32 12, i32 14>
@@ -337,17 +337,17 @@
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm3[2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX1-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: hadd_reverse_v16f32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vhaddps %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[2,0,3,1]
 ; AVX2-NEXT:    vhaddps %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,0,3,1]
 ; AVX2-NEXT:    vmovaps %ymm3, %ymm0
 ; AVX2-NEXT:    retq
@@ -380,20 +380,20 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vhaddps %ymm3, %ymm1, %ymm1
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    vhaddps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    vmovaps %ymm3, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: hadd_reverse2_v16f32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vhaddps %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[2,3,0,1]
 ; AVX2-NEXT:    vhaddps %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vmovaps %ymm3, %ymm0
 ; AVX2-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/haddsub-undef.ll b/llvm/test/CodeGen/X86/haddsub-undef.ll
--- a/llvm/test/CodeGen/X86/haddsub-undef.ll
+++ b/llvm/test/CodeGen/X86/haddsub-undef.ll
@@ -208,7 +208,7 @@
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
 ; AVX-SLOW-NEXT:    retq
@@ -216,7 +216,7 @@
 ; AVX-FAST-LABEL: test8_undef:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; AVX-FAST-NEXT:    retq
   %vecext = extractelement <4 x float> %a, i32 0
   %vecext1 = extractelement <4 x float> %a, i32 1
@@ -377,7 +377,7 @@
 ; AVX512-SLOW-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; AVX512-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX512-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; AVX512-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX512-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512-SLOW-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX512-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
 ; AVX512-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
@@ -385,7 +385,7 @@
 ; AVX512-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm2
 ; AVX512-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0],xmm1[3]
 ; AVX512-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; AVX512-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-SLOW-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX512-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
 ; AVX512-SLOW-NEXT:    retq
@@ -520,15 +520,15 @@
 ;
 ; AVX-SLOW-LABEL: add_ps_030:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,2,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0,2,3]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: add_ps_030:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; AVX-FAST-NEXT:    retq
   %l = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 2>
   %r = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 1, i32 3>
@@ -591,7 +591,7 @@
 ; AVX-LABEL: add_ps_016:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vhaddps %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,3,3]
 ; AVX-NEXT:    retq
   %3 = shufflevector <4 x float> %1, <4 x float> %0, <2 x i32> <i32 0, i32 6>
   %4 = shufflevector <4 x float> %1, <4 x float> %0, <2 x i32> <i32 1, i32 7>
@@ -620,8 +620,8 @@
 ;
 ; AVX-SLOW-LABEL: add_ps_017:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
@@ -943,8 +943,8 @@
 ;
 ; AVX-SLOW-LABEL: PR45747_1:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,2,2,2]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,2,2,2]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
@@ -976,7 +976,7 @@
 ; AVX-SLOW-LABEL: PR45747_2:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,1,1]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,1,1]
 ; AVX-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/haddsub.ll b/llvm/test/CodeGen/X86/haddsub.ll
--- a/llvm/test/CodeGen/X86/haddsub.ll
+++ b/llvm/test/CodeGen/X86/haddsub.ll
@@ -502,7 +502,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v4f32_fadd_f32:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
@@ -563,7 +563,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v4f32_fadd_f32_commute:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
@@ -683,7 +683,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v4f32_fsub_f32:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vsubss %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
@@ -729,7 +729,7 @@
 ; AVX-LABEL: extract_extract23_v4f32_fsub_f32_commute:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vsubss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %x0 = extractelement <4 x float> %x, i32 2
@@ -837,7 +837,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v8f32_fadd_f32:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
@@ -873,7 +873,7 @@
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
@@ -939,7 +939,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v8f32_fadd_f32_commute:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
@@ -975,7 +975,7 @@
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
@@ -1172,7 +1172,7 @@
 ; AVX-SLOW-LABEL: extract_extract23_v8f32_fsub_f32:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vsubss %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll b/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll
--- a/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll
+++ b/llvm/test/CodeGen/X86/horizontal-reduce-smax.ll
@@ -457,7 +457,7 @@
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
@@ -470,7 +470,7 @@
 ; X86-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
@@ -531,7 +531,7 @@
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vmovq %xmm0, %rax
@@ -543,7 +543,7 @@
 ; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX2-NEXT:    vmovq %xmm0, %rax
@@ -1122,7 +1122,7 @@
 ; X86-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; X86-AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm2, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
@@ -1137,7 +1137,7 @@
 ; X86-AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
@@ -1240,7 +1240,7 @@
 ; X64-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; X64-AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm2, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vmovq %xmm0, %rax
@@ -1254,7 +1254,7 @@
 ; X64-AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX2-NEXT:    vmovq %xmm0, %rax
diff --git a/llvm/test/CodeGen/X86/horizontal-reduce-smin.ll b/llvm/test/CodeGen/X86/horizontal-reduce-smin.ll
--- a/llvm/test/CodeGen/X86/horizontal-reduce-smin.ll
+++ b/llvm/test/CodeGen/X86/horizontal-reduce-smin.ll
@@ -460,7 +460,7 @@
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
@@ -473,7 +473,7 @@
 ; X86-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
@@ -535,7 +535,7 @@
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vmovq %xmm0, %rax
@@ -547,7 +547,7 @@
 ; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX2-NEXT:    vmovq %xmm0, %rax
@@ -1126,7 +1126,7 @@
 ; X86-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; X86-AVX1-NEXT:    vblendvpd %xmm1, %xmm2, %xmm0, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX1-NEXT:    vmovd %xmm0, %eax
@@ -1141,7 +1141,7 @@
 ; X86-AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X86-AVX2-NEXT:    vmovd %xmm0, %eax
@@ -1244,7 +1244,7 @@
 ; X64-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; X64-AVX1-NEXT:    vblendvpd %xmm1, %xmm2, %xmm0, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX1-NEXT:    vmovq %xmm0, %rax
@@ -1258,7 +1258,7 @@
 ; X64-AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; X64-AVX2-NEXT:    vmovq %xmm0, %rax
diff --git a/llvm/test/CodeGen/X86/horizontal-reduce-umax.ll b/llvm/test/CodeGen/X86/horizontal-reduce-umax.ll
--- a/llvm/test/CodeGen/X86/horizontal-reduce-umax.ll
+++ b/llvm/test/CodeGen/X86/horizontal-reduce-umax.ll
@@ -52,7 +52,7 @@
 ;
 ; X86-AVX1-LABEL: test_reduce_v2i64:
 ; X86-AVX1:       ## %bb.0:
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
 ; X86-AVX1-NEXT:    ## xmm2 = mem[0,0]
 ; X86-AVX1-NEXT:    vxorps %xmm2, %xmm0, %xmm3
@@ -531,7 +531,7 @@
 ; X86-AVX1-NEXT:    vxorps %xmm2, %xmm0, %xmm4
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X86-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -549,7 +549,7 @@
 ; X86-AVX2-NEXT:    vpxor %xmm2, %xmm0, %xmm4
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X86-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -621,7 +621,7 @@
 ; X64-AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm4
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X64-AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -638,7 +638,7 @@
 ; X64-AVX2-NEXT:    vpxor %xmm2, %xmm0, %xmm4
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X64-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -1236,7 +1236,7 @@
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm3, %xmm4
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; X86-AVX1-NEXT:    vblendvpd %xmm1, %xmm3, %xmm0, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -1258,7 +1258,7 @@
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm4
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X86-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -1381,7 +1381,7 @@
 ; X64-AVX1-NEXT:    vxorpd %xmm2, %xmm3, %xmm4
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; X64-AVX1-NEXT:    vblendvpd %xmm1, %xmm3, %xmm0, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -1402,7 +1402,7 @@
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm4
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X64-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
diff --git a/llvm/test/CodeGen/X86/horizontal-reduce-umin.ll b/llvm/test/CodeGen/X86/horizontal-reduce-umin.ll
--- a/llvm/test/CodeGen/X86/horizontal-reduce-umin.ll
+++ b/llvm/test/CodeGen/X86/horizontal-reduce-umin.ll
@@ -53,7 +53,7 @@
 ;
 ; X86-AVX1-LABEL: test_reduce_v2i64:
 ; X86-AVX1:       ## %bb.0:
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808]
 ; X86-AVX1-NEXT:    ## xmm2 = mem[0,0]
 ; X86-AVX1-NEXT:    vxorps %xmm2, %xmm0, %xmm3
@@ -475,7 +475,7 @@
 ; X86-AVX1-NEXT:    vxorps %xmm1, %xmm3, %xmm4
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm4, %xmm2
 ; X86-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm3, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vxorpd %xmm1, %xmm0, %xmm3
 ; X86-AVX1-NEXT:    vxorpd %xmm1, %xmm2, %xmm1
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
@@ -493,7 +493,7 @@
 ; X86-AVX2-NEXT:    vpxor %xmm2, %xmm1, %xmm4
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X86-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -567,7 +567,7 @@
 ; X64-AVX1-NEXT:    vpxor %xmm1, %xmm3, %xmm4
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm4, %xmm2
 ; X64-AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm3, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vxorpd %xmm1, %xmm0, %xmm3
 ; X64-AVX1-NEXT:    vxorpd %xmm1, %xmm2, %xmm1
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
@@ -584,7 +584,7 @@
 ; X64-AVX2-NEXT:    vpxor %xmm2, %xmm1, %xmm4
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X64-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -1152,7 +1152,7 @@
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm3, %xmm4
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; X86-AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm3, %xmm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -1174,7 +1174,7 @@
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm4
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X86-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X86-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X86-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -1299,7 +1299,7 @@
 ; X64-AVX1-NEXT:    vxorpd %xmm3, %xmm2, %xmm4
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; X64-AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm2, %xmm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX1-NEXT:    vxorpd %xmm3, %xmm0, %xmm2
 ; X64-AVX1-NEXT:    vxorpd %xmm3, %xmm1, %xmm3
 ; X64-AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -1320,7 +1320,7 @@
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm4
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; X64-AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; X64-AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; X64-AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/horizontal-shuffle-2.ll b/llvm/test/CodeGen/X86/horizontal-shuffle-2.ll
--- a/llvm/test/CodeGen/X86/horizontal-shuffle-2.ll
+++ b/llvm/test/CodeGen/X86/horizontal-shuffle-2.ll
@@ -14,7 +14,7 @@
 ; AVX-LABEL: test_unpacklo_hadd_v4f32:
 ; AVX:       ## %bb.0:
 ; AVX-NEXT:    vhaddps %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    ret{{[l|q]}}
   %5 = tail call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %0, <4 x float> %1) #4
   %6 = tail call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %2, <4 x float> %3) #4
@@ -33,7 +33,7 @@
 ; AVX-LABEL: test_unpackhi_hadd_v4f32:
 ; AVX:       ## %bb.0:
 ; AVX-NEXT:    vhaddps %xmm3, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    ret{{[l|q]}}
   %5 = tail call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %0, <4 x float> %1) #4
   %6 = tail call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %2, <4 x float> %3) #4
@@ -51,7 +51,7 @@
 ; AVX-LABEL: test_unpacklo_hsub_v4f32:
 ; AVX:       ## %bb.0:
 ; AVX-NEXT:    vhsubps %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    ret{{[l|q]}}
   %5 = tail call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %0, <4 x float> %1) #4
   %6 = tail call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %2, <4 x float> %3) #4
@@ -70,7 +70,7 @@
 ; AVX-LABEL: test_unpackhi_hsub_v4f32:
 ; AVX:       ## %bb.0:
 ; AVX-NEXT:    vhsubps %xmm3, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    ret{{[l|q]}}
   %5 = tail call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %0, <4 x float> %1) #4
   %6 = tail call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %2, <4 x float> %3) #4
@@ -164,7 +164,7 @@
 ; AVX-LABEL: test_unpacklo_hadd_v4f32_unary:
 ; AVX:       ## %bb.0:
 ; AVX-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX-NEXT:    ret{{[l|q]}}
   %2 = tail call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %0, <4 x float> %0) #4
   %3 = shufflevector <4 x float> %2, <4 x float> %2, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
diff --git a/llvm/test/CodeGen/X86/horizontal-shuffle-3.ll b/llvm/test/CodeGen/X86/horizontal-shuffle-3.ll
--- a/llvm/test/CodeGen/X86/horizontal-shuffle-3.ll
+++ b/llvm/test/CodeGen/X86/horizontal-shuffle-3.ll
@@ -6,7 +6,7 @@
 ; CHECK-LABEL: test_unpacklo_hadd_v8f32:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vhaddps %ymm2, %ymm0, %ymm0
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %5 = tail call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %0, <8 x float> %1) #4
   %6 = tail call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %2, <8 x float> %3) #4
@@ -18,7 +18,7 @@
 ; CHECK-LABEL: test_unpackhi_hadd_v8f32:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vhaddps %ymm3, %ymm1, %ymm0
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %5 = tail call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %0, <8 x float> %1) #4
   %6 = tail call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %2, <8 x float> %3) #4
@@ -30,7 +30,7 @@
 ; CHECK-LABEL: test_unpacklo_hsub_v8f32:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vhsubps %ymm2, %ymm0, %ymm0
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %5 = tail call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %0, <8 x float> %1) #4
   %6 = tail call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %2, <8 x float> %3) #4
@@ -42,7 +42,7 @@
 ; CHECK-LABEL: test_unpackhi_hsub_v8f32:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vhsubps %ymm3, %ymm1, %ymm0
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %5 = tail call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %0, <8 x float> %1) #4
   %6 = tail call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %2, <8 x float> %3) #4
diff --git a/llvm/test/CodeGen/X86/horizontal-shuffle-4.ll b/llvm/test/CodeGen/X86/horizontal-shuffle-4.ll
--- a/llvm/test/CodeGen/X86/horizontal-shuffle-4.ll
+++ b/llvm/test/CodeGen/X86/horizontal-shuffle-4.ll
@@ -61,7 +61,7 @@
 ; CHECK-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vhaddps %ymm3, %ymm2, %ymm1
 ; CHECK-NEXT:    vhaddps %ymm1, %ymm0, %ymm0
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %a2, <8 x float> %a3)
diff --git a/llvm/test/CodeGen/X86/horizontal-sum.ll b/llvm/test/CodeGen/X86/horizontal-sum.ll
--- a/llvm/test/CodeGen/X86/horizontal-sum.ll
+++ b/llvm/test/CodeGen/X86/horizontal-sum.ll
@@ -218,8 +218,8 @@
 ; AVX1-SLOW-LABEL: pair_sum_v8f32_v4f32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,3,1,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vhaddps %xmm4, %xmm4, %xmm1
 ; AVX1-SLOW-NEXT:    vhaddps %xmm5, %xmm5, %xmm4
@@ -262,8 +262,8 @@
 ; AVX2-SLOW-LABEL: pair_sum_v8f32_v4f32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,3,1,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX2-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vhaddps %xmm4, %xmm4, %xmm1
 ; AVX2-SLOW-NEXT:    vhaddps %xmm5, %xmm5, %xmm4
@@ -893,27 +893,27 @@
 ; AVX-SLOW-NEXT:    vaddss %xmm4, %xmm0, %xmm4
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm0[1,0]
 ; AVX-SLOW-NEXT:    vaddss %xmm5, %xmm4, %xmm4
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm0, %xmm4, %xmm0
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm4, %xmm1, %xmm4
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
 ; AVX-SLOW-NEXT:    vaddss %xmm5, %xmm4, %xmm4
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm4, %xmm1
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm2, %xmm1
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
 ; AVX-SLOW-NEXT:    vaddss %xmm4, %xmm1, %xmm1
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm3[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm1, %xmm3, %xmm1
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[3,3,3,3]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
 ; AVX-SLOW-NEXT:    retq
@@ -923,24 +923,24 @@
 ; AVX-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm4
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm0[1,0]
 ; AVX-FAST-NEXT:    vaddss %xmm5, %xmm4, %xmm4
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-FAST-NEXT:    vaddss %xmm0, %xmm4, %xmm0
 ; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm1, %xmm4
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
 ; AVX-FAST-NEXT:    vaddss %xmm5, %xmm4, %xmm4
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-FAST-NEXT:    vaddss %xmm1, %xmm4, %xmm1
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
 ; AVX-FAST-NEXT:    vhaddps %xmm2, %xmm2, %xmm1
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
 ; AVX-FAST-NEXT:    vaddss %xmm4, %xmm1, %xmm1
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
 ; AVX-FAST-NEXT:    vhaddps %xmm3, %xmm3, %xmm1
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[3,3,3,3]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
 ; AVX-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
 ; AVX-FAST-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/i64-to-float.ll b/llvm/test/CodeGen/X86/i64-to-float.ll
--- a/llvm/test/CodeGen/X86/i64-to-float.ll
+++ b/llvm/test/CodeGen/X86/i64-to-float.ll
@@ -299,7 +299,7 @@
 ; X86-AVX-NEXT:    # xmm1 = mem[0,0]
 ; X86-AVX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X86-AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X86-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X86-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; X86-AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; X86-AVX-NEXT:    retl
 ;
@@ -359,7 +359,7 @@
 ; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255]
 ; X64-AVX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; X64-AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; X64-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; X64-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; X64-AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; X64-AVX-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/insertelement-var-index.ll b/llvm/test/CodeGen/X86/insertelement-var-index.ll
--- a/llvm/test/CodeGen/X86/insertelement-var-index.ll
+++ b/llvm/test/CodeGen/X86/insertelement-var-index.ll
@@ -187,7 +187,7 @@
 ;
 ; AVX1-LABEL: arg_f32_v4f32_undef:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: arg_f32_v4f32_undef:
@@ -578,7 +578,7 @@
 ;
 ; AVX1-LABEL: arg_f32_v8f32_undef:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -1053,7 +1053,7 @@
 ;
 ; AVX1-LABEL: arg_f32_v4f32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vmovd %edi, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; AVX1-NEXT:    vpcmpeqd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
@@ -1777,7 +1777,7 @@
 ;
 ; AVX1-LABEL: arg_f32_v8f32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-NEXT:    vmovd %edi, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
diff --git a/llvm/test/CodeGen/X86/known-bits-vector.ll b/llvm/test/CodeGen/X86/known-bits-vector.ll
--- a/llvm/test/CodeGen/X86/known-bits-vector.ll
+++ b/llvm/test/CodeGen/X86/known-bits-vector.ll
@@ -136,14 +136,14 @@
 define <4 x float> @knownbits_mask_shuffle_uitofp(<4 x i32> %a0) nounwind {
 ; X86-LABEL: knownbits_mask_shuffle_uitofp:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; X86-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: knownbits_mask_shuffle_uitofp:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; X64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    retq
@@ -173,7 +173,7 @@
 define <4 x float> @knownbits_mask_xor_shuffle_uitofp(<4 x i32> %a0) nounwind {
 ; X86-LABEL: knownbits_mask_xor_shuffle_uitofp:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; X86-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-NEXT:    vxorps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
@@ -181,7 +181,7 @@
 ;
 ; X64-LABEL: knownbits_mask_xor_shuffle_uitofp:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; X64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-NEXT:    vxorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
@@ -384,10 +384,10 @@
 define <8 x float> @knownbits_mask_concat_uitofp(<4 x i32> %a0, <4 x i32> %a1) nounwind {
 ; X86-LABEL: knownbits_mask_concat_uitofp:
 ; X86:       # %bb.0:
-; X86-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,3,1,3]
+; X86-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,1,3]
 ; X86-NEXT:    vmovaps {{.*#+}} xmm2 = [131071,131071,131071,131071]
 ; X86-NEXT:    vandps %xmm2, %xmm1, %xmm1
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,2]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,0,2]
 ; X86-NEXT:    vandps %xmm2, %xmm0, %xmm0
 ; X86-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; X86-NEXT:    vcvtdq2ps %ymm0, %ymm0
@@ -395,10 +395,10 @@
 ;
 ; X64-LABEL: knownbits_mask_concat_uitofp:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,3,1,3]
+; X64-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,1,3]
 ; X64-NEXT:    vmovaps {{.*#+}} xmm2 = [131071,131071,131071,131071]
 ; X64-NEXT:    vandps %xmm2, %xmm1, %xmm1
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,2]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,0,2]
 ; X64-NEXT:    vandps %xmm2, %xmm0, %xmm0
 ; X64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; X64-NEXT:    vcvtdq2ps %ymm0, %ymm0
@@ -601,7 +601,7 @@
 ; X86-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3, %xmm3
 ; X86-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vblendvps %xmm0, %xmm2, %xmm3, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,2]
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X86-NEXT:    movl %ebp, %esp
 ; X86-NEXT:    popl %ebp
@@ -613,7 +613,7 @@
 ; X64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
 ; X64-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vblendvps %xmm0, %xmm2, %xmm3, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,2]
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    retq
   %1 = and <4 x i32> %a2, <i32 65535, i32 -1, i32 255, i32 -1>
@@ -637,7 +637,7 @@
 ; X86-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3, %xmm3
 ; X86-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vblendvps %xmm0, %xmm2, %xmm3, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,2]
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X86-NEXT:    movl %ebp, %esp
 ; X86-NEXT:    popl %ebp
@@ -649,7 +649,7 @@
 ; X64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
 ; X64-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vblendvps %xmm0, %xmm2, %xmm3, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,2]
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    retq
   %1 = lshr <4 x i32> %a2, <i32 5, i32 1, i32 5, i32 1>
diff --git a/llvm/test/CodeGen/X86/known-signbits-vector.ll b/llvm/test/CodeGen/X86/known-signbits-vector.ll
--- a/llvm/test/CodeGen/X86/known-signbits-vector.ll
+++ b/llvm/test/CodeGen/X86/known-signbits-vector.ll
@@ -148,7 +148,7 @@
 ; X86-LABEL: signbits_ashr_extract_sitofp_0:
 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -157,7 +157,7 @@
 ;
 ; X64-LABEL: signbits_ashr_extract_sitofp_0:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    retq
   %1 = ashr <2 x i64> %a0, <i64 32, i64 32>
@@ -170,7 +170,7 @@
 ; X86-LABEL: signbits_ashr_extract_sitofp_1:
 ; X86:       # %bb.0:
 ; X86-NEXT:    pushl %eax
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X86-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -179,7 +179,7 @@
 ;
 ; X64-LABEL: signbits_ashr_extract_sitofp_1:
 ; X64:       # %bb.0:
-; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; X64-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; X64-NEXT:    retq
   %1 = ashr <2 x i64> %a0, <i64 32, i64 63>
@@ -322,7 +322,7 @@
 define <2 x double> @signbits_ashr_concat_ashr_extract_sitofp(<2 x i64> %a0, <4 x i64> %a1) nounwind {
 ; CHECK-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,2,3]
 ; CHECK-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = ashr <2 x i64> %a0, <i64 16, i64 16>
@@ -415,7 +415,7 @@
 ; X86-NEXT:    vpsrad $1, %xmm2, %xmm2
 ; X86-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; X86-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3],xmm2[4,5],xmm5[6,7]
-; X86-NEXT:    vpermilps {{.*#+}} xmm5 = xmm3[2,2,3,3]
+; X86-NEXT:    vshufps {{.*#+}} xmm5 = xmm3[2,2,3,3]
 ; X86-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm6
 ; X86-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; X86-NEXT:    vextractf128 $1, %ymm0, %xmm0
@@ -443,7 +443,7 @@
 ; X64-AVX1-NEXT:    vpsrad $1, %xmm2, %xmm2
 ; X64-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; X64-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3],xmm2[4,5],xmm5[6,7]
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} xmm5 = xmm3[2,2,3,3]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} xmm5 = xmm3[2,2,3,3]
 ; X64-AVX1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm6
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
@@ -465,7 +465,7 @@
 ; X64-AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
 ; X64-AVX2-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
 ; X64-AVX2-NEXT:    vblendvpd %ymm0, %ymm2, %ymm3, %ymm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; X64-AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; X64-AVX2-NEXT:    vcvtdq2ps %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/masked_store.ll b/llvm/test/CodeGen/X86/masked_store.ll
--- a/llvm/test/CodeGen/X86/masked_store.ll
+++ b/llvm/test/CodeGen/X86/masked_store.ll
@@ -4958,7 +4958,7 @@
 ; X86-AVX512-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; X86-AVX512-NEXT:    vmovlps %xmm0, 48(%eax)
 ; X86-AVX512-NEXT:    vextractf128 $1, %ymm1, %xmm0
-; X86-AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X86-AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; X86-AVX512-NEXT:    vmovlps %xmm0, 88(%eax)
 ; X86-AVX512-NEXT:    vzeroupper
 ; X86-AVX512-NEXT:    retl
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc.ll b/llvm/test/CodeGen/X86/masked_store_trunc.ll
--- a/llvm/test/CodeGen/X86/masked_store_trunc.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc.ll
@@ -1649,7 +1649,7 @@
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,2],zero,zero
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vmaskmovps %xmm0, %xmm1, (%rdi)
 ; AVX1-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
--- a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
@@ -2440,7 +2440,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm3
 ; AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vmaskmovps %xmm0, %xmm1, (%rdi)
 ; AVX1-NEXT:    retq
 ;
@@ -2457,7 +2457,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-NEXT:    vpmaskmovd %xmm0, %xmm1, (%rdi)
 ; AVX2-NEXT:    retq
 ;
@@ -2597,7 +2597,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [18446744073709518848,18446744073709518848]
 ; AVX-NEXT:    vpcmpgtq %xmm3, %xmm0, %xmm4
 ; AVX-NEXT:    vblendvpd %xmm4, %xmm0, %xmm3, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX-NEXT:    vpcmpeqq %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vmovmskpd %xmm1, %eax
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
--- a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
@@ -2142,7 +2142,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [9223372041149743103,9223372041149743103]
 ; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vmaskmovps %xmm0, %xmm1, (%rdi)
 ; AVX1-NEXT:    retq
 ;
@@ -2158,7 +2158,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = [9223372041149743103,9223372041149743103]
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-NEXT:    vpmaskmovd %xmm0, %xmm1, (%rdi)
 ; AVX2-NEXT:    retq
 ;
@@ -2279,7 +2279,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343]
 ; AVX-NEXT:    vpcmpgtq %xmm4, %xmm5, %xmm4
 ; AVX-NEXT:    vblendvpd %xmm4, %xmm0, %xmm3, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX-NEXT:    vpcmpeqq %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vmovmskpd %xmm1, %eax
diff --git a/llvm/test/CodeGen/X86/matrix-multiply.ll b/llvm/test/CodeGen/X86/matrix-multiply.ll
--- a/llvm/test/CodeGen/X86/matrix-multiply.ll
+++ b/llvm/test/CodeGen/X86/matrix-multiply.ll
@@ -43,9 +43,9 @@
 ; AVX1-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vmulps %xmm4, %xmm2, %xmm4
 ; AVX1-NEXT:    vaddps %xmm4, %xmm3, %xmm3
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,2,2,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[2,2,2,2]
 ; AVX1-NEXT:    vmulps %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vmulps %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm0[0]
@@ -59,9 +59,9 @@
 ; AVX2-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
 ; AVX2-NEXT:    vmulps %xmm4, %xmm2, %xmm4
 ; AVX2-NEXT:    vaddps %xmm4, %xmm3, %xmm3
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,2,2,2]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[2,2,2,2]
 ; AVX2-NEXT:    vmulps %xmm4, %xmm0, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-NEXT:    vmulps %xmm1, %xmm2, %xmm1
 ; AVX2-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm0[0]
@@ -75,9 +75,9 @@
 ; AVX512-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
 ; AVX512-NEXT:    vmulps %xmm4, %xmm2, %xmm4
 ; AVX512-NEXT:    vaddps %xmm4, %xmm3, %xmm3
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,2,2,2]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[2,2,2,2]
 ; AVX512-NEXT:    vmulps %xmm4, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulps %xmm1, %xmm2, %xmm1
 ; AVX512-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm0[0]
@@ -287,7 +287,7 @@
 ; AVX1-NEXT:    vmulss %xmm3, %xmm8, %xmm3
 ; AVX1-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm7[0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[1,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[1,2,2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -365,9 +365,9 @@
 ; AVX512F-NEXT:    vmulps %xmm6, %xmm2, %xmm4
 ; AVX512F-NEXT:    vaddps %xmm4, %xmm3, %xmm4
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm5[1,0]
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm7 = xmm1[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm7 = xmm1[3,3,3,3]
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm8 = xmm1[1,0]
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm9 = xmm1[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm9 = xmm1[2,2,2,2]
 ; AVX512F-NEXT:    vmulps %xmm3, %xmm9, %xmm9
 ; AVX512F-NEXT:    vaddps %xmm4, %xmm9, %xmm9
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
@@ -393,9 +393,9 @@
 ; AVX512F-NEXT:    vmulss %xmm11, %xmm10, %xmm11
 ; AVX512F-NEXT:    vaddss %xmm7, %xmm11, %xmm7
 ; AVX512F-NEXT:    vinsertps {{.*#+}} xmm7 = xmm8[0,1],xmm7[0],xmm8[3]
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm9[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm9[3,3,3,3]
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[2,2,2,2]
 ; AVX512F-NEXT:    vmulps %xmm0, %xmm9, %xmm0
 ; AVX512F-NEXT:    vmulps %xmm2, %xmm8, %xmm2
 ; AVX512F-NEXT:    vaddps %xmm2, %xmm0, %xmm0
@@ -424,9 +424,9 @@
 ; AVX512VL-NEXT:    vmulps %xmm5, %xmm2, %xmm6
 ; AVX512VL-NEXT:    vaddps %xmm6, %xmm3, %xmm3
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm4[1,0]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm7 = xmm1[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm7 = xmm1[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm8 = xmm1[1,0]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm9 = xmm1[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm9 = xmm1[2,2,2,2]
 ; AVX512VL-NEXT:    vmulps %xmm6, %xmm9, %xmm9
 ; AVX512VL-NEXT:    vaddps %xmm3, %xmm9, %xmm3
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm9 = xmm0[1,0]
@@ -452,9 +452,9 @@
 ; AVX512VL-NEXT:    vmulss %xmm11, %xmm10, %xmm11
 ; AVX512VL-NEXT:    vaddss %xmm7, %xmm11, %xmm7
 ; AVX512VL-NEXT:    vinsertps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0],xmm5[3]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm7 = xmm8[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm7 = xmm8[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm11 = xmm8[1,0]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[2,2,2,2]
 ; AVX512VL-NEXT:    vmulps %xmm0, %xmm8, %xmm0
 ; AVX512VL-NEXT:    vmulps %xmm7, %xmm2, %xmm2
 ; AVX512VL-NEXT:    vaddps %xmm2, %xmm0, %xmm0
@@ -1008,60 +1008,60 @@
 ;
 ; AVX1-LABEL: test_mul4x4_f32:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm4 = ymm2[1,1,1,1,5,5,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    vmulps %ymm4, %ymm5, %ymm4
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm6 = ymm2[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
 ; AVX1-NEXT:    vmulps %ymm6, %ymm7, %ymm0
 ; AVX1-NEXT:    vaddps %ymm4, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm4 = ymm2[2,2,2,2,6,6,6,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[2,2,2,2,6,6,6,6]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm6
 ; AVX1-NEXT:    vmulps %ymm4, %ymm6, %ymm4
 ; AVX1-NEXT:    vaddps %ymm4, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[3,3,3,3,7,7,7,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,3,3,3,7,7,7,7]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-NEXT:    vmulps %ymm2, %ymm1, %ymm2
 ; AVX1-NEXT:    vaddps %ymm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[1,1,1,1,5,5,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[1,1,1,1,5,5,5,5]
 ; AVX1-NEXT:    vmulps %ymm2, %ymm5, %ymm2
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm4 = ymm3[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    vmulps %ymm4, %ymm7, %ymm4
 ; AVX1-NEXT:    vaddps %ymm2, %ymm4, %ymm2
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm4 = ymm3[2,2,2,2,6,6,6,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[2,2,2,2,6,6,6,6]
 ; AVX1-NEXT:    vmulps %ymm4, %ymm6, %ymm4
 ; AVX1-NEXT:    vaddps %ymm4, %ymm2, %ymm2
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,3,3,3,7,7,7,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,3,3,3,7,7,7,7]
 ; AVX1-NEXT:    vmulps %ymm3, %ymm1, %ymm1
 ; AVX1-NEXT:    vaddps %ymm1, %ymm2, %ymm1
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_mul4x4_f32:
 ; AVX2:       # %bb.0: # %entry
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm4 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm0[2,3,2,3]
 ; AVX2-NEXT:    vmulps %ymm4, %ymm5, %ymm4
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[0,0,0,0,4,4,4,4]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm2[0,0,0,0,4,4,4,4]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm7 = ymm0[0,1,0,1]
 ; AVX2-NEXT:    vmulps %ymm6, %ymm7, %ymm0
 ; AVX2-NEXT:    vaddps %ymm4, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm4 = ymm2[2,2,2,2,6,6,6,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[2,2,2,2,6,6,6,6]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm6 = ymm1[0,1,0,1]
 ; AVX2-NEXT:    vmulps %ymm4, %ymm6, %ymm4
 ; AVX2-NEXT:    vaddps %ymm4, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[3,3,3,3,7,7,7,7]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,3,3,3,7,7,7,7]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX2-NEXT:    vmulps %ymm2, %ymm1, %ymm2
 ; AVX2-NEXT:    vaddps %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[1,1,1,1,5,5,5,5]
 ; AVX2-NEXT:    vmulps %ymm2, %ymm5, %ymm2
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm4 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-NEXT:    vmulps %ymm4, %ymm7, %ymm4
 ; AVX2-NEXT:    vaddps %ymm2, %ymm4, %ymm2
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm4 = ymm3[2,2,2,2,6,6,6,6]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[2,2,2,2,6,6,6,6]
 ; AVX2-NEXT:    vmulps %ymm4, %ymm6, %ymm4
 ; AVX2-NEXT:    vaddps %ymm4, %ymm2, %ymm2
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,3,3,3,7,7,7,7]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,3,3,3,7,7,7,7]
 ; AVX2-NEXT:    vmulps %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    vaddps %ymm1, %ymm2, %ymm1
 ; AVX2-NEXT:    retq
@@ -1074,20 +1074,20 @@
 ; AVX512-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
 ; AVX512-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX512-NEXT:    vinsertf64x4 $1, %ymm3, %zmm1, %zmm2
-; AVX512-NEXT:    vpermilps {{.*#+}} zmm4 = zmm2[1,1,1,1,5,5,5,5,9,9,9,9,13,13,13,13]
+; AVX512-NEXT:    vshufps {{.*#+}} zmm4 = zmm2[1,1,1,1,5,5,5,5,9,9,9,9,13,13,13,13]
 ; AVX512-NEXT:    vshuff64x2 {{.*#+}} zmm5 = zmm0[2,3,2,3,2,3,2,3]
 ; AVX512-NEXT:    vmulps %zmm4, %zmm5, %zmm4
-; AVX512-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
-; AVX512-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,0,0,4,4,4,4]
+; AVX512-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,0,0,4,4,4,4]
+; AVX512-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX512-NEXT:    vinsertf64x4 $1, %ymm3, %zmm1, %zmm1
 ; AVX512-NEXT:    vshuff64x2 {{.*#+}} zmm3 = zmm0[0,1,0,1,0,1,0,1]
 ; AVX512-NEXT:    vmulps %zmm1, %zmm3, %zmm1
 ; AVX512-NEXT:    vaddps %zmm4, %zmm1, %zmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} zmm3 = zmm2[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512-NEXT:    vshufps {{.*#+}} zmm3 = zmm2[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512-NEXT:    vshuff64x2 {{.*#+}} zmm4 = zmm0[4,5,4,5,4,5,4,5]
 ; AVX512-NEXT:    vmulps %zmm3, %zmm4, %zmm3
 ; AVX512-NEXT:    vaddps %zmm3, %zmm1, %zmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} zmm2 = zmm2[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512-NEXT:    vshufps {{.*#+}} zmm2 = zmm2[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512-NEXT:    vshuff64x2 {{.*#+}} zmm0 = zmm0[6,7,6,7,6,7,6,7]
 ; AVX512-NEXT:    vmulps %zmm2, %zmm0, %zmm0
 ; AVX512-NEXT:    vaddps %zmm0, %zmm1, %zmm0
@@ -2400,10 +2400,10 @@
 ;
 ; AVX512F-LABEL: test_mul8x8_f32:
 ; AVX512F:       # %bb.0: # %entry
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm4[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm4[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm9
 ; AVX512F-NEXT:    vextractf64x4 $1, %zmm4, %ymm8
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm10 = xmm8[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm10 = xmm8[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm10, %ymm10
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm10, %zmm9, %zmm16
 ; AVX512F-NEXT:    vbroadcastss %xmm4, %ymm10
@@ -2415,20 +2415,20 @@
 ; AVX512F-NEXT:    vmovshdup {{.*#+}} xmm12 = xmm8[1,1,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm12, %ymm12
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm12, %zmm11, %zmm18
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm12 = xmm4[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm12 = xmm4[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm12, %ymm12
 ; AVX512F-NEXT:    vextractf128 $1, %ymm4, %xmm13
 ; AVX512F-NEXT:    vbroadcastss %xmm13, %ymm13
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm8
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm8, %zmm12, %zmm19
 ; AVX512F-NEXT:    vextractf32x4 $3, %zmm4, %xmm12
 ; AVX512F-NEXT:    vbroadcastss %xmm12, %ymm12
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm14 = xmm5[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm14 = xmm5[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm14, %ymm14
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm12, %zmm13, %zmm20
 ; AVX512F-NEXT:    vextractf64x4 $1, %zmm5, %ymm13
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm15 = xmm13[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm15 = xmm13[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm15, %ymm15
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm15, %zmm14, %zmm21
 ; AVX512F-NEXT:    vbroadcastss %xmm5, %ymm15
@@ -2440,20 +2440,20 @@
 ; AVX512F-NEXT:    vmovshdup {{.*#+}} xmm10 = xmm13[1,1,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm10, %ymm10
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm10, %zmm15, %zmm10
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm15 = xmm5[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm15 = xmm5[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm15, %ymm15
 ; AVX512F-NEXT:    vextractf128 $1, %ymm5, %xmm11
 ; AVX512F-NEXT:    vbroadcastss %xmm11, %ymm11
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm13, %ymm13
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm13, %zmm15, %zmm15
 ; AVX512F-NEXT:    vextractf32x4 $3, %zmm5, %xmm13
 ; AVX512F-NEXT:    vbroadcastss %xmm13, %ymm13
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm6[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm6[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm8
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm13, %zmm11, %zmm11
 ; AVX512F-NEXT:    vextractf64x4 $1, %zmm6, %ymm13
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm12 = xmm13[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm12 = xmm13[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm12, %ymm12
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm12, %zmm8, %zmm8
 ; AVX512F-NEXT:    vbroadcastss %xmm6, %ymm12
@@ -2472,8 +2472,8 @@
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm2, %zmm2, %zmm26
 ; AVX512F-NEXT:    vshuff64x2 {{.*#+}} zmm27 = zmm2[4,5,6,7,4,5,6,7]
 ; AVX512F-NEXT:    vmovshdup {{.*#+}} zmm0 = zmm4[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm1 = zmm4[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm2 = zmm4[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm1 = zmm4[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm2 = zmm4[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm3, %zmm3, %zmm4
 ; AVX512F-NEXT:    vshuff64x2 {{.*#+}} zmm3 = zmm3[4,5,6,7,4,5,6,7]
 ; AVX512F-NEXT:    vmulps %zmm17, %zmm14, %zmm17
@@ -2507,11 +2507,11 @@
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm2 = zmm2[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm2, %zmm27, %zmm2
 ; AVX512F-NEXT:    vaddps %zmm2, %zmm1, %zmm1
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm2 = zmm5[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm2 = zmm5[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm2 = zmm2[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm2, %zmm4, %zmm2
 ; AVX512F-NEXT:    vaddps %zmm2, %zmm1, %zmm1
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm2 = zmm5[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm2 = zmm5[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm2 = zmm2[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm2, %zmm3, %zmm2
 ; AVX512F-NEXT:    vaddps %zmm2, %zmm1, %zmm1
@@ -2520,9 +2520,9 @@
 ; AVX512F-NEXT:    vaddps %zmm5, %zmm2, %zmm2
 ; AVX512F-NEXT:    vmulps %zmm8, %zmm24, %zmm5
 ; AVX512F-NEXT:    vaddps %zmm5, %zmm2, %zmm2
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm5, %ymm5
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm13[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm13[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm8
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm8, %zmm5, %zmm5
 ; AVX512F-NEXT:    vextractf128 $1, %ymm6, %xmm8
@@ -2532,7 +2532,7 @@
 ; AVX512F-NEXT:    vextractf32x4 $3, %zmm6, %xmm5
 ; AVX512F-NEXT:    vbroadcastss %xmm5, %ymm5
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm5, %zmm8, %zmm5
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm7[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm7[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm8
 ; AVX512F-NEXT:    vmulps %zmm5, %zmm26, %zmm5
 ; AVX512F-NEXT:    vaddps %zmm5, %zmm2, %zmm2
@@ -2540,16 +2540,16 @@
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm5, %zmm27, %zmm5
 ; AVX512F-NEXT:    vaddps %zmm5, %zmm2, %zmm2
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm5 = zmm6[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm5 = zmm6[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm5, %zmm4, %zmm5
 ; AVX512F-NEXT:    vaddps %zmm5, %zmm2, %zmm2
 ; AVX512F-NEXT:    vextractf64x4 $1, %zmm7, %ymm5
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm6 = zmm6[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm6 = zmm6[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm6 = zmm6[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm6, %zmm3, %zmm6
 ; AVX512F-NEXT:    vaddps %zmm6, %zmm2, %zmm2
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[2,2,2,2]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[2,2,2,2]
 ; AVX512F-NEXT:    vbroadcastsd %xmm6, %ymm6
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm6, %zmm8, %zmm6
 ; AVX512F-NEXT:    vbroadcastss %xmm7, %ymm8
@@ -2566,9 +2566,9 @@
 ; AVX512F-NEXT:    vmulps %zmm9, %zmm23, %zmm9
 ; AVX512F-NEXT:    vaddps %zmm9, %zmm8, %zmm8
 ; AVX512F-NEXT:    vaddps %zmm6, %zmm8, %zmm6
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm8 = xmm7[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm8 = xmm7[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm8, %ymm8
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,3,3,3]
 ; AVX512F-NEXT:    vbroadcastsd %xmm5, %ymm5
 ; AVX512F-NEXT:    vinsertf64x4 $1, %ymm5, %zmm8, %zmm5
 ; AVX512F-NEXT:    vmulps %zmm5, %zmm25, %zmm5
@@ -2584,11 +2584,11 @@
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm6 = zmm6[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm6, %zmm27, %zmm6
 ; AVX512F-NEXT:    vaddps %zmm6, %zmm5, %zmm5
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm6 = zmm7[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm6 = zmm7[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm6 = zmm6[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm6, %zmm4, %zmm4
 ; AVX512F-NEXT:    vaddps %zmm4, %zmm5, %zmm4
-; AVX512F-NEXT:    vpermilps {{.*#+}} zmm5 = zmm7[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512F-NEXT:    vshufps {{.*#+}} zmm5 = zmm7[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512F-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512F-NEXT:    vmulps %zmm5, %zmm3, %zmm3
 ; AVX512F-NEXT:    vaddps %zmm3, %zmm4, %zmm3
@@ -2599,9 +2599,9 @@
 ; AVX512VL-NEXT:    vbroadcastss %xmm4, %ymm11
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm8 = xmm4[1,1,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm8, %ymm10
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm8 = xmm4[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm8 = xmm4[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm8, %ymm14
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm8 = xmm4[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm8 = xmm4[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm8, %ymm9
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm4, %xmm8
 ; AVX512VL-NEXT:    vbroadcastss %xmm8, %ymm8
@@ -2610,9 +2610,9 @@
 ; AVX512VL-NEXT:    vextractf64x4 $1, %zmm4, %ymm12
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm13 = xmm12[1,1,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm13, %ymm13
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm16 = xmm12[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm16 = xmm12[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm16, %ymm16
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm12, %ymm12
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm16, %zmm14, %zmm16
 ; AVX512VL-NEXT:    vextractf32x4 $3, %zmm4, %xmm14
@@ -2631,7 +2631,7 @@
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm13 = xmm5[1,1,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm13, %ymm16
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm12, %zmm9, %zmm12
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm9 = xmm5[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm9 = xmm5[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm9, %ymm18
 ; AVX512VL-NEXT:    vshuff64x2 {{.*#+}} zmm9 = zmm1[4,5,6,7,4,5,6,7]
 ; AVX512VL-NEXT:    vmulps %zmm12, %zmm9, %zmm1
@@ -2645,18 +2645,18 @@
 ; AVX512VL-NEXT:    vshuff64x2 {{.*#+}} zmm12 = zmm2[4,5,6,7,4,5,6,7]
 ; AVX512VL-NEXT:    vmulps %zmm1, %zmm12, %zmm1
 ; AVX512VL-NEXT:    vaddps %zmm1, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm1 = zmm4[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm1 = zmm4[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm1 = zmm1[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm3, %zmm3, %zmm13
 ; AVX512VL-NEXT:    vmulps %zmm1, %zmm13, %zmm1
 ; AVX512VL-NEXT:    vaddps %zmm1, %zmm0, %zmm0
 ; AVX512VL-NEXT:    vextractf64x4 $1, %zmm5, %ymm1
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm2 = zmm4[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm2 = zmm4[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm2 = zmm2[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vshuff64x2 {{.*#+}} zmm3 = zmm3[4,5,6,7,4,5,6,7]
 ; AVX512VL-NEXT:    vmulps %zmm2, %zmm3, %zmm2
 ; AVX512VL-NEXT:    vaddps %zmm2, %zmm0, %zmm0
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm2, %ymm2
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm18, %zmm2
 ; AVX512VL-NEXT:    vextractf32x4 $2, %zmm5, %xmm4
@@ -2665,14 +2665,14 @@
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm15 = xmm1[1,1,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm15, %ymm15
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm15, %zmm16, %zmm15
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm16 = xmm5[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm16 = xmm5[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm16, %ymm16
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm11, %zmm4
 ; AVX512VL-NEXT:    vmulps %zmm15, %zmm10, %zmm15
 ; AVX512VL-NEXT:    vaddps %zmm15, %zmm4, %zmm4
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm5, %xmm15
 ; AVX512VL-NEXT:    vbroadcastss %xmm15, %ymm15
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX512VL-NEXT:    vmulps %zmm2, %zmm14, %zmm2
 ; AVX512VL-NEXT:    vaddps %zmm2, %zmm4, %zmm2
@@ -2685,7 +2685,7 @@
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm6[1,1,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm2, %ymm2
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm4, %zmm15, %zmm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm15 = xmm6[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm15 = xmm6[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm15, %ymm15
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm8, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm1, %zmm1
@@ -2693,16 +2693,16 @@
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm4 = zmm4[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm12, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm1, %zmm1
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm4 = zmm5[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm4 = zmm5[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm4 = zmm4[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm13, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm1, %zmm1
 ; AVX512VL-NEXT:    vextractf64x4 $1, %zmm6, %ymm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm5 = zmm5[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm5 = zmm5[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm5, %zmm3, %zmm5
 ; AVX512VL-NEXT:    vaddps %zmm5, %zmm1, %zmm1
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm4[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm4[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm5, %ymm5
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm5, %zmm15, %zmm5
 ; AVX512VL-NEXT:    vextractf32x4 $2, %zmm6, %xmm15
@@ -2716,9 +2716,9 @@
 ; AVX512VL-NEXT:    vaddps %zmm2, %zmm15, %zmm2
 ; AVX512VL-NEXT:    vmulps %zmm5, %zmm14, %zmm5
 ; AVX512VL-NEXT:    vaddps %zmm5, %zmm2, %zmm2
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm5, %ymm5
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm4, %ymm4
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm4, %zmm5, %zmm4
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm6, %xmm5
@@ -2728,7 +2728,7 @@
 ; AVX512VL-NEXT:    vextractf32x4 $3, %zmm6, %xmm4
 ; AVX512VL-NEXT:    vbroadcastss %xmm4, %ymm4
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm4, %zmm5, %zmm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm7[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm7[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm5, %ymm5
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm8, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm2, %zmm2
@@ -2736,16 +2736,16 @@
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm4 = zmm4[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm12, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm2, %zmm2
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm4 = zmm6[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm4 = zmm6[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm4 = zmm4[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm13, %zmm4
 ; AVX512VL-NEXT:    vaddps %zmm4, %zmm2, %zmm2
 ; AVX512VL-NEXT:    vextractf64x4 $1, %zmm7, %ymm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm6 = zmm6[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm6 = zmm6[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm6 = zmm6[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm6, %zmm3, %zmm6
 ; AVX512VL-NEXT:    vaddps %zmm6, %zmm2, %zmm2
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm6 = xmm4[2,2,2,2]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm6 = xmm4[2,2,2,2]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm6, %ymm6
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm6, %zmm5, %zmm5
 ; AVX512VL-NEXT:    vbroadcastss %xmm7, %ymm6
@@ -2762,9 +2762,9 @@
 ; AVX512VL-NEXT:    vmulps %zmm11, %zmm10, %zmm10
 ; AVX512VL-NEXT:    vaddps %zmm10, %zmm6, %zmm6
 ; AVX512VL-NEXT:    vaddps %zmm5, %zmm6, %zmm5
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm6, %ymm6
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
 ; AVX512VL-NEXT:    vbroadcastsd %xmm4, %ymm4
 ; AVX512VL-NEXT:    vinsertf64x4 $1, %ymm4, %zmm6, %zmm4
 ; AVX512VL-NEXT:    vmulps %zmm4, %zmm9, %zmm4
@@ -2780,11 +2780,11 @@
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm5, %zmm12, %zmm5
 ; AVX512VL-NEXT:    vaddps %zmm5, %zmm4, %zmm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm5 = zmm7[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm5 = zmm7[2,2,2,2,6,6,6,6,10,10,10,10,14,14,14,14]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm5, %zmm13, %zmm5
 ; AVX512VL-NEXT:    vaddps %zmm5, %zmm4, %zmm4
-; AVX512VL-NEXT:    vpermilps {{.*#+}} zmm5 = zmm7[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
+; AVX512VL-NEXT:    vshufps {{.*#+}} zmm5 = zmm7[3,3,3,3,7,7,7,7,11,11,11,11,15,15,15,15]
 ; AVX512VL-NEXT:    vpermpd {{.*#+}} zmm5 = zmm5[2,2,2,2,6,6,6,6]
 ; AVX512VL-NEXT:    vmulps %zmm5, %zmm3, %zmm3
 ; AVX512VL-NEXT:    vaddps %zmm3, %zmm4, %zmm3
diff --git a/llvm/test/CodeGen/X86/oddshuffles.ll b/llvm/test/CodeGen/X86/oddshuffles.ll
--- a/llvm/test/CodeGen/X86/oddshuffles.ll
+++ b/llvm/test/CodeGen/X86/oddshuffles.ll
@@ -129,7 +129,7 @@
 ;
 ; AVX-LABEL: v5i32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,2,2,3]
 ; AVX-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX-NEXT:    vextractps $3, %xmm0, 16(%rdi)
 ; AVX-NEXT:    vmovaps %xmm1, (%rdi)
@@ -161,7 +161,7 @@
 ; AVX-LABEL: v5f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,1],xmm1[1,2]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; AVX-NEXT:    vextractps $3, %xmm0, 16(%rdi)
 ; AVX-NEXT:    vmovaps %xmm1, (%rdi)
 ; AVX-NEXT:    retq
@@ -313,9 +313,9 @@
 ; AVX-LABEL: v7i32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,2,3,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2,3,2]
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,2,3]
 ; AVX-NEXT:    vmovss %xmm1, 24(%rdi)
 ; AVX-NEXT:    vmovlps %xmm0, 16(%rdi)
 ; AVX-NEXT:    vmovaps %xmm2, (%rdi)
@@ -513,12 +513,12 @@
 ; AVX1-NEXT:    vmovsldup {{.*#+}} ymm2 = ymm2[0,0,2,2,4,4,6,6]
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[0,u,u,1,5,u,u,6]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4,5],ymm2[6],ymm3[7]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm3
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; AVX1-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; AVX1-NEXT:    vmovaps %ymm2, (%rdi)
@@ -533,7 +533,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rdi)
@@ -548,7 +548,7 @@
 ; AVX2-FAST-ALL-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX2-FAST-ALL-NEXT:    vmovaps {{.*#+}} ymm3 = <u,3,7,u,u,u,u,u>
 ; AVX2-FAST-ALL-NEXT:    vpermps %ymm0, %ymm3, %ymm0
-; AVX2-FAST-ALL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-FAST-ALL-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-FAST-ALL-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; AVX2-FAST-ALL-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, (%rdi)
@@ -563,7 +563,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, (%rdi)
@@ -574,12 +574,12 @@
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
 ; XOP-NEXT:    vpermil2ps {{.*#+}} ymm2 = ymm0[0],ymm2[0],ymm0[u,1,5,u],ymm2[6],ymm0[6]
-; XOP-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[0,1,0,1]
+; XOP-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[0,1,0,1]
 ; XOP-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm3
 ; XOP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; XOP-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOP-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; XOP-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; XOP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; XOP-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; XOP-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; XOP-NEXT:    vmovaps %ymm2, (%rdi)
@@ -1479,14 +1479,14 @@
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm4[0,2],ymm3[4,7],ymm4[4,6]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[1,0],ymm4[2,0],ymm0[5,4],ymm4[6,4]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[2,0],ymm4[3,0],ymm0[6,4],ymm4[7,4]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm5 = ymm4[0,0],ymm5[2,0],ymm4[4,4],ymm5[6,4]
 ; AVX1-NEXT:    vmovups 16(%rdi), %xmm6
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[2,0],ymm6[5,4],ymm2[6,4]
@@ -1520,7 +1520,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, (%rsi)
@@ -1580,7 +1580,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm1, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, (%rsi)
@@ -1600,14 +1600,14 @@
 ; XOP-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm4[0,2],ymm3[4,7],ymm4[4,6]
 ; XOP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1]
 ; XOP-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[1,0],ymm4[2,0],ymm0[5,4],ymm4[6,4]
-; XOP-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; XOP-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; XOP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; XOP-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[2,0],ymm4[3,0],ymm0[6,4],ymm4[7,4]
 ; XOP-NEXT:    vshufps {{.*#+}} ymm5 = ymm4[0,0],ymm5[2,0],ymm4[4,4],ymm5[6,4]
 ; XOP-NEXT:    vmovups 16(%rdi), %xmm6
 ; XOP-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; XOP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7]
-; XOP-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
+; XOP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
 ; XOP-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7]
 ; XOP-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
 ; XOP-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[2,0],ymm6[5,4],ymm2[6,4]
@@ -1746,7 +1746,7 @@
 ; AVX2-SLOW-NEXT:    vmovups (%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups (%rcx), %ymm2
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%rsi), %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3]
@@ -1757,7 +1757,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd (%rcx), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm2[1,1,2,2]
@@ -1786,7 +1786,7 @@
 ; AVX2-FAST-ALL-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
 ; AVX2-FAST-ALL-NEXT:    vbroadcastsd (%rcx), %ymm5
 ; AVX2-FAST-ALL-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
-; AVX2-FAST-ALL-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-ALL-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm2[1,1,2,2]
@@ -1803,7 +1803,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups (%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups (%rcx), %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%rsi), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3]
@@ -1814,7 +1814,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd (%rcx), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm2[1,1,2,2]
@@ -2409,7 +2409,7 @@
 ; AVX1-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm0
@@ -2420,7 +2420,7 @@
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,1,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-NEXT:    vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
@@ -2481,7 +2481,7 @@
 ; XOP-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
 ; XOP-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; XOP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; XOP-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; XOP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; XOP-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
 ; XOP-NEXT:    vpsrld $16, %xmm0, %xmm0
@@ -2492,7 +2492,7 @@
 ; XOP-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,1,3,3]
 ; XOP-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; XOP-NEXT:    vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7]
-; XOP-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
+; XOP-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
 ; XOP-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
 ; XOP-NEXT:    vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1]
 ; XOP-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
diff --git a/llvm/test/CodeGen/X86/opt-pipeline.ll b/llvm/test/CodeGen/X86/opt-pipeline.ll
--- a/llvm/test/CodeGen/X86/opt-pipeline.ll
+++ b/llvm/test/CodeGen/X86/opt-pipeline.ll
@@ -202,6 +202,7 @@
 ; CHECK-NEXT:       Lazy Machine Block Frequency Analysis
 ; CHECK-NEXT:       X86 Atom pad short functions
 ; CHECK-NEXT:       X86 LEA Fixup
+; CHECK-NEXT:       X86 Fixup Inst Tuning
 ; CHECK-NEXT:       Compressing EVEX instrs to VEX encoding when possible
 ; CHECK-NEXT:       X86 Discriminate Memory Operands
 ; CHECK-NEXT:       X86 Insert Cache Prefetches
diff --git a/llvm/test/CodeGen/X86/packss.ll b/llvm/test/CodeGen/X86/packss.ll
--- a/llvm/test/CodeGen/X86/packss.ll
+++ b/llvm/test/CodeGen/X86/packss.ll
@@ -185,7 +185,7 @@
 ; X86-AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; X86-AVX1-NEXT:    vpaddq %xmm3, %xmm1, %xmm1
 ; X86-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X86-AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
 ; X86-AVX1-NEXT:    vzeroupper
@@ -236,7 +236,7 @@
 ; X64-AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; X64-AVX1-NEXT:    vpsubq %xmm2, %xmm1, %xmm1
 ; X64-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; X64-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; X64-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; X64-AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
 ; X64-AVX1-NEXT:    vzeroupper
diff --git a/llvm/test/CodeGen/X86/palignr.ll b/llvm/test/CodeGen/X86/palignr.ll
--- a/llvm/test/CodeGen/X86/palignr.ll
+++ b/llvm/test/CodeGen/X86/palignr.ll
@@ -11,7 +11,7 @@
 ;
 ; CHECK-AVX-LABEL: test1:
 ; CHECK-AVX:       # %bb.0:
-; CHECK-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,0]
+; CHECK-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2,3,0]
 ; CHECK-AVX-NEXT:    retl
   %C = shufflevector <4 x i32> %A, <4 x i32> undef, <4 x i32> < i32 1, i32 2, i32 3, i32 0 >
   ret <4 x i32> %C
diff --git a/llvm/test/CodeGen/X86/pr31956.ll b/llvm/test/CodeGen/X86/pr31956.ll
--- a/llvm/test/CodeGen/X86/pr31956.ll
+++ b/llvm/test/CodeGen/X86/pr31956.ll
@@ -12,7 +12,7 @@
 ; CHECK-NEXT:    vmovaps G2(%rip), %xmm0
 ; CHECK-NEXT:    vmovlps {{.*#+}} xmm0 = mem[0,1],xmm0[2,3]
 ; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],mem[0,2]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,3,1]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0,3,1]
 ; CHECK-NEXT:    retq
 entry:
   %V = load <2 x float>, ptr @G1, align 8
diff --git a/llvm/test/CodeGen/X86/pr40730.ll b/llvm/test/CodeGen/X86/pr40730.ll
--- a/llvm/test/CodeGen/X86/pr40730.ll
+++ b/llvm/test/CodeGen/X86/pr40730.ll
@@ -6,7 +6,7 @@
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm0[1,2,3]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,1,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,1,1,0]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; CHECK-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
@@ -27,7 +27,7 @@
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm0[1,2,3]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,1,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,1,1,0]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],mem[1,2,3],ymm0[4],mem[5],ymm0[6,7]
 ; CHECK-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/pr40811.ll b/llvm/test/CodeGen/X86/pr40811.ll
--- a/llvm/test/CodeGen/X86/pr40811.ll
+++ b/llvm/test/CodeGen/X86/pr40811.ll
@@ -6,8 +6,8 @@
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vmovaps (%rdi), %xmm0
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0],mem[1],xmm0[2,3]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,1,0]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2,1,0]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; CHECK-NEXT:    retq
 entry:
diff --git a/llvm/test/CodeGen/X86/pr50609.ll b/llvm/test/CodeGen/X86/pr50609.ll
--- a/llvm/test/CodeGen/X86/pr50609.ll
+++ b/llvm/test/CodeGen/X86/pr50609.ll
@@ -13,7 +13,7 @@
 ; CHECK-NEXT:    vpaddd %xmm3, %xmm2, %xmm2
 ; CHECK-NEXT:    vpsrad $2, %xmm2, %xmm2
 ; CHECK-NEXT:    vcvtdq2ps %ymm2, %ymm2
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm2, %ymm2, %ymm2
 ; CHECK-NEXT:    vmaskmovps %ymm2, %ymm0, (%rdi)
 ; CHECK-NEXT:    vmaskmovps %ymm2, %ymm1, 32(%rdi)
diff --git a/llvm/test/CodeGen/X86/rotate_vec.ll b/llvm/test/CodeGen/X86/rotate_vec.ll
--- a/llvm/test/CodeGen/X86/rotate_vec.ll
+++ b/llvm/test/CodeGen/X86/rotate_vec.ll
@@ -80,7 +80,7 @@
 define <4 x i32> @rot_v4i32_zero_non_splat(<4 x i32> %x) {
 ; XOPAVX1-LABEL: rot_v4i32_zero_non_splat:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: rot_v4i32_zero_non_splat:
diff --git a/llvm/test/CodeGen/X86/scalarize-fp.ll b/llvm/test/CodeGen/X86/scalarize-fp.ll
--- a/llvm/test/CodeGen/X86/scalarize-fp.ll
+++ b/llvm/test/CodeGen/X86/scalarize-fp.ll
@@ -394,7 +394,7 @@
 ; AVX-LABEL: fmul_splat_splat_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    retq
   %splatx = shufflevector <4 x float> %vx, <4 x float> undef, <4 x i32> zeroinitializer
   %splaty = shufflevector <4 x float> %vy, <4 x float> undef, <4 x i32> zeroinitializer
@@ -413,7 +413,7 @@
 ; AVX-LABEL: fdiv_splat_splat_v8f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vdivss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %splatx = shufflevector <8 x float> %vx, <8 x float> undef, <8 x i32> zeroinitializer
@@ -555,7 +555,7 @@
 ; AVX-LABEL: fmul_splat_const_op1_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    retq
   %splatx = shufflevector <4 x float> %vx, <4 x float> undef, <4 x i32> zeroinitializer
   %r = fmul fast <4 x float> %splatx, <float 17.0, float 17.0, float 17.0, float 17.0>
@@ -575,7 +575,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vdivss %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %splatx = shufflevector <8 x float> <float 4.5, float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0>, <8 x float> undef, <8 x i32> zeroinitializer
@@ -597,7 +597,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vdivss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %splatx = shufflevector <8 x float> %vx, <8 x float> undef, <8 x i32> zeroinitializer
@@ -654,7 +654,7 @@
 ; AVX-LABEL: splat0_fmul_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    retq
   %b = fmul fast <4 x float> %vx, %vy
   %r = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
@@ -672,7 +672,7 @@
 ; AVX-LABEL: splat0_fdiv_v8f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vdivss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %b = fdiv fast <8 x float> %vx, %vy
@@ -729,7 +729,7 @@
 ; AVX-LABEL: splat0_fmul_const_op1_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    retq
   %b = fmul fast <4 x float> %vx, <float 6.0, float -1.0, float 1.0, float 7.0>
   %r = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
@@ -745,7 +745,7 @@
 ;
 ; AVX-LABEL: splat0_fdiv_const_op1_v8f32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %b = fdiv fast <8 x float> %vx, <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>
@@ -766,7 +766,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vdivss %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    retq
   %b = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %vx
@@ -786,7 +786,7 @@
 ; AVX-LABEL: multi_use_binop:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/shuffle-of-shift.ll b/llvm/test/CodeGen/X86/shuffle-of-shift.ll
--- a/llvm/test/CodeGen/X86/shuffle-of-shift.ll
+++ b/llvm/test/CodeGen/X86/shuffle-of-shift.ll
@@ -156,7 +156,7 @@
 ; X64-AVX2-NEXT:    pushq %rax
 ; X64-AVX2-NEXT:    movl $63, %edi
 ; X64-AVX2-NEXT:    callq llvm.x86.sse2.psrai.q@PLT
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; X64-AVX2-NEXT:    popq %rax
 ; X64-AVX2-NEXT:    retq
 ;
@@ -173,7 +173,7 @@
 ; X86-AVX2-NEXT:    pushl $63
 ; X86-AVX2-NEXT:    calll llvm.x86.sse2.psrai.q@PLT
 ; X86-AVX2-NEXT:    addl $4, %esp
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; X86-AVX2-NEXT:    retl
   %i1 = tail call <2 x i64> @llvm.x86.sse2.psrai.q(<2 x i64> %x, i32 63)
   %i2 = bitcast <2 x i64> %i1 to <4 x i32>
@@ -336,7 +336,7 @@
 ; X64-AVX2-NEXT:    pushq %rax
 ; X64-AVX2-NEXT:    movl $63, %edi
 ; X64-AVX2-NEXT:    callq llvm.x86.sse2.psrai.q@PLT
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; X64-AVX2-NEXT:    popq %rax
 ; X64-AVX2-NEXT:    retq
 ;
@@ -353,7 +353,7 @@
 ; X86-AVX2-NEXT:    pushl $63
 ; X86-AVX2-NEXT:    calll llvm.x86.sse2.psrai.q@PLT
 ; X86-AVX2-NEXT:    addl $4, %esp
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X86-AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; X86-AVX2-NEXT:    retl
   %i1 = tail call <2 x i64> @llvm.x86.sse2.psrai.q(<2 x i64> %x, i32 63)
   %i2 = bitcast <2 x i64> %i1 to <2 x i64>
diff --git a/llvm/test/CodeGen/X86/shuffle-of-splat-multiuses.ll b/llvm/test/CodeGen/X86/shuffle-of-splat-multiuses.ll
--- a/llvm/test/CodeGen/X86/shuffle-of-splat-multiuses.ll
+++ b/llvm/test/CodeGen/X86/shuffle-of-splat-multiuses.ll
@@ -58,7 +58,7 @@
 define <4 x i32> @undef_splatmask(<4 x i32> %v) nounwind {
 ; AVX2-LABEL: undef_splatmask:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX2-NEXT:    retq
   %res = shufflevector <4 x i32> %v, <4 x i32> undef, <4 x i32> <i32 2, i32 undef, i32 2, i32 undef>
   %res1 = shufflevector <4 x i32> %res, <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 undef, i32 undef>
@@ -68,7 +68,7 @@
 define <4 x i32> @undef_splatmask2(<4 x i32> %v) nounwind {
 ; AVX2-LABEL: undef_splatmask2:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX2-NEXT:    retq
   %res = shufflevector <4 x i32> %v, <4 x i32> undef, <4 x i32> <i32 2, i32 1, i32 2, i32 undef>
   %res1 = shufflevector <4 x i32> %res, <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 undef, i32 undef>
@@ -78,7 +78,7 @@
 define <4 x i32> @undef_splatmask3(<4 x i32> %v) nounwind {
 ; AVX2-LABEL: undef_splatmask3:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX2-NEXT:    retq
   %res = shufflevector <4 x i32> %v, <4 x i32> undef, <4 x i32> <i32 2, i32 undef, i32 2, i32 undef>
   %res1 = shufflevector <4 x i32> %res, <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 undef, i32 3>
@@ -88,8 +88,8 @@
 define <4 x i32> @undef_splatmask4(<4 x i32> %v, ptr %p) nounwind {
 ; AVX2-LABEL: undef_splatmask4:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,2,3,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vmovaps %xmm0, (%rdi)
 ; AVX2-NEXT:    vmovaps %xmm1, %xmm0
 ; AVX2-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/sse-fsignum.ll b/llvm/test/CodeGen/X86/sse-fsignum.ll
--- a/llvm/test/CodeGen/X86/sse-fsignum.ll
+++ b/llvm/test/CodeGen/X86/sse-fsignum.ll
@@ -38,10 +38,10 @@
 ; AVX-NEXT:    vmovapd (%rdi), %xmm0
 ; AVX-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpltpd %xmm1, %xmm0, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; AVX-NEXT:    vcvtdq2pd %xmm2, %xmm2
 ; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX-NEXT:    vsubpd %xmm0, %xmm2, %xmm0
 ; AVX-NEXT:    vmovapd %xmm0, (%rdi)
diff --git a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse-intrinsics-fast-isel.ll
@@ -2104,7 +2104,7 @@
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vmovss {{[0-9]+}}(%esp), %xmm0 # encoding: [0xc5,0xfa,0x10,0x44,0x24,0x04]
 ; X86-AVX1-NEXT:    # xmm0 = mem[0],zero,zero,zero
-; X86-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X86-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
 ;
@@ -2123,7 +2123,7 @@
 ;
 ; X64-AVX1-LABEL: test_mm_set_ps1:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X64-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X64-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
 ;
@@ -2265,7 +2265,7 @@
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vmovss {{[0-9]+}}(%esp), %xmm0 # encoding: [0xc5,0xfa,0x10,0x44,0x24,0x04]
 ; X86-AVX1-NEXT:    # xmm0 = mem[0],zero,zero,zero
-; X86-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X86-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
 ;
@@ -2284,7 +2284,7 @@
 ;
 ; X64-AVX1-LABEL: test_mm_set1_ps:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X64-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X64-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
 ;
@@ -2623,7 +2623,7 @@
 ; X86-AVX1-LABEL: test_mm_store_ps1:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X86-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X86-AVX1-NEXT:    vmovaps %xmm0, (%eax) # encoding: [0xc5,0xf8,0x29,0x00]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
@@ -2644,7 +2644,7 @@
 ;
 ; X64-AVX1-LABEL: test_mm_store_ps1:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X64-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X64-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X64-AVX1-NEXT:    vmovaps %xmm0, (%rdi) # encoding: [0xc5,0xf8,0x29,0x07]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
@@ -2710,7 +2710,7 @@
 ; X86-AVX1-LABEL: test_mm_store1_ps:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X86-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X86-AVX1-NEXT:    vmovaps %xmm0, (%eax) # encoding: [0xc5,0xf8,0x29,0x00]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
@@ -2731,7 +2731,7 @@
 ;
 ; X64-AVX1-LABEL: test_mm_store1_ps:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X64-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X64-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X64-AVX1-NEXT:    vmovaps %xmm0, (%rdi) # encoding: [0xc5,0xf8,0x29,0x07]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
@@ -2972,7 +2972,7 @@
 ; X86-AVX1-LABEL: test_mm_storer_ps:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT:    vpermilps $27, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; X86-AVX1-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[3,2,1,0]
 ; X86-AVX1-NEXT:    vmovaps %xmm0, (%eax) # encoding: [0xc5,0xf8,0x29,0x00]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
@@ -2980,7 +2980,7 @@
 ; X86-AVX512-LABEL: test_mm_storer_ps:
 ; X86-AVX512:       # %bb.0:
 ; X86-AVX512-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX512-NEXT:    vpermilps $27, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; X86-AVX512-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; X86-AVX512-NEXT:    # xmm0 = xmm0[3,2,1,0]
 ; X86-AVX512-NEXT:    vmovaps %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x29,0x00]
 ; X86-AVX512-NEXT:    retl # encoding: [0xc3]
@@ -2994,14 +2994,14 @@
 ;
 ; X64-AVX1-LABEL: test_mm_storer_ps:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vpermilps $27, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; X64-AVX1-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; X64-AVX1-NEXT:    # xmm0 = xmm0[3,2,1,0]
 ; X64-AVX1-NEXT:    vmovaps %xmm0, (%rdi) # encoding: [0xc5,0xf8,0x29,0x07]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
 ;
 ; X64-AVX512-LABEL: test_mm_storer_ps:
 ; X64-AVX512:       # %bb.0:
-; X64-AVX512-NEXT:    vpermilps $27, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; X64-AVX512-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; X64-AVX512-NEXT:    # xmm0 = xmm0[3,2,1,0]
 ; X64-AVX512-NEXT:    vmovaps %xmm0, (%rdi) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x29,0x07]
 ; X64-AVX512-NEXT:    retq # encoding: [0xc3]
diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
@@ -4523,7 +4523,7 @@
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vmovss {{[0-9]+}}(%esp), %xmm0 # encoding: [0xc5,0xfa,0x10,0x44,0x24,0x04]
 ; X86-AVX1-NEXT:    # xmm0 = mem[0],zero,zero,zero
-; X86-AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; X86-AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; X86-AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; X86-AVX1-NEXT:    retl # encoding: [0xc3]
 ;
@@ -5636,7 +5636,7 @@
 ;
 ; AVX1-LABEL: test_mm_shuffle_epi32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; AVX1-NEXT:    # xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
 ;
diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-x86-upgrade.ll
@@ -402,13 +402,13 @@
 ;
 ; AVX1-LABEL: test_x86_sse2_pshuf_d:
 ; AVX1:       ## %bb.0: ## %entry
-; AVX1-NEXT:    vpermilps $27, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; AVX1-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; AVX1-NEXT:    ## xmm0 = xmm0[3,2,1,0]
 ; AVX1-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
 ;
 ; AVX512-LABEL: test_x86_sse2_pshuf_d:
 ; AVX512:       ## %bb.0: ## %entry
-; AVX512-NEXT:    vpermilps $27, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x1b]
+; AVX512-NEXT:    vshufps $27, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0x1b]
 ; AVX512-NEXT:    ## xmm0 = xmm0[3,2,1,0]
 ; AVX512-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
 entry:
diff --git a/llvm/test/CodeGen/X86/sse2.ll b/llvm/test/CodeGen/X86/sse2.ll
--- a/llvm/test/CodeGen/X86/sse2.ll
+++ b/llvm/test/CodeGen/X86/sse2.ll
@@ -144,7 +144,7 @@
 ; X86-AVX-LABEL: test4:
 ; X86-AVX:       # %bb.0:
 ; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; X86-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,3,3]
 ; X86-AVX-NEXT:    vmovaps %xmm0, (%eax)
 ; X86-AVX-NEXT:    retl
 ;
@@ -156,7 +156,7 @@
 ;
 ; X64-AVX-LABEL: test4:
 ; X64-AVX:       # %bb.0:
-; X64-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; X64-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,3,3]
 ; X64-AVX-NEXT:    vmovaps %xmm0, (%rdi)
 ; X64-AVX-NEXT:    retq
 	%tmp5 = shufflevector <4 x float> %X, <4 x float> undef, <4 x i32> < i32 2, i32 6, i32 3, i32 7 >		; <<4 x float>> [#uses=1]
@@ -448,7 +448,7 @@
 ; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-AVX-NEXT:    vmovaps (%edx), %xmm0
 ; X86-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1],mem[0,1]
-; X86-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; X86-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; X86-AVX-NEXT:    vmovaps %xmm0, (%eax)
 ; X86-AVX-NEXT:    retl
 ;
@@ -464,7 +464,7 @@
 ; X64-AVX:       # %bb.0:
 ; X64-AVX-NEXT:    vmovaps (%rdx), %xmm0
 ; X64-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1],mem[0,1]
-; X64-AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; X64-AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; X64-AVX-NEXT:    vmovaps %xmm0, (%rdi)
 ; X64-AVX-NEXT:    retq
   %tmp3 = load <4 x float>, ptr %B            ; <<4 x float>> [#uses=1]
diff --git a/llvm/test/CodeGen/X86/sse3-avx-addsub-2.ll b/llvm/test/CodeGen/X86/sse3-avx-addsub-2.ll
--- a/llvm/test/CodeGen/X86/sse3-avx-addsub-2.ll
+++ b/llvm/test/CodeGen/X86/sse3-avx-addsub-2.ll
@@ -338,16 +338,16 @@
 ;
 ; AVX1-LABEL: test13:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX512-LABEL: test13:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vbroadcastss %xmm0, %xmm0
 ; AVX512-NEXT:    retq
@@ -407,8 +407,8 @@
 ; AVX1-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovsldup {{.*#+}} xmm1 = xmm2[0,0,2,2]
 ; AVX1-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
@@ -419,8 +419,8 @@
 ; AVX512-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; AVX512-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vbroadcastss %xmm2, %xmm1
 ; AVX512-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
@@ -469,8 +469,8 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm5, %xmm2
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[2,3]
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm4[0],xmm2[3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/sse41.ll b/llvm/test/CodeGen/X86/sse41.ll
--- a/llvm/test/CodeGen/X86/sse41.ll
+++ b/llvm/test/CodeGen/X86/sse41.ll
@@ -194,7 +194,7 @@
 ; X86-AVX1-LABEL: ext_1:
 ; X86-AVX1:       ## %bb.0:
 ; X86-AVX1-NEXT:    pushl %eax ## encoding: [0x50]
-; X86-AVX1-NEXT:    vpermilps $255, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X86-AVX1-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X86-AVX1-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X86-AVX1-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0 ## encoding: [0xc5,0xfa,0x58,0x05,A,A,A,A]
 ; X86-AVX1-NEXT:    ## fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
@@ -206,7 +206,7 @@
 ; X86-AVX512-LABEL: ext_1:
 ; X86-AVX512:       ## %bb.0:
 ; X86-AVX512-NEXT:    pushl %eax ## encoding: [0x50]
-; X86-AVX512-NEXT:    vpermilps $255, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X86-AVX512-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X86-AVX512-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X86-AVX512-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x58,0x05,A,A,A,A]
 ; X86-AVX512-NEXT:    ## fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
@@ -225,7 +225,7 @@
 ;
 ; X64-AVX1-LABEL: ext_1:
 ; X64-AVX1:       ## %bb.0:
-; X64-AVX1-NEXT:    vpermilps $255, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X64-AVX1-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X64-AVX1-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X64-AVX1-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 ## encoding: [0xc5,0xfa,0x58,0x05,A,A,A,A]
 ; X64-AVX1-NEXT:    ## fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
@@ -233,7 +233,7 @@
 ;
 ; X64-AVX512-LABEL: ext_1:
 ; X64-AVX512:       ## %bb.0:
-; X64-AVX512-NEXT:    vpermilps $255, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X64-AVX512-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X64-AVX512-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X64-AVX512-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x58,0x05,A,A,A,A]
 ; X64-AVX512-NEXT:    ## fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
@@ -257,7 +257,7 @@
 ; X86-AVX1-LABEL: ext_2:
 ; X86-AVX1:       ## %bb.0:
 ; X86-AVX1-NEXT:    pushl %eax ## encoding: [0x50]
-; X86-AVX1-NEXT:    vpermilps $255, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X86-AVX1-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X86-AVX1-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X86-AVX1-NEXT:    vmovss %xmm0, (%esp) ## encoding: [0xc5,0xfa,0x11,0x04,0x24]
 ; X86-AVX1-NEXT:    flds (%esp) ## encoding: [0xd9,0x04,0x24]
@@ -267,7 +267,7 @@
 ; X86-AVX512-LABEL: ext_2:
 ; X86-AVX512:       ## %bb.0:
 ; X86-AVX512-NEXT:    pushl %eax ## encoding: [0x50]
-; X86-AVX512-NEXT:    vpermilps $255, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X86-AVX512-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X86-AVX512-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X86-AVX512-NEXT:    vmovss %xmm0, (%esp) ## EVEX TO VEX Compression encoding: [0xc5,0xfa,0x11,0x04,0x24]
 ; X86-AVX512-NEXT:    flds (%esp) ## encoding: [0xd9,0x04,0x24]
@@ -282,13 +282,13 @@
 ;
 ; X64-AVX1-LABEL: ext_2:
 ; X64-AVX1:       ## %bb.0:
-; X64-AVX1-NEXT:    vpermilps $255, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X64-AVX1-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X64-AVX1-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X64-AVX1-NEXT:    retq ## encoding: [0xc3]
 ;
 ; X64-AVX512-LABEL: ext_2:
 ; X64-AVX512:       ## %bb.0:
-; X64-AVX512-NEXT:    vpermilps $255, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; X64-AVX512-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; X64-AVX512-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; X64-AVX512-NEXT:    retq ## encoding: [0xc3]
   %s = extractelement <4 x float> %v, i32 3
@@ -696,7 +696,7 @@
 ;
 ; AVX1-LABEL: insertps_from_shufflevector_i32_2:
 ; AVX1:       ## %bb.0: ## %entry
-; AVX1-NEXT:    vpermilps $238, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; AVX1-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; AVX1-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -704,7 +704,7 @@
 ;
 ; AVX512-LABEL: insertps_from_shufflevector_i32_2:
 ; AVX512:       ## %bb.0: ## %entry
-; AVX512-NEXT:    vpermilps $238, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; AVX512-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; AVX512-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; AVX512-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; AVX512-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -1096,7 +1096,7 @@
 ;
 ; AVX1-LABEL: i32_shuf_XYY0:
 ; AVX1:       ## %bb.0:
-; AVX1-NEXT:    vpermilps $212, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xd4]
+; AVX1-NEXT:    vshufps $212, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xd4]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0,1,1,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX1-NEXT:    vblendps $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x08]
@@ -1105,7 +1105,7 @@
 ;
 ; AVX512-LABEL: i32_shuf_XYY0:
 ; AVX512:       ## %bb.0:
-; AVX512-NEXT:    vpermilps $212, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xd4]
+; AVX512-NEXT:    vshufps $212, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xd4]
 ; AVX512-NEXT:    ## xmm0 = xmm0[0,1,1,3]
 ; AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX512-NEXT:    vblendps $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x08]
@@ -1132,7 +1132,7 @@
 ;
 ; AVX1-LABEL: i32_shuf_XYW0:
 ; AVX1:       ## %bb.0:
-; AVX1-NEXT:    vpermilps $244, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xf4]
+; AVX1-NEXT:    vshufps $244, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xf4]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0,1,3,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX1-NEXT:    vblendps $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x08]
@@ -1141,7 +1141,7 @@
 ;
 ; AVX512-LABEL: i32_shuf_XYW0:
 ; AVX512:       ## %bb.0:
-; AVX512-NEXT:    vpermilps $244, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xf4]
+; AVX512-NEXT:    vshufps $244, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xf4]
 ; AVX512-NEXT:    ## xmm0 = xmm0[0,1,3,3]
 ; AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX512-NEXT:    vblendps $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x08]
@@ -1169,7 +1169,7 @@
 ;
 ; AVX1-LABEL: i32_shuf_W00W:
 ; AVX1:       ## %bb.0:
-; AVX1-NEXT:    vpermilps $255, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; AVX1-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; AVX1-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX1-NEXT:    vblendps $6, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x06]
@@ -1178,7 +1178,7 @@
 ;
 ; AVX512-LABEL: i32_shuf_W00W:
 ; AVX512:       ## %bb.0:
-; AVX512-NEXT:    vpermilps $255, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc0,0xff]
+; AVX512-NEXT:    vshufps $255, %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xc0,0xff]
 ; AVX512-NEXT:    ## xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0x57,0xc9]
 ; AVX512-NEXT:    vblendps $6, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x06]
@@ -1209,7 +1209,7 @@
 ; AVX1-NEXT:    vxorps %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe8,0x57,0xd2]
 ; AVX1-NEXT:    vblendps $1, %xmm0, %xmm2, %xmm0 ## encoding: [0xc4,0xe3,0x69,0x0c,0xc0,0x01]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0],xmm2[1,2,3]
-; AVX1-NEXT:    vpermilps $0, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x04,0xc9,0x00]
+; AVX1-NEXT:    vshufps $0, %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0xc6,0xc9,0x00]
 ; AVX1-NEXT:    ## xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vblendps $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x08]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0,1,2],xmm1[3]
@@ -1245,7 +1245,7 @@
 ; AVX1-LABEL: i32_shuf_X00X:
 ; AVX1:       ## %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
-; AVX1-NEXT:    vpermilps $0, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x04,0xc0,0x00]
+; AVX1-NEXT:    vshufps $0, %xmm0, %xmm0, %xmm0 ## encoding: [0xc5,0xf8,0xc6,0xc0,0x00]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vblendps $6, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x06]
 ; AVX1-NEXT:    ## xmm0 = xmm0[0],xmm1[1,2],xmm0[3]
@@ -1880,7 +1880,7 @@
 ; X86-AVX1-LABEL: insertps_pr20411:
 ; X86-AVX1:       ## %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT:    vpermilps $238, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; X86-AVX1-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; X86-AVX1-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; X86-AVX1-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; X86-AVX1-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -1890,7 +1890,7 @@
 ; X86-AVX512-LABEL: insertps_pr20411:
 ; X86-AVX512:       ## %bb.0:
 ; X86-AVX512-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX512-NEXT:    vpermilps $238, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; X86-AVX512-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; X86-AVX512-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; X86-AVX512-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; X86-AVX512-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -1908,7 +1908,7 @@
 ;
 ; X64-AVX1-LABEL: insertps_pr20411:
 ; X64-AVX1:       ## %bb.0:
-; X64-AVX1-NEXT:    vpermilps $238, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; X64-AVX1-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; X64-AVX1-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; X64-AVX1-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; X64-AVX1-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -1917,7 +1917,7 @@
 ;
 ; X64-AVX512-LABEL: insertps_pr20411:
 ; X64-AVX512:       ## %bb.0:
-; X64-AVX512-NEXT:    vpermilps $238, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xc9,0xee]
+; X64-AVX512-NEXT:    vshufps $238, %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf0,0xc6,0xc9,0xee]
 ; X64-AVX512-NEXT:    ## xmm1 = xmm1[2,3,2,3]
 ; X64-AVX512-NEXT:    vblendps $2, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0c,0xc1,0x02]
 ; X64-AVX512-NEXT:    ## xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
diff --git a/llvm/test/CodeGen/X86/swizzle-avx2.ll b/llvm/test/CodeGen/X86/swizzle-avx2.ll
--- a/llvm/test/CodeGen/X86/swizzle-avx2.ll
+++ b/llvm/test/CodeGen/X86/swizzle-avx2.ll
@@ -25,7 +25,7 @@
 define <8 x i32> @swizzle_2(<8 x i32> %v) {
 ; CHECK-LABEL: swizzle_2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; CHECK-NEXT:    retq
   %1 = shufflevector <8 x i32> %v, <8 x i32> undef, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 0, i32 1, i32 2, i32 3>
   %2 = shufflevector <8 x i32> %1, <8 x i32> undef, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 0, i32 1, i32 2, i32 3>
@@ -35,7 +35,7 @@
 define <8 x i32> @swizzle_3(<8 x i32> %v) {
 ; CHECK-LABEL: swizzle_3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; CHECK-NEXT:    retq
   %1 = shufflevector <8 x i32> %v, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
   %2 = shufflevector <8 x i32> %1, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
diff --git a/llvm/test/CodeGen/X86/tuning-shuffle-permilps-avx512.ll b/llvm/test/CodeGen/X86/tuning-shuffle-permilps-avx512.ll
--- a/llvm/test/CodeGen/X86/tuning-shuffle-permilps-avx512.ll
+++ b/llvm/test/CodeGen/X86/tuning-shuffle-permilps-avx512.ll
@@ -7,7 +7,7 @@
 define <16 x float> @transform_VPERMILPSZrr(<16 x float> %a) nounwind {
 ; CHECK-LABEL: transform_VPERMILPSZrr:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; CHECK-NEXT:    retq
   %shufp = shufflevector <16 x float> %a, <16 x float> poison, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   ret <16 x float> %shufp
@@ -16,7 +16,7 @@
 define <8 x float> @transform_VPERMILPSYrr(<8 x float> %a) nounwind {
 ; CHECK-LABEL: transform_VPERMILPSYrr:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; CHECK-NEXT:    retq
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
@@ -25,7 +25,7 @@
 define <4 x float> @transform_VPERMILPSrr(<4 x float> %a) nounwind {
 ; CHECK-LABEL: transform_VPERMILPSrr:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; CHECK-NEXT:    retq
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
@@ -107,30 +107,75 @@
 }
 
 define <16 x float> @transform_VPERMILPSZrm(ptr %ap) nounwind {
-; CHECK-LABEL: transform_VPERMILPSZrm:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
-; CHECK-NEXT:    retq
+; CHECK-ICX-LABEL: transform_VPERMILPSZrm:
+; CHECK-ICX:       # %bb.0:
+; CHECK-ICX-NEXT:    vpshufd {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-ICX-NEXT:    retq
+;
+; CHECK-V4-LABEL: transform_VPERMILPSZrm:
+; CHECK-V4:       # %bb.0:
+; CHECK-V4-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-V4-NEXT:    retq
+;
+; CHECK-AVX512-LABEL: transform_VPERMILPSZrm:
+; CHECK-AVX512:       # %bb.0:
+; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-AVX512-NEXT:    retq
+;
+; CHECK-ZNVER4-LABEL: transform_VPERMILPSZrm:
+; CHECK-ZNVER4:       # %bb.0:
+; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-ZNVER4-NEXT:    retq
   %a = load <16 x float>, ptr %ap
   %shufp = shufflevector <16 x float> %a, <16 x float> poison, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   ret <16 x float> %shufp
 }
 
 define <8 x float> @transform_VPERMILPSYrm(ptr %ap) nounwind {
-; CHECK-LABEL: transform_VPERMILPSYrm:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
-; CHECK-NEXT:    retq
+; CHECK-ICX-LABEL: transform_VPERMILPSYrm:
+; CHECK-ICX:       # %bb.0:
+; CHECK-ICX-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ICX-NEXT:    retq
+;
+; CHECK-V4-LABEL: transform_VPERMILPSYrm:
+; CHECK-V4:       # %bb.0:
+; CHECK-V4-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-V4-NEXT:    retq
+;
+; CHECK-AVX512-LABEL: transform_VPERMILPSYrm:
+; CHECK-AVX512:       # %bb.0:
+; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX512-NEXT:    retq
+;
+; CHECK-ZNVER4-LABEL: transform_VPERMILPSYrm:
+; CHECK-ZNVER4:       # %bb.0:
+; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-ZNVER4-NEXT:    retq
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
 }
 
 define <4 x float> @transform_VPERMILPSrm(ptr %ap) nounwind {
-; CHECK-LABEL: transform_VPERMILPSrm:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
-; CHECK-NEXT:    retq
+; CHECK-ICX-LABEL: transform_VPERMILPSrm:
+; CHECK-ICX:       # %bb.0:
+; CHECK-ICX-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ICX-NEXT:    retq
+;
+; CHECK-V4-LABEL: transform_VPERMILPSrm:
+; CHECK-V4:       # %bb.0:
+; CHECK-V4-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-V4-NEXT:    retq
+;
+; CHECK-AVX512-LABEL: transform_VPERMILPSrm:
+; CHECK-AVX512:       # %bb.0:
+; CHECK-AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX512-NEXT:    retq
+;
+; CHECK-ZNVER4-LABEL: transform_VPERMILPSrm:
+; CHECK-ZNVER4:       # %bb.0:
+; CHECK-ZNVER4-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-ZNVER4-NEXT:    retq
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
@@ -213,9 +258,3 @@
   %res = select <4 x i1> %mask, <4 x float> %shufp, <4 x float> %b
   ret <4 x float> %res
 }
-
-;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; CHECK-AVX512: {{.*}}
-; CHECK-ICX: {{.*}}
-; CHECK-V4: {{.*}}
-; CHECK-ZNVER4: {{.*}}
diff --git a/llvm/test/CodeGen/X86/tuning-shuffle-permilps.ll b/llvm/test/CodeGen/X86/tuning-shuffle-permilps.ll
--- a/llvm/test/CodeGen/X86/tuning-shuffle-permilps.ll
+++ b/llvm/test/CodeGen/X86/tuning-shuffle-permilps.ll
@@ -7,7 +7,7 @@
 define <8 x float> @transform_VPERMILPSYrr(<8 x float> %a) nounwind {
 ; CHECK-LABEL: transform_VPERMILPSYrr:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; CHECK-NEXT:    retq
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
@@ -16,33 +16,58 @@
 define <4 x float> @transform_VPERMILPSrr(<4 x float> %a) nounwind {
 ; CHECK-LABEL: transform_VPERMILPSrr:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; CHECK-NEXT:    retq
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
 }
 
 define <8 x float> @transform_VPERMILPSYrm(ptr %ap) nounwind {
-; CHECK-LABEL: transform_VPERMILPSYrm:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
-; CHECK-NEXT:    retq
+; CHECK-AVX1-LABEL: transform_VPERMILPSYrm:
+; CHECK-AVX1:       # %bb.0:
+; CHECK-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX1-NEXT:    retq
+;
+; CHECK-AVX1-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-AVX1-DELAY:       # %bb.0:
+; CHECK-AVX1-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX1-DELAY-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: transform_VPERMILPSYrm:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-AVX2-DELAY-LABEL: transform_VPERMILPSYrm:
+; CHECK-AVX2-DELAY:       # %bb.0:
+; CHECK-AVX2-DELAY-NEXT:    vpermilps {{.*#+}} ymm0 = mem[3,2,1,0,7,6,5,4]
+; CHECK-AVX2-DELAY-NEXT:    retq
   %a = load <8 x float>, ptr %ap
   %shufp = shufflevector <8 x float> %a, <8 x float> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shufp
 }
 
 define <4 x float> @transform_VPERMILPSrm(ptr %ap) nounwind {
-; CHECK-LABEL: transform_VPERMILPSrm:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
-; CHECK-NEXT:    retq
+; CHECK-AVX1-LABEL: transform_VPERMILPSrm:
+; CHECK-AVX1:       # %bb.0:
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX1-NEXT:    retq
+;
+; CHECK-AVX1-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-AVX1-DELAY:       # %bb.0:
+; CHECK-AVX1-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX1-DELAY-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: transform_VPERMILPSrm:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-AVX2-DELAY-LABEL: transform_VPERMILPSrm:
+; CHECK-AVX2-DELAY:       # %bb.0:
+; CHECK-AVX2-DELAY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,2,1,0]
+; CHECK-AVX2-DELAY-NEXT:    retq
   %a = load <4 x float>, ptr %ap
   %shufp = shufflevector <4 x float> %a, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shufp
 }
-;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; CHECK-AVX1: {{.*}}
-; CHECK-AVX1-DELAY: {{.*}}
-; CHECK-AVX2: {{.*}}
-; CHECK-AVX2-DELAY: {{.*}}
diff --git a/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll b/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
--- a/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
@@ -607,7 +607,7 @@
 ;
 ; AVX-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64:
 ; AVX-64:       # %bb.0:
-; AVX-64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX-64-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX-64-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX-64-NEXT:    vmovq %rax, %xmm1
 ; AVX-64-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -661,7 +661,7 @@
 ;
 ; AVX512F-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64:
 ; AVX512F-64:       # %bb.0:
-; AVX512F-64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512F-64-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512F-64-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX512F-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512F-64-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -715,7 +715,7 @@
 ;
 ; AVX512VL-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64:
 ; AVX512VL-64:       # %bb.0:
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -774,7 +774,7 @@
 ; AVX-32-NEXT:    movzbl %al, %eax
 ; AVX-32-NEXT:    shll $31, %eax
 ; AVX-32-NEXT:    xorl {{[0-9]+}}(%esp), %eax
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX-32-NEXT:    vcomiss %xmm1, %xmm2
 ; AVX-32-NEXT:    vmovaps %xmm1, %xmm3
 ; AVX-32-NEXT:    jae .LBB3_4
@@ -836,7 +836,7 @@
 ;
 ; AVX-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64:
 ; AVX-64:       # %bb.0:
-; AVX-64-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-64-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-64-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX-64-NEXT:    vcomiss %xmm1, %xmm3
 ; AVX-64-NEXT:    vxorps %xmm2, %xmm2, %xmm2
@@ -908,7 +908,7 @@
 ; AVX512F-32-NEXT:    andl $-8, %esp
 ; AVX512F-32-NEXT:    subl $40, %esp
 ; AVX512F-32-NEXT:    .cfi_offset %ebx, -12
-; AVX512F-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512F-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512F-32-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX512F-32-NEXT:    xorl %eax, %eax
 ; AVX512F-32-NEXT:    vcomiss %xmm1, %xmm2
@@ -974,7 +974,7 @@
 ;
 ; AVX512F-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64:
 ; AVX512F-64:       # %bb.0:
-; AVX512F-64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512F-64-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512F-64-NEXT:    vcvttss2usi %xmm1, %rax
 ; AVX512F-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512F-64-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1001,7 +1001,7 @@
 ; AVX512VL-32-NEXT:    andl $-8, %esp
 ; AVX512VL-32-NEXT:    subl $40, %esp
 ; AVX512VL-32-NEXT:    .cfi_offset %ebx, -12
-; AVX512VL-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512VL-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512VL-32-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX512VL-32-NEXT:    xorl %eax, %eax
 ; AVX512VL-32-NEXT:    vcomiss %xmm1, %xmm2
@@ -1067,7 +1067,7 @@
 ;
 ; AVX512VL-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64:
 ; AVX512VL-64:       # %bb.0:
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2usi %xmm1, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
diff --git a/llvm/test/CodeGen/X86/vec-strict-fptoint-512.ll b/llvm/test/CodeGen/X86/vec-strict-fptoint-512.ll
--- a/llvm/test/CodeGen/X86/vec-strict-fptoint-512.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-fptoint-512.ll
@@ -385,7 +385,7 @@
 ; AVX512VL-64-LABEL: strict_vector_fptosi_v8f32_to_v8i64:
 ; AVX512VL-64:       # %bb.0:
 ; AVX512VL-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2si %xmm2, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm2
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
@@ -399,7 +399,7 @@
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-64-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm3[0],xmm1[0]
 ; AVX512VL-64-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2si %xmm2, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm2
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
@@ -442,7 +442,7 @@
 ; AVX512VL-32-NEXT:    .cfi_offset %edi, -16
 ; AVX512VL-32-NEXT:    .cfi_offset %ebx, -12
 ; AVX512VL-32-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX512VL-32-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[3,3,3,3]
+; AVX512VL-32-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,3,3,3]
 ; AVX512VL-32-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX512VL-32-NEXT:    xorl %eax, %eax
 ; AVX512VL-32-NEXT:    vcomiss %xmm1, %xmm3
@@ -477,7 +477,7 @@
 ; AVX512VL-32-NEXT:    vmovss %xmm1, %xmm1, %xmm3 {%k1} {z}
 ; AVX512VL-32-NEXT:    vsubss %xmm3, %xmm2, %xmm2
 ; AVX512VL-32-NEXT:    vmovss %xmm2, {{[0-9]+}}(%esp)
-; AVX512VL-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512VL-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512VL-32-NEXT:    xorl %eax, %eax
 ; AVX512VL-32-NEXT:    vcomiss %xmm1, %xmm2
 ; AVX512VL-32-NEXT:    setae %al
@@ -574,7 +574,7 @@
 ; AVX512VL-64-LABEL: strict_vector_fptoui_v8f32_to_v8i64:
 ; AVX512VL-64:       # %bb.0:
 ; AVX512VL-64-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2usi %xmm2, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm2
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
@@ -588,7 +588,7 @@
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-64-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm3[0],xmm1[0]
 ; AVX512VL-64-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
-; AVX512VL-64-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512VL-64-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512VL-64-NEXT:    vcvttss2usi %xmm2, %rax
 ; AVX512VL-64-NEXT:    vmovq %rax, %xmm2
 ; AVX512VL-64-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
@@ -209,7 +209,7 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $24, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fstps (%esp)
@@ -236,7 +236,7 @@
 ; AVX512DQ-32:       # %bb.0:
 ; AVX512DQ-32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; AVX512DQ-32-NEXT:    vcvtqq2ps %zmm0, %ymm1
-; AVX512DQ-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX512DQ-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX512DQ-32-NEXT:    vcvtqq2ps %zmm0, %ymm0
 ; AVX512DQ-32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],zero,zero
 ; AVX512DQ-32-NEXT:    vzeroupper
@@ -410,7 +410,7 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $24, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
 ; AVX-32-NEXT:    shrl $31, %eax
@@ -471,7 +471,7 @@
 ; AVX512DQ-32:       # %bb.0:
 ; AVX512DQ-32-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
 ; AVX512DQ-32-NEXT:    vcvtuqq2ps %zmm0, %ymm1
-; AVX512DQ-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX512DQ-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX512DQ-32-NEXT:    vcvtuqq2ps %zmm0, %ymm0
 ; AVX512DQ-32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],zero,zero
 ; AVX512DQ-32-NEXT:    vzeroupper
@@ -887,21 +887,21 @@
 ;
 ; AVX1-32-LABEL: uitofp_v2i1_v2f64:
 ; AVX1-32:       # %bb.0:
-; AVX1-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-32-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
 ; AVX1-32-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX1-32-NEXT:    retl
 ;
 ; AVX1-64-LABEL: uitofp_v2i1_v2f64:
 ; AVX1-64:       # %bb.0:
-; AVX1-64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-64-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX1-64-NEXT:    retq
 ;
 ; AVX512F-LABEL: uitofp_v2i1_v2f64:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX512F-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1,1,1,1]
 ; AVX512F-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX512F-NEXT:    vcvtdq2pd %xmm0, %xmm0
@@ -923,7 +923,7 @@
 ;
 ; AVX512DQ-LABEL: uitofp_v2i1_v2f64:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512DQ-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX512DQ-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1,1,1,1]
 ; AVX512DQ-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vcvtdq2pd %xmm0, %xmm0
@@ -931,14 +931,14 @@
 ;
 ; AVX512DQVL-32-LABEL: uitofp_v2i1_v2f64:
 ; AVX512DQVL-32:       # %bb.0:
-; AVX512DQVL-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512DQVL-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX512DQVL-32-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %xmm0, %xmm0
 ; AVX512DQVL-32-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX512DQVL-32-NEXT:    retl
 ;
 ; AVX512DQVL-64-LABEL: uitofp_v2i1_v2f64:
 ; AVX512DQVL-64:       # %bb.0:
-; AVX512DQVL-64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512DQVL-64-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX512DQVL-64-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
 ; AVX512DQVL-64-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX512DQVL-64-NEXT:    retq
@@ -1218,7 +1218,7 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $32, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
@@ -1406,7 +1406,7 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $32, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
 ; AVX-32-NEXT:    shrl $31, %eax
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
@@ -641,11 +641,11 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $64, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
@@ -758,11 +758,11 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $64, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
 ; AVX-32-NEXT:    shrl $31, %eax
@@ -919,11 +919,11 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $48, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    fstps {{[0-9]+}}(%esp)
@@ -1042,11 +1042,11 @@
 ; AVX-32-NEXT:    andl $-8, %esp
 ; AVX-32-NEXT:    subl $48, %esp
 ; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; AVX-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX-32-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
 ; AVX-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
 ; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
 ; AVX-32-NEXT:    shrl $31, %eax
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
@@ -273,18 +273,18 @@
 ; NODQ-32-NEXT:    subl $128, %esp
 ; NODQ-32-NEXT:    vextractf32x4 $2, %zmm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $3, %zmm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
@@ -371,18 +371,18 @@
 ; NODQ-32-NEXT:    subl $128, %esp
 ; NODQ-32-NEXT:    vextractf32x4 $2, %zmm0, %xmm3
 ; NODQ-32-NEXT:    vmovlps %xmm3, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $3, %zmm0, %xmm2
 ; NODQ-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm4, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractps $1, %xmm3, %eax
 ; NODQ-32-NEXT:    shrl $31, %eax
@@ -499,19 +499,19 @@
 ; NODQ-32-NEXT:    andl $-8, %esp
 ; NODQ-32-NEXT:    subl $96, %esp
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $2, %zmm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    fstps {{[0-9]+}}(%esp)
@@ -595,19 +595,19 @@
 ; NODQ-32-NEXT:    andl $-8, %esp
 ; NODQ-32-NEXT:    subl $96, %esp
 ; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; NODQ-32-NEXT:    vmovlps %xmm3, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
 ; NODQ-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractf32x4 $3, %zmm0, %xmm1
 ; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
-; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[2,3,2,3]
 ; NODQ-32-NEXT:    vmovlps %xmm4, {{[0-9]+}}(%esp)
 ; NODQ-32-NEXT:    vextractps $1, %xmm0, %eax
 ; NODQ-32-NEXT:    shrl $31, %eax
diff --git a/llvm/test/CodeGen/X86/vec_fp_to_int.ll b/llvm/test/CodeGen/X86/vec_fp_to_int.ll
--- a/llvm/test/CodeGen/X86/vec_fp_to_int.ll
+++ b/llvm/test/CodeGen/X86/vec_fp_to_int.ll
@@ -954,7 +954,7 @@
 ;
 ; AVX1-LABEL: fptosi_4f32_to_4i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX1-NEXT:    vmovq %rax, %xmm1
 ; AVX1-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -972,7 +972,7 @@
 ;
 ; AVX2-LABEL: fptosi_4f32_to_4i64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX2-NEXT:    vmovq %rax, %xmm1
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -990,7 +990,7 @@
 ;
 ; AVX512F-LABEL: fptosi_4f32_to_4i64:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX512F-NEXT:    vmovq %rax, %xmm1
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1008,7 +1008,7 @@
 ;
 ; AVX512VL-LABEL: fptosi_4f32_to_4i64:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX512VL-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1062,7 +1062,7 @@
 ;
 ; AVX1-LABEL: fptosi_8f32_to_4i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX1-NEXT:    vmovq %rax, %xmm1
 ; AVX1-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1080,7 +1080,7 @@
 ;
 ; AVX2-LABEL: fptosi_8f32_to_4i64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vcvttss2si %xmm1, %rax
 ; AVX2-NEXT:    vmovq %rax, %xmm1
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1103,7 +1103,7 @@
 ; AVX512F-NEXT:    vcvttss2si %xmm0, %rcx
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; AVX512F-NEXT:    vcvttss2si %xmm1, %rdx
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvttss2si %xmm0, %rsi
 ; AVX512F-NEXT:    vmovq %rsi, %xmm0
 ; AVX512F-NEXT:    vmovq %rdx, %xmm1
@@ -1121,7 +1121,7 @@
 ; AVX512VL-NEXT:    vcvttss2si %xmm0, %rcx
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; AVX512VL-NEXT:    vcvttss2si %xmm1, %rdx
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvttss2si %xmm0, %rsi
 ; AVX512VL-NEXT:    vmovq %rsi, %xmm0
 ; AVX512VL-NEXT:    vmovq %rdx, %xmm1
@@ -1566,7 +1566,7 @@
 ;
 ; AVX1-LABEL: fptoui_4f32_to_4i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX1-NEXT:    vsubss %xmm1, %xmm2, %xmm3
 ; AVX1-NEXT:    vcvttss2si %xmm3, %rax
@@ -1609,7 +1609,7 @@
 ;
 ; AVX2-LABEL: fptoui_4f32_to_4i64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX2-NEXT:    vsubss %xmm1, %xmm2, %xmm3
 ; AVX2-NEXT:    vcvttss2si %xmm3, %rax
@@ -1652,7 +1652,7 @@
 ;
 ; AVX512F-LABEL: fptoui_4f32_to_4i64:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvttss2usi %xmm1, %rax
 ; AVX512F-NEXT:    vmovq %rax, %xmm1
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1670,7 +1670,7 @@
 ;
 ; AVX512VL-LABEL: fptoui_4f32_to_4i64:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvttss2usi %xmm1, %rax
 ; AVX512VL-NEXT:    vmovq %rax, %xmm1
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
@@ -1750,7 +1750,7 @@
 ;
 ; AVX1-LABEL: fptoui_8f32_to_4i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX1-NEXT:    vsubss %xmm1, %xmm2, %xmm3
 ; AVX1-NEXT:    vcvttss2si %xmm3, %rax
@@ -1793,7 +1793,7 @@
 ;
 ; AVX2-LABEL: fptoui_8f32_to_4i64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; AVX2-NEXT:    vsubss %xmm1, %xmm2, %xmm3
 ; AVX2-NEXT:    vcvttss2si %xmm3, %rax
@@ -1841,7 +1841,7 @@
 ; AVX512F-NEXT:    vcvttss2usi %xmm0, %rcx
 ; AVX512F-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; AVX512F-NEXT:    vcvttss2usi %xmm1, %rdx
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvttss2usi %xmm0, %rsi
 ; AVX512F-NEXT:    vmovq %rsi, %xmm0
 ; AVX512F-NEXT:    vmovq %rdx, %xmm1
@@ -1859,7 +1859,7 @@
 ; AVX512VL-NEXT:    vcvttss2usi %xmm0, %rcx
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; AVX512VL-NEXT:    vcvttss2usi %xmm1, %rdx
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvttss2usi %xmm0, %rsi
 ; AVX512VL-NEXT:    vmovq %rsi, %xmm0
 ; AVX512VL-NEXT:    vmovq %rdx, %xmm1
diff --git a/llvm/test/CodeGen/X86/vec_int_to_fp.ll b/llvm/test/CodeGen/X86/vec_int_to_fp.ll
--- a/llvm/test/CodeGen/X86/vec_int_to_fp.ll
+++ b/llvm/test/CodeGen/X86/vec_int_to_fp.ll
@@ -933,7 +933,7 @@
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-NEXT:    vorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    vorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vaddpd %ymm0, %ymm2, %ymm0
@@ -3631,7 +3631,7 @@
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-NEXT:    vorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm2
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,3],ymm0[1,3],ymm1[5,7],ymm0[5,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
 ; AVX1-NEXT:    vorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vaddpd %ymm0, %ymm2, %ymm0
@@ -5431,7 +5431,7 @@
 ;
 ; AVX-LABEL: extract3_sitofp_v4i32_f32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vcvtdq2ps %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %e = extractelement <4 x i32> %x, i32 3
@@ -5457,7 +5457,7 @@
 ;
 ; AVX-LABEL: extract3_sitofp_v4i32_f64:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %e = extractelement <4 x i32> %x, i32 3
@@ -5489,7 +5489,7 @@
 ;
 ; AVX512F-LABEL: extract3_uitofp_v4i32_f32:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvtudq2ps %zmm0, %zmm0
 ; AVX512F-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512F-NEXT:    vzeroupper
@@ -5497,13 +5497,13 @@
 ;
 ; AVX512VL-LABEL: extract3_uitofp_v4i32_f32:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvtudq2ps %xmm0, %xmm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: extract3_uitofp_v4i32_f32:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512DQ-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512DQ-NEXT:    vcvtudq2ps %zmm0, %zmm0
 ; AVX512DQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512DQ-NEXT:    vzeroupper
@@ -5511,7 +5511,7 @@
 ;
 ; AVX512VLDQ-LABEL: extract3_uitofp_v4i32_f32:
 ; AVX512VLDQ:       # %bb.0:
-; AVX512VLDQ-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VLDQ-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VLDQ-NEXT:    vcvtudq2ps %xmm0, %xmm0
 ; AVX512VLDQ-NEXT:    retq
   %e = extractelement <4 x i32> %x, i32 3
@@ -5543,7 +5543,7 @@
 ;
 ; AVX512F-LABEL: extract3_uitofp_v4i32_f64:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512F-NEXT:    vcvtudq2pd %ymm0, %zmm0
 ; AVX512F-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512F-NEXT:    vzeroupper
@@ -5551,13 +5551,13 @@
 ;
 ; AVX512VL-LABEL: extract3_uitofp_v4i32_f64:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vcvtudq2pd %xmm0, %xmm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: extract3_uitofp_v4i32_f64:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512DQ-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512DQ-NEXT:    vcvtudq2pd %ymm0, %zmm0
 ; AVX512DQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512DQ-NEXT:    vzeroupper
@@ -5565,7 +5565,7 @@
 ;
 ; AVX512VLDQ-LABEL: extract3_uitofp_v4i32_f64:
 ; AVX512VLDQ:       # %bb.0:
-; AVX512VLDQ-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512VLDQ-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512VLDQ-NEXT:    vcvtudq2pd %xmm0, %xmm0
 ; AVX512VLDQ-NEXT:    retq
   %e = extractelement <4 x i32> %x, i32 3
diff --git a/llvm/test/CodeGen/X86/vec_umulo.ll b/llvm/test/CodeGen/X86/vec_umulo.ll
--- a/llvm/test/CodeGen/X86/vec_umulo.ll
+++ b/llvm/test/CodeGen/X86/vec_umulo.ll
@@ -2748,8 +2748,8 @@
 ; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm2 = [16777215,16777215,16777215,16777215]
 ; AVX1-NEXT:    vandps %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vandps %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
diff --git a/llvm/test/CodeGen/X86/vector-fshr-256.ll b/llvm/test/CodeGen/X86/vector-fshr-256.ll
--- a/llvm/test/CodeGen/X86/vector-fshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-256.ll
@@ -43,7 +43,7 @@
 ; AVX1-NEXT:    vpsllq %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vpsllq %xmm3, %xmm5, %xmm3
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4,5,6,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm4 = xmm2[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,3,2,3]
 ; AVX1-NEXT:    vpaddq %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsllq %xmm4, %xmm0, %xmm4
 ; AVX1-NEXT:    vpsllq %xmm2, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-half-conversions.ll b/llvm/test/CodeGen/X86/vector-half-conversions.ll
--- a/llvm/test/CodeGen/X86/vector-half-conversions.ll
+++ b/llvm/test/CodeGen/X86/vector-half-conversions.ll
@@ -2252,7 +2252,7 @@
 ; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    callq __truncsfhf2@PLT
 ; AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2324,7 +2324,7 @@
 ; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2373,7 +2373,7 @@
 ; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2429,7 +2429,7 @@
 ; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2478,7 +2478,7 @@
 ; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2729,7 +2729,7 @@
 ; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    callq __truncsfhf2@PLT
 ; AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2806,7 +2806,7 @@
 ; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2855,7 +2855,7 @@
 ; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2916,7 +2916,7 @@
 ; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -2965,7 +2965,7 @@
 ; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -4432,7 +4432,7 @@
 ; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -4481,7 +4481,7 @@
 ; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -4531,7 +4531,7 @@
 ; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4580,7 +4580,7 @@
 ; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    callq __truncsfhf2@PLT
 ; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4645,7 +4645,7 @@
 ; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -4694,7 +4694,7 @@
 ; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
@@ -4744,7 +4744,7 @@
 ; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4793,7 +4793,7 @@
 ; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    callq __truncsfhf2@PLT
 ; AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
diff --git a/llvm/test/CodeGen/X86/vector-interleave.ll b/llvm/test/CodeGen/X86/vector-interleave.ll
--- a/llvm/test/CodeGen/X86/vector-interleave.ll
+++ b/llvm/test/CodeGen/X86/vector-interleave.ll
@@ -538,10 +538,10 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovups (%rdi), %xmm0
 ; AVX1-NEXT:    vmovups 16(%rdi), %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX1-NEXT:    vmovups %xmm1, 48(%rsi)
 ; AVX1-NEXT:    vmovups %xmm3, 32(%rsi)
 ; AVX1-NEXT:    vmovups %xmm0, 16(%rsi)
@@ -551,8 +551,8 @@
 ; AVX2-LABEL: splat2_i32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpermpd $216, (%rdi), %ymm0 # ymm0 = mem[0,2,1,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
 ; AVX2-NEXT:    vmovups %ymm0, 32(%rsi)
 ; AVX2-NEXT:    vmovups %ymm1, (%rsi)
 ; AVX2-NEXT:    vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -1170,7 +1170,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{.*#+}} ymm11 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,65535,65535,65535]
 ; AVX1-ONLY-NEXT:    vandps %ymm11, %ymm9, %ymm12
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm13, %ymm11, %ymm13
 ; AVX1-ONLY-NEXT:    vorps %ymm13, %ymm12, %ymm12
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm12, %ymm10
@@ -2278,7 +2278,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm3, %ymm14, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm14, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm0
@@ -2321,7 +2321,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm14, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm8, %ymm14, %ymm8
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm8, %ymm0
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm0
@@ -4701,7 +4701,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm6, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm6, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm0
@@ -4744,7 +4744,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm6, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm6, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm0
@@ -4787,7 +4787,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm6, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm6, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm0
@@ -4830,7 +4830,7 @@
 ; AVX1-ONLY-NEXT:    vandps %ymm6, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vandnps %ymm11, %ymm6, %ymm11
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm11, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm0
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
@@ -4080,7 +4080,7 @@
 ; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3]
 ; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[0,1,0,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[0,1,0,3]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm13[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vpunpckhqdq {{.*#+}} xmm4 = xmm1[1],xmm4[1]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
@@ -2520,7 +2520,7 @@
 ; AVX1-ONLY-NEXT:    vmovdqa %xmm11, %xmm1
 ; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm7[2],xmm11[2],xmm7[3],xmm11[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm12[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
@@ -5302,7 +5302,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
@@ -5326,12 +5326,12 @@
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = mem[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
@@ -5383,14 +5383,14 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm14[0],xmm3[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm11[2],mem[2],xmm11[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm12[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm12[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm11[1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
@@ -5409,7 +5409,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = mem[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
@@ -5618,7 +5618,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm8[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -5649,7 +5649,7 @@
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm4[0],mem[0],xmm4[1],mem[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm4[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm14 = xmm14[0],mem[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
@@ -5666,7 +5666,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
@@ -5690,7 +5690,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm7[0,1,2],xmm0[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
@@ -5705,7 +5705,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm7[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm7[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -5716,7 +5716,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm4[2],xmm9[2],xmm4[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm3[0,1,2],xmm14[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
@@ -5728,7 +5728,7 @@
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm11[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -5763,12 +5763,12 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm2[2],xmm7[2],xmm2[3],xmm7[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm6[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0],xmm1[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3]
@@ -5809,8 +5809,8 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm13[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm12[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
@@ -28,8 +28,8 @@
 ; AVX-LABEL: load_i32_stride2_vf2:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps (%rdi), %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,2,2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,2,3]
 ; AVX-NEXT:    vmovlps %xmm1, (%rsi)
 ; AVX-NEXT:    vmovlps %xmm0, (%rdx)
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -35,10 +35,10 @@
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,0,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovlps %xmm2, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovlps %xmm3, (%rdx)
@@ -49,9 +49,9 @@
 ; AVX2-ONLY:       # %bb.0:
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,3,2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; AVX2-ONLY-NEXT:    vbroadcastss 8(%rdi), %xmm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
 ; AVX2-ONLY-NEXT:    vmovlps %xmm2, (%rsi)
@@ -63,9 +63,9 @@
 ; AVX512F-SLOW:       # %bb.0:
 ; AVX512F-SLOW-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX512F-SLOW-NEXT:    vmovaps 16(%rdi), %xmm1
-; AVX512F-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX512F-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,3,2,3]
 ; AVX512F-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX512F-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX512F-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; AVX512F-SLOW-NEXT:    vbroadcastss 8(%rdi), %xmm3
 ; AVX512F-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
 ; AVX512F-SLOW-NEXT:    vmovlps %xmm2, (%rsi)
@@ -91,9 +91,9 @@
 ; AVX512BW-SLOW:       # %bb.0:
 ; AVX512BW-SLOW-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX512BW-SLOW-NEXT:    vmovaps 16(%rdi), %xmm1
-; AVX512BW-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX512BW-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,3,2,3]
 ; AVX512BW-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX512BW-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX512BW-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; AVX512BW-SLOW-NEXT:    vbroadcastss 8(%rdi), %xmm3
 ; AVX512BW-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
 ; AVX512BW-SLOW-NEXT:    vmovlps %xmm2, (%rsi)
@@ -153,14 +153,14 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0],xmm2[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2],xmm3[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,3,2,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,3,2,1]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = mem[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1,2],xmm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,0,3,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,0,3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, (%rdx)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rcx)
@@ -260,14 +260,14 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm4[0,2],ymm3[4,7],ymm4[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[1,0],ymm4[2,0],ymm0[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[2,0],ymm4[3,0],ymm0[6,4],ymm4[7,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm4[0,0],ymm5[2,0],ymm4[4,4],ymm5[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[2,0],ymm6[5,4],ymm2[6,4]
@@ -301,7 +301,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm3, (%rsi)
@@ -361,7 +361,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm1, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm3, (%rsi)
@@ -506,7 +506,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm7[0,2],ymm5[4,7],ymm7[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm3[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,0],ymm7[2,0],ymm3[5,4],ymm7[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm9
@@ -514,21 +514,21 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,3],ymm9[0,2],ymm8[4,7],ymm9[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm0[1,0],ymm9[2,0],ymm0[5,4],ymm9[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm3[2,0],ymm7[3,0],ymm3[6,4],ymm7[7,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0],ymm10[2,0],ymm7[4,4],ymm10[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm6[0,1],ymm4[2],ymm6[3,4],ymm4[5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[1,2],ymm11[0,3],ymm12[5,6],ymm11[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm0[2,0],ymm9[3,0],ymm0[6,4],ymm9[7,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm9[0,0],ymm12[2,0],ymm9[4,4],ymm12[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[1,2],ymm13[0,3],ymm14[5,6],ymm13[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4],ymm12[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm4[0,1],ymm6[2],ymm4[3,4],ymm6[5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm11[1,0],ymm6[2,0],ymm11[5,4],ymm6[6,4]
@@ -581,12 +581,12 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm3 = <2,5,0,3,6,u,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm3, %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm3, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm6, 32(%rsi)
@@ -678,12 +678,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm3 = <2,5,0,3,6,u,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm3, %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm6, 32(%rsi)
@@ -968,7 +968,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm8[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,0],ymm10[2,0],ymm8[5,4],ymm10[6,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0],ymm5[1],ymm7[2,3],ymm5[4],ymm7[5,6],ymm5[7]
@@ -979,7 +979,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm6[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[1,0],ymm14[2,0],ymm6[5,4],ymm14[6,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -990,7 +990,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm12[2,0],ymm2[5,4],ymm12[6,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm15
@@ -1002,7 +1002,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm4[1,0],ymm1[2,0],ymm4[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1012,7 +1012,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = mem[0,1],ymm9[2],mem[3,4],ymm9[5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[1,2],ymm13[0,3],ymm8[5,6],ymm13[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[2,0],ymm14[3,0],ymm6[6,4],ymm14[7,4]
@@ -1020,7 +1020,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1],ymm5[2],ymm7[3,4],ymm5[5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[1,2],ymm8[0,3],ymm10[5,6],ymm8[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[2,0],ymm1[3,0],ymm4[6,4],ymm1[7,4]
@@ -1028,7 +1028,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,2],ymm10[0,3],ymm6[5,6],ymm10[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm12[3,0],ymm2[6,4],ymm12[7,4]
@@ -1038,7 +1038,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm7 # 32-byte Folded Reload
@@ -1155,22 +1155,22 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm15[2],ymm10[3,4],ymm15[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm10 = <2,5,0,3,6,u,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm10, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm11[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2],ymm4[3,4],ymm7[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm10, %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1],ymm13[2],ymm3[3,4],ymm13[5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm5, %ymm10, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1],ymm14[2],ymm2[3,4],ymm14[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm5, %ymm10, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm6[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -1360,22 +1360,22 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm15[2],ymm10[3,4],ymm15[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm10 = <2,5,0,3,6,u,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm10, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm11[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2],ymm4[3,4],ymm7[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm10, %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1],ymm13[2],ymm3[3,4],ymm13[5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm5, %ymm10, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1],ymm14[2],ymm2[3,4],ymm14[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm5, %ymm10, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm6[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -1930,7 +1930,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,0],ymm1[2,0],ymm8[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5,6],ymm6[7]
@@ -1939,7 +1939,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm1[0,2],ymm0[4,7],ymm1[4,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm15[2,0],ymm5[5,4],ymm15[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
@@ -1949,7 +1949,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm5[2,0],ymm2[5,4],ymm5[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %ymm2
@@ -1964,7 +1964,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm14[2,0],ymm1[5,4],ymm14[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm8
@@ -1978,7 +1978,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm1[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm13[2,0],ymm1[5,4],ymm13[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm9
@@ -1993,7 +1993,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm12[2,0],ymm2[5,4],ymm12[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm4
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm10
@@ -2008,7 +2008,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm7[2,0],ymm2[5,4],ymm7[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm6
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm11
@@ -2023,7 +2023,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[1,0],ymm1[2,0],ymm0[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -2036,7 +2036,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm1[2],mem[3,4],ymm1[5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,2],ymm2[0,3],ymm1[5,6],ymm2[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2049,7 +2049,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm1[2],mem[3,4],ymm1[5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,2],ymm2[0,3],ymm1[5,6],ymm2[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2062,7 +2062,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm1[2],mem[3,4],ymm1[5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,2],ymm2[0,3],ymm1[5,6],ymm2[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2075,7 +2075,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm1[2],mem[3,4],ymm1[5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,2],ymm2[0,3],ymm1[5,6],ymm2[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2086,7 +2086,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm8[2],ymm0[3,4],ymm8[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,2],ymm3[0,3],ymm2[5,6],ymm3[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2096,7 +2096,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm9[2],ymm1[3,4],ymm9[5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,2],ymm4[0,3],ymm5[5,6],ymm4[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2106,7 +2106,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm2[0,1],ymm10[2],ymm2[3,4],ymm10[5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[1,2],ymm6[0,3],ymm15[5,6],ymm6[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0,1,2,3,4],ymm5[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
@@ -2117,7 +2117,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm5[0,1],ymm11[2],ymm5[3,4],ymm11[5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[1,2],ymm7[0,3],ymm14[5,6],ymm7[4,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm15[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm15 # 32-byte Reload
@@ -2383,12 +2383,12 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm12 = <2,5,0,3,6,u,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm1 # 32-byte Folded Reload
@@ -2414,7 +2414,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm12, %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm11[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -2428,7 +2428,7 @@
 ; AVX2-SLOW-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm6 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1],ymm14[2],mem[3,4],ymm14[5],mem[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm12, %ymm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm15[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm15[0,1,0,3,4,5,4,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
@@ -2865,12 +2865,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm12 = <2,5,0,3,6,u,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm12, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm1 # 32-byte Folded Reload
@@ -2896,7 +2896,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm12, %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm11[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -2910,7 +2910,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1],ymm14[2],mem[3,4],ymm14[5],mem[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm12, %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm15[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm15[0,1,0,3,4,5,4,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -227,7 +227,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm6
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
@@ -434,7 +434,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm9[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2],xmm11[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm11[7]
@@ -447,7 +447,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm0[1,0],ymm11[0,0],ymm0[5,4],ymm11[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm2[0,0],ymm1[3,0],ymm2[4,4],ymm1[7,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,0],ymm1[2,2],ymm12[6,4],ymm1[6,6]
@@ -455,7 +455,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm12[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm0[2,0],ymm11[1,0],ymm0[6,4],ymm11[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
@@ -863,7 +863,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
@@ -876,7 +876,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm9[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 304(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
@@ -891,7 +891,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm1[1,0],ymm4[0,0],ymm1[5,4],ymm4[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm2[2,0],ymm0[7,4],ymm2[6,4]
@@ -905,7 +905,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[1,0],ymm3[0,0],ymm0[5,4],ymm3[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm6[0,0],ymm5[3,0],ymm6[4,4],ymm5[7,4]
@@ -913,7 +913,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm14[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[2,0],ymm4[1,0],ymm1[6,4],ymm4[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm11[0,0],ymm2[3,0],ymm11[4,4],ymm2[7,4]
@@ -921,7 +921,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[2,0],ymm3[1,0],ymm0[6,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6,7]
@@ -1793,7 +1793,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 304(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -1809,7 +1809,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 624(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -1825,7 +1825,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -1841,7 +1841,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 464(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -1859,7 +1859,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm1[0,0],ymm5[5,4],ymm1[4,4]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm11
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm7[2,0],ymm13[7,4],ymm7[6,4]
@@ -1874,7 +1874,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm6
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[3,0],ymm14[2,0],ymm10[7,4],ymm14[6,4]
@@ -1888,7 +1888,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm9
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm9[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm13[0,0],ymm9[5,4],ymm13[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1904,7 +1904,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[1,0],ymm10[0,0],ymm4[5,4],ymm10[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1916,7 +1916,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm5[2,0],ymm11[1,0],ymm5[6,4],ymm11[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1928,7 +1928,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1940,7 +1940,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[2,0],ymm13[1,0],ymm9[6,4],ymm13[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1952,7 +1952,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm10[1,0],ymm4[6,4],ymm10[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
@@ -3734,7 +3734,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 304(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3751,7 +3751,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 624(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3768,7 +3768,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 944(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3785,7 +3785,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 1264(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3801,7 +3801,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 144(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3817,7 +3817,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 464(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3834,7 +3834,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 784(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3851,7 +3851,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 1104(%rdi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
@@ -3872,7 +3872,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm1[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[1,0],ymm10[0,0],ymm1[5,4],ymm10[4,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -3891,7 +3891,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm1[0,0],ymm2[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -3910,7 +3910,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm1[0,0],ymm2[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3929,7 +3929,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm1[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm7[0,0],ymm1[5,4],ymm7[4,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3947,7 +3947,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm1[0,0],ymm2[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3964,7 +3964,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm3[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[1,0],ymm9[0,0],ymm3[5,4],ymm9[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3980,7 +3980,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm6
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3998,7 +3998,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm1[0,0],ymm8[5,4],ymm1[4,4]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4012,7 +4012,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm10[1,0],ymm1[6,4],ymm10[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4027,7 +4027,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4042,7 +4042,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4056,7 +4056,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm7[1,0],ymm1[6,4],ymm7[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4071,7 +4071,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4083,7 +4083,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm2[1,0],ymm8[6,4],ymm2[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4095,7 +4095,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4107,7 +4107,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[2,0],ymm9[1,0],ymm3[6,4],ymm9[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -51,18 +51,18 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0],xmm3[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0],xmm6[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovlps %xmm3, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovlps %xmm4, (%rdx)
 ; AVX1-ONLY-NEXT:    vmovlps %xmm5, (%rcx)
@@ -78,13 +78,13 @@
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm2[2,2,3,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm3 = <4,2,u,u>
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
@@ -307,16 +307,16 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[3,3]
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm6[1]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm1[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm1[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0],xmm7[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm5[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,0,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,0,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
 ; AVX1-ONLY-NEXT:    vpmovzxdq {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero
 ; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm6[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rdx)
@@ -558,7 +558,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm3[0,1]
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm11 = ymm10[0],ymm4[0],ymm10[3],ymm4[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[2,0],ymm5[2,3],ymm11[6,4],ymm5[6,7]
@@ -566,7 +566,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,2],xmm9[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm10[3,1],ymm4[1,3],ymm10[7,5],ymm4[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm4[0,1],ymm3[2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm8[2,3,0,1]
@@ -589,7 +589,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[2,0],ymm4[0,0],ymm3[6,4],ymm4[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[0,2],ymm8[2,0],ymm4[4,6],ymm8[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm12
@@ -600,7 +600,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0],ymm4[1,0],ymm3[7,4],ymm4[5,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,3],ymm3[2,0],ymm4[4,7],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm11[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,1],ymm1[1,3],ymm12[7,5],ymm1[5,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[2,0],ymm0[5,5],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
@@ -628,7 +628,7 @@
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
@@ -636,42 +636,42 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm10 = <1,7,5,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm10, %ymm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm8, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm9
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm11 = <4,2,u,u>
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm11, %ymm4
@@ -681,7 +681,7 @@
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm5 = <5,3,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm5, %ymm3
@@ -713,7 +713,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm2, %ymm2
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
@@ -721,7 +721,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm10 = <1,7,5,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm10, %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm8, %ymm8
@@ -730,33 +730,33 @@
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm9 = <u,u,u,4,2,u,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm9, %ymm8
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm9
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm11 = <4,2,u,u>
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm11, %ymm4
@@ -766,7 +766,7 @@
 ; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm5 = <5,3,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm5, %ymm3
@@ -798,7 +798,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm2, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm7[0,1],ymm5[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [8589934596,8589934596,8589934596,8589934596]
@@ -806,42 +806,42 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm10 = <1,7,5,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm10, %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm8, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm9
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm11[1,2,3],ymm10[4],ymm11[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm5[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm4[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm11[1,2,3],ymm9[4],ymm11[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3],ymm11[4,5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm11 = <4,2,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm11, %ymm4
@@ -851,7 +851,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm10[3],ymm4[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm5 = <5,3,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm5, %ymm3
@@ -1253,7 +1253,7 @@
 ; AVX1-ONLY-NEXT:    vmovapd %ymm0, %ymm10
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm11 = ymm7[0],ymm14[0],ymm7[3],ymm14[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rdi), %ymm3, %ymm1
@@ -1270,7 +1270,7 @@
 ; AVX1-ONLY-NEXT:    vmovapd 320(%rdi), %ymm12
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm5[0,1]
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm11 = ymm5[0],ymm12[0],ymm5[3],ymm12[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm13[3,0],ymm6[1,0],ymm13[7,4],ymm6[5,4]
@@ -1280,7 +1280,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm9[0,2],xmm2[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm6[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[3,1],ymm14[1,3],ymm7[7,5],ymm14[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %ymm8, %ymm4
@@ -1290,7 +1290,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm0[1,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,1],ymm12[1,3],ymm5[7,5],ymm12[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -1341,7 +1341,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm7
@@ -1356,7 +1356,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm11[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm1[0],xmm12[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 272(%rdi), %xmm2
@@ -1373,7 +1373,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $215, (%rsp), %ymm7, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm7[3,1],mem[1,3],ymm7[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm14[1,1],ymm5[2,0],ymm14[5,5],ymm5[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[3,0],ymm12[1,0],ymm9[7,4],ymm12[5,4]
@@ -1381,7 +1381,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm3[1,3],ymm2[7,5],ymm3[5,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[1,1],ymm2[2,0],ymm13[5,5],ymm2[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -1431,7 +1431,7 @@
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm1, %ymm6
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
@@ -1443,7 +1443,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 256(%rdi), %ymm5
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %ymm14
 ; AVX2-SLOW-NEXT:    vmovaps 352(%rdi), %ymm15
@@ -1453,73 +1453,73 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm0, %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm4, %ymm12
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm4, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %xmm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -1528,7 +1528,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 272(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm5 = <4,2,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm5, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
@@ -1542,13 +1542,13 @@
 ; AVX2-SLOW-NEXT:    # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm11
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm14, %ymm10, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm4 = <5,3,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm4, %ymm1
@@ -1557,7 +1557,7 @@
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm2, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm4, %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
@@ -1607,7 +1607,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm1, %ymm6
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
@@ -1619,7 +1619,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 256(%rdi), %ymm5
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 320(%rdi), %ymm14
 ; AVX2-FAST-NEXT:    vmovaps 352(%rdi), %ymm15
@@ -1629,21 +1629,21 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm0, %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm4, %ymm12
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm4, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm3 = [2,0,6,4,2,0,6,7]
@@ -1651,9 +1651,9 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
@@ -1661,40 +1661,40 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm3, %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -1703,7 +1703,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 272(%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm5 = <4,2,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm5, %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
@@ -1717,13 +1717,13 @@
 ; AVX2-FAST-NEXT:    # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm5, %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm14, %ymm10, %ymm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm4 = <5,3,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm4, %ymm1
@@ -1732,7 +1732,7 @@
 ; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm2, %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm4, %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
@@ -1782,7 +1782,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm1, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm7[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [8589934596,8589934596,8589934596,8589934596]
@@ -1794,7 +1794,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 256(%rdi), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[0,1],ymm8[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2],ymm15[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 352(%rdi), %ymm15
@@ -1804,73 +1804,73 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm4 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm4, %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm4, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -1879,7 +1879,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 272(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm5 = <4,2,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm5, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7]
@@ -1893,13 +1893,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm11[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm11
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm12[0,1],ymm6[2,3],ymm12[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm14, %ymm10, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm4 = <5,3,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm4, %ymm1
@@ -1908,7 +1908,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm2, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm11[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm4, %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3,4,5,6,7]
@@ -2782,7 +2782,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[0,1]
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm12[0],ymm5[0],ymm12[3],ymm5[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rdi), %ymm3, %ymm5
@@ -2803,7 +2803,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
@@ -2829,7 +2829,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[3],ymm7[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %ymm1
@@ -2854,7 +2854,7 @@
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm13[0,1]
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm13 = ymm15[0],ymm0[0],ymm15[3],ymm0[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -2865,7 +2865,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm12[3,1],ymm11[1,3],ymm12[7,5],ymm11[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
@@ -2878,7 +2878,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[3,1],ymm8[1,3],ymm0[7,5],ymm8[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2890,7 +2890,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm0[3,1],mem[1,3],ymm0[7,5],mem[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -2901,7 +2901,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[3,1],ymm7[1,3],ymm15[7,5],ymm7[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3017,7 +3017,7 @@
 ; AVX1-ONLY-NEXT:    # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -3036,7 +3036,7 @@
 ; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3]
@@ -3055,7 +3055,7 @@
 ; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 272(%rdi), %xmm12
@@ -3071,7 +3071,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm10[0],xmm5[1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vmovapd 656(%rdi), %xmm5
@@ -3094,7 +3094,7 @@
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm15[1,1],ymm8[2,0],ymm15[5,5],ymm8[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm3[1,0],ymm6[7,4],ymm3[5,4]
@@ -3107,7 +3107,7 @@
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm6[3,1],mem[1,3],ymm6[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[3,0],ymm0[1,0],ymm9[7,4],ymm0[5,4]
@@ -3118,7 +3118,7 @@
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm12[3,1],mem[1,3],ymm12[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm14[1,0],ymm4[7,4],ymm14[5,4]
@@ -3127,7 +3127,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm5[3,1],mem[1,3],ymm5[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[1,1],ymm5[2,0],ymm7[5,5],ymm5[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -3205,7 +3205,7 @@
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm14, %ymm2
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
@@ -3219,7 +3219,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm14, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 512(%rdi), %ymm0
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3241,7 +3241,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm14, %ymm15
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %ymm0
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
@@ -3263,7 +3263,7 @@
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
 ; AVX2-SLOW-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3275,59 +3275,59 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-SLOW-NEXT:    vpermps %ymm10, %ymm11, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm11, %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm0, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm11, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm14, %ymm11, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 384(%rdi), %xmm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
@@ -3341,13 +3341,13 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 576(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
@@ -3360,69 +3360,69 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vpermilps $224, (%rsp), %ymm6 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm6 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm8, %ymm13
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -3430,10 +3430,10 @@
 ; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
@@ -3442,7 +3442,7 @@
 ; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -3454,7 +3454,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 464(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm1, %ymm4
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm9
@@ -3474,7 +3474,7 @@
 ; AVX2-SLOW-NEXT:    # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 272(%rdi), %xmm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm9, %ymm1
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm9, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
@@ -3488,7 +3488,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 656(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm1, %ymm11
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
@@ -3505,7 +3505,7 @@
 ; AVX2-SLOW-NEXT:    # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm3, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -3513,7 +3513,7 @@
 ; AVX2-SLOW-NEXT:    # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm14 = <5,3,u,u>
@@ -3523,19 +3523,19 @@
 ; AVX2-SLOW-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-SLOW-NEXT:    vpermps %ymm7, %ymm3, %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm10, %ymm3, %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm9, %ymm14, %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm15, %ymm3, %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm14, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -3615,7 +3615,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm2
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
@@ -3629,7 +3629,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm14, %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 512(%rdi), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3651,7 +3651,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm14, %ymm15
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 320(%rdi), %ymm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3673,7 +3673,7 @@
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
 ; AVX2-FAST-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3685,34 +3685,34 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm11, %ymm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm11, %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm0, %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm11, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm13, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm14, %ymm11, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
@@ -3722,17 +3722,17 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 384(%rdi), %xmm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[2,3,2,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps $51, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
@@ -3750,7 +3750,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 576(%rdi), %xmm3
 ; AVX2-FAST-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm5 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
@@ -3772,64 +3772,64 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm7, %ymm4
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %xmm13
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm13[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[2,3,2,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm6 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm5 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm13[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vmovaps %ymm15, %ymm13
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -3837,10 +3837,10 @@
 ; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
@@ -3849,7 +3849,7 @@
 ; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -3862,7 +3862,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 464(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3,4,5,6,7]
@@ -3880,7 +3880,7 @@
 ; AVX2-FAST-NEXT:    # ymm13 = mem[0,1,2,3],ymm13[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 272(%rdi), %xmm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
@@ -3893,7 +3893,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 656(%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm1, %ymm11
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
@@ -3910,7 +3910,7 @@
 ; AVX2-FAST-NEXT:    # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm3, %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -3918,7 +3918,7 @@
 ; AVX2-FAST-NEXT:    # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm2, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm14 = <5,3,u,u>
@@ -3928,19 +3928,19 @@
 ; AVX2-FAST-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm3, %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm3, %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm14, %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm15, %ymm3, %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm14, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -4020,7 +4020,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm14, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm3[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [8589934596,8589934596,8589934596,8589934596]
@@ -4034,7 +4034,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm14, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 512(%rdi), %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4056,7 +4056,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm14, %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
@@ -4078,7 +4078,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm15[0,1],ymm1[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4090,59 +4090,59 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm0 = <1,7,5,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm11[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd {{.*#+}} ymm11 = [12884901893,12884901893,12884901893,12884901893]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm10, %ymm11, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm12[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm11, %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm11, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm14, %ymm11, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 384(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
@@ -4156,13 +4156,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 576(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3],ymm0[4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
@@ -4175,69 +4175,69 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, (%rsp), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm8, %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3],ymm4[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -4245,10 +4245,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
@@ -4257,7 +4257,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -4269,7 +4269,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 464(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm1[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm1 = <4,2,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm1, %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm9
@@ -4289,7 +4289,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 272(%rdi), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm9, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm9, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4,5,6,7]
@@ -4303,7 +4303,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm12[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 656(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm4[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm1, %ymm11
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2,3,4,5,6,7]
@@ -4320,7 +4320,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm0[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm3, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm11[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -4328,7 +4328,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm11 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm2, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0,1,2],mem[3],ymm2[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm14 = <5,3,u,u>
@@ -4338,19 +4338,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm7, %ymm3, %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm13[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm10, %ymm3, %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm9, %ymm14, %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm15, %ymm3, %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm14, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -6048,7 +6048,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm4[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[3],ymm5[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 672(%rdi), %ymm3, %ymm1
@@ -6070,7 +6070,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %ymm1
@@ -6097,7 +6097,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm1
@@ -6124,7 +6124,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
@@ -6151,7 +6151,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm1
@@ -6178,7 +6178,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm1
@@ -6205,7 +6205,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 1248(%rdi), %ymm1
@@ -6233,7 +6233,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm15[0,1]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm2[0],ymm0[0],ymm2[3],ymm0[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6248,7 +6248,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6262,7 +6262,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[3,1],ymm15[1,3],ymm1[7,5],ymm15[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6275,7 +6275,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm1[3,1],ymm13[1,3],ymm1[7,5],ymm13[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6288,7 +6288,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm1[3,1],ymm14[1,3],ymm1[7,5],ymm14[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6301,7 +6301,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm7 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6314,7 +6314,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
@@ -6327,7 +6327,7 @@
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -6341,7 +6341,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,1],ymm2[1,3],ymm1[7,5],ymm2[5,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6591,7 +6591,7 @@
 ; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6611,7 +6611,7 @@
 ; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6635,7 +6635,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6654,7 +6654,7 @@
 ; AVX1-ONLY-NEXT:    # ymm12 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 592(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6674,7 +6674,7 @@
 ; AVX1-ONLY-NEXT:    # ymm11 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 784(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6693,7 +6693,7 @@
 ; AVX1-ONLY-NEXT:    # ymm10 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6714,7 +6714,7 @@
 ; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 1168(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
@@ -6734,7 +6734,7 @@
 ; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-ONLY-NEXT:    vmovaps 1360(%rdi), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
@@ -6761,7 +6761,7 @@
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm15[3,1],mem[1,3],ymm15[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm14[1,1],ymm15[2,0],ymm14[5,5],ymm15[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3,4],ymm8[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm8 # 32-byte Reload
@@ -6775,7 +6775,7 @@
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4],ymm6[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
@@ -6789,7 +6789,7 @@
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
@@ -6802,7 +6802,7 @@
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm8[1,1],ymm13[2,0],ymm8[5,5],ymm13[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4],ymm4[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
@@ -6815,7 +6815,7 @@
 ; AVX1-ONLY-NEXT:    # ymm12 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm8[1,1],ymm12[2,0],ymm8[5,5],ymm12[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm2[1,0],ymm10[7,4],ymm2[5,4]
@@ -6828,7 +6828,7 @@
 ; AVX1-ONLY-NEXT:    # ymm11 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm8[1,1],ymm11[2,0],ymm8[5,5],ymm11[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0],ymm1[1,0],ymm9[7,4],ymm1[5,4]
@@ -6841,7 +6841,7 @@
 ; AVX1-ONLY-NEXT:    # ymm10 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm8[1,1],ymm10[2,0],ymm8[5,5],ymm10[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[3,0],ymm0[1,0],ymm7[7,4],ymm0[5,4]
@@ -6854,7 +6854,7 @@
 ; AVX1-ONLY-NEXT:    # ymm9 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[1,1],ymm9[2,0],ymm8[5,5],ymm9[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
@@ -6976,7 +6976,7 @@
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6993,7 +6993,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7018,7 +7018,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1088(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
@@ -7043,7 +7043,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1472(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7066,7 +7066,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm12, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7088,7 +7088,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm7, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 512(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7110,7 +7110,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm4, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 896(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7132,7 +7132,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1280(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7173,43 +7173,43 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm12, %ymm9, %ymm12
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm0, %ymm11
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm7, %ymm9, %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm8, %ymm0, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm3, %ymm9, %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm5, %ymm0, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm9, %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %xmm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -7218,14 +7218,14 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 576(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -7239,19 +7239,19 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 960(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -7260,34 +7260,34 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 1344(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 1152(%rdi), %xmm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -7301,14 +7301,14 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -7322,14 +7322,14 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 384(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm8 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
@@ -7342,26 +7342,26 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 768(%rdi), %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = mem[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm8 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
@@ -7372,7 +7372,7 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
@@ -7384,7 +7384,7 @@
 ; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
@@ -7399,7 +7399,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm4 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = mem[3,3,3,3,7,7,7,7]
@@ -7408,7 +7408,7 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
@@ -7420,24 +7420,24 @@
 ; AVX2-SLOW-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
@@ -7451,19 +7451,19 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
@@ -7471,19 +7471,19 @@
 ; AVX2-SLOW-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -7519,7 +7519,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 272(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm8 = <4,2,u,u>
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm8, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -7539,7 +7539,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 464(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
@@ -7558,7 +7558,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 656(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -7577,7 +7577,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 848(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
@@ -7597,7 +7597,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 1040(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
@@ -7613,7 +7613,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 1232(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -7630,7 +7630,7 @@
 ; AVX2-SLOW-NEXT:    # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1424(%rdi), %xmm12
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm8, %ymm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -7647,7 +7647,7 @@
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm8, %ymm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -7703,20 +7703,20 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm14, %ymm0, %ymm14
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm10, %ymm0, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm13, %ymm1, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm0, %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermps %ymm4, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
@@ -7845,7 +7845,7 @@
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7862,7 +7862,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7887,7 +7887,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1088(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7912,7 +7912,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1472(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7935,7 +7935,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7957,7 +7957,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 512(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7979,7 +7979,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm4, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 896(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8001,7 +8001,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1280(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8042,31 +8042,31 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm12, %ymm10, %ymm12
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm0, %ymm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm7, %ymm10, %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm0, %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm10, %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm0, %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm10, %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %xmm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -8086,7 +8086,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 576(%rdi), %xmm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm3 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
@@ -8105,7 +8105,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 960(%rdi), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm4 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
@@ -8116,7 +8116,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm13[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm13[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm5 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
@@ -8125,7 +8125,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 1344(%rdi), %xmm3
 ; AVX2-FAST-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm5 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
@@ -8136,16 +8136,16 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm15[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 1152(%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm5 = xmm4[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm7 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1,2,3],ymm5[4],ymm7[5,6,7]
@@ -8156,7 +8156,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm8 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
@@ -8165,7 +8165,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm5 = xmm4[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm8 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4],ymm8[5,6,7]
@@ -8185,7 +8185,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 384(%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm8 = xmm4[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm9 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
@@ -8208,21 +8208,21 @@
 ; AVX2-FAST-NEXT:    # ymm8 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm8, %ymm0, %ymm8
 ; AVX2-FAST-NEXT:    vmovaps 768(%rdi), %xmm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm9 = xmm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm9 = xmm0[2,3,2,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm5[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm8 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
@@ -8241,7 +8241,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6,7]
@@ -8260,7 +8260,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
@@ -8271,7 +8271,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2],ymm2[3],ymm6[4,5,6],ymm2[7]
@@ -8283,13 +8283,13 @@
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
@@ -8297,7 +8297,7 @@
 ; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm1 = mem[3,3,3,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = mem[3,3,3,3,7,7,7,7]
@@ -8306,26 +8306,26 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm3 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vmovaps %ymm7, %ymm12
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -8333,19 +8333,19 @@
 ; AVX2-FAST-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -8381,7 +8381,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 272(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm8 = <4,2,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm8, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -8402,7 +8402,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 464(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
@@ -8421,7 +8421,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 656(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8439,7 +8439,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 848(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8460,7 +8460,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 1040(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8477,7 +8477,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 1232(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8495,7 +8495,7 @@
 ; AVX2-FAST-NEXT:    # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1424(%rdi), %xmm13
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm8, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8511,7 +8511,7 @@
 ; AVX2-FAST-NEXT:    # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm8, %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -8566,20 +8566,20 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm14, %ymm0, %ymm14
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm10, %ymm0, %ymm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm13[3],ymm7[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm0, %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
@@ -8707,7 +8707,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[0,1],ymm8[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8724,7 +8724,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 704(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8749,7 +8749,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1088(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
@@ -8774,7 +8774,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm0, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1472(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8797,7 +8797,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm12, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8819,7 +8819,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm7, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 512(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8841,7 +8841,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 896(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8863,7 +8863,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[0,1],ymm4[0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1280(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8904,43 +8904,43 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm12, %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm0, %ymm11
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm7, %ymm9, %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm8, %ymm0, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm3, %ymm9, %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm5, %ymm0, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm9, %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -8949,14 +8949,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 576(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -8970,19 +8970,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 960(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -8991,34 +8991,34 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1344(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1152(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -9032,14 +9032,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -9053,14 +9053,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 384(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
@@ -9073,26 +9073,26 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 768(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
@@ -9103,7 +9103,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
@@ -9115,7 +9115,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
@@ -9130,7 +9130,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm4 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[3,3,3,3,7,7,7,7]
@@ -9139,7 +9139,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
@@ -9151,24 +9151,24 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
@@ -9182,19 +9182,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
@@ -9202,19 +9202,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = mem[3,3,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -9250,7 +9250,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 272(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm8 = <4,2,u,u>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm8, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
@@ -9270,7 +9270,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 464(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
@@ -9289,7 +9289,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 656(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -9308,7 +9308,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 848(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
@@ -9328,7 +9328,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1040(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
@@ -9344,7 +9344,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1232(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm2, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -9361,7 +9361,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1424(%rdi), %xmm12
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm8, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -9378,7 +9378,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm8, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -9434,20 +9434,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm14, %ymm0, %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm10, %ymm0, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm13, %ymm1, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm0, %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm4, %ymm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
@@ -58,21 +58,21 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0],xmm5[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm4[0],xmm2[1],xmm4[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[1,0,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,2,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[0,0],ymm0[1,0],ymm1[4,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[2,0],ymm1[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vmovlps %xmm5, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovlps %xmm6, (%rdx)
@@ -95,18 +95,18 @@
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0],xmm3[1],xmm4[2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[1,0,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[1,0,2,3]
 ; AVX2-ONLY-NEXT:    vbroadcastss 8(%rdi), %xmm6
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{.*#+}} xmm4 = <4,3,u,u>
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vpermps %ymm7, %ymm4, %ymm4
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,2,3,5,4,6,7]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,0,2,3,5,4,6,7]
 ; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-ONLY-NEXT:    vmovlps %xmm2, (%rsi)
@@ -346,37 +346,37 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm7[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm4[0],xmm3[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[1,0],mem[3,3]
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm8 = xmm8[0,1,2],xmm6[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0],xmm4[1],xmm9[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm9[0],xmm5[1],xmm9[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,1,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,1,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm5[0,1,2],xmm10[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm9[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm0[0,0],ymm1[1,0],ymm0[4,4],ymm1[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm10[0,1],xmm5[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[2,0],ymm0[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsi)
@@ -401,9 +401,9 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss 84(%rdi), %xmm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rdi), %xmm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 8(%rdi), %xmm6
@@ -415,9 +415,9 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,1,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 100(%rdi), %xmm9
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
@@ -426,13 +426,13 @@
 ; AVX2-SLOW-NEXT:    vpermps %ymm11, %ymm10, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 80(%rdi), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
@@ -461,7 +461,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm4, %ymm3, %ymm3
 ; AVX2-FAST-NEXT:    vmovaps 80(%rdi), %xmm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],xmm6[3]
 ; AVX2-FAST-NEXT:    vbroadcastss 8(%rdi), %xmm6
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %xmm7
@@ -472,9 +472,9 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,1,0]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastss 100(%rdi), %xmm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
@@ -483,13 +483,13 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm11, %ymm10, %ymm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastss 80(%rdi), %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
@@ -515,9 +515,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 84(%rdi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 8(%rdi), %xmm6
@@ -529,9 +529,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,1,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 100(%rdi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
@@ -540,13 +540,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm11, %ymm10, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 80(%rdi), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
@@ -761,7 +761,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm13[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm6
@@ -781,7 +781,7 @@
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm10[2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm9[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[3,1],ymm11[0,3],ymm2[7,5],ymm11[4,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4]
@@ -796,7 +796,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,0],ymm12[0,0],ymm11[5,4],ymm12[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[3,1],ymm11[0,2],ymm12[7,5],ymm11[4,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm9[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm14[0,1],ymm4[1,3],ymm14[4,5],ymm4[5,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm8[0,2],ymm12[2,0],ymm8[4,6],ymm12[6,4]
@@ -806,9 +806,9 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm13[0,0],ymm8[7,4],ymm13[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0],ymm12[2,0],ymm8[6,4],ymm12[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm12[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = mem[0],xmm9[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
@@ -818,20 +818,20 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm4[2,0],ymm9[5,4],ymm4[6,4]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm12[0,1,2],xmm15[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm0[0,0],ymm1[1,0],ymm0[4,4],ymm1[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm13[0,1],xmm9[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm14[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm14[3,0],ymm9[0,0],ymm14[7,4],ymm9[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1],ymm9[2,0],ymm5[4,5],ymm9[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm12[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[2,0],ymm0[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -1576,7 +1576,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1595,7 +1595,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1639,7 +1639,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm15[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm10[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm5[0,3],ymm14[7,5],ymm5[4,7]
@@ -1651,7 +1651,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm12[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm11[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[3,1],ymm4[0,3],ymm12[7,5],ymm4[4,7]
@@ -1666,7 +1666,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm7[0,0],ymm5[5,4],ymm7[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[3,1],ymm2[0,2],ymm7[7,5],ymm2[4,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm10[0,1,2],xmm15[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,1],ymm8[1,3],ymm7[4,5],ymm8[5,7]
@@ -1676,7 +1676,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm13[0,0],ymm4[5,4],ymm13[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[3,1],ymm2[0,2],ymm13[7,5],ymm2[4,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm11[0,1,2],xmm9[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[0,1],ymm3[1,3],ymm5[4,5],ymm3[5,7]
@@ -1689,9 +1689,9 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm10[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
@@ -1701,10 +1701,10 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm3[2,0],ymm5[5,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm14[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm10[0,1,2],xmm6[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = mem[0],xmm11[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
@@ -1719,7 +1719,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm13[0,0],ymm12[1,0],ymm13[4,4],ymm12[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm8[0,1],xmm2[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3],ymm6[4,5,6,7]
@@ -1733,13 +1733,13 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm10[0,0],ymm11[1,0],ymm10[4,4],ymm11[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm8[0,1],xmm6[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm7[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[3,0],ymm6[0,0],ymm7[7,4],ymm6[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],xmm1[1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,0],ymm7[4,5],ymm6[6,4]
@@ -1747,13 +1747,13 @@
 ; AVX1-ONLY-NEXT:    # xmm7 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm13[1,0],ymm12[2,0],ymm13[5,4],ymm12[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm5[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,0],ymm7[0,0],ymm5[7,4],ymm7[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2,0],ymm7[4,5],ymm5[6,4]
@@ -1761,7 +1761,7 @@
 ; AVX1-ONLY-NEXT:    # xmm7 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm7[0,1,2],xmm14[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm10[1,0],ymm11[2,0],ymm10[5,4],ymm11[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
@@ -3460,7 +3460,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3481,7 +3481,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3504,7 +3504,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm14
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm12
 ; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3532,7 +3532,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps 752(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3636,7 +3636,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm10[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
@@ -3652,7 +3652,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
@@ -3669,7 +3669,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm6[0,3],ymm15[7,5],ymm6[4,7]
@@ -3683,7 +3683,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4],ymm6[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm13[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm0[1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -3705,7 +3705,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm10[3,1],ymm6[0,2],ymm10[7,5],ymm6[4,6]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm10 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm10 = xmm5[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3719,7 +3719,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm8[3,1],ymm6[0,2],ymm8[7,5],ymm6[4,6]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm9 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm9 = xmm1[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm6[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm6
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -3730,7 +3730,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,0],ymm7[0,0],ymm3[5,4],ymm7[4,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[3,1],ymm8[0,2],ymm7[7,5],ymm8[4,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm0[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm9
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm15[1,3],ymm9[4,5],ymm15[5,7]
@@ -3745,7 +3745,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm15
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -3762,7 +3762,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm7 = mem[2,3,2,3]
@@ -3777,7 +3777,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm3[2,0],ymm2[6,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
@@ -3794,7 +3794,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
@@ -3809,7 +3809,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm11[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm11[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
@@ -3831,7 +3831,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm13[0,1],xmm10[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
@@ -3848,7 +3848,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm3[0,0],ymm4[1,0],ymm3[4,4],ymm4[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm13[0,1],xmm12[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0,1,2,3],ymm10[4,5,6,7]
@@ -3864,7 +3864,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,0],ymm14[1,0],ymm10[4,4],ymm14[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[0,1],xmm1[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
@@ -3880,13 +3880,13 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm11[0,0],mem[1,0],ymm11[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm15[0,1],xmm13[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm13[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm1[0,0],ymm6[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm2[1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
@@ -3894,7 +3894,7 @@
 ; AVX1-ONLY-NEXT:    # xmm6 = mem[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm3[1,0],ymm4[2,0],ymm3[5,4],ymm4[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -3914,14 +3914,14 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm9[1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
@@ -3930,7 +3930,7 @@
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0,1,2],mem[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[1,0],ymm14[2,0],ymm10[5,4],ymm14[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
@@ -3948,7 +3948,7 @@
 ; AVX1-ONLY-NEXT:    # xmm7 = xmm8[0,1,2],mem[3]
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm11[1,0],mem[2,0],ymm11[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -4285,8 +4285,8 @@
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovdqa %ymm9, %ymm14
@@ -4358,13 +4358,13 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 216(%rdi), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm14
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
@@ -4372,41 +4372,41 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 544(%rdi), %xmm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 664(%rdi), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %xmm11
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm11[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 440(%rdi), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 768(%rdi), %xmm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm7[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 888(%rdi), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
@@ -4796,8 +4796,8 @@
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
@@ -5308,8 +5308,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm9, %ymm14
@@ -5381,13 +5381,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 216(%rdi), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm14
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
@@ -5395,41 +5395,41 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 544(%rdi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 664(%rdi), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %xmm11
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm11[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 440(%rdi), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 768(%rdi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 888(%rdi), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
@@ -7429,7 +7429,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7452,7 +7452,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 672(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 832(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7474,7 +7474,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vmovaps 1200(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7501,7 +7501,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 1648(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7528,7 +7528,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7555,7 +7555,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps 528(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7583,7 +7583,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 896(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %ymm6
 ; AVX1-ONLY-NEXT:    vmovaps 976(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7610,7 +7610,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1344(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 1424(%rdi), %xmm1
@@ -7805,7 +7805,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm15[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
@@ -7822,7 +7822,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm10[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
@@ -7839,7 +7839,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -7857,7 +7857,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm9[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -7875,7 +7875,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm4[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm6[0,3],ymm14[7,5],ymm6[4,7]
@@ -7951,7 +7951,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm13 = mem[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7968,7 +7968,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm14 = mem[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7982,7 +7982,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm14 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm14 = xmm11[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7996,7 +7996,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm14 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm14 = xmm9[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1760(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8012,7 +8012,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1536(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8026,7 +8026,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm2[0,2],ymm3[7,5],ymm2[4,6]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm1[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1088(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps %ymm6, %ymm14
@@ -8044,7 +8044,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm9[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 640(%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -8062,7 +8062,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm4[0,1,2],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm13
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -8079,9 +8079,9 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
@@ -8095,7 +8095,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,0],ymm0[6,4],ymm4[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
@@ -8110,13 +8110,13 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm3[2,0],ymm12[5,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 512(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -8130,7 +8130,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 768(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 736(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
@@ -8146,7 +8146,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 992(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 960(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
@@ -8165,7 +8165,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 1216(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 1184(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
@@ -8183,7 +8183,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 1440(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 1408(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm1[0,1,2],xmm10[3]
 ; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
@@ -8200,7 +8200,7 @@
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm2[1,0],mem[2,0],ymm2[5,4],mem[6,4]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
 ; AVX1-ONLY-NEXT:    vmovaps 1664(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm0[0,1,2],xmm10[3]
 ; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
@@ -8222,7 +8222,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm14[0,0],ymm15[1,0],ymm14[4,4],ymm15[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm11[0,1],xmm10[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
@@ -8239,7 +8239,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm10 = ymm10[0,0],mem[1,0],ymm10[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm10[0,1],xmm8[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
@@ -8257,7 +8257,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[0,0],mem[1,0],ymm8[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm8[0,1],xmm7[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
@@ -8275,7 +8275,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm7 = ymm7[0,0],mem[1,0],ymm7[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[0,1],xmm6[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
@@ -8294,7 +8294,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm6[0,0],mem[1,0],ymm6[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[0,1],xmm5[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
@@ -8312,7 +8312,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0,0],mem[1,0],ymm5[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -8330,7 +8330,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = ymm4[0,0],mem[1,0],ymm4[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,1],xmm1[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
@@ -8346,14 +8346,14 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[0,1],xmm0[3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm0[0,0],ymm13[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -8363,14 +8363,14 @@
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0,1,2],mem[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,0],ymm15[2,0],ymm14[5,4],ymm15[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm11[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm2 = xmm2[0],mem[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -8382,7 +8382,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm3[1,0],mem[2,0],ymm3[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -8401,14 +8401,14 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = ymm4[1,0],mem[2,0],ymm4[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[0,0],ymm4[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm4 = xmm4[0],mem[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
@@ -8420,7 +8420,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -8440,14 +8440,14 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[3,0],ymm3[0,0],ymm6[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0],mem[1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
@@ -8459,7 +8459,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm7 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm3[4,5,6,7]
@@ -8479,7 +8479,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[1,0],mem[2,0],ymm8[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
@@ -8498,7 +8498,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm9 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
@@ -9092,8 +9092,8 @@
 ; AVX2-SLOW-NEXT:    # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 432(%rdi), %ymm14
@@ -9105,8 +9105,8 @@
 ; AVX2-SLOW-NEXT:    # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
@@ -9119,8 +9119,8 @@
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -9134,8 +9134,8 @@
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
@@ -9175,8 +9175,8 @@
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovdqa %ymm11, %ymm7
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
@@ -9189,8 +9189,8 @@
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -9321,13 +9321,13 @@
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 216(%rdi), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm8, %xmm8
@@ -9336,14 +9336,14 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 320(%rdi), %xmm0
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 440(%rdi), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
@@ -9351,7 +9351,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 544(%rdi), %xmm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm2
@@ -9359,7 +9359,7 @@
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 664(%rdi), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
@@ -9367,7 +9367,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 768(%rdi), %xmm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm2
@@ -9375,7 +9375,7 @@
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 888(%rdi), %ymm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
@@ -9383,7 +9383,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 992(%rdi), %xmm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm3
@@ -9391,14 +9391,14 @@
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 1112(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1216(%rdi), %xmm2
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm4, %xmm4
@@ -9406,7 +9406,7 @@
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 1336(%rdi), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
@@ -9414,27 +9414,27 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 1440(%rdi), %xmm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 1560(%rdi), %ymm13
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 1664(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
 ; AVX2-SLOW-NEXT:    vbroadcastss 1784(%rdi), %ymm12
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
@@ -10134,8 +10134,8 @@
 ; AVX2-FAST-NEXT:    # ymm13 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm4[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm15[0,2],ymm3[1,3],ymm15[4,6],ymm3[5,7]
@@ -10161,8 +10161,8 @@
 ; AVX2-FAST-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
@@ -10176,8 +10176,8 @@
 ; AVX2-FAST-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
@@ -10217,8 +10217,8 @@
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,2],ymm7[1,3],ymm9[4,6],ymm7[5,7]
@@ -10231,8 +10231,8 @@
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups (%rsp), %ymm4 # 32-byte Reload
@@ -10370,7 +10370,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %xmm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm6 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm6, %xmm6
@@ -10379,7 +10379,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 320(%rdi), %xmm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm1[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm1 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
@@ -10394,7 +10394,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 544(%rdi), %xmm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm5[0,1,2],xmm2[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm1 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
@@ -10409,7 +10409,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 768(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm3[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm2, %xmm2
@@ -10423,7 +10423,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 992(%rdi), %xmm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm3, %xmm3
@@ -10438,7 +10438,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 1216(%rdi), %xmm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm4, %xmm4
@@ -10452,7 +10452,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1440(%rdi), %xmm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm8[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm8, %xmm8
@@ -10465,7 +10465,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 1664(%rdi), %xmm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-FAST-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-NEXT:    vextractf128 $1, %ymm15, %xmm15
@@ -11183,8 +11183,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 432(%rdi), %ymm14
@@ -11196,8 +11196,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
@@ -11210,8 +11210,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
@@ -11225,8 +11225,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
@@ -11266,8 +11266,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm11, %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
@@ -11280,8 +11280,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -11412,13 +11412,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 216(%rdi), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm8, %xmm8
@@ -11427,14 +11427,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 320(%rdi), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 440(%rdi), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
@@ -11442,7 +11442,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 544(%rdi), %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm2, %xmm2
@@ -11450,7 +11450,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 664(%rdi), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
@@ -11458,7 +11458,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 768(%rdi), %xmm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm2, %xmm2
@@ -11466,7 +11466,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 888(%rdi), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
@@ -11474,7 +11474,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 992(%rdi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm3 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm3, %xmm3
@@ -11482,14 +11482,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1112(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1216(%rdi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm4, %xmm4
@@ -11497,7 +11497,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1336(%rdi), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
@@ -11505,27 +11505,27 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1440(%rdi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm7 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1560(%rdi), %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 1664(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[1,0,2,3,5,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 1784(%rdi), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
@@ -61,12 +61,12 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vmovq %xmm4, (%rsi)
 ; AVX1-ONLY-NEXT:    vmovq %xmm5, (%rdx)
@@ -95,12 +95,12 @@
 ; AVX2-ONLY-NEXT:    vbroadcastss 48(%rdi), %xmm3
 ; AVX2-ONLY-NEXT:    vbroadcastss 16(%rdi), %ymm6
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm0[1],ymm6[2,3,4],ymm0[5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
 ; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-ONLY-NEXT:    vmovq %xmm4, (%rsi)
 ; AVX2-ONLY-NEXT:    vmovq %xmm5, (%rdx)
@@ -128,12 +128,12 @@
 ; AVX512F-SLOW-NEXT:    vmovaps (%rdi), %ymm4
 ; AVX512F-SLOW-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
 ; AVX512F-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
-; AVX512F-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm4[1,1,1,1,5,5,5,5]
+; AVX512F-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX512F-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1],ymm6[2,3,4],ymm1[5],ymm6[6,7]
 ; AVX512F-SLOW-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX512F-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
 ; AVX512F-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX512F-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX512F-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX512F-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX512F-SLOW-NEXT:    vmovq %xmm2, (%rsi)
 ; AVX512F-SLOW-NEXT:    vmovq %xmm3, (%rdx)
@@ -195,12 +195,12 @@
 ; AVX512BW-SLOW-NEXT:    vmovaps (%rdi), %ymm4
 ; AVX512BW-SLOW-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
 ; AVX512BW-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm5
-; AVX512BW-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm4[1,1,1,1,5,5,5,5]
+; AVX512BW-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX512BW-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1],ymm6[2,3,4],ymm1[5],ymm6[6,7]
 ; AVX512BW-SLOW-NEXT:    vextractf128 $1, %ymm6, %xmm6
 ; AVX512BW-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
 ; AVX512BW-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX512BW-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX512BW-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX512BW-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX512BW-SLOW-NEXT:    vmovq %xmm2, (%rsi)
 ; AVX512BW-SLOW-NEXT:    vmovq %xmm3, (%rdx)
@@ -330,17 +330,17 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm9[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm6[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm6[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm5[1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm7[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm7[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm8[0,1,2],xmm10[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
@@ -348,11 +348,11 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
@@ -360,7 +360,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, (%rsi)
@@ -393,10 +393,10 @@
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm7[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm7[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm6[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm9[0,1,2],xmm10[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm7[0,1],xmm10[2,3]
@@ -605,18 +605,18 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm14
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm15
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm7[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm8[1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm13[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm13[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0],xmm12[1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm6[6,7]
@@ -626,11 +626,11 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm14[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm14[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm9[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm9[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm11[0,1,2],xmm13[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm8[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm10[4,5,6,7]
@@ -640,7 +640,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3],ymm9[4,5,6,7]
@@ -648,7 +648,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm12 = ymm7[0],ymm6[0],ymm7[1],ymm6[1],ymm7[4],ymm6[4],ymm7[5],ymm6[5]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,0],ymm12[4,5],ymm8[6,4]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm14 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
@@ -660,7 +660,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm14 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2,3],ymm8[4,5,6,7]
@@ -668,7 +668,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm7[2],ymm6[2],ymm7[3],ymm6[3],ymm7[6],ymm6[6],ymm7[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,0],ymm15[4,5],ymm14[6,4]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
@@ -680,7 +680,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
@@ -728,12 +728,12 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm14[0],xmm11[0],xmm14[1],xmm11[1]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm12[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm12[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm13[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm10[0],xmm8[0],xmm10[1],xmm8[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm7[1],xmm15[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],ymm6[6,7]
@@ -741,13 +741,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm15[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm15 = xmm9[2],xmm7[2],xmm9[3],xmm7[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm8[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm8[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm10[0,1,2],xmm9[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm9
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm11[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm0[0,1],xmm12[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm12[0,1,2,3],ymm7[4,5,6,7]
@@ -755,7 +755,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
 ; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm13
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm15[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm15[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm14[2],xmm11[2],xmm14[3],xmm11[3]
@@ -1146,7 +1146,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -1160,7 +1160,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm4[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm4[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm15
@@ -1172,22 +1172,22 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm9[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm8[1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm15[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm11[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm11[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm10[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
@@ -1196,14 +1196,14 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm9[2],xmm3[2],xmm9[3],xmm3[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm12[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm12[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
@@ -1212,13 +1212,13 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm7[2],xmm15[2],xmm7[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm14[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
@@ -1229,7 +1229,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm4
@@ -1247,7 +1247,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %ymm7
 ; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -1261,7 +1261,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm10
 ; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
@@ -1280,7 +1280,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm14
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm13
@@ -1300,7 +1300,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm0[1,0],ymm2[1,0],ymm0[5,4],ymm2[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
@@ -1313,7 +1313,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm8[0],ymm6[1],ymm8[1],ymm6[4],ymm8[4],ymm6[5],ymm8[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm13[1,0],ymm14[1,0],ymm13[5,4],ymm14[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7]
@@ -1326,7 +1326,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm3[4,5,6,7]
@@ -1336,7 +1336,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm6[1],ymm8[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm12 = ymm14[2],ymm13[2],ymm14[3],ymm13[3],ymm14[6],ymm13[6],ymm14[7],ymm13[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
@@ -1352,7 +1352,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = ymm5[3,0],mem[3,0],ymm5[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1362,7 +1362,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm6[2],ymm8[2],ymm6[3],ymm8[3],ymm6[6],ymm8[6],ymm6[7],ymm8[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[3,0],ymm14[3,0],ymm13[7,4],ymm14[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -1461,19 +1461,19 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm14[0],xmm12[0],xmm14[1],xmm12[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm4[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm4[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm5[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm15[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm15, %xmm10
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm5[1],xmm0[2,3]
@@ -1485,7 +1485,7 @@
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm9[0],xmm2[1],xmm9[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm13[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
@@ -1494,12 +1494,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm5[2],xmm10[3],xmm5[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm13[2],xmm4[2],xmm13[3],xmm4[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm9[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm9[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm2[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm13
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm15[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm3[0,1,2],xmm13[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm3, %xmm9
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm0[0,1],xmm13[2,3]
@@ -1509,13 +1509,13 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm4 = xmm2[2],mem[2],xmm2[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps (%rsp), %xmm10 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm10[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm10[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm11[0,1,2],xmm5[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm12[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm14[0,1,2],xmm3[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
@@ -1528,7 +1528,7 @@
 ; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %ymm6
@@ -1540,7 +1540,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
 ; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm13
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm7
@@ -2473,7 +2473,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm15[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -2490,7 +2490,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
@@ -2519,7 +2519,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm11
@@ -2546,7 +2546,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 704(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm12[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm7
@@ -2560,19 +2560,19 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm3[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm14[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm9[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm8[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
@@ -2586,7 +2586,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm4[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm11[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
@@ -2598,7 +2598,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
@@ -2622,7 +2622,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3]
@@ -2644,7 +2644,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
@@ -2658,14 +2658,14 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm13[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
@@ -2677,14 +2677,14 @@
 ; AVX1-ONLY-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm11[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm11[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm9[0,1,2],xmm7[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm2[0,1,2],xmm8[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
@@ -2759,7 +2759,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, %ymm14
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2786,7 +2786,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2813,7 +2813,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 608(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm10[4,5,6,7]
@@ -2834,7 +2834,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2854,7 +2854,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[1,0],ymm6[1,0],ymm7[5,4],ymm6[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2871,7 +2871,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[1,0],ymm9[1,0],ymm7[5,4],ymm9[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2888,7 +2888,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,0],ymm2[1,0],ymm5[5,4],ymm2[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2906,7 +2906,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm15[1,0],mem[1,0],ymm15[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2922,7 +2922,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm3[1],ymm12[1],ymm3[3],ymm12[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2936,7 +2936,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2952,7 +2952,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2966,7 +2966,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm5[1],mem[1],ymm5[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm9[2],ymm15[2],ymm9[3],ymm15[3],ymm9[6],ymm15[6],ymm9[7],ymm15[7]
@@ -2986,7 +2986,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm11 = ymm11[3,0],mem[3,0],ymm11[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2999,7 +2999,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = ymm2[3,0],mem[3,0],ymm2[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -3017,7 +3017,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm3[3,0],mem[3,0],ymm3[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -3029,7 +3029,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[6],ymm5[6],ymm3[7],ymm5[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm15[3,0],ymm9[3,0],ymm15[7,4],ymm9[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
@@ -3225,7 +3225,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm11[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm13[0],xmm10[1],xmm13[1]
@@ -3261,21 +3261,21 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm14[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm8[0],mem[0],xmm8[1],mem[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm7[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm7[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -3287,7 +3287,7 @@
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm6[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm6[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -3301,13 +3301,13 @@
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm15[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm1[0,1,2],xmm15[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm14[0,1],xmm15[2,3]
@@ -3320,13 +3320,13 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = xmm11[2],mem[2],xmm11[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm12[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm12[0,1,2],xmm14[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm13
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm10[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm10[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm0[0,1,2],xmm14[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
@@ -3334,12 +3334,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm3[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm4[0,1,2],xmm8[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm14
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm9[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm9[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm3[0,1,2],xmm14[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm13[0,1],xmm14[2,3]
@@ -3349,7 +3349,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm9 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm9 = xmm4[2],mem[2],xmm4[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm14[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm14[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
@@ -3359,7 +3359,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm7 = xmm4[2],mem[2],xmm4[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm8[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm11[0,1,2],xmm4[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
@@ -3397,14 +3397,14 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm10[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm10[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm14[2],xmm5[3],xmm14[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm9[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm11[2],xmm8[2],xmm11[3],xmm8[3]
@@ -5428,7 +5428,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm14[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -5445,7 +5445,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 800(%rdi), %xmm1
@@ -5474,7 +5474,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1312(%rdi), %xmm1
@@ -5502,7 +5502,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1824(%rdi), %xmm1
@@ -5530,7 +5530,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm6
@@ -5559,7 +5559,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 544(%rdi), %xmm4
@@ -5588,7 +5588,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1056(%rdi), %xmm12
@@ -5616,7 +5616,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 1568(%rdi), %xmm8
@@ -5632,7 +5632,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm15, %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm15[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
@@ -5668,7 +5668,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm15[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm15[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm13 = xmm13[0],mem[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
@@ -5717,7 +5717,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm12[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
@@ -5730,11 +5730,11 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm8[1],xmm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm9[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm9[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -5860,7 +5860,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm1[0,1,2],xmm14[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm2[0,1],xmm14[2,3]
@@ -5882,7 +5882,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm7[0,1,2],xmm12[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm6[0,1],xmm12[2,3]
@@ -5894,14 +5894,14 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm9[2],xmm5[2],xmm9[3],xmm5[3]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm12[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm12[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm11[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm5[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm0[0,1],xmm9[2,3]
@@ -6035,7 +6035,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm12[0],ymm2[2],ymm12[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6062,7 +6062,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm9[0],ymm2[2],ymm9[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6089,7 +6089,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 864(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6116,7 +6116,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1120(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6143,7 +6143,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1376(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6170,7 +6170,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1632(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
@@ -6197,7 +6197,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 1888(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
@@ -6218,7 +6218,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6242,7 +6242,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6259,7 +6259,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[1,0],ymm5[1,0],ymm1[5,4],ymm5[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6278,7 +6278,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6297,7 +6297,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6316,7 +6316,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6334,7 +6334,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,0],mem[1,0],ymm0[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm2[4,5,6,7]
@@ -6353,7 +6353,7 @@
 ; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,0],mem[1,0],ymm0[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm1[4,5,6,7]
@@ -6372,7 +6372,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $17, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm14[1,0],mem[1,0],ymm14[5,4],mem[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6389,7 +6389,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm10[1],ymm6[1],ymm10[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6407,7 +6407,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm3[1],ymm10[1],ymm3[3],ymm10[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6423,7 +6423,7 @@
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm7[1],ymm11[1],ymm7[3],ymm11[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6440,7 +6440,7 @@
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm5[1],ymm12[1],ymm5[3],ymm12[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6458,7 +6458,7 @@
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6475,7 +6475,7 @@
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm13[1],ymm15[1],ymm13[3],ymm15[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6494,7 +6494,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm1[1],mem[1],ymm1[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm14, %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
@@ -6509,7 +6509,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
@@ -6530,7 +6530,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm6, %xmm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
@@ -6547,7 +6547,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
@@ -6564,7 +6564,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
@@ -6582,7 +6582,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
@@ -6599,7 +6599,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm1[3,0],mem[3,0],ymm1[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm8[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
@@ -6617,7 +6617,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,0],mem[3,0],ymm8[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6633,7 +6633,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,0],mem[3,0],ymm8[7,4],mem[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
@@ -6647,7 +6647,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0],ymm15[3,0],ymm9[7,4],ymm15[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
@@ -7020,7 +7020,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm12[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps $2, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -7058,7 +7058,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm13[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -7077,7 +7077,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm15[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -7096,13 +7096,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm14[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm8[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
@@ -7176,7 +7176,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm12[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -7197,14 +7197,14 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm11[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm11[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = mem[0,1,2],xmm1[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
@@ -7218,7 +7218,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm13[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -7239,7 +7239,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm15[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -7259,13 +7259,13 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm8[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
@@ -7279,14 +7279,14 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm4[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm4[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -7299,13 +7299,13 @@
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm7[0,1,2],xmm0[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm1[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm9[0,1,2],xmm15[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm8[0,1],xmm15[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
@@ -169,7 +169,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 80(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
@@ -298,11 +298,11 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 80(%rdi), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 16(%rdi), %xmm1
@@ -543,19 +543,19 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm11[2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rdi), %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 368(%rdi), %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 80(%rdi), %ymm9
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 272(%rdi), %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 112(%rdi), %xmm7
@@ -1070,39 +1070,39 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 368(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 560(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm4[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 752(%rdi), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm4[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm4 = ymm4[0,1],mem[2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 80(%rdi), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm4[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm15[2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 272(%rdi), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm4[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 464(%rdi), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 656(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
@@ -2193,97 +2193,97 @@
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 368(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 560(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 752(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 944(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm10 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1136(%rdi), %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1],ymm9[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1328(%rdi), %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1520(%rdi), %ymm9
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 80(%rdi), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1],ymm6[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 272(%rdi), %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm14[0,1],ymm5[2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 464(%rdi), %ymm6
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1],ymm4[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 656(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, (%rsp) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 848(%rdi), %ymm4
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3],ymm15[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1040(%rdi), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1232(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 1424(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
@@ -267,7 +267,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm4[0,1,2,0,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vpunpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm7[0]
 ; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3,4],xmm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,3,1,3,4,5,6,7]
@@ -1692,8 +1692,8 @@
 ; AVX1-ONLY-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,xmm4[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm12[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5],xmm2[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm4[0,1,0,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm12[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm4[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm12[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
@@ -1740,8 +1740,8 @@
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm4, %ymm2
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm14[0,1,0,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm13[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vpsrld $16, %xmm5, %xmm4
 ; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm11[2],xmm4[2],xmm11[3],xmm4[3]
@@ -3842,7 +3842,7 @@
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm5, %ymm4
 ; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm13[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3]
 ; AVX1-ONLY-NEXT:    vpsrld $16, %xmm2, %xmm6
 ; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
@@ -8117,7 +8117,7 @@
 ; AVX1-ONLY-NEXT:    vorps %ymm12, %ymm13, %ymm12
 ; AVX1-ONLY-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm13 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm11[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm11[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1,2],xmm14[3]
 ; AVX1-ONLY-NEXT:    vpsrld $16, %xmm1, %xmm14
 ; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm14 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-8.ll
@@ -1174,8 +1174,8 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4],ymm3[5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm3[0,0,0,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vmovdqu (%rsp), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm5[0,1,0,1]
@@ -1183,10 +1183,10 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm5[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[0,0,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[0,0,1,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-2.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-2.ll
@@ -57,8 +57,8 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rdx)
 ; AVX1-ONLY-NEXT:    vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
@@ -39,7 +39,7 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,1,0,3,7,5,4,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,0,3,7,5,4,7]
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,u,1,u,5,u,u]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm1
@@ -273,12 +273,12 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd (%rdx), %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%rdi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5,6],ymm5[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm2[1,1,2,2]
@@ -301,7 +301,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd (%rdx), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm2[1,1,2,2]
@@ -330,12 +330,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd (%rdx), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm2[1,1,2,2]
@@ -526,7 +526,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd (%rdx), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 56(%rdi), %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm1[2,1,3,3]
@@ -538,17 +538,17 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 32(%rdx), %ymm9
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%rdi), %ymm9
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm5[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm6[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6],ymm10[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm6[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
@@ -570,7 +570,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm4
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2],ymm7[3,4],ymm0[5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm5[1,1,2,2]
@@ -582,7 +582,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm8[1],ymm1[2,3],ymm8[4],ymm1[5,6],ymm8[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd (%rdx), %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm8[2],ymm1[3,4],ymm8[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm4[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm2[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3,4],ymm8[5],ymm9[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm6[1,1,2,2]
@@ -627,7 +627,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd (%rdx), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm1[2,1,3,3]
@@ -639,17 +639,17 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 32(%rdx), %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm5[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm6[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm6[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
@@ -1024,7 +1024,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3,4],ymm4[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 88(%rdi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm9[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm4[2],ymm11[3,4],ymm4[5],ymm11[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[2,1,3,3]
@@ -1038,7 +1038,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 56(%rdi), %ymm12
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm7[2,1,3,3]
@@ -1052,7 +1052,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 120(%rdi), %ymm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4],ymm14[5],ymm15[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm3[2,1,3,3]
@@ -1065,29 +1065,29 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 96(%rdx), %ymm15
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4],ymm15[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm15
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm15[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm15[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%rdi), %ymm12
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm12
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm12[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm13[1],ymm11[2,3],ymm13[4],ymm11[5,6],ymm13[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm15[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm15[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2],ymm8[3,4],ymm13[5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0],ymm8[1,2],ymm12[3],ymm8[4,5],ymm12[6],ymm8[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm9[2],ymm6[3,4],ymm9[5],ymm6[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm10[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2],ymm9[3],ymm6[4,5],ymm9[6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm7[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0],ymm1[1,2],ymm5[3],ymm1[4,5],ymm5[6],ymm1[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm3[1,1,2,2]
@@ -1125,7 +1125,7 @@
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm2
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm12
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm14
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm4[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2],ymm5[3,4],ymm0[5],ymm5[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[1,1,2,2]
@@ -1150,12 +1150,12 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0],ymm7[1],ymm15[2,3],ymm7[4],ymm15[5,6],ymm7[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 64(%rdx), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2],ymm7[3,4],ymm15[5],ymm7[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3,4],ymm13[5],ymm11[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm13 = ymm14[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm13[0],ymm11[1,2],ymm13[3],ymm11[4,5],ymm13[6],ymm11[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm13 = ymm10[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm6[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2],ymm14[3,4],ymm13[5],ymm14[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm12[1,1,2,2]
@@ -1171,7 +1171,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm10[1],ymm6[2,3],ymm10[4],ymm6[5,6],ymm10[7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 32(%rdx), %ymm10
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm10[2],ymm6[3,4],ymm10[5],ymm6[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm15[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm15[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm3[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1],ymm10[2],ymm14[3,4],ymm10[5],ymm14[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm14
@@ -1229,7 +1229,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 88(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm9[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm4[2],ymm11[3,4],ymm4[5],ymm11[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[2,1,3,3]
@@ -1243,7 +1243,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm7[2,1,3,3]
@@ -1257,7 +1257,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 120(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2],ymm15[3,4],ymm14[5],ymm15[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm3[2,1,3,3]
@@ -1270,29 +1270,29 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 96(%rdx), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4],ymm15[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm15
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm15[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm15[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%rdi), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm12[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm13[1],ymm11[2,3],ymm13[4],ymm11[5,6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm15[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm15[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2],ymm8[3,4],ymm13[5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0],ymm8[1,2],ymm12[3],ymm8[4,5],ymm12[6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm9[2],ymm6[3,4],ymm9[5],ymm6[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm10[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1,2],ymm9[3],ymm6[4,5],ymm9[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm7[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0],ymm1[1,2],ymm5[3],ymm1[4,5],ymm5[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm3[1,1,2,2]
@@ -2048,7 +2048,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%rdi), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[2,1,3,3]
@@ -2062,7 +2062,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 56(%rdi), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[2,1,3,3]
@@ -2076,7 +2076,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastsd 88(%rdi), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm5[2,1,3,3]
@@ -2092,7 +2092,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm13
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 120(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2110,7 +2110,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %ymm10
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 152(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2127,7 +2127,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 184(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2144,7 +2144,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 216(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2161,7 +2161,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 248(%rdi), %ymm11
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1],ymm11[2],ymm0[3,4],ymm11[5],ymm0[6,7]
@@ -2193,29 +2193,29 @@
 ; AVX2-SLOW-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm13[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm13 = mem[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2],ymm13[3,4],ymm15[5],ymm13[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0],ymm13[1,2],ymm15[3],ymm13[4,5],ymm15[6],ymm13[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2],ymm12[3,4],ymm10[5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2],ymm9[3],ymm10[4,5],ymm9[6],ymm10[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2],ymm3[3],ymm4[4,5],ymm3[6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
@@ -2288,7 +2288,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd (%rdx), %ymm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2],ymm7[3,4],ymm11[5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2],ymm4[3,4],ymm7[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm10[1,1,2,2]
@@ -2307,7 +2307,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 32(%rdx), %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm8[2],ymm4[3,4],ymm8[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm3[1,1,2,2]
@@ -2325,7 +2325,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 64(%rdx), %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm0[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm9[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[1,1,2,2]
@@ -2343,7 +2343,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 96(%rdx), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm12[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm14[1,1,2,2]
@@ -2362,7 +2362,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 128(%rdx), %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm0[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdx), %ymm2
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm2[1,1,2,2]
@@ -2380,7 +2380,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 160(%rdx), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdx), %ymm4
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm4[1,1,2,2]
@@ -2398,7 +2398,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 192(%rdx), %ymm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm9[2],ymm0[3,4],ymm9[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm5[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm9[2],ymm4[3,4],ymm9[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdx), %ymm9
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm9[1,1,2,2]
@@ -2416,7 +2416,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 224(%rdx), %ymm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm11[2],ymm6[3,4],ymm11[5],ymm6[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[1,1,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm9[0,0,3,3,4,4,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm11
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[1,1,2,2]
@@ -2493,7 +2493,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[2,1,3,3]
@@ -2507,7 +2507,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[2,1,3,3]
@@ -2521,7 +2521,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 88(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm5[2,1,3,3]
@@ -2537,7 +2537,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 120(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2555,7 +2555,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %ymm10
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 152(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2572,7 +2572,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 184(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2589,7 +2589,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 216(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
@@ -2606,7 +2606,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,2,3,3,5,6,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 248(%rdi), %ymm11
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1],ymm11[2],ymm0[3,4],ymm11[5],ymm0[6,7]
@@ -2638,29 +2638,29 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm13[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm13[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm13 = mem[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2],ymm13[3,4],ymm15[5],ymm13[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0],ymm13[1,2],ymm15[3],ymm13[4,5],ymm15[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2],ymm12[3,4],ymm10[5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2],ymm9[3],ymm10[4,5],ymm9[6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2],ymm3[3],ymm4[4,5],ymm3[6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
@@ -41,8 +41,8 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,0,2,5,7,4,6]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3,0,2,5,7,4,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r8)
 ; AVX1-ONLY-NEXT:    vzeroupper
@@ -57,7 +57,7 @@
 ; AVX2-SLOW-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX2-SLOW-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%r8)
 ; AVX2-SLOW-NEXT:    vzeroupper
@@ -87,7 +87,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX2-FAST-PERLANE-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, (%r8)
 ; AVX2-FAST-PERLANE-NEXT:    vzeroupper
@@ -102,7 +102,7 @@
 ; AVX512F-SLOW-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX512F-SLOW-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX512F-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512F-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX512F-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX512F-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX512F-SLOW-NEXT:    vmovaps %ymm0, (%r8)
 ; AVX512F-SLOW-NEXT:    vzeroupper
@@ -132,7 +132,7 @@
 ; AVX512BW-SLOW-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX512BW-SLOW-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX512BW-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512BW-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX512BW-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX512BW-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX512BW-SLOW-NEXT:    vmovaps %ymm0, (%r8)
 ; AVX512BW-SLOW-NEXT:    vzeroupper
@@ -198,16 +198,16 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm2[0,1,1,0,4,5,5,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,1,1,0,4,5,5,4]
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} ymm6 = ymm5[0,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2],ymm3[3],ymm6[4,5],ymm3[6],ymm6[7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[1,0,2,3,5,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[1,0,2,3,5,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,3,2,4,5,7,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,3,2,4,5,7,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5],ymm2[6],ymm5[7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,2,3,7,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,2,3,7,6,6,7]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm4[1,0,3,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -321,7 +321,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm5[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm9[2,3],ymm3[4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm9 = xmm0[1],xmm1[1],zero,zero
@@ -329,7 +329,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm6[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3],ymm9[4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm7[2],xmm5[2],xmm7[3],xmm5[3]
@@ -337,7 +337,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[3,0],xmm2[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm7, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3],ymm2[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
@@ -345,7 +345,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm4
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
@@ -529,7 +529,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm15[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm13, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -539,7 +539,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm11[0],xmm12[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm13, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -549,7 +549,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm6[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -562,7 +562,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm10, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
@@ -571,7 +571,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm7[2],xmm9[2],xmm7[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[3,0],xmm7[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
@@ -580,7 +580,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm7[2],xmm5[2],xmm7[3],xmm5[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[3,0],xmm7[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
@@ -590,7 +590,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[3,0],xmm7[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
@@ -598,7 +598,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm13[2],xmm14[2],xmm13[3],xmm14[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[3,0],xmm13[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
@@ -987,7 +987,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1004,7 +1004,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1019,7 +1019,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm12[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1034,7 +1034,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm8
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm10[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1049,7 +1049,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm7
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1064,7 +1064,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm6[0],xmm5[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1079,7 +1079,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1093,7 +1093,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm11[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm15, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm14[2,3],ymm0[4,5],ymm14[6,7]
@@ -1107,7 +1107,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm1[2,3],ymm14[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1120,7 +1120,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm1[2],xmm15[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm1[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm0[2,3],ymm14[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -1131,7 +1131,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm14[2],xmm1[2],xmm14[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm1[3,0],xmm14[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm12, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm8[2],xmm10[2],xmm8[3],xmm10[3]
@@ -1141,7 +1141,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm1[2],xmm12[2],xmm1[3],xmm12[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[3,0],xmm1[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm10, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1],ymm8[2,3],ymm10[4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm7[2],xmm9[2],xmm7[3],xmm9[3]
@@ -1151,7 +1151,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[3,0],xmm1[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
@@ -1161,7 +1161,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm9[2],xmm1[2],xmm9[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm1[3,0],xmm9[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
@@ -1171,7 +1171,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm1[3,0],xmm6[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm2[2],xmm11[2],xmm2[3],xmm11[3]
@@ -1180,7 +1180,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm13[3,0],xmm4[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
@@ -1941,7 +1941,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm11[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1954,7 +1954,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm9[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1966,7 +1966,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -1983,7 +1983,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2000,7 +2000,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2017,7 +2017,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2034,7 +2034,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2051,7 +2051,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2068,7 +2068,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2085,7 +2085,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2102,7 +2102,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2118,7 +2118,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 176(%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm12[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2133,7 +2133,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm8
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2148,7 +2148,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2162,7 +2162,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -2175,7 +2175,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm6[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm14, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
@@ -2189,7 +2189,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2202,7 +2202,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2215,7 +2215,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2228,7 +2228,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2241,7 +2241,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2254,7 +2254,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2267,7 +2267,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2280,7 +2280,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2293,7 +2293,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2306,7 +2306,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2319,7 +2319,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -2330,7 +2330,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm0[3,0],xmm14[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm13, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm13 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
@@ -2340,7 +2340,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm13[2],xmm0[2],xmm13[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[3,0],xmm13[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
@@ -2350,7 +2350,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm0[3,0],xmm9[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
@@ -2359,7 +2359,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
@@ -2367,7 +2367,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm10[2],xmm7[2],xmm10[3],xmm7[3]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm7[3,0],xmm10[3,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,0,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,0,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%r8)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
@@ -49,13 +49,13 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,1,4,6,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,1,4,6,6,5]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[u,u,0,2,u,5,7,u]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovlps %xmm1, 32(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%r9)
@@ -246,7 +246,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5],ymm5[6],ymm4[7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm5 = <u,3,7,u,u,u,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm5, %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps %ymm4, 32(%r9)
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, (%r9)
@@ -410,7 +410,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm14
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm13, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3],ymm5[4,5,6],ymm13[7]
@@ -441,9 +441,9 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm10[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4],ymm8[5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
@@ -466,29 +466,29 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm6
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm7
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm8
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm9
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3],ymm11[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2,3],ymm10[4],ymm5[5,6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3],ymm6[4,5],ymm8[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm0[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4,5],ymm7[6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm2[1],ymm7[2,3,4],ymm2[5],ymm7[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm1[1,2],ymm8[3,4],ymm1[5,6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4],ymm8[5,6,7]
@@ -496,17 +496,17 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3,4],ymm8[5],ymm7[6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2],ymm10[3,4],ymm9[5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%r8), %ymm9
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
@@ -532,7 +532,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm8, %ymm5
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm9
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm10
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm10[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm11 = xmm10[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm9[2],xmm11[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3],ymm11[4,5],ymm5[6,7]
@@ -546,10 +546,10 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6],ymm8[7]
 ; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm6, %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm2[1],ymm7[2,3,4],ymm2[5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm1[1,2],ymm8[3,4],ymm1[5,6],ymm8[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4],ymm8[5,6,7]
@@ -557,17 +557,17 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3,4],ymm8[5],ymm7[6,7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2],ymm10[3,4],ymm9[5,6],ymm10[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
@@ -589,29 +589,29 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm7
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm8
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm10[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3],ymm11[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm10, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0],ymm5[1,2,3],ymm10[4],ymm5[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3],ymm6[4,5],ymm8[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm0[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3,4,5],ymm7[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm2[1],ymm7[2,3,4],ymm2[5],ymm7[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm1[1,2],ymm8[3,4],ymm1[5,6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4],ymm8[5,6,7]
@@ -619,17 +619,17 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3,4],ymm8[5],ymm7[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1,2],ymm10[3,4],ymm9[5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4],ymm9[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%r8), %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
@@ -875,7 +875,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm12[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm9[2,3],ymm4[4,5,6],ymm9[7]
@@ -886,7 +886,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm14
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm14[0],xmm13[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm15
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm15 = xmm8[0],xmm7[0],zero,zero
@@ -959,16 +959,16 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm0[2],ymm8[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm6[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm12[3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm14[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm14[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm9 = ymm13[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm3[2],ymm9[3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
@@ -1005,11 +1005,11 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm13
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %xmm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm13[2],xmm15[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
@@ -1019,37 +1019,37 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm15
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %xmm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm9[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3],ymm10[4,5,6],ymm13[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, (%r8), %ymm9, %ymm9
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm14[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm14[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2],xmm13[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm13[1],ymm10[2,3,4,5],ymm13[6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6],ymm13[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm11
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1,2,3],ymm11[4],ymm12[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm4[1],ymm12[2,3,4],ymm4[5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,2,2,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm3[1,2],ymm13[3,4],ymm3[5,6],ymm13[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
@@ -1057,15 +1057,15 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4],ymm14[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 56(%r8), %ymm14
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm15
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm0
@@ -1075,26 +1075,26 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 24(%r8), %ymm14
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm14[2],ymm8[3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1,2],ymm9[3,4],ymm15[5,6],ymm9[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3,4],ymm9[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm14
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0],ymm9[1,2,3,4],ymm14[5],ymm9[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -1132,7 +1132,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm13
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm14
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm15
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm12[2],xmm5[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
@@ -1141,7 +1141,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm7[2],xmm9[3],xmm7[3]
 ; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm11, %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm8 = xmm15[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm8 = xmm15[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm13[2],xmm8[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
@@ -1158,17 +1158,17 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm12
 ; AVX2-FAST-NEXT:    vinsertf128 $1, (%r8), %ymm6, %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm10[1,2,3],ymm6[4],ymm10[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm8[1],ymm10[2,3,4],ymm8[5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1,2],ymm14[3,4],ymm1[5,6],ymm14[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2],ymm10[3,4],ymm14[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm14
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0],ymm10[1,2,3,4],ymm14[5],ymm10[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm14 = ymm12[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm14[1,2],ymm4[3,4],ymm14[5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm14 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[3,3,3,3]
@@ -1185,8 +1185,8 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vinsertf128 $1, 32(%r8), %ymm7, %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0],ymm0[1,2,3],ymm4[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm9[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm9[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3,4],ymm0[5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm4
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm13
@@ -1196,26 +1196,26 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 24(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm11[1],ymm15[2,3,4],ymm11[5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,2]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm15[3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0],ymm5[1,2,3,4],ymm15[5],ymm5[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm2[2],ymm8[3,4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4,5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
@@ -1251,11 +1251,11 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm13
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %xmm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm13[2],xmm15[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
@@ -1265,37 +1265,37 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm15
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %xmm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm9[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3],ymm10[4,5,6],ymm13[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, (%r8), %ymm9, %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm10 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm14[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm14[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm13[1],ymm10[2,3,4,5],ymm13[6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6],ymm13[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm11, %ymm11
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1,2,3],ymm11[4],ymm12[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm12 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm4[1],ymm12[2,3,4],ymm4[5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm3[1,2],ymm13[3,4],ymm3[5,6],ymm13[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4],ymm13[5,6,7]
@@ -1303,15 +1303,15 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,1,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3,4],ymm14[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 56(%r8), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2],ymm15[3,4],ymm14[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm0
@@ -1321,26 +1321,26 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4],ymm14[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 24(%r8), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm14[2],ymm8[3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1,2],ymm9[3,4],ymm15[5,6],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm14[3,4],ymm9[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0],ymm9[1,2,3,4],ymm14[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -1921,7 +1921,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm12
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm12[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6],ymm2[7]
@@ -1934,7 +1934,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6],ymm2[7]
@@ -1947,7 +1947,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm8
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 68(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm13 = zero,xmm1[1],xmm2[1],zero
@@ -1962,7 +1962,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm6, %ymm13
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
@@ -2136,7 +2136,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
@@ -2147,16 +2147,16 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm12[1,2,3],ymm1[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm9[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm9 = ymm11[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2],ymm9[3,4,5,6],ymm7[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm15[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm15[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = ymm14[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
@@ -2215,14 +2215,14 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm6
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm2
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
@@ -2231,9 +2231,9 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm5
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
@@ -2241,19 +2241,19 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
@@ -2261,19 +2261,19 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -2283,12 +2283,12 @@
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm6
@@ -2298,23 +2298,23 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm14
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm13
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -2322,8 +2322,8 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm6[2],ymm13[3],ymm6[3],ymm13[6],ymm6[6],ymm13[7],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2334,20 +2334,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm12
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm11
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm10
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm9
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2358,20 +2358,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %ymm8
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %ymm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm6
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2382,20 +2382,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %ymm4
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm2[1,2],ymm15[3,4],ymm2[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4],ymm15[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 112(%r8), %ymm15
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2,3,4],ymm15[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
@@ -2403,36 +2403,36 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 120(%r8), %ymm15
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
 ; AVX2-SLOW-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2],ymm14[3,4,5,6],ymm13[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1,2,3],ymm0[4,5],ymm13[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4,5],ymm6[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -2497,7 +2497,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm8
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm8[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm8[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm7[2],xmm3[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
@@ -2506,7 +2506,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm9[2],xmm6[2],xmm9[3],xmm6[3]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2],xmm3[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
@@ -2517,7 +2517,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm14, %ymm10
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm3[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm11 = xmm3[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2],xmm11[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
@@ -2536,7 +2536,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm11
 ; AVX2-FAST-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm12[2],xmm11[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
@@ -2556,7 +2556,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm7, %ymm8
 ; AVX2-FAST-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7]
@@ -2565,15 +2565,15 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm13
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm12
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm12[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm12[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm13[1,2],ymm5[3,4],ymm13[5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3,4],ymm5[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm12[2],ymm13[2],ymm12[3],ymm13[3],ymm12[6],ymm13[6],ymm12[7],ymm13[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
@@ -2592,20 +2592,20 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm10
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm9
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1],ymm1[2,3,4],ymm10[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm8
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm7[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm8[1,2],ymm2[3,4],ymm8[5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4],ymm2[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4],ymm2[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm10[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[3,3,3,3]
@@ -2625,20 +2625,20 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm6
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1],ymm0[2,3,4],ymm6[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2661,20 +2661,20 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm14
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1,2],ymm11[3,4],ymm15[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 112(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2,3,4],ymm15[5],ymm11[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0],ymm11[1,2],ymm15[3,4],ymm11[5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
@@ -2682,7 +2682,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm15[3,4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 120(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4],ymm12[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm13 = mem[0,1,3,0,4,5,7,4]
@@ -2692,26 +2692,26 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1,2,3],ymm12[4,5],ymm13[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm12 = ymm12[0,1,2],mem[3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm9[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3,4,5,6],ymm8[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm5[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4,5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -2771,14 +2771,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
@@ -2787,9 +2787,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
@@ -2797,19 +2797,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
@@ -2817,19 +2817,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -2839,12 +2839,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm6
@@ -2854,23 +2854,23 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm14
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -2878,8 +2878,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm6[2],ymm13[3],ymm6[3],ymm13[6],ymm6[6],ymm13[7],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2890,20 +2890,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2914,20 +2914,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %ymm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3,4],ymm6[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -2938,20 +2938,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm2[1,2],ymm15[3,4],ymm2[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4],ymm15[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 112(%r8), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2,3,4],ymm15[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm4[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0],ymm0[1,2],ymm15[3,4],ymm0[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
@@ -2959,36 +2959,36 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 120(%r8), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2],ymm0[3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm14[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $78, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2],ymm14[3,4,5,6],ymm13[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1,2,3],ymm0[4,5],ymm13[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2],ymm11[3,4,5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4,5],ymm6[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -4107,7 +4107,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 4(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5,6],ymm4[7]
@@ -4122,7 +4122,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 36(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5,6],ymm4[7]
@@ -4136,7 +4136,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 68(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertps {{.*#+}} xmm3 = zero,xmm13[1],xmm10[1],zero
@@ -4151,7 +4151,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 100(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm11
@@ -4168,7 +4168,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 132(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm0
@@ -4186,7 +4186,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 164(%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm6
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
@@ -4204,7 +4204,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 196(%rdx), %xmm6
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm0
@@ -4222,7 +4222,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vbroadcastss 228(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
@@ -4580,7 +4580,7 @@
 ; AVX1-ONLY-NEXT:    # ymm3 = mem[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm14[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm14[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm10 = mem[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2],ymm10[3,4,5,6],ymm6[7]
@@ -4633,17 +4633,17 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0],ymm12[1,2,3],ymm14[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm12 = ymm12[0,1,2],mem[3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4],ymm7[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm11[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm11[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm11 = mem[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1],ymm8[2],ymm11[3,4,5,6],ymm8[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1,2,3],ymm7[4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3],ymm7[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[0,1,3,0,4,5,7,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,1,3,0,4,5,7,4]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm5[1,0,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
@@ -4741,14 +4741,14 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm6
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm2
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm12
@@ -4757,9 +4757,9 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %xmm5
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
@@ -4767,19 +4767,19 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
@@ -4787,19 +4787,19 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -4809,12 +4809,12 @@
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %ymm6
@@ -4823,9 +4823,9 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -4835,23 +4835,23 @@
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 128(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-SLOW-NEXT:    vmovaps 128(%r8), %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -4861,21 +4861,21 @@
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 160(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -4887,21 +4887,21 @@
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 192(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -4913,21 +4913,21 @@
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %xmm3
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %xmm4
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, 224(%r8), %ymm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -4938,7 +4938,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %ymm14
 ; AVX2-SLOW-NEXT:    vmovaps (%rcx), %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
@@ -4946,7 +4946,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2],ymm1[3,4],ymm3[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -4954,8 +4954,8 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -4966,7 +4966,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -4976,7 +4976,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -4984,8 +4984,8 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -4996,7 +4996,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -5006,7 +5006,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -5014,8 +5014,8 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5026,7 +5026,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -5036,7 +5036,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -5044,8 +5044,8 @@
 ; AVX2-SLOW-NEXT:    vbroadcastsd 112(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5056,7 +5056,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -5066,15 +5066,15 @@
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %ymm13
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 144(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5085,20 +5085,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rdx), %ymm12
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %ymm11
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rdi), %ymm9
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %ymm8
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 176(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5109,20 +5109,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdx), %ymm7
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rcx), %ymm6
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rdi), %ymm5
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 208(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5133,20 +5133,20 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm3
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rcx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm1
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0,1,2],ymm10[3,4],ymm15[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 240(%r8), %ymm15
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2,3,4],ymm15[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm10, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2],ymm15[3,4],ymm10[5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
@@ -5161,7 +5161,7 @@
 ; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
@@ -5205,7 +5205,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm13 = ymm13[0,1,2,3],mem[4],ymm13[5,6,7]
 ; AVX2-SLOW-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
@@ -5216,26 +5216,26 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2,3],ymm13[4,5],ymm14[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4],ymm8[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2],ymm11[3,4,5,6],ymm9[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5],ymm9[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -5340,7 +5340,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm8
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %xmm14
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %xmm9
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm14[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm14[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
@@ -5349,7 +5349,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm9[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
@@ -5360,7 +5360,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %xmm6
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
@@ -5379,7 +5379,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm12 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm12 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],xmm0[2],xmm12[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2,3],ymm12[4,5],ymm15[6,7]
@@ -5398,7 +5398,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
@@ -5417,7 +5417,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rcx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
@@ -5436,7 +5436,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
@@ -5455,7 +5455,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm0[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,1,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3],ymm15[4,5],ymm12[6,7]
@@ -5475,7 +5475,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovaps (%rcx), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm13
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1],ymm10[2,3,4],ymm0[5],ymm10[6,7]
@@ -5484,7 +5484,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %ymm15
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm15[1,2],ymm11[3,4],ymm15[5,6],ymm11[7]
@@ -5493,8 +5493,8 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%r8), %ymm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0],ymm10[1,2,3,4],ymm11[5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm13[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm13[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2],ymm11[3,4],ymm10[5,6],ymm11[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm1[2],ymm15[2],ymm1[3],ymm15[3],ymm1[6],ymm15[6],ymm1[7],ymm15[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
@@ -5513,7 +5513,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 32(%rcx), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm8
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3,4],ymm0[5],ymm4[6,7]
@@ -5523,7 +5523,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdi), %ymm9
 ; AVX2-FAST-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm10
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm9[1,2],ymm5[3,4],ymm9[5,6],ymm5[7]
@@ -5532,8 +5532,8 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%r8), %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0],ymm4[1,2,3,4],ymm5[5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3,4],ymm4[5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
@@ -5553,7 +5553,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm5
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
@@ -5563,7 +5563,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 64(%rdi), %ymm6
 ; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovaps %ymm0, %ymm7
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm6[1,2],ymm4[3,4],ymm6[5,6],ymm4[7]
@@ -5572,8 +5572,8 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 80(%r8), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1,2,3,4],ymm4[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm5[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2],ymm4[3,4],ymm3[5,6],ymm4[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm7[2],ymm6[2],ymm7[3],ymm6[3],ymm7[6],ymm6[6],ymm7[7],ymm6[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
@@ -5596,7 +5596,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm3, %ymm4
 ; AVX2-FAST-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -5606,7 +5606,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm5
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm6[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -5614,8 +5614,8 @@
 ; AVX2-FAST-NEXT:    vbroadcastsd 112(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm4[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[6],ymm5[6],ymm6[7],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5638,7 +5638,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdx), %ymm13
 ; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
@@ -5648,15 +5648,15 @@
 ; AVX2-FAST-NEXT:    vmovaps 128(%rdi), %ymm4
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rsi), %ymm13
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 144(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5679,22 +5679,22 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdx), %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 160(%rcx), %ymm12
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovaps %ymm1, %ymm3
 ; AVX2-FAST-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 160(%rdi), %ymm10
 ; AVX2-FAST-NEXT:    vmovaps 160(%rsi), %ymm9
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1,2],ymm1[3,4],ymm10[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 176(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm12[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm12[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5717,20 +5717,20 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdx), %ymm7
 ; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %ymm6
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 192(%rdi), %ymm5
 ; AVX2-FAST-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 208(%r8), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -5753,20 +5753,20 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm3
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %ymm1
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm1[1,2],ymm11[3,4],ymm1[5,6],ymm11[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastsd 240(%r8), %ymm11
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0],ymm8[1,2,3,4],ymm11[5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm11 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm11 = ymm3[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0],ymm8[1,2],ymm11[3,4],ymm8[5,6],ymm11[7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
@@ -5824,7 +5824,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm14[1,2,3],ymm8[4,5],ymm14[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm14 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm14 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm14 = ymm14[0,1,2,3],mem[4],ymm14[5,6,7]
 ; AVX2-FAST-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
@@ -5835,27 +5835,27 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3],ymm14[4,5],ymm13[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4],ymm9[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm12[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm12[0,1,3,0,4,5,7,4]
 ; AVX2-FAST-NEXT:    vpermilps $78, (%rsp), %ymm12 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm12 = mem[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2],ymm12[3,4,5,6],ymm10[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4,5],ymm10[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm9 = ymm9[0,1,2],mem[3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -5954,14 +5954,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm6
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm10
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm11 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm11[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1],ymm5[2,3],ymm12[4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm12
@@ -5970,9 +5970,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2],xmm7[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
@@ -5980,19 +5980,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5],ymm8[6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm8[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 32(%r8), %ymm8, %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3],ymm8[4],ymm7[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
@@ -6000,19 +6000,19 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm4[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 64(%r8), %ymm4, %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3],ymm4[4],ymm3[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -6022,12 +6022,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %ymm6
@@ -6036,9 +6036,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
@@ -6048,23 +6048,23 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 128(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%r8), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -6074,21 +6074,21 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 160(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -6100,21 +6100,21 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 192(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -6126,21 +6126,21 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %xmm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3],ymm6[4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, 224(%r8), %ymm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3],ymm5[4],ymm2[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
@@ -6151,7 +6151,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %ymm14
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rcx), %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
@@ -6159,7 +6159,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2],ymm1[3,4],ymm3[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -6167,8 +6167,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6179,7 +6179,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -6189,7 +6189,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -6197,8 +6197,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 48(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6209,7 +6209,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -6219,7 +6219,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -6227,8 +6227,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 80(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6239,7 +6239,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -6249,7 +6249,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
@@ -6257,8 +6257,8 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 112(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6269,7 +6269,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -6279,15 +6279,15 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %ymm13
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3,4],ymm4[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 144(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm13[2],ymm4[2],ymm13[3],ymm4[3],ymm13[6],ymm4[6],ymm13[7],ymm4[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6298,20 +6298,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdx), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %ymm11
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rdi), %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm9[1,2],ymm1[3,4],ymm9[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 176(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6322,20 +6322,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdx), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rcx), %ymm6
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rdi), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1,2],ymm1[3,4],ymm5[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 208(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -6346,20 +6346,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm1[1,2],ymm15[3,4],ymm1[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0,1,2],ymm10[3,4],ymm15[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 240(%r8), %ymm15
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2,3,4],ymm15[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[0,2,3,3,4,6,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm3[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0],ymm10[1,2],ymm15[3,4],ymm10[5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
@@ -6374,7 +6374,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2,3],mem[4],ymm10[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm15 = mem[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6],ymm15[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
@@ -6418,7 +6418,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm14[1,2,3],ymm10[4,5],ymm14[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm10 = ymm10[0,1,2],mem[3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm13[0,1,2,3],mem[4],ymm13[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
@@ -6429,26 +6429,26 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2,3],ymm13[4,5],ymm14[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm13 = ymm13[0,1,2],mem[3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4],ymm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm11[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm9[2],ymm11[3,4,5,6],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4,5],ymm9[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = ymm8[0,1,2],mem[3],ymm8[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4,5,6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,0,4,5,7,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
@@ -56,7 +56,7 @@
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[u,u,0,2,u,u,5,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[1,3],xmm4[1,3]
@@ -78,9 +78,9 @@
 ; AVX2-SLOW-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX2-SLOW-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
 ; AVX2-SLOW-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[0,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
@@ -127,9 +127,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
 ; AVX2-FAST-PERLANE-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
 ; AVX2-FAST-PERLANE-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[0,2,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
@@ -228,24 +228,24 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[1],ymm8[1],ymm9[4],ymm8[4],ymm9[5],ymm8[5]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm11
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm6[0],ymm11[0],ymm6[2],ymm11[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[0,0],xmm1[0,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm6 = ymm11[2],ymm6[2],ymm11[3],ymm6[3],ymm11[6],ymm6[6],ymm11[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm11
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm11[1,2],ymm7[5,6],ymm11[5,6]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3],ymm6[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3],ymm2[3,3],ymm0[7,7],ymm2[7,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, 32(%rax)
@@ -285,7 +285,7 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [2,6,3,7,2,6,3,7]
 ; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,0,1]
 ; AVX2-ONLY-NEXT:    vpermps %ymm8, %ymm1, %ymm1
@@ -414,7 +414,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[1],ymm8[1],ymm7[4],ymm8[4],ymm7[5],ymm8[5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
@@ -423,7 +423,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm10
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm10[1,2],xmm9[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
@@ -437,30 +437,30 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm8 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm2[1,2],ymm3[1,2],ymm2[5,6],ymm3[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r8), %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm12[2,3],ymm7[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r9), %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm11, %ymm9
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm10, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3,4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm10[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm10[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4,5,6],ymm10[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm2[3,0],ymm3[7,4],ymm2[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm8[2,3],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss (%rcx), %xmm2
 ; AVX1-ONLY-NEXT:    vbroadcastss (%rdx), %xmm3
@@ -944,7 +944,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm6[1,2],xmm8[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -961,7 +961,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm4[0],ymm13[0],ymm4[1],ymm13[1],ymm4[4],ymm13[4],ymm4[5],ymm13[5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],ymm7[0],ymm9[2],ymm7[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5],ymm0[6,7]
@@ -969,7 +969,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm10[1,2],xmm3[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
@@ -988,7 +988,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm11
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm12
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1],ymm10[2,3],ymm14[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm14
@@ -999,7 +999,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm13 = ymm4[2],ymm13[2],ymm4[3],ymm13[3],ymm4[6],ymm13[6],ymm4[7],ymm13[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm7[1,2],ymm9[1,2],ymm7[5,6],ymm9[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm13[4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 52(%r8), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm10[2,3],ymm4[4,5,6,7]
@@ -1008,20 +1008,20 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[1,2],ymm12[1,2],ymm11[5,6],ymm12[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r8), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r9), %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0],ymm5[1],ymm0[2,3,4,5,6],ymm5[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 32(%rcx), %xmm0
@@ -1035,25 +1035,25 @@
 ; AVX1-ONLY-NEXT:    vbroadcastss 32(%r9), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm9[3,0],ymm7[3,0],ymm9[7,4],ymm7[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm15, %ymm3
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm7, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3,4,5,6],ymm6[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss (%rcx), %xmm6
@@ -1069,14 +1069,14 @@
 ; AVX1-ONLY-NEXT:    vbroadcastss (%r9), %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm12[3,0],ymm11[3,0],ymm12[7,4],ymm11[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm7[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm7[1],ymm4[2,3,4,5,6],ymm7[7]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, 160(%rax)
@@ -2351,7 +2351,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2370,7 +2370,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm10[0],ymm4[0],ymm10[2],ymm4[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
@@ -2378,7 +2378,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,2],xmm2[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
@@ -2402,7 +2402,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm1
@@ -2416,7 +2416,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,2],xmm0[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2441,7 +2441,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
@@ -2455,7 +2455,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm15[1,2]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm4
@@ -2476,7 +2476,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm11
 ; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm11[2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm1
@@ -2490,7 +2490,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[1,2],ymm10[1,2],ymm0[5,6],ymm10[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r8), %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3],ymm9[4,5,6,7]
@@ -2503,7 +2503,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm10[1,2],ymm9[1,2],ymm10[5,6],ymm9[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 52(%r8), %xmm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
@@ -2514,7 +2514,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm1[2],ymm5[2],ymm1[3],ymm5[3],ymm1[6],ymm5[6],ymm1[7],ymm5[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[1,2],ymm14[1,2],ymm6[5,6],ymm14[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 84(%r8), %xmm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5,6,7]
@@ -2522,7 +2522,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2],ymm6[3],ymm5[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm13[0],xmm4[1],xmm13[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm15[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm15[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = mem[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
@@ -2536,7 +2536,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,2],ymm13[1,2],ymm12[5,6],ymm13[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 116(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
@@ -2559,26 +2559,26 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm2[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 32(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vbroadcastss 32(%rdx), %xmm2
@@ -2595,26 +2595,26 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm3[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm3[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0],ymm3[1],ymm1[2,3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[3,0],ymm10[3,0],ymm9[7,4],ymm10[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3,4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 64(%rcx), %xmm3
 ; AVX1-ONLY-NEXT:    vbroadcastss 64(%rdx), %xmm5
@@ -2631,50 +2631,50 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm8, %ymm5
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm8[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm8[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm9, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1],ymm5[2,3,4,5,6],ymm8[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = ymm8[3,0],mem[3,0],ymm8[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm8[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm8 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3,4,5,6],ymm8[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm9, %ymm8
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3,4,5,6],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm13[3,0],ymm12[3,0],ymm13[7,4],ymm12[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm9 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1],ymm4[2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm9[1],ymm4[2,3,4,5,6],ymm9[7]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps %ymm4, 736(%rax)
@@ -2930,7 +2930,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm13 = xmm10[2],mem[2],xmm10[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm13, %ymm10, %ymm13
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
@@ -2963,7 +2963,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm8, %ymm1
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
@@ -2996,7 +2996,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm9, %ymm3
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
@@ -3031,7 +3031,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm10, %ymm5
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
@@ -3685,7 +3685,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm13 = xmm10[2],mem[2],xmm10[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm13, %ymm10, %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
@@ -3718,7 +3718,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm8, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
@@ -3751,7 +3751,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm3, %ymm9, %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
@@ -3786,7 +3786,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm10, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
@@ -5147,7 +5147,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5165,7 +5165,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5],ymm0[6,7]
@@ -5173,7 +5173,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm3[1,2],xmm2[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5199,7 +5199,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm1
@@ -5213,7 +5213,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,2],xmm0[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5239,7 +5239,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
@@ -5253,7 +5253,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5278,7 +5278,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm1
@@ -5292,7 +5292,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5317,7 +5317,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm1
@@ -5331,7 +5331,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm6
@@ -5355,7 +5355,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm1
@@ -5369,7 +5369,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,2],xmm0[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm2
@@ -5391,7 +5391,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm1
@@ -5405,7 +5405,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[1,2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5429,7 +5429,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm5[0],ymm1[0],ymm5[2],ymm1[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm15, %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm15[2,3],ymm12[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm0
@@ -5445,7 +5445,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm15 = ymm0[1,2],mem[1,2],ymm0[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm15[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm12[4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 20(%r8), %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3],ymm15[4,5,6,7]
@@ -5460,7 +5460,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,2],mem[1,2],ymm0[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 52(%r8), %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5,6,7]
@@ -5475,7 +5475,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,2],mem[1,2],ymm0[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 84(%r8), %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5,6,7]
@@ -5503,7 +5503,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm0[1,2],mem[1,2],ymm0[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm13[4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 116(%r8), %xmm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5,6,7]
@@ -5517,7 +5517,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm11 = ymm11[1,2],mem[1,2],ymm11[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm0[4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 148(%r8), %xmm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6,7]
@@ -5526,9 +5526,9 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm15[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm15[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm14[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm14[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm11[0],xmm7[0],xmm11[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3],ymm11[4,5,6,7]
@@ -5545,7 +5545,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = ymm0[1,2],mem[1,2],ymm0[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 180(%r8), %xmm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
@@ -5556,7 +5556,7 @@
 ; AVX1-ONLY-NEXT:    vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm3 = mem[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[0,0,0,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[0,0,0,0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3],ymm6[4,5,6,7]
@@ -5570,7 +5570,7 @@
 ; AVX1-ONLY-NEXT:    vshufps $153, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = ymm3[1,2],mem[1,2],ymm3[5,6],mem[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 212(%r8), %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
@@ -5581,7 +5581,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,2],ymm5[1,2],ymm1[5,6],ymm5[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 244(%r8), %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
@@ -5604,28 +5604,28 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vbroadcastss 32(%rcx), %xmm0
@@ -5644,29 +5644,29 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vbroadcastss 64(%rcx), %xmm1
@@ -5685,53 +5685,53 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm2[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = mem[2,3],ymm2[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0],ymm4[1],ymm2[2,3,4,5,6],ymm4[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm4[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm4[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3,4,5,6],ymm4[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm4 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm13[2,3],ymm4[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm4[0],ymm5[1],ymm4[2,3,4,5,6],ymm5[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 128(%rcx), %xmm4
 ; AVX1-ONLY-NEXT:    vbroadcastss 128(%rdx), %xmm11
@@ -5748,75 +5748,75 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm11 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm12, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm12[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm12[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm13, %ymm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1],ymm11[2,3,4,5,6],ymm12[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm12 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm10 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm10 = mem[2,3],ymm12[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm12 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm12[1],ymm10[2,3,4,5,6],ymm12[7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm13, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm14, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4,5,6],ymm13[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = ymm0[3,0],mem[3,0],ymm0[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm9 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm9 = mem[2,3],ymm13[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm13 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm13[0,1],ymm9[2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm13[1],ymm9[2,3,4,5,6],ymm13[7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm13 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm13 = xmm7[2],mem[2],xmm7[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm14, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm15, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4,5,6],ymm14[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm3[3,0],ymm0[7,4],ymm3[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3],ymm14[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm14[1],ymm3[2,3,4,5,6],ymm14[7]
 ; AVX1-ONLY-NEXT:    vbroadcastss 224(%rcx), %xmm14
 ; AVX1-ONLY-NEXT:    vbroadcastss 224(%rdx), %xmm15
@@ -5833,28 +5833,28 @@
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm14 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm14 = xmm7[2],mem[2],xmm7[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm7, %ymm14
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = mem[2,1,3,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm15, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%r9), %xmm15
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm15[0,2,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm15[0,2,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm14[0],ymm7[1],ymm14[2,3,4,5,6],ymm7[7]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = ymm6[3,0],mem[3,0],ymm6[7,4],mem[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = mem[2,3],ymm14[2,3]
 ; AVX1-ONLY-NEXT:    vperm2f128 $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm14 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0,1],ymm6[2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm14[1],ymm6[2,3,4,5,6],ymm14[7]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps %ymm6, 1504(%rax)
@@ -6380,7 +6380,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm14, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
@@ -6407,10 +6407,10 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[2,1,3,3,6,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[2,1,3,3,6,5,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = mem[0,2,2,3,4,6,6,7]
@@ -6420,7 +6420,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -6459,7 +6459,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -6497,7 +6497,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -6534,7 +6534,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -6569,7 +6569,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -6604,7 +6604,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
 ; AVX2-SLOW-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
@@ -6639,7 +6639,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm13 # 16-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # xmm13 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm13, %ymm3, %ymm13
 ; AVX2-SLOW-NEXT:    vpermilps $250, (%rsp), %xmm14 # 16-byte Folded Reload
@@ -6664,10 +6664,10 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss 240(%r9), %ymm3
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4],ymm3[5],ymm14[6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm14 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-SLOW-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm14 = mem[2,3],ymm14[2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[2,1,3,3,6,5,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,1,3,3,6,5,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5],ymm15[6,7]
 ; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm15 = mem[0,2,2,3,4,6,6,7]
@@ -7975,7 +7975,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm14, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
@@ -8002,10 +8002,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm15[2,1,3,3,6,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[2,1,3,3,6,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = mem[0,2,2,3,4,6,6,7]
@@ -8015,7 +8015,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8054,7 +8054,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8092,7 +8092,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8129,7 +8129,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8164,7 +8164,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8199,7 +8199,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
@@ -8234,7 +8234,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm13 # 16-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # xmm13 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm13, %ymm3, %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps $250, (%rsp), %xmm14 # 16-byte Folded Reload
@@ -8259,10 +8259,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%r9), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4],ymm3[5],ymm14[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm14 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm14 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,2,3,6,7,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm14 = mem[2,3],ymm14[2,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = ymm15[2,1,3,3,6,5,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[2,1,3,3,6,5,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5],ymm15[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm15 = mem[0,2,2,3,4,6,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
@@ -96,9 +96,9 @@
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} xmm3 = <3,5,7,u>
@@ -166,9 +166,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} xmm3 = <3,5,7,u>
@@ -393,7 +393,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm7[0],ymm5[0],ymm7[1],ymm5[1],ymm7[4],ymm5[4],ymm7[5],ymm5[5]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[0,0],xmm3[0,0]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,1,2,0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6],ymm3[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3],xmm2[3,3]
@@ -423,15 +423,15 @@
 ; AVX2-SLOW-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = [2,6,0,3,2,6,0,3]
 ; AVX2-SLOW-NEXT:    # ymm8 = mem[0,1,0,1]
 ; AVX2-SLOW-NEXT:    vpermps %ymm6, %ymm8, %ymm8
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[3,3,3,3,7,7,7,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm9
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm9[2],ymm3[2],ymm9[3],ymm3[3],ymm9[6],ymm3[6],ymm9[7],ymm3[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm9
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm5[1],xmm0[1],zero
 ; AVX2-SLOW-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [5,0,2,6,5,0,2,6]
@@ -474,7 +474,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm3
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm6
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm8 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm8 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4],ymm7[5],ymm8[6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm8 = <u,1,5,u,u,u,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm8, %ymm8
@@ -491,7 +491,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5,6],ymm9[7]
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm4, %ymm4
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm5 = <u,u,0,4>
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm5, %ymm3
@@ -506,7 +506,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm3 = <u,3,7,u,u,u,u,u>
 ; AVX2-FAST-NEXT:    vpermps %ymm6, %ymm3, %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps %ymm2, (%rax)
 ; AVX2-FAST-NEXT:    vmovaps %ymm4, 64(%rax)
@@ -532,15 +532,15 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = [2,6,0,3,2,6,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    # ymm8 = mem[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm6, %ymm8, %ymm8
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm4[3,3,3,3,7,7,7,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm9[2],ymm3[2],ymm9[3],ymm3[3],ymm9[6],ymm3[6],ymm9[7],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm5 = zero,xmm5[1],xmm0[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [5,0,2,6,5,0,2,6]
@@ -792,12 +792,12 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[2,0],xmm13[2,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm12, %ymm12
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1],ymm9[2,3],ymm12[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6],ymm9[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm9[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
@@ -820,7 +820,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm9 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm9, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5,6],ymm9[7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm5[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rax), %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5,6,7]
@@ -859,7 +859,7 @@
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm9
 ; AVX2-SLOW-NEXT:    vmovaps (%r8), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %xmm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm5[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm10 = xmm5[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm4[1],xmm10[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm10, %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5],ymm10[6,7]
@@ -868,7 +868,7 @@
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm10 = zero,xmm13[1],xmm12[1],zero
 ; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm14
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm15
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm15[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm15[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm14[2],xmm11[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2],ymm11[3,4,5,6,7]
@@ -878,7 +878,7 @@
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm6[2],ymm11[3,4,5],ymm6[6],ymm11[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 16(%rax), %xmm9
@@ -888,10 +888,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm9 = xmm14[3,3],xmm15[3,3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm11 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5,6],ymm11[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm11 = xmm5[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm11 = xmm5[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 8(%rax), %ymm10
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
@@ -900,7 +900,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm13, %xmm10
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
@@ -909,12 +909,12 @@
 ; AVX2-SLOW-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm12[0],ymm10[2],ymm12[2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm9[0,1,2,3],ymm10[4,5,6],ymm9[7]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1],ymm10[2,3,4],ymm0[5],ymm10[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6],ymm10[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm13 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm13 = ymm6[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0],ymm10[1],ymm13[2,3,4],ymm10[5],ymm13[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%rax), %ymm13
@@ -931,7 +931,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1,2,3,4,5,6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[4],ymm2[4],ymm0[5],ymm2[5]
 ; AVX2-SLOW-NEXT:    vbroadcastss 16(%rdx), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[3,3],xmm5[3,3]
@@ -964,7 +964,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm6
 ; AVX2-FAST-NEXT:    vmovaps (%r8), %xmm4
 ; AVX2-FAST-NEXT:    vmovaps (%r9), %xmm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm9 = xmm5[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm9 = xmm5[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0],xmm4[1],xmm9[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm9, %ymm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3,4],ymm6[5],ymm9[6,7]
@@ -973,7 +973,7 @@
 ; AVX2-FAST-NEXT:    vinsertps {{.*#+}} xmm9 = zero,xmm12[1],xmm11[1],zero
 ; AVX2-FAST-NEXT:    vmovaps (%rdi), %xmm13
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm14
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm13[2],xmm15[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,2,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm15[0],ymm9[1,2],ymm15[3,4,5,6,7]
@@ -983,20 +983,20 @@
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm8[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm7[2],ymm9[3,4,5],ymm7[6],ymm9[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 16(%rax), %xmm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1],ymm9[2,3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm6[2,3,4,5],ymm9[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1],ymm6[2,3,4],ymm1[5],ymm6[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[2,2,2,2]
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm10[1,1],ymm3[1,1],ymm10[5,5],ymm3[5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm15[5,6],ymm6[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm8[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm8[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm7[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0],ymm15[1],ymm9[2,3,4],ymm15[5],ymm9[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[2,1,2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%rax), %ymm15
@@ -1010,7 +1010,7 @@
 ; AVX2-FAST-NEXT:    # ymm0 = mem[0,1,0,1]
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm0, %ymm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3,4],ymm6[5,6],ymm9[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm9 = xmm5[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm9 = xmm5[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm9 = xmm4[0,1,2],xmm9[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 8(%rax), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5,6,7]
@@ -1038,7 +1038,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6],ymm0[7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
 ; AVX2-FAST-NEXT:    vbroadcastss 16(%rdx), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[3,3],xmm5[3,3]
@@ -1071,7 +1071,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r8), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %xmm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm5[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm10 = xmm5[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm4[1],xmm10[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm10, %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5],ymm10[6,7]
@@ -1080,7 +1080,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm10 = zero,xmm13[1],xmm12[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdi), %xmm14
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm15
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm15[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm15[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm14[2],xmm11[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2],ymm11[3,4,5,6,7]
@@ -1090,7 +1090,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[6],ymm2[6],ymm0[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm11 = ymm7[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm6[2],ymm11[3,4,5],ymm6[6],ymm11[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 16(%rax), %xmm9
@@ -1100,10 +1100,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm9 = xmm14[3,3],xmm15[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm11 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5,6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm11 = xmm5[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm11 = xmm5[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 8(%rax), %ymm10
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
@@ -1112,7 +1112,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm13, %xmm10
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm10 = xmm10[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
@@ -1121,12 +1121,12 @@
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm12[0],ymm10[2],ymm12[2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm9[0,1,2,3],ymm10[4,5,6],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0],ymm0[1],ymm10[2,3,4],ymm0[5],ymm10[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm7[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm13 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm7[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm13 = ymm6[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm13[0],ymm10[1],ymm13[2,3,4],ymm10[5],ymm13[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%rax), %ymm13
@@ -1143,7 +1143,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1,2,3,4,5,6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[4],ymm2[4],ymm0[5],ymm2[5]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 16(%rdx), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[3,3],xmm5[3,3]
@@ -1583,7 +1583,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[1,1,2,2,5,5,6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,1,2,2,5,5,6,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3,4,5,6],ymm4[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm2[2,3],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4],ymm4[5],ymm3[6,7]
@@ -1606,7 +1606,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, %xmm14
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
@@ -1668,7 +1668,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm12[2,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1691,7 +1691,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6],ymm0[7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm7[2],ymm6[2],ymm7[3],ymm6[3],ymm7[6],ymm6[6],ymm7[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm3[2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
@@ -1724,7 +1724,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[3,3],ymm1[3,3],ymm2[7,7],ymm1[7,7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[2,3],ymm1[1,2],ymm4[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1737,7 +1737,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm5[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rax), %ymm2
@@ -1750,7 +1750,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm15[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm14[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rax), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
@@ -1821,7 +1821,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -1835,18 +1835,18 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdi), %xmm8
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm12
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm7
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0],ymm5[1,2],ymm6[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm4
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm13[1],xmm5[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm5, %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm12[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm2[2],xmm5[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps (%rdx), %xmm5
@@ -1869,7 +1869,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm0[2],ymm6[3,4,5],ymm0[6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 16(%rax), %xmm14
@@ -1889,7 +1889,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %ymm2
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[1,2,2,3,5,6,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0],ymm1[1,2,3,4,5,6],ymm10[7]
 ; AVX2-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
@@ -1910,11 +1910,11 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[3,3],xmm7[3,3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 40(%rax), %ymm10
@@ -1925,7 +1925,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm9, %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3],ymm1[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
@@ -1940,11 +1940,11 @@
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm5[2],xmm9[2],xmm5[3],xmm9[3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 8(%rax), %ymm2
@@ -1955,7 +1955,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm5, %xmm1
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm12[0],xmm10[1],xmm12[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
@@ -1967,7 +1967,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm6[1,1],ymm4[5,5],ymm6[5,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
@@ -1981,7 +1981,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm2[0,1,2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[2,2,2,2]
@@ -1989,9 +1989,9 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[1,1],ymm13[1,1],ymm2[5,5],ymm13[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4],ymm3[5,6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3,4],ymm3[5],ymm4[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%rax), %ymm4
@@ -2010,7 +2010,7 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3,4],ymm4[5,6,7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm14[0],ymm15[0],ymm14[1],ymm15[1],ymm14[4],ymm15[4],ymm14[5],ymm15[5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vbroadcastss 48(%rdx), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5],ymm5[6,7]
@@ -2020,7 +2020,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm12[0],ymm2[0],ymm12[1],ymm2[1],ymm12[4],ymm2[4],ymm12[5],ymm2[5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm13[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm13[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vbroadcastss 16(%rdx), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
@@ -2072,7 +2072,7 @@
 ; AVX2-FAST-NEXT:    vmovaps (%r9), %xmm8
 ; AVX2-FAST-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%r9), %xmm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; AVX2-FAST-NEXT:    vmovaps %xmm4, %xmm14
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
@@ -2088,18 +2088,18 @@
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm10
 ; AVX2-FAST-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm4 = xmm3[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0],ymm1[1,2],ymm4[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm7[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm10[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm10[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps (%rdx), %xmm2
@@ -2124,7 +2124,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 16(%rax), %xmm4
@@ -2169,7 +2169,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4],ymm1[5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vmovaps %xmm14, %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm15 = xmm8[0,1,2],xmm15[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 40(%rax), %ymm14
@@ -2190,12 +2190,12 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm11[1,1],ymm2[5,5],ymm11[5,5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm4[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm10[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4],ymm2[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,1,2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 48(%rax), %ymm3
@@ -2212,7 +2212,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm3, %ymm7, %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6],ymm3[7]
 ; AVX2-FAST-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm10[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps (%rsp), %xmm14 # 16-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm3 = xmm14[0,1,2],xmm3[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 8(%rax), %ymm4
@@ -2233,7 +2233,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6],ymm0[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3,4],ymm6[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm3[2,2,2,2]
@@ -2242,9 +2242,9 @@
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[1,1],ymm1[1,1],ymm3[5,5],ymm1[5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm9 = ymm2[0,1,2,3,4],ymm4[5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm4 = ymm7[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm4 = ymm7[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm2[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm2[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3,4],ymm4[5],ymm5[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%rax), %ymm5
@@ -2264,7 +2264,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,1,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm13[0],ymm12[0],ymm13[1],ymm12[1],ymm13[4],ymm12[4],ymm13[5],ymm12[5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm6 = ymm11[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm6 = ymm11[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vbroadcastss 48(%rdx), %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6],ymm6[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
@@ -2273,7 +2273,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1,2],xmm15[3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm7 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm7 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vbroadcastss 16(%rdx), %ymm8
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6],ymm7[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
@@ -2327,7 +2327,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -2341,18 +2341,18 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdi), %xmm8
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm12
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm7
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm7[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0],ymm5[1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm4
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm13[1],xmm5[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm5, %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm12[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm2[2],xmm5[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm5[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rdx), %xmm5
@@ -2375,7 +2375,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm0[2],ymm6[3,4,5],ymm0[6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 16(%rax), %xmm14
@@ -2395,7 +2395,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm10 = ymm2[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm10 = ymm2[1,2,2,3,5,6,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0],ymm1[1,2,3,4,5,6],ymm10[7]
 ; AVX2-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
@@ -2416,11 +2416,11 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[3,3],xmm7[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 40(%rax), %ymm10
@@ -2431,7 +2431,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm9, %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1],ymm3[2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
@@ -2446,11 +2446,11 @@
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm5[2],xmm9[2],xmm5[3],xmm9[3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 8(%rax), %ymm2
@@ -2461,7 +2461,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm5, %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm12[0],xmm10[1],xmm12[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
@@ -2473,7 +2473,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm6[1,1],ymm4[5,5],ymm6[5,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
@@ -2487,7 +2487,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm2[0,1,2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[2,2,2,2]
@@ -2495,9 +2495,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[1,1],ymm13[1,1],ymm2[5,5],ymm13[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3,4],ymm3[5,6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm1[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm1[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm4 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm4 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3,4],ymm3[5],ymm4[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%rax), %ymm4
@@ -2516,7 +2516,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm14[0],ymm15[0],ymm14[1],ymm15[1],ymm14[4],ymm15[4],ymm14[5],ymm15[5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 48(%rdx), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5],ymm5[6,7]
@@ -2526,7 +2526,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm12[0],ymm2[0],ymm12[1],ymm2[1],ymm12[4],ymm2[4],ymm12[5],ymm2[5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm13[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm13[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 16(%rdx), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
@@ -3445,7 +3445,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -3469,7 +3469,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -3526,7 +3526,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3589,7 +3589,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3649,7 +3649,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm14[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3675,7 +3675,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
@@ -3691,7 +3691,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm5[2],ymm8[3],ymm5[3],ymm8[6],ymm5[6],ymm8[7],ymm5[7]
@@ -3706,7 +3706,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm15[1],ymm0[3],ymm15[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3785,7 +3785,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3],ymm1[1,2],ymm2[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -3798,7 +3798,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm15[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rax), %ymm13
@@ -3817,7 +3817,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm2[2,3],ymm11[1,2],ymm2[6,7],ymm11[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0],ymm1[1,2,3,4],ymm11[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -3830,7 +3830,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6],ymm8[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm12[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm12[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rax), %ymm10
@@ -3852,7 +3852,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[2,3],ymm4[1,2],ymm2[6,7],ymm4[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm9[1,2,3,4],ymm4[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
@@ -3973,7 +3973,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %xmm8
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %xmm9
 ; AVX2-SLOW-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -3988,18 +3988,18 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm12
 ; AVX2-SLOW-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm4[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2],xmm2[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm9[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm6
@@ -4011,7 +4011,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rax), %xmm1
@@ -4022,7 +4022,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %xmm13
@@ -4034,7 +4034,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r8), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r9), %xmm11
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm11[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rax), %xmm1
@@ -4045,7 +4045,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm3
@@ -4072,7 +4072,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 16(%rax), %xmm2
@@ -4095,7 +4095,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 48(%rax), %xmm2
@@ -4118,7 +4118,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rax), %xmm2
@@ -4127,7 +4127,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdi), %ymm3
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rdx), %ymm9
@@ -4146,7 +4146,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm7, %xmm15
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3],ymm15[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4161,10 +4161,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[3,3],xmm4[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm7[2],xmm10[2],xmm7[3],xmm10[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm8[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 8(%rax), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -4177,7 +4177,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm10[0],xmm14[0],xmm10[1],xmm14[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
@@ -4191,10 +4191,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm10[3,3],xmm14[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm9[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 40(%rax), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -4206,7 +4206,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -4220,10 +4220,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm9[3,3],xmm8[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 72(%rax), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -4237,7 +4237,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -4250,23 +4250,23 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[3,3],xmm7[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm4 = xmm11[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 104(%rax), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastss 112(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6],ymm5[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 108(%r8), %ymm5
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[2,2,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
@@ -4296,7 +4296,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
@@ -4305,9 +4305,9 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[1,1],ymm4[1,1],ymm9[5,5],ymm4[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm13[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4],ymm2[5],ymm3[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%rax), %ymm3
@@ -4315,7 +4315,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm12[1],ymm2[2,3,4],ymm12[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
@@ -4353,7 +4353,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm3[3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 16(%rdx), %ymm3
 ; AVX2-SLOW-NEXT:    vmovaps %ymm4, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6],ymm7[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5],ymm3[6,7]
@@ -4374,7 +4374,7 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm3[1,2,3,4],ymm8[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 48(%rdx), %ymm3
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm3[6],ymm9[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm12[0],ymm14[0],ymm12[1],ymm14[1],ymm12[4],ymm14[4],ymm12[5],ymm14[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5],ymm3[6,7]
@@ -4399,7 +4399,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0],ymm3[1,2,3,4],ymm11[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 80(%rdx), %ymm11
 ; AVX2-SLOW-NEXT:    vmovaps %ymm15, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm12 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm12 = ymm15[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6],ymm12[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -4495,7 +4495,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%r9), %xmm5
 ; AVX2-FAST-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -4513,18 +4513,18 @@
 ; AVX2-FAST-NEXT:    vmovaps (%rsi), %xmm13
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm8
 ; AVX2-FAST-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm13[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm6 = xmm13[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm9[2],xmm6[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0],ymm1[1,2],ymm6[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm5[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm5[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm8[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm2
@@ -4537,7 +4537,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 64(%rax), %xmm1
@@ -4548,7 +4548,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm2
@@ -4562,7 +4562,7 @@
 ; AVX2-FAST-NEXT:    vmovaps 96(%r8), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%r9), %xmm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[1,1,1,1]
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, %xmm12
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
@@ -4575,7 +4575,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm2
@@ -4600,7 +4600,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm6
 ; AVX2-FAST-NEXT:    vmovaps (%r9), %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm6[2],ymm1[3,4,5],ymm6[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4625,7 +4625,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 48(%rax), %xmm15
@@ -4648,7 +4648,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rax), %xmm15
@@ -4657,7 +4657,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdi), %ymm3
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0],ymm3[1],ymm15[2,3,4],ymm3[5],ymm15[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rdx), %ymm4
@@ -4672,13 +4672,13 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vbroadcastss 112(%rdx), %ymm14
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6],ymm15[7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm15 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 108(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0],ymm15[1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm12[2,2,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm12[2,2,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3],ymm14[4,5,6,7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[2,2,2,2]
@@ -4730,20 +4730,20 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm7, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm11[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm11[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 8(%rax), %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm2[2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm10[1],ymm1[2,3,4],ymm10[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[1,1],ymm8[1,1],ymm9[5,5],ymm8[5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm2 = ymm5[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm3 = ymm6[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm3 = ymm6[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4],ymm2[5],ymm3[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,1,2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 16(%rax), %ymm3
@@ -4773,7 +4773,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vpermps %ymm2, %ymm7, %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6],ymm2[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm8[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 40(%rax), %ymm3
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
@@ -4821,14 +4821,14 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
 ; AVX2-FAST-NEXT:    vpermps %ymm5, %ymm7, %ymm5
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6],ymm5[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm5 = xmm1[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm5 = xmm1[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 72(%rax), %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4],ymm4[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
@@ -4868,14 +4868,14 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm9, %ymm7, %ymm0
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 104(%rax), %ymm9
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm6[2,3,4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 16(%rdx), %ymm6
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3,4,5],ymm6[6],ymm9[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm10[0],ymm1[0],ymm10[1],ymm1[1],ymm10[4],ymm1[4],ymm10[5],ymm1[5]
@@ -4900,7 +4900,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,1,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1,2,3,4],ymm7[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 48(%rdx), %ymm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm10 = ymm11[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm10 = ymm11[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6],ymm10[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm10 = ymm12[0],ymm0[0],ymm12[1],ymm0[1],ymm12[4],ymm0[4],ymm12[5],ymm0[5]
@@ -4925,7 +4925,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1,2,3,4],ymm11[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 80(%rdx), %ymm11
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm13[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm12 = ymm13[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6],ymm12[7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm12 = ymm14[0],ymm15[0],ymm14[1],ymm15[1],ymm14[4],ymm15[4],ymm14[5],ymm15[5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
@@ -5016,7 +5016,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %xmm8
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %xmm9
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -5031,18 +5031,18 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm12
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm4[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2],xmm2[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm9[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm6
@@ -5054,7 +5054,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rax), %xmm1
@@ -5065,7 +5065,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %xmm13
@@ -5077,7 +5077,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r8), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r9), %xmm11
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm11[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rax), %xmm1
@@ -5088,7 +5088,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm3
@@ -5115,7 +5115,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 16(%rax), %xmm2
@@ -5138,7 +5138,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 48(%rax), %xmm2
@@ -5161,7 +5161,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rax), %xmm2
@@ -5170,7 +5170,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdi), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rdx), %ymm9
@@ -5189,7 +5189,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm7, %xmm15
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3],ymm15[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5204,10 +5204,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm5[3,3],xmm4[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm7[2],xmm10[2],xmm7[3],xmm10[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm8[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm8[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 8(%rax), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -5220,7 +5220,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm10[0],xmm14[0],xmm10[1],xmm14[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
@@ -5234,10 +5234,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm10[3,3],xmm14[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm9[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 40(%rax), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -5249,7 +5249,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -5263,10 +5263,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm9[3,3],xmm8[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 72(%rax), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
@@ -5280,7 +5280,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -5293,23 +5293,23 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm8[3,3],xmm7[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm4 = xmm11[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm4 = xmm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 104(%rax), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3,4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 112(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm5 = ymm1[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6],ymm5[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 108(%r8), %ymm5
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[2,2,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
@@ -5339,7 +5339,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm8[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
@@ -5348,9 +5348,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[1,1],ymm4[1,1],ymm9[5,5],ymm4[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm13[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm3 = ymm11[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4],ymm2[5],ymm3[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%rax), %ymm3
@@ -5358,7 +5358,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm1[3,4,5,6],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm12[1],ymm2[2,3,4],ymm12[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
@@ -5396,7 +5396,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm3[3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 16(%rdx), %ymm3
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm4, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm7 = ymm4[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6],ymm7[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5],ymm3[6,7]
@@ -5417,7 +5417,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0],ymm3[1,2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 48(%rdx), %ymm3
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm3[6],ymm9[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm12[0],ymm14[0],ymm12[1],ymm14[1],ymm12[4],ymm14[4],ymm12[5],ymm14[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5],ymm3[6,7]
@@ -5442,7 +5442,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0],ymm3[1,2,3,4],ymm11[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 80(%rdx), %ymm11
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm15, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm12 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm12 = ymm15[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6],ymm12[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7232,7 +7232,7 @@
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7]
@@ -7255,7 +7255,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -7312,7 +7312,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
@@ -7376,7 +7376,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7439,7 +7439,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7502,7 +7502,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7566,7 +7566,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7627,7 +7627,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7672,7 +7672,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
@@ -7689,7 +7689,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm8[2],ymm1[3],ymm8[3],ymm1[6],ymm8[6],ymm1[7],ymm8[7]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
@@ -7705,7 +7705,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm8[2],mem[2],ymm8[3],mem[3],ymm8[6],mem[6],ymm8[7],mem[7]
@@ -7722,7 +7722,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
@@ -7739,7 +7739,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
@@ -7756,7 +7756,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovaps %ymm2, %ymm10
@@ -7773,7 +7773,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm14[2],ymm7[3],ymm14[3],ymm7[6],ymm14[6],ymm7[7],ymm14[7]
@@ -7829,7 +7829,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm11[2,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm1, %ymm1
@@ -7909,7 +7909,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3],ymm1[1,2],ymm2[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -7944,7 +7944,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3],ymm1[1,2],ymm2[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -7978,7 +7978,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3],ymm1[1,2],ymm2[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8012,7 +8012,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,3],ymm1[1,2],ymm2[6,7],ymm1[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8047,7 +8047,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,3],ymm2[1,2],ymm3[6,7],ymm2[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8060,7 +8060,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm15[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm13[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rax), %ymm3
@@ -8081,7 +8081,7 @@
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[2,3],ymm2[1,2],ymm3[6,7],ymm2[5,6]
 ; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,3,1,4,6,7,5]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8094,7 +8094,7 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6],ymm2[7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rax), %ymm3
@@ -8313,7 +8313,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -8328,7 +8328,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-SLOW-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2],xmm2[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4,5,6,7]
@@ -8337,11 +8337,11 @@
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rax), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm11[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 32(%rdx), %xmm7
@@ -8353,7 +8353,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rax), %xmm1
@@ -8364,7 +8364,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 64(%rcx), %xmm2
@@ -8378,7 +8378,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r9), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rax), %xmm1
@@ -8389,7 +8389,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 96(%rcx), %xmm3
@@ -8404,7 +8404,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%r9), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rax), %xmm1
@@ -8415,7 +8415,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 128(%rcx), %xmm3
@@ -8430,7 +8430,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%r9), %xmm0
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rax), %xmm1
@@ -8441,7 +8441,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 160(%rcx), %xmm3
@@ -8456,7 +8456,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%r8), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rax), %xmm1
@@ -8467,7 +8467,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rsi), %xmm1
 ; AVX2-SLOW-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vmovaps 192(%rcx), %xmm3
@@ -8494,7 +8494,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 16(%rax), %xmm2
@@ -8517,7 +8517,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 48(%rax), %xmm2
@@ -8540,7 +8540,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 80(%rax), %xmm2
@@ -8563,7 +8563,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 96(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 112(%rax), %xmm2
@@ -8586,7 +8586,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 128(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 144(%rax), %xmm2
@@ -8609,7 +8609,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 160(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 176(%rax), %xmm2
@@ -8632,7 +8632,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 192(%r9), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vmovaps 208(%rax), %xmm2
@@ -8647,7 +8647,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %xmm4
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %xmm2
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3],ymm15[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 224(%r8), %ymm15
@@ -8659,14 +8659,14 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm15, %ymm13
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm13[6],ymm14[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm4[1,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm4[1,1,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm2[2],xmm13[3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 228(%r8), %ymm12
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm13
@@ -8675,20 +8675,20 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovaps 224(%r8), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdi), %ymm4
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rsi), %ymm2
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vmovaps 224(%rdx), %ymm10
@@ -8706,7 +8706,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastss %xmm8, %xmm13
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
@@ -8720,10 +8720,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 8(%rax), %ymm8
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
@@ -8736,7 +8736,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -8750,10 +8750,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm15[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm15[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 40(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8766,7 +8766,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8780,10 +8780,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 72(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8797,7 +8797,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -8811,10 +8811,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 104(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8828,7 +8828,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -8842,10 +8842,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 136(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8859,7 +8859,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -8873,10 +8873,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 168(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8890,7 +8890,7 @@
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -8904,10 +8904,10 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 200(%rax), %ymm7
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8930,13 +8930,13 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastss 240(%rdx), %ymm5
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 236(%r8), %ymm6
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,2,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
@@ -8966,7 +8966,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -8975,9 +8975,9 @@
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vbroadcastsd 16(%rax), %ymm2
@@ -8985,7 +8985,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -9004,7 +9004,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-SLOW-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -9081,7 +9081,7 @@
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
@@ -9096,7 +9096,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-SLOW-NEXT:    vbroadcastss 16(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
@@ -9118,7 +9118,7 @@
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 48(%rdx), %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
@@ -9143,7 +9143,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 80(%rdx), %ymm0
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
@@ -9169,7 +9169,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 112(%rdx), %ymm2
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -9196,7 +9196,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 144(%rdx), %ymm1
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm1[6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -9223,7 +9223,7 @@
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0],ymm0[1,2,3,4],ymm7[5,6,7]
 ; AVX2-SLOW-NEXT:    vbroadcastss 176(%rdx), %ymm7
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
@@ -9381,7 +9381,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%r9), %xmm4
 ; AVX2-FAST-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -9400,7 +9400,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%rsi), %xmm7
 ; AVX2-FAST-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2],xmm2[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4,5,6,7]
@@ -9409,11 +9409,11 @@
 ; AVX2-FAST-NEXT:    vmovaps 32(%rax), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 32(%rdx), %xmm2
@@ -9426,7 +9426,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 64(%rax), %xmm1
@@ -9437,7 +9437,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 64(%rcx), %xmm3
@@ -9452,7 +9452,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%r9), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 96(%rax), %xmm1
@@ -9463,7 +9463,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 96(%rcx), %xmm3
@@ -9478,7 +9478,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%r9), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 128(%rax), %xmm1
@@ -9489,7 +9489,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 128(%rcx), %xmm3
@@ -9504,7 +9504,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%r9), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 160(%rax), %xmm1
@@ -9515,7 +9515,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 160(%rcx), %xmm3
@@ -9530,7 +9530,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%r8), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps 192(%rax), %xmm1
@@ -9541,7 +9541,7 @@
 ; AVX2-FAST-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%rsi), %xmm1
 ; AVX2-FAST-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 192(%rcx), %xmm3
@@ -9567,7 +9567,7 @@
 ; AVX2-FAST-NEXT:    vmovaps (%r8), %ymm2
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps (%r9), %ymm12
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
@@ -9591,7 +9591,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 48(%rax), %xmm2
@@ -9614,7 +9614,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 80(%rax), %xmm2
@@ -9637,7 +9637,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 96(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 112(%rax), %xmm2
@@ -9660,7 +9660,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 128(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 144(%rax), %xmm2
@@ -9683,7 +9683,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 160(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 176(%rax), %xmm2
@@ -9707,7 +9707,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 192(%r9), %ymm1
 ; AVX2-FAST-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-NEXT:    vmovaps 208(%rax), %xmm2
@@ -9716,7 +9716,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %xmm0
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %xmm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[1,1,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[1,1,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2],xmm2[3]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rcx), %xmm3
@@ -9726,7 +9726,7 @@
 ; AVX2-FAST-NEXT:    vbroadcastss 228(%r8), %ymm4
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%r9), %xmm7
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm15 = xmm7[1,1,1,1]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm15 = xmm7[1,1,1,1]
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rax), %xmm15
@@ -9755,14 +9755,14 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovaps 224(%r8), %ymm6
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdi), %ymm2
 ; AVX2-FAST-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm4 = ymm0[2,2,2,2]
 ; AVX2-FAST-NEXT:    vmovaps 224(%rdx), %ymm3
@@ -9789,13 +9789,13 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0],ymm4[1,2,3,4,5,6],ymm15[7]
 ; AVX2-FAST-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vbroadcastss 240(%rdx), %ymm4
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm15 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm15[0,1,2,3,4,5],ymm4[6],ymm15[7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm15[4,5],ymm4[6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 236(%r8), %ymm15
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm15[1],ymm4[2,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm7 = xmm7[2,2,3,3]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[2,2,3,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5,6,7]
 ; AVX2-FAST-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
@@ -9846,7 +9846,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 8(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -9860,7 +9860,7 @@
 ; AVX2-FAST-NEXT:    vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm1 = ymm14[1,1],mem[1,1],ymm14[5,5],mem[5,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-NEXT:    vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm2 = mem[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
@@ -9892,7 +9892,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 40(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -9940,14 +9940,14 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 72(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -9988,7 +9988,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 104(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -10036,7 +10036,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 136(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -10084,7 +10084,7 @@
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 168(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -10132,7 +10132,7 @@
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm10, %ymm1
 ; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,2,2,2]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[2,2,2,2]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
 ; AVX2-FAST-NEXT:    vbroadcastsd 200(%rax), %ymm2
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -10142,7 +10142,7 @@
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm13[1,1],ymm0[5,5],ymm13[5,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
@@ -10159,7 +10159,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vbroadcastss 16(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -10188,7 +10188,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vbroadcastss 48(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -10217,7 +10217,7 @@
 ; AVX2-FAST-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-NEXT:    vbroadcastss 80(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovaps %ymm15, %ymm3
@@ -10244,7 +10244,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 112(%rdx), %ymm0
 ; AVX2-FAST-NEXT:    vmovaps %ymm11, %ymm3
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm5 = ymm11[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6],ymm5[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm5 = ymm12[0],ymm2[0],ymm12[1],ymm2[1],ymm12[4],ymm2[4],ymm12[5],ymm2[5]
@@ -10270,7 +10270,7 @@
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0],ymm0[1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 144(%rdx), %ymm5
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm12 = ymm2[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm12 = ymm2[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm12[0,1,2,3,4,5],ymm5[6],ymm12[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm12 = ymm9[0],ymm1[0],ymm9[1],ymm1[1],ymm9[4],ymm1[4],ymm9[5],ymm1[5]
@@ -10295,7 +10295,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,1,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm0[1,2,3,4],ymm5[5,6,7]
 ; AVX2-FAST-NEXT:    vbroadcastss 176(%rdx), %ymm5
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm14[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm14[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6],ymm9[7]
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -10321,7 +10321,7 @@
 ; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[2,1,2,3]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4],ymm8[5,6,7]
 ; AVX2-FAST-NEXT:    vunpcklps {{.*#+}} ymm8 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
-; AVX2-FAST-NEXT:    vpermilps {{.*#+}} ymm9 = ymm13[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} ymm9 = ymm13[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-NEXT:    vbroadcastss 208(%rdx), %ymm7
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6],ymm9[7]
 ; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5],ymm7[6,7]
@@ -10454,7 +10454,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
@@ -10469,7 +10469,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rsi), %xmm11
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2],xmm2[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4,5,6,7]
@@ -10478,11 +10478,11 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rax), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm11[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm11[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%rdx), %xmm7
@@ -10494,7 +10494,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rax), %xmm1
@@ -10505,7 +10505,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%rcx), %xmm2
@@ -10519,7 +10519,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r9), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rax), %xmm1
@@ -10530,7 +10530,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%rcx), %xmm3
@@ -10545,7 +10545,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%r9), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rax), %xmm1
@@ -10556,7 +10556,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%rcx), %xmm3
@@ -10571,7 +10571,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%r9), %xmm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rax), %xmm1
@@ -10582,7 +10582,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%rcx), %xmm3
@@ -10597,7 +10597,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%r8), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rax), %xmm1
@@ -10608,7 +10608,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rsi), %xmm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%rcx), %xmm3
@@ -10635,7 +10635,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps (%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 16(%rax), %xmm2
@@ -10658,7 +10658,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 32(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 48(%rax), %xmm2
@@ -10681,7 +10681,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 64(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 80(%rax), %xmm2
@@ -10704,7 +10704,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 96(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 112(%rax), %xmm2
@@ -10727,7 +10727,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 128(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 144(%rax), %xmm2
@@ -10750,7 +10750,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 160(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 176(%rax), %xmm2
@@ -10773,7 +10773,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 192(%r9), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,1,2,2,5,5,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 208(%rax), %xmm2
@@ -10788,7 +10788,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %xmm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %xmm2
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3],ymm15[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 224(%r8), %ymm15
@@ -10800,14 +10800,14 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm15, %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm13[6],ymm14[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm4[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm4[1,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm2[2],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 228(%r8), %ymm12
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm15, %ymm14, %ymm13
@@ -10816,20 +10816,20 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%r8), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 232(%rax), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdi), %ymm4
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rsi), %ymm2
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps 224(%rdx), %ymm10
@@ -10847,7 +10847,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss %xmm8, %xmm13
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
@@ -10861,10 +10861,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 8(%rax), %ymm8
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
@@ -10877,7 +10877,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -10891,10 +10891,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm15[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm15[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 40(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10907,7 +10907,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -10921,10 +10921,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 72(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10938,7 +10938,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -10952,10 +10952,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 104(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10969,7 +10969,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -10983,10 +10983,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 136(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -11000,7 +11000,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
@@ -11014,10 +11014,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm11[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm11[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 168(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -11031,7 +11031,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -11045,10 +11045,10 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
 ; AVX2-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 200(%rax), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -11071,13 +11071,13 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 240(%rdx), %ymm5
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 236(%r8), %ymm6
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,2,3,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
@@ -11107,7 +11107,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -11116,9 +11116,9 @@
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastsd 16(%rax), %ymm2
@@ -11126,7 +11126,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -11145,7 +11145,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-FAST-PERLANE-NEXT:    # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -11222,7 +11222,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
@@ -11237,7 +11237,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 16(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
@@ -11259,7 +11259,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 48(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
@@ -11284,7 +11284,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 80(%rdx), %ymm0
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
@@ -11310,7 +11310,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 112(%rdx), %ymm2
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -11337,7 +11337,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 144(%rdx), %ymm1
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm1[6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -11364,7 +11364,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0],ymm0[1,2,3,4],ymm7[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastss 176(%rdx), %ymm7
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1,2,0,7,5,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
 ; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
@@ -495,7 +495,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm6[0],ymm4[1],ymm6[1],ymm4[4],ymm6[4],ymm4[5],ymm6[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm3[1,0],ymm5[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -506,7 +506,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm2 = ymm4[2],ymm6[2],ymm4[3],ymm6[3],ymm4[6],ymm6[6],ymm4[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm5[3,0],ymm3[3,0],ymm5[7,4],ymm3[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm11[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -517,7 +517,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm11 = ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[6],ymm5[6],ymm3[7],ymm5[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm11, %xmm11
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm6[1],ymm4[1],ymm6[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm12, %xmm12
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm12[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm2[4,5,6,7]
@@ -528,7 +528,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm3 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[4],ymm5[4],ymm3[5],ymm5[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm6[0],ymm4[0],ymm6[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
@@ -543,18 +543,18 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm13
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm14[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm13[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm11[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm11[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm10[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0,1],xmm9[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3],ymm8[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm15 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm15 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
@@ -565,14 +565,14 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm10[0],xmm9[0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm12[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
@@ -1079,7 +1079,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm10 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
@@ -1090,7 +1090,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm10 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm13[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
@@ -1101,7 +1101,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm2[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm2
@@ -1117,7 +1117,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
@@ -1128,7 +1128,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm2[1],ymm9[1],ymm2[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1139,7 +1139,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[3,0],ymm7[3,0],ymm8[7,4],ymm7[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1150,7 +1150,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm7[0],ymm8[0],ymm7[1],ymm8[1],ymm7[4],ymm8[4],ymm7[5],ymm8[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm9[0],ymm2[2],ymm9[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1161,7 +1161,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm9[0],ymm2[0],ymm9[1],ymm2[1],ymm9[4],ymm2[4],ymm9[5],ymm2[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm8[1,0],ymm7[1,0],ymm8[5,4],ymm7[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1175,12 +1175,12 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm15
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm13[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm13[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm15[1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm11[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm11[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm0
@@ -1194,14 +1194,14 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm12[0]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r10), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm9[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm9[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm7
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm6
@@ -1211,7 +1211,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm0[0,1,2],xmm12[3]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
@@ -1222,7 +1222,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm12 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm12[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
@@ -1230,16 +1230,16 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm8[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm9 = xmm8[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm9[0],xmm4[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm5[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm5[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm6[1],xmm3[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
@@ -1249,14 +1249,14 @@
 ; AVX1-ONLY-NEXT:    # xmm3 = xmm11[2],mem[2],xmm11[3],mem[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm13[2],xmm15[2],xmm13[3],xmm15[3]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm9[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm9[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm14[0,1,2],xmm6[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
@@ -1264,7 +1264,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm6[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
@@ -1413,7 +1413,7 @@
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rsi), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm4
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, %xmm5
@@ -1425,7 +1425,7 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vmovaps (%r9), %xmm6
 ; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm14
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm14[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm6[1],xmm2[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -1448,7 +1448,7 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm8
 ; AVX2-ONLY-NEXT:    vmovaps 32(%r10), %xmm5
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm8[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm8[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %xmm10
@@ -1458,7 +1458,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rcx), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm1[0,1,2],xmm13[3]
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %xmm4
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm3
@@ -1470,7 +1470,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm7 = xmm5[2],xmm8[2],xmm5[3],xmm8[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm6[4,5,6,7]
@@ -1488,11 +1488,11 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm6[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm10[1],xmm2[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -1503,13 +1503,13 @@
 ; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm2 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # xmm2 = xmm14[2],mem[2],xmm14[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm11[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm11[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm12[0,1,2],xmm3[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm5[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm5[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm15[0,1,2],xmm4[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0,1],xmm4[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -1517,7 +1517,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm12[2],xmm11[2],xmm12[3],xmm11[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -2335,7 +2335,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
@@ -2346,7 +2346,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm9 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm4[1,0],ymm3[1,0],ymm4[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm9 = xmm13[0,1],xmm9[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
@@ -2357,7 +2357,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm6[1],ymm5[1],ymm6[3],ymm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm7[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm7
@@ -2373,7 +2373,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[3,0],ymm4[7,4],ymm3[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
@@ -2384,7 +2384,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm7[0],ymm2[0],ymm7[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -2395,7 +2395,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm2[0],ymm7[0],ymm2[1],ymm7[1],ymm2[4],ymm7[4],ymm2[5],ymm7[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -2406,7 +2406,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm7[1],ymm2[1],ymm7[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -2423,7 +2423,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %ymm0
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
@@ -2436,7 +2436,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -2447,7 +2447,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm4[1,0],ymm3[1,0],ymm4[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -2458,7 +2458,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -2471,7 +2471,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[3,0],ymm4[7,4],ymm3[7,4]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %ymm1
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r10), %ymm4
@@ -2488,7 +2488,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -2499,7 +2499,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm3[1,0],ymm2[1,0],ymm3[5,4],ymm2[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -2510,7 +2510,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -2521,7 +2521,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm2[3,0],ymm3[7,4],ymm2[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -2533,7 +2533,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm10
 ; AVX1-ONLY-NEXT:    vmovaps (%r10), %xmm1
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm6
@@ -2545,11 +2545,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm9[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm2[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
@@ -2557,12 +2557,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -2571,7 +2571,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -2583,7 +2583,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r10), %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm10
@@ -2595,11 +2595,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm15[0],xmm6[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm4[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0,1,2,3,4,5],ymm5[6,7]
@@ -2607,12 +2607,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm8[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm7[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm10[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm10[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -2621,7 +2621,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -2633,7 +2633,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r10), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm10 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm10[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm10[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm15
@@ -2644,23 +2644,23 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm14[0],xmm2[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm14[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm4[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm10[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm15[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm1[4,5,6,7]
@@ -2668,7 +2668,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -2679,7 +2679,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r10), %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm8[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm8[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm14
@@ -2690,23 +2690,23 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm13[0],xmm2[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm1[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm13[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm5[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm13[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm7[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm0[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
@@ -2714,7 +2714,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -3010,13 +3010,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -3024,12 +3024,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -3038,7 +3038,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -3065,13 +3065,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -3079,12 +3079,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -3093,7 +3093,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -3119,25 +3119,25 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm4[1],xmm8[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm15[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm7[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm7[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm0[1],xmm14[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3],ymm14[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm5[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3],ymm5[4,5,6,7]
@@ -3145,7 +3145,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm1[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -3170,25 +3170,25 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm5[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm6[1],xmm5[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm12[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm0[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm0[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0],xmm15[1],xmm11[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm11[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm6[2],xmm14[3],xmm6[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm1[0,1,2],xmm11[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm4[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm3[0,1,2],xmm12[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm6[0,1],xmm12[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
@@ -3196,7 +3196,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm6[1],xmm3[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
@@ -4905,7 +4905,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm10 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
@@ -4916,7 +4916,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm10 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm4[1,0],ymm3[1,0],ymm4[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm13[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
@@ -4927,7 +4927,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm9 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm9, %xmm9
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm6[1],ymm5[1],ymm6[3],ymm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm13, %xmm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm9[0,1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %ymm9
@@ -4943,7 +4943,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm5 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[3,0],ymm4[7,4],ymm3[7,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
@@ -4954,7 +4954,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm9[0],ymm2[0],ymm9[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -4965,7 +4965,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm4 = ymm2[0],ymm9[0],ymm2[1],ymm9[1],ymm2[4],ymm9[4],ymm2[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -4976,7 +4976,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm4 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm4, %xmm4
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
@@ -4993,7 +4993,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %ymm0
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm8, %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7]
@@ -5006,7 +5006,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -5017,7 +5017,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm4[1,0],ymm3[1,0],ymm4[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -5028,7 +5028,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
@@ -5041,7 +5041,7 @@
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[3,0],ymm4[7,4],ymm3[7,4]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %ymm1
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm5, %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r10), %ymm4
@@ -5058,7 +5058,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5069,7 +5069,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm3[1,0],ymm2[1,0],ymm3[5,4],ymm2[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5080,7 +5080,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5091,7 +5091,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm2[3,0],ymm3[7,4],ymm2[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -5110,7 +5110,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5121,7 +5121,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5132,7 +5132,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5143,7 +5143,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -5162,7 +5162,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5173,7 +5173,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5184,7 +5184,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5195,7 +5195,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -5214,7 +5214,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5225,7 +5225,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5236,7 +5236,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5247,7 +5247,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -5266,7 +5266,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %ymm8
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %ymm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5277,7 +5277,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[1,0],ymm1[5,4],ymm0[5,4]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5288,7 +5288,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm7 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm7, %xmm7
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,2,0,4,5,6,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm10, %xmm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -5299,7 +5299,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
 ; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[3,0],ymm1[7,4],ymm0[7,4]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm3, %xmm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
 ; AVX1-ONLY-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -5311,7 +5311,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps (%r10), %xmm1
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %xmm6
@@ -5323,11 +5323,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm9[1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm2[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
@@ -5335,12 +5335,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm4[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
@@ -5349,7 +5349,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
@@ -5361,7 +5361,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r10), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rcx), %xmm6
@@ -5373,11 +5373,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm1[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm0[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
@@ -5385,12 +5385,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
@@ -5399,7 +5399,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -5411,7 +5411,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r10), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %xmm6
@@ -5423,11 +5423,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm1[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm0[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
@@ -5435,12 +5435,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
@@ -5449,7 +5449,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -5461,7 +5461,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r10), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm6
@@ -5473,11 +5473,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm10[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm1[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm1[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm0[1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
@@ -5485,12 +5485,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
@@ -5499,7 +5499,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -5511,7 +5511,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r10), %xmm4
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %xmm6
@@ -5523,11 +5523,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm2[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm2[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm1[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm1[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm0[1],xmm10[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1,2,3,4,5],ymm5[6,7]
@@ -5535,12 +5535,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
@@ -5549,7 +5549,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -5561,7 +5561,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %xmm9
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r10), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3,4,5],ymm6[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm3
@@ -5573,11 +5573,11 @@
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm15[0],xmm6[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm15[0,1,2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm4[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm7[1],xmm15[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0,1,2,3,4,5],ymm5[6,7]
@@ -5585,12 +5585,12 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm9[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm8[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm3[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm0[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
@@ -5599,7 +5599,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -5611,7 +5611,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm7
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r10), %xmm8
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm9[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm15
@@ -5622,23 +5622,23 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm14 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm14[0],xmm2[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm14[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm14[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm4[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm9[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm15[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm4[0,1,2,3],ymm1[4,5,6,7]
@@ -5646,7 +5646,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm2[4,5,6,7]
@@ -5657,7 +5657,7 @@
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rax), %xmm4
 ; AVX1-ONLY-NEXT:    vmovaps 224(%r10), %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm8 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm8[0,1,0,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm8[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm14[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rcx), %xmm14
@@ -5668,23 +5668,23 @@
 ; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm13 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm13 = xmm13[0],xmm2[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm0[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm1[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm13[0,1],xmm2[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0],xmm5[1],xmm13[2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm13[0,1,2,3,4,5],ymm8[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm4[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm7[0,1,2],xmm3[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm5
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[2,2,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[2,2,2,2]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm0[0,1],xmm5[2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
@@ -5692,7 +5692,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
 ; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6256,13 +6256,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm4[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm4[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm3[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6270,12 +6270,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm3[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6284,7 +6284,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm4[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6311,13 +6311,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm8[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6325,12 +6325,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6339,7 +6339,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6366,13 +6366,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm8[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm8[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6380,12 +6380,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6394,7 +6394,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6421,13 +6421,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6435,12 +6435,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6449,7 +6449,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6476,13 +6476,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6490,12 +6490,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6504,7 +6504,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6531,13 +6531,13 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6545,12 +6545,12 @@
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm0[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
@@ -6559,7 +6559,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6585,25 +6585,25 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm5[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0],xmm4[1],xmm8[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm15[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm15 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm7[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm7[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm0[1],xmm14[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3],ymm14[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm5[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3],ymm5[4,5,6,7]
@@ -6611,7 +6611,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm1[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -6636,25 +6636,25 @@
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm13
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm5[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0],xmm6[1],xmm5[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm12[2,3]
 ; AVX2-ONLY-NEXT:    vunpcklps {{.*#+}} xmm12 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm0[1,1,1,1]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm0[1,1,1,1]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0],xmm15[1],xmm11[2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm11[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm6[2],xmm14[3],xmm6[3]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm2[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm2[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm11 = xmm1[0,1,2],xmm11[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm12 = xmm4[2,2,2,2]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm3[0,1,2],xmm12[3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm6[0,1],xmm12[2,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
@@ -6662,7 +6662,7 @@
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm6[1],xmm3[1]
 ; AVX2-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
@@ -139,7 +139,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd (%rdx), %ymm4
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
@@ -270,7 +270,7 @@
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm3
 ; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm4
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm5
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm4[2,3],ymm6[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
@@ -278,7 +278,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3],ymm8[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd (%rdx), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm5[2,3],ymm8[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
@@ -547,11 +547,11 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 64(%rdx), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm12
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm13
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm11[2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm10[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm6[4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm6[1],ymm10[1],ymm6[3],ymm10[3]
@@ -563,7 +563,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm11[2,3],ymm6[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rdx), %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm11[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm5[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm5[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm3[4,5],ymm11[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3],ymm11[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
@@ -575,7 +575,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm12[2,3],ymm3[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rdx), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm0[4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm1
@@ -1087,7 +1087,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd (%rdx), %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm2[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm2[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1102,7 +1102,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rdx), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm6[2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
@@ -1117,7 +1117,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 64(%rdx), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3],ymm0[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1131,7 +1131,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rdx), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm14[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1],ymm15[2,3],ymm0[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
@@ -1143,7 +1143,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 128(%rdx), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm10[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm10
@@ -1158,7 +1158,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rdx), %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm15[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm1
@@ -1173,7 +1173,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 192(%rdx), %ymm4
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm4
@@ -1188,7 +1188,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rdx), %ymm7
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm7
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm9 = ymm7[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm9 = ymm7[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm4[4,5],ymm9[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm7[1],ymm4[3],ymm7[3]
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm7
@@ -2224,7 +2224,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd (%rdx), %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm14 = ymm11[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm14 = ymm11[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm8[4,5],ymm14[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm12[2,3],ymm14[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2239,7 +2239,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rdx), %ymm11
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm9[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm5[4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3],ymm8[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2254,7 +2254,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 64(%rdx), %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm6[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm6[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm3[4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2269,7 +2269,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rdx), %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm2[4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm7[2,3],ymm3[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2284,7 +2284,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 128(%rdx), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm1
@@ -2302,7 +2302,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm1
@@ -2320,7 +2320,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
@@ -2338,7 +2338,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm1
@@ -2356,7 +2356,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %ymm1
@@ -2374,7 +2374,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 288(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %ymm1
@@ -2392,7 +2392,7 @@
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 320(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %ymm1
@@ -2409,7 +2409,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 352(%rdx), %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 352(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %ymm1
@@ -2424,7 +2424,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 384(%rdx), %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 384(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %ymm1
@@ -2439,7 +2439,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 416(%rdx), %ymm2
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 416(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm15 = ymm1[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm15 = ymm1[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5],ymm15[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %ymm1
@@ -2454,7 +2454,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 448(%rdx), %ymm14
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm14[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rsi), %ymm14
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm13 = ymm14[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm13 = ymm14[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm1[4,5],ymm13[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm14[1],ymm1[3],ymm14[3]
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %ymm14
@@ -2469,7 +2469,7 @@
 ; AVX2-ONLY-NEXT:    vbroadcastsd 480(%rdx), %ymm12
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 480(%rsi), %ymm12
-; AVX2-ONLY-NEXT:    vpermilps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[2,3,0,1,6,7,4,5]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm14[4,5],ymm11[6,7]
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm14[1],ymm12[1],ymm14[3],ymm12[3]
 ; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %ymm14
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
@@ -803,7 +803,7 @@
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],mem[0],ymm8[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovapd 96(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
@@ -1737,7 +1737,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
@@ -1749,13 +1749,13 @@
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],mem[0],ymm11[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovapd 160(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],mem[0],ymm4[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm0
@@ -3719,7 +3719,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
@@ -3733,14 +3733,14 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],mem[0],ymm4[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm0
@@ -3763,7 +3763,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
@@ -3772,7 +3772,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %ymm0
@@ -3788,7 +3788,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %ymm0
@@ -3810,7 +3810,7 @@
 ; AVX1-ONLY-NEXT:    vmovapd %ymm1, %ymm15
 ; AVX1-ONLY-NEXT:    vmovapd 480(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm9
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
@@ -570,12 +570,12 @@
 ; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm10
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm10[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm10[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm13
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm13[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2,3,4,5,6,7]
@@ -2030,7 +2030,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
@@ -2041,7 +2041,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
@@ -2079,7 +2079,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm8
@@ -2096,7 +2096,7 @@
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm15
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm14 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm15[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -2113,7 +2113,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm13 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm13
@@ -4756,7 +4756,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm7[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm11
 ; AVX1-ONLY-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm11 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm7[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm7
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
@@ -4778,7 +4778,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm6[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
@@ -4824,7 +4824,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 48(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4841,7 +4841,7 @@
 ; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
@@ -4868,7 +4868,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4916,7 +4916,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 112(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4933,7 +4933,7 @@
 ; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm3
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm14[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
@@ -4960,7 +4960,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 144(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5008,7 +5008,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 176(%r8), %xmm0
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5025,7 +5025,7 @@
 ; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm4[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm8[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
@@ -10400,7 +10400,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
@@ -10412,7 +10412,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm2[1],ymm7[3],ymm2[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
@@ -10456,7 +10456,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm1
@@ -10476,7 +10476,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10496,7 +10496,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
@@ -10538,7 +10538,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm1
@@ -10558,7 +10558,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10578,7 +10578,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 144(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
@@ -10620,7 +10620,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 176(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm1
@@ -10640,7 +10640,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10660,7 +10660,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm0
@@ -10702,7 +10702,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 256(%r9), %xmm1
@@ -10722,7 +10722,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10742,7 +10742,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 272(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %xmm0
@@ -10784,7 +10784,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 304(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%r9), %xmm1
@@ -10804,7 +10804,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10824,7 +10824,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 336(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %xmm0
@@ -10863,7 +10863,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm0[1],ymm14[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 368(%rcx), %xmm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 384(%r9), %xmm13
@@ -10881,7 +10881,7 @@
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
@@ -10898,7 +10898,7 @@
 ; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm0[1],ymm8[3],ymm0[3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps 400(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm7[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rsi), %xmm3
@@ -10925,7 +10925,7 @@
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 448(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rcx), %ymm1
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovapd 448(%rdx), %ymm5
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-6.ll
@@ -1181,8 +1181,8 @@
 ; AVX1-ONLY-NEXT:    vpor %xmm9, %xmm8, %xmm8
 ; AVX1-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm9[0,0,1,1]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[1,1,2,2]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm8 = xmm9[0,0,1,1]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[1,1,2,2]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm10 = xmm13[1,0,2,2,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm10 = xmm10[0,1,0,1]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll
@@ -1598,7 +1598,7 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[0,0,2,1,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm0[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [281474976710655,281474976710655,281474976710655,281474976710655]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm0, %ymm2
 ; AVX1-ONLY-NEXT:    vmovdqa (%r9), %xmm7
@@ -1607,7 +1607,7 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm12[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm12[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm3, %ymm5
 ; AVX1-ONLY-NEXT:    vmovdqa (%rsi), %xmm2
@@ -1639,9 +1639,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vmovdqa 16(%r10), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm5, %ymm1
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm15[0,1,2,3,4,4,6,5]
@@ -1653,7 +1653,7 @@
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
 ; AVX1-ONLY-NEXT:    vmovdqa 16(%rax), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm12 = ymm12[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm12, %ymm2, %ymm12
 ; AVX1-ONLY-NEXT:    vandps %ymm2, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm12, %ymm3
@@ -1674,9 +1674,9 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm4
 ; AVX1-ONLY-NEXT:    vmovdqa 16(%r8), %xmm7
 ; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm4, %ymm3
 ; AVX1-ONLY-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm11[8],xmm10[8],xmm11[9],xmm10[9],xmm11[10],xmm10[10],xmm11[11],xmm10[11],xmm11[12],xmm10[12],xmm11[13],xmm10[13],xmm11[14],xmm10[14],xmm11[15],xmm10[15]
@@ -1684,7 +1684,7 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm1[0,1,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm2, %ymm4
 ; AVX1-ONLY-NEXT:    vpunpckhbw {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm1 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # xmm1 = xmm6[8],mem[8],xmm6[9],mem[9],xmm6[10],mem[10],xmm6[11],mem[11],xmm6[12],mem[12],xmm6[13],mem[13],xmm6[14],mem[14],xmm6[15],mem[15]
@@ -1707,9 +1707,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm12[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm12[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm4, %ymm4
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm4, %ymm5
 ; AVX1-ONLY-NEXT:    vmovdqa 16(%rcx), %xmm7
@@ -1726,7 +1726,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm15 = xmm14[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm15 = xmm15[0],zero,zero,zero,xmm15[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm13, %ymm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vandps %ymm2, %ymm13, %ymm13
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm13, %ymm1
@@ -1741,9 +1741,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm15[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm15[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm8, %ymm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
@@ -1756,7 +1756,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm4[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,xmm7[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vandps %ymm2, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm6, %ymm3
@@ -1767,9 +1767,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm15[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm15[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vorps %ymm5, %ymm6, %ymm5
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm6 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero
@@ -1791,9 +1791,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm12[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm12[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm4 = ymm4[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm5, %ymm5
 ; AVX1-ONLY-NEXT:    vorps %ymm4, %ymm5, %ymm4
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm5 = xmm14[0],zero,zero,zero,xmm14[1],zero,zero,zero
@@ -1817,9 +1817,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm7[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm0, %ymm6, %ymm0
 ; AVX1-ONLY-NEXT:    vorps %ymm5, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -4157,7 +4157,7 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm5 = [281474976710655,281474976710655,281474976710655,281474976710655]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm5, %ymm0
 ; AVX1-ONLY-NEXT:    vmovdqa (%r9), %xmm3
@@ -4168,7 +4168,7 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm3, %ymm0
 ; AVX1-ONLY-NEXT:    vmovdqa (%rcx), %xmm4
@@ -4189,7 +4189,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm8 = xmm6[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm8 = xmm8[0],zero,zero,zero,xmm8[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm4[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vbroadcastsd {{.*#+}} ymm4 = [18446744069414649855,18446744069414649855,18446744069414649855,18446744069414649855]
 ; AVX1-ONLY-NEXT:    vandnps %ymm8, %ymm4, %ymm8
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm7, %ymm7
@@ -4204,9 +4204,9 @@
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
 ; AVX1-ONLY-NEXT:    vmovdqa 48(%r10), %xmm1
 ; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm2, %ymm0
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm2 = xmm6[0],zero,zero,zero,xmm6[1],zero,zero,zero
@@ -4235,9 +4235,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm12[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm8 = xmm12[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm6, %ymm5, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vorps %ymm6, %ymm7, %ymm13
 ; AVX1-ONLY-NEXT:    vmovdqa 48(%rsi), %xmm8
@@ -4266,9 +4266,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm12[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm11 = xmm12[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm15[0,1,2,3,4,4,6,5]
@@ -4279,7 +4279,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm10 = xmm10[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm10 = xmm10[0],zero,zero,zero,xmm10[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm4, %ymm2
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm10, %ymm10
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm10, %ymm2
@@ -4295,9 +4295,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm0[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm10 = xmm0[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
@@ -4322,9 +4322,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm5, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,4,6,5]
@@ -4335,7 +4335,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm4, %ymm2
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm1, %ymm1
@@ -4355,9 +4355,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm12[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm12[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm3, %ymm13
 ; AVX1-ONLY-NEXT:    vmovdqa 32(%rsi), %xmm7
@@ -4386,9 +4386,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm12[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm11 = xmm12[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm15[0,1,2,3,4,4,6,5]
@@ -4399,7 +4399,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm10 = xmm10[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm10 = xmm10[0],zero,zero,zero,xmm10[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm11, %ymm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm4, %ymm2
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm10, %ymm10
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm10, %ymm2
@@ -4415,9 +4415,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm8 = xmm2[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm10 = xmm2[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm8, %ymm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm8 = ymm8[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm8, %ymm8
 ; AVX1-ONLY-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm9[0],xmm7[0],xmm9[1],xmm7[1],xmm9[2],xmm7[2],xmm9[3],xmm7[3],xmm9[4],xmm7[4],xmm9[5],xmm7[5],xmm9[6],xmm7[6],xmm9[7],xmm7[7]
@@ -4442,9 +4442,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm2[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm6, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm5, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,4,6,5]
@@ -4455,7 +4455,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm4, %ymm0
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm1, %ymm0
@@ -4475,9 +4475,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm1[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm8 = xmm1[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm5, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm6, %ymm6
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm6, %ymm13
 ; AVX1-ONLY-NEXT:    vmovdqa 16(%rsi), %xmm11
@@ -4505,9 +4505,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm2, %ymm5, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vorps %ymm2, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,6,5]
@@ -4518,7 +4518,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm10[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm4, %ymm0
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm2, %ymm0
@@ -4533,9 +4533,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm3[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm3[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vorps %ymm1, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3],xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
@@ -4559,9 +4559,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm3[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm3, %ymm0
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,4,4,6,5]
@@ -4572,7 +4572,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm6, %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm4, %ymm3
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm1, %ymm1
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm1, %ymm1
@@ -4589,9 +4589,9 @@
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm6[0,1,1,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vpshuflw {{.*#+}} xmm8 = xmm6[2,1,3,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandnps %ymm3, %ymm5, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm7 = ymm7[0,0,2,1,4,4,6,5]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,0,2,1,4,4,6,5]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vorps %ymm3, %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -4619,9 +4619,9 @@
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm6[0,1,2,3,4,5,5,7]
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm8, %ymm6
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm6 = ymm6[0,2,2,3,4,6,6,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,2,2,3,4,6,6,7]
 ; AVX1-ONLY-NEXT:    vandps %ymm5, %ymm6, %ymm5
 ; AVX1-ONLY-NEXT:    vorps %ymm0, %ymm5, %ymm0
 ; AVX1-ONLY-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm9[0,1,2,3,4,4,6,5]
@@ -4632,7 +4632,7 @@
 ; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,3,3,3]
 ; AVX1-ONLY-NEXT:    vpmovzxwq {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm3
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} ymm5 = ymm5[2,1,3,3,6,5,7,7]
+; AVX1-ONLY-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,1,3,3,6,5,7,7]
 ; AVX1-ONLY-NEXT:    vandnps %ymm5, %ymm4, %ymm5
 ; AVX1-ONLY-NEXT:    vandps %ymm4, %ymm3, %ymm3
 ; AVX1-ONLY-NEXT:    vorps %ymm5, %ymm3, %ymm3
diff --git a/llvm/test/CodeGen/X86/vector-reduce-add-mask.ll b/llvm/test/CodeGen/X86/vector-reduce-add-mask.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-add-mask.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-add-mask.ll
@@ -446,7 +446,7 @@
 ; AVX1-SLOW-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-SLOW-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-reduce-and-cmp.ll b/llvm/test/CodeGen/X86/vector-reduce-and-cmp.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-and-cmp.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-and-cmp.ll
@@ -102,7 +102,7 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    testq %rax, %rax
@@ -165,7 +165,7 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    testq %rax, %rax
@@ -340,9 +340,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    testl %eax, %eax
@@ -411,9 +411,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    testl %eax, %eax
@@ -638,9 +638,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -718,9 +718,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -1002,9 +1002,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -1091,9 +1091,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-reduce-and.ll b/llvm/test/CodeGen/X86/vector-reduce-and.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-and.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-and.ll
@@ -86,7 +86,7 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -140,7 +140,7 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -287,9 +287,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -349,9 +349,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -554,9 +554,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -629,9 +629,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -891,9 +891,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
@@ -975,9 +975,9 @@
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vandps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-reduce-fadd.ll b/llvm/test/CodeGen/X86/vector-reduce-fadd.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-fadd.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-fadd.ll
@@ -76,7 +76,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
@@ -87,7 +87,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fadd.f32.v4f32(float %a0, <4 x float> %a1)
@@ -144,7 +144,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -152,7 +152,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -164,7 +164,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -172,7 +172,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -264,7 +264,7 @@
 ; AVX-NEXT:    vaddss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm3, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -272,14 +272,14 @@
 ; AVX-NEXT:    vaddss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm3, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -287,7 +287,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -299,7 +299,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
@@ -307,7 +307,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm1, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
@@ -315,7 +315,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm1, %xmm1
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -323,7 +323,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -405,7 +405,7 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX1-SLOW-NEXT:    retq
 ;
@@ -414,7 +414,7 @@
 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm1
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX1-FAST-NEXT:    retq
 ;
@@ -424,7 +424,7 @@
 ; AVX2-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    retq
 ;
@@ -434,7 +434,7 @@
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fadd.f32.v4f32(float -0.0, <4 x float> %a0)
@@ -488,7 +488,7 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -496,7 +496,7 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX1-SLOW-NEXT:    vzeroupper
 ; AVX1-SLOW-NEXT:    retq
@@ -506,7 +506,7 @@
 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm1
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX1-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-FAST-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -514,7 +514,7 @@
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX1-FAST-NEXT:    vzeroupper
 ; AVX1-FAST-NEXT:    retq
@@ -525,7 +525,7 @@
 ; AVX2-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -533,7 +533,7 @@
 ; AVX2-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -544,7 +544,7 @@
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -552,7 +552,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -641,7 +641,7 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm2
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-SLOW-NEXT:    vaddss %xmm0, %xmm2, %xmm2
@@ -649,14 +649,14 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX1-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -664,7 +664,7 @@
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX1-SLOW-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vzeroupper
 ; AVX1-SLOW-NEXT:    retq
@@ -674,7 +674,7 @@
 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm2
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX1-FAST-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-FAST-NEXT:    vaddss %xmm0, %xmm2, %xmm2
@@ -682,14 +682,14 @@
 ; AVX1-FAST-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX1-FAST-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-FAST-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -697,7 +697,7 @@
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX1-FAST-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-FAST-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vzeroupper
 ; AVX1-FAST-NEXT:    retq
@@ -708,7 +708,7 @@
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm2
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX2-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vaddss %xmm0, %xmm2, %xmm2
@@ -716,14 +716,14 @@
 ; AVX2-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX2-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX2-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX2-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -731,7 +731,7 @@
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX2-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX2-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -742,7 +742,7 @@
 ; AVX512-NEXT:    vaddss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
@@ -750,7 +750,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
@@ -758,7 +758,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -766,7 +766,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -836,7 +836,7 @@
 ; AVX-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    retq
 ;
@@ -846,7 +846,7 @@
 ; AVX512-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fadd.f32.v4f32(float undef, <4 x float> %a0)
@@ -900,7 +900,7 @@
 ; AVX-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -908,7 +908,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -919,7 +919,7 @@
 ; AVX512-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -927,7 +927,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -1016,7 +1016,7 @@
 ; AVX-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vaddss %xmm0, %xmm2, %xmm2
@@ -1024,14 +1024,14 @@
 ; AVX-NEXT:    vaddss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vaddss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm0, %xmm2, %xmm0
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
@@ -1039,7 +1039,7 @@
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -1050,7 +1050,7 @@
 ; AVX512-NEXT:    vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
@@ -1058,7 +1058,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
@@ -1066,7 +1066,7 @@
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm1
@@ -1074,7 +1074,7 @@
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vaddss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-reduce-fmax.ll b/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
@@ -197,7 +197,7 @@
 ;
 ; AVX-LABEL: test_v4f32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX-NEXT:    vmaxss %xmm0, %xmm3, %xmm4
@@ -213,7 +213,7 @@
 ;
 ; AVX512-LABEL: test_v4f32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX512-NEXT:    vmaxss %xmm0, %xmm3, %xmm4
@@ -302,10 +302,10 @@
 ; AVX-LABEL: test_v8f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX-NEXT:    vmaxss %xmm0, %xmm7, %xmm8
@@ -335,10 +335,10 @@
 ; AVX512BW-LABEL: test_v8f32:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vextractf128 $1, %ymm0, %xmm3
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
 ; AVX512BW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX512BW-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX512BW-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX512BW-NEXT:    vmaxss %xmm0, %xmm7, %xmm8
@@ -368,10 +368,10 @@
 ; AVX512VL-LABEL: test_v8f32:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX512VL-NEXT:    vmaxss %xmm0, %xmm7, %xmm8
@@ -506,7 +506,7 @@
 ; AVX-NEXT:    vmaxss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmaxss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
@@ -522,7 +522,7 @@
 ; AVX-NEXT:    vmaxss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmaxss %xmm1, %xmm0, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vzeroupper
@@ -539,7 +539,7 @@
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
@@ -555,7 +555,7 @@
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
 ; AVX512BW-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm1, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm3, %xmm3 {%k1}
@@ -571,7 +571,7 @@
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm1, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm3, %xmm3 {%k1}
 ; AVX512BW-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm3, %xmm1, %xmm2
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm2, %xmm2 {%k1}
@@ -586,7 +586,7 @@
 ; AVX512BW-NEXT:    vcmpunordss %xmm2, %xmm2, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm2, %xmm1, %xmm2
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm2, %xmm2 {%k1}
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm2, %xmm2, %k1
 ; AVX512BW-NEXT:    vmaxss %xmm2, %xmm1, %xmm0
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm0, %xmm0 {%k1}
@@ -596,18 +596,18 @@
 ; AVX512VL-LABEL: test_v16f32:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vextractf32x4 $3, %zmm0, %xmm3
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; AVX512VL-NEXT:    vextractf32x4 $2, %zmm0, %xmm6
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm7 = xmm6[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm8 = xmm6[1,1,3,3]
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm0, %xmm9
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm12 = xmm9[1,1,3,3]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm13 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm14 = xmm0[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm15 = xmm0[1,1,3,3]
 ; AVX512VL-NEXT:    vmaxss %xmm0, %xmm15, %xmm16
diff --git a/llvm/test/CodeGen/X86/vector-reduce-fmin.ll b/llvm/test/CodeGen/X86/vector-reduce-fmin.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-fmin.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-fmin.ll
@@ -130,7 +130,7 @@
 ;
 ; AVX-LABEL: test_v4f32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX-NEXT:    vminss %xmm0, %xmm3, %xmm4
@@ -146,7 +146,7 @@
 ;
 ; AVX512-LABEL: test_v4f32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; AVX512-NEXT:    vminss %xmm0, %xmm3, %xmm4
@@ -235,10 +235,10 @@
 ; AVX-LABEL: test_v8f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX-NEXT:    vminss %xmm0, %xmm7, %xmm8
@@ -268,10 +268,10 @@
 ; AVX512BW-LABEL: test_v8f32:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vextractf128 $1, %ymm0, %xmm3
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
 ; AVX512BW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX512BW-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX512BW-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX512BW-NEXT:    vminss %xmm0, %xmm7, %xmm8
@@ -301,10 +301,10 @@
 ; AVX512VL-LABEL: test_v8f32:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
 ; AVX512VL-NEXT:    vminss %xmm0, %xmm7, %xmm8
@@ -439,7 +439,7 @@
 ; AVX-NEXT:    vminss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vminss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
@@ -455,7 +455,7 @@
 ; AVX-NEXT:    vminss %xmm1, %xmm3, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm3, %xmm1, %xmm1
 ; AVX-NEXT:    vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vminss %xmm1, %xmm0, %xmm1
 ; AVX-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vzeroupper
@@ -472,7 +472,7 @@
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
@@ -488,7 +488,7 @@
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm2, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm2, %xmm3, %xmm3 {%k1}
 ; AVX512BW-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm1, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm3, %xmm3 {%k1}
@@ -504,7 +504,7 @@
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm1, %xmm3
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm3, %xmm3 {%k1}
 ; AVX512BW-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm3, %xmm3, %k1
 ; AVX512BW-NEXT:    vminss %xmm3, %xmm1, %xmm2
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm2, %xmm2 {%k1}
@@ -519,7 +519,7 @@
 ; AVX512BW-NEXT:    vcmpunordss %xmm2, %xmm2, %k1
 ; AVX512BW-NEXT:    vminss %xmm2, %xmm1, %xmm2
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm2, %xmm2 {%k1}
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vcmpunordss %xmm2, %xmm2, %k1
 ; AVX512BW-NEXT:    vminss %xmm2, %xmm1, %xmm0
 ; AVX512BW-NEXT:    vmovss %xmm1, %xmm0, %xmm0 {%k1}
@@ -529,18 +529,18 @@
 ; AVX512VL-LABEL: test_v16f32:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vextractf32x4 $3, %zmm0, %xmm3
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; AVX512VL-NEXT:    vextractf32x4 $2, %zmm0, %xmm6
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm7 = xmm6[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm8 = xmm6[1,1,3,3]
 ; AVX512VL-NEXT:    vextractf128 $1, %ymm0, %xmm9
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm10 = xmm9[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm12 = xmm9[1,1,3,3]
-; AVX512VL-NEXT:    vpermilps {{.*#+}} xmm13 = xmm0[3,3,3,3]
+; AVX512VL-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[3,3,3,3]
 ; AVX512VL-NEXT:    vpermilpd {{.*#+}} xmm14 = xmm0[1,0]
 ; AVX512VL-NEXT:    vmovshdup {{.*#+}} xmm15 = xmm0[1,1,3,3]
 ; AVX512VL-NEXT:    vminss %xmm0, %xmm15, %xmm16
diff --git a/llvm/test/CodeGen/X86/vector-reduce-fmul.ll b/llvm/test/CodeGen/X86/vector-reduce-fmul.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-fmul.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-fmul.ll
@@ -75,7 +75,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
 ;
@@ -86,7 +86,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fmul.f32.v4f32(float %a0, <4 x float> %a1)
@@ -143,7 +143,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -151,7 +151,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -163,7 +163,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -171,7 +171,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -263,7 +263,7 @@
 ; AVX-NEXT:    vmulss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -271,14 +271,14 @@
 ; AVX-NEXT:    vmulss %xmm3, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -286,7 +286,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -298,7 +298,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
@@ -306,7 +306,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm1, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
@@ -314,7 +314,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm1, %xmm1
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -322,7 +322,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -393,7 +393,7 @@
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    retq
 ;
@@ -403,7 +403,7 @@
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fmul.f32.v4f32(float 1.0, <4 x float> %a0)
@@ -457,7 +457,7 @@
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -465,7 +465,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -476,7 +476,7 @@
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -484,7 +484,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -573,7 +573,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmulss %xmm0, %xmm2, %xmm2
@@ -581,14 +581,14 @@
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm2, %xmm0
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -596,7 +596,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -607,7 +607,7 @@
 ; AVX512-NEXT:    vmulss %xmm1, %xmm0, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
@@ -615,7 +615,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
@@ -623,7 +623,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -631,7 +631,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -701,7 +701,7 @@
 ; AVX-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    retq
 ;
@@ -711,7 +711,7 @@
 ; AVX512-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    retq
   %1 = call float @llvm.vector.reduce.fmul.f32.v4f32(float undef, <4 x float> %a0)
@@ -765,7 +765,7 @@
 ; AVX-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -773,7 +773,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -784,7 +784,7 @@
 ; AVX512-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -792,7 +792,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
@@ -881,7 +881,7 @@
 ; AVX-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX-NEXT:    vmulss %xmm0, %xmm2, %xmm2
@@ -889,14 +889,14 @@
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
 ; AVX-NEXT:    vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm0, %xmm2, %xmm0
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
@@ -904,7 +904,7 @@
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
 ; AVX-NEXT:    vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-NEXT:    vmulss %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
@@ -915,7 +915,7 @@
 ; AVX512-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
@@ -923,7 +923,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
@@ -931,7 +931,7 @@
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
 ; AVX512-NEXT:    vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vextractf32x4 $3, %zmm0, %xmm0
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm1
@@ -939,7 +939,7 @@
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
 ; AVX512-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
 ; AVX512-NEXT:    vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX512-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX512-NEXT:    vmulss %xmm0, %xmm1, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-reduce-or.ll b/llvm/test/CodeGen/X86/vector-reduce-or.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-or.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-or.ll
@@ -86,7 +86,7 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -140,7 +140,7 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -287,9 +287,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -349,9 +349,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -554,9 +554,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
@@ -629,9 +629,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
@@ -891,9 +891,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
@@ -975,9 +975,9 @@
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-reduce-smax.ll b/llvm/test/CodeGen/X86/vector-reduce-smax.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-smax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-smax.ll
@@ -170,7 +170,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -182,7 +182,7 @@
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
@@ -360,7 +360,7 @@
 ; AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -374,7 +374,7 @@
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
@@ -681,7 +681,7 @@
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm5, %xmm4, %xmm1
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -699,7 +699,7 @@
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
diff --git a/llvm/test/CodeGen/X86/vector-reduce-smin.ll b/llvm/test/CodeGen/X86/vector-reduce-smin.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-smin.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-smin.ll
@@ -170,7 +170,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -182,7 +182,7 @@
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
@@ -360,7 +360,7 @@
 ; AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm1
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -374,7 +374,7 @@
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
@@ -681,7 +681,7 @@
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm4, %xmm5, %xmm1
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
@@ -699,7 +699,7 @@
 ; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
diff --git a/llvm/test/CodeGen/X86/vector-reduce-umax.ll b/llvm/test/CodeGen/X86/vector-reduce-umax.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-umax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-umax.ll
@@ -185,7 +185,7 @@
 ; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm4
 ; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX1-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -202,7 +202,7 @@
 ; AVX2-NEXT:    vpxor %xmm2, %xmm0, %xmm4
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -400,7 +400,7 @@
 ; AVX1-NEXT:    vxorpd %xmm2, %xmm3, %xmm4
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX1-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -421,7 +421,7 @@
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm4
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -769,7 +769,7 @@
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm0, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm0, %xmm2
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm1, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -798,7 +798,7 @@
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm0, %xmm3
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm0, %xmm2
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm1, %xmm3
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-reduce-umin.ll b/llvm/test/CodeGen/X86/vector-reduce-umin.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-umin.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-umin.ll
@@ -186,7 +186,7 @@
 ; AVX1-NEXT:    vpxor %xmm1, %xmm3, %xmm4
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm4, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm0, %xmm3
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
@@ -203,7 +203,7 @@
 ; AVX2-NEXT:    vpxor %xmm2, %xmm1, %xmm4
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -403,7 +403,7 @@
 ; AVX1-NEXT:    vxorpd %xmm3, %xmm2, %xmm4
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
 ; AVX1-NEXT:    vblendvpd %xmm1, %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm3, %xmm0, %xmm2
 ; AVX1-NEXT:    vxorpd %xmm3, %xmm1, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -424,7 +424,7 @@
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm4
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm4, %xmm3
 ; AVX2-NEXT:    vblendvpd %xmm3, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm0, %xmm3
 ; AVX2-NEXT:    vxorpd %xmm2, %xmm1, %xmm2
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
@@ -773,7 +773,7 @@
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm0, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm0, %xmm2
 ; AVX1-NEXT:    vxorpd %xmm4, %xmm1, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
@@ -802,7 +802,7 @@
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm1, %xmm3
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm0, %xmm2
 ; AVX2-NEXT:    vxorpd %xmm4, %xmm1, %xmm3
 ; AVX2-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-reduce-xor.ll b/llvm/test/CodeGen/X86/vector-reduce-xor.ll
--- a/llvm/test/CodeGen/X86/vector-reduce-xor.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-xor.ll
@@ -86,7 +86,7 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -140,7 +140,7 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
@@ -287,9 +287,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -349,9 +349,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vzeroupper
@@ -554,9 +554,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
@@ -629,9 +629,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
@@ -891,9 +891,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
@@ -975,9 +975,9 @@
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-sext.ll b/llvm/test/CodeGen/X86/vector-sext.ll
--- a/llvm/test/CodeGen/X86/vector-sext.ll
+++ b/llvm/test/CodeGen/X86/vector-sext.ll
@@ -2681,7 +2681,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm2 = [9241421688590303745,9241421688590303745]
diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
@@ -1526,8 +1526,8 @@
 ; XOPAVX1-LABEL: vector_variable_shift_right:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vpslld $31, %xmm0, %xmm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; XOPAVX1-NEXT:    vblendvps %xmm0, %xmm1, %xmm2, %xmm0
 ; XOPAVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
@@ -1712,7 +1712,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shift32_v4i64:
@@ -1724,7 +1724,7 @@
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shift32_v4i64:
@@ -1746,7 +1746,7 @@
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; X86-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; X86-AVX1-NEXT:    retl
 ;
 ; X86-AVX2-LABEL: shift32_v4i64:
diff --git a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
--- a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
@@ -1600,7 +1600,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[0,2],ymm1[4,6],ymm0[4,6]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shift32_v4i64:
@@ -1612,7 +1612,7 @@
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[0,2],ymm1[4,6],ymm0[4,6]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shift32_v4i64:
@@ -1634,7 +1634,7 @@
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; X86-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[0,2],ymm1[4,6],ymm0[4,6]
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; X86-AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; X86-AVX1-NEXT:    retl
 ;
 ; X86-AVX2-LABEL: shift32_v4i64:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v2.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v2.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v2.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v2.ll
@@ -15,7 +15,7 @@
 ;
 ; AVX1-LABEL: shuffle_v2i64_00:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v2i64_00:
@@ -38,7 +38,7 @@
 ;
 ; AVX-LABEL: shuffle_v2i64_10:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 1, i32 0>
   ret <2 x i64> %shuffle
@@ -51,7 +51,7 @@
 ;
 ; AVX-LABEL: shuffle_v2i64_11:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 1, i32 1>
   ret <2 x i64> %shuffle
@@ -64,7 +64,7 @@
 ;
 ; AVX1-LABEL: shuffle_v2i64_22:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v2i64_22:
@@ -87,7 +87,7 @@
 ;
 ; AVX-LABEL: shuffle_v2i64_32:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[2,3,0,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 3, i32 2>
   ret <2 x i64> %shuffle
@@ -100,7 +100,7 @@
 ;
 ; AVX-LABEL: shuffle_v2i64_33:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 3, i32 3>
   ret <2 x i64> %shuffle
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll
@@ -18,7 +18,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_0001:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
   ret <4 x i32> %shuffle
@@ -31,7 +31,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_0020:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 0>
   ret <4 x i32> %shuffle
@@ -44,7 +44,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_0112:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,2]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 1, i32 1, i32 2>
   ret <4 x i32> %shuffle
@@ -57,7 +57,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_0300:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 3, i32 0, i32 0>
   ret <4 x i32> %shuffle
@@ -70,7 +70,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_1000:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
   ret <4 x i32> %shuffle
@@ -83,7 +83,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_2200:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 2, i32 2, i32 0, i32 0>
   ret <4 x i32> %shuffle
@@ -96,7 +96,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_3330:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
   ret <4 x i32> %shuffle
@@ -109,7 +109,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_3210:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x i32> %shuffle
@@ -123,7 +123,7 @@
 ;
 ; AVX-LABEL: shuffle_v4i32_2121:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,2,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,2,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 2, i32 1, i32 2, i32 1>
   ret <4 x i32> %shuffle
@@ -137,7 +137,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_0001:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
   ret <4 x float> %shuffle
@@ -150,7 +150,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_0020:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,2,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,2,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 0>
   ret <4 x float> %shuffle
@@ -163,7 +163,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_0300:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 3, i32 0, i32 0>
   ret <4 x float> %shuffle
@@ -176,7 +176,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_1000:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
   ret <4 x float> %shuffle
@@ -189,7 +189,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_2200:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,0,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 2, i32 2, i32 0, i32 0>
   ret <4 x float> %shuffle
@@ -202,7 +202,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_3330:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
   ret <4 x float> %shuffle
@@ -215,7 +215,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_3210:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   ret <4 x float> %shuffle
@@ -228,7 +228,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_0011:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
   ret <4 x float> %shuffle
@@ -241,7 +241,7 @@
 ;
 ; AVX-LABEL: shuffle_v4f32_2233:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
   ret <4 x float> %shuffle
@@ -358,7 +358,7 @@
 ;
 ; AVX1-LABEL: shuffle_v4i32_0124:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
 ; AVX1-NEXT:    retq
 ;
@@ -404,15 +404,15 @@
 ;
 ; AVX1-LABEL: shuffle_v4i32_0142:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,2]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i32_0142:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastss %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,2]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,2,2]
 ; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
 ; AVX2-NEXT:    retq
 ;
@@ -455,15 +455,15 @@
 ;
 ; AVX1-LABEL: shuffle_v4i32_0412:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,2]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i32_0412:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastss %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,2]
 ; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-NEXT:    retq
 ;
@@ -505,7 +505,7 @@
 ;
 ; AVX1OR2-LABEL: shuffle_v4i32_4012:
 ; AVX1OR2:       # %bb.0:
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,2]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,2]
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
 ; AVX1OR2-NEXT:    retq
 ;
@@ -540,7 +540,7 @@
 ; AVX1OR2-LABEL: shuffle_v4i32_0451:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i32_0451:
@@ -575,7 +575,7 @@
 ; AVX1OR2-LABEL: shuffle_v4i32_4015:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i32_4015:
@@ -1203,14 +1203,14 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v4i32_z4zz:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v4i32_z4zz:
@@ -1259,14 +1259,14 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v4i32_zz4z:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,0,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v4i32_zz4z:
@@ -1327,14 +1327,14 @@
 ;
 ; AVX1-LABEL: shuffle_v4i32_z6zz:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v4i32_z6zz:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX2-SLOW-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
 ; AVX2-SLOW-NEXT:    retq
@@ -2409,7 +2409,7 @@
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vmovaps %xmm1, (%rsi)
 ; AVX1-NEXT:    retq
 ;
@@ -2517,7 +2517,7 @@
 ; AVX1OR2-LABEL: shuffle_mem_v4f32_0624:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],mem[0,2]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,3,1]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0,3,1]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_mem_v4f32_0624:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll
@@ -20,7 +20,7 @@
 ;
 ; AVX-LABEL: shuffle_v8i16_01012323:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3>
   ret <8 x i16> %shuffle
@@ -33,7 +33,7 @@
 ;
 ; AVX-LABEL: shuffle_v8i16_67452301:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 2, i32 3, i32 0, i32 1>
   ret <8 x i16> %shuffle
@@ -240,7 +240,7 @@
 ;
 ; AVX-LABEL: shuffle_v8i16_23016745:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,3,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,3,2]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5>
   ret <8 x i16> %shuffle
@@ -1382,7 +1382,7 @@
 ;
 ; AVX-LABEL: shuffle_v8i16_XXXdXXXX:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[2,2,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[2,2,3,3]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 13, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -691,7 +691,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
@@ -722,7 +722,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
@@ -741,7 +741,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
@@ -772,7 +772,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
@@ -1542,7 +1542,7 @@
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
@@ -1564,7 +1564,7 @@
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
@@ -1728,7 +1728,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,1,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_08:
@@ -1759,7 +1759,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,1,0,4,5,6,7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_08:
@@ -1778,7 +1778,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,2,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_08:
@@ -1809,7 +1809,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,2,4,5,6,7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_08:
@@ -1828,7 +1828,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_08:
@@ -1859,7 +1859,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,0,4,5,6,7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_08:
@@ -2990,7 +2990,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v16i16_02_03_zz_zz_06_07_zz_zz_10_11_zz_zz_14_15_zz_zz:
@@ -3002,7 +3002,7 @@
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_02_03_zz_zz_06_07_zz_zz_10_11_zz_zz_14_15_zz_zz:
@@ -3363,7 +3363,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_09:
@@ -3384,7 +3384,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; XOPAVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
 ; XOPAVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_09:
@@ -3638,7 +3638,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
@@ -3747,7 +3747,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_13:
@@ -3768,7 +3768,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; XOPAVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; XOPAVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_13:
@@ -3789,7 +3789,7 @@
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[2,3,0,2,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
@@ -5788,7 +5788,7 @@
 define <16 x i16> @shuffle_v16i16_uu_uu_uu_21_uu_uu_uu_uu_uu_uu_uu_29_uu_uu_uu_uu(<16 x i16> %a, <16 x i16> %b) {
 ; ALL-LABEL: shuffle_v16i16_uu_uu_uu_21_uu_uu_uu_uu_uu_uu_uu_29_uu_uu_uu_uu:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm1[2,2,2,2,6,6,6,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,2,2,2,6,6,6,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 21, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 29, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
@@ -6979,7 +6979,7 @@
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25:
@@ -7001,7 +7001,7 @@
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25:
@@ -7171,7 +7171,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,7,5]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_04_06_07_uu_uu_06_07_05_12_14_15_uu_uu_14_15_13:
@@ -7202,7 +7202,7 @@
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; XOPAVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,7,5]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; XOPAVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; XOPAVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_04_06_07_uu_uu_06_07_05_12_14_15_uu_uu_14_15_13:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -2436,7 +2436,7 @@
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -794,7 +794,7 @@
 define <4 x i64> @shuffle_v4i64_0000(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_0000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -814,7 +814,7 @@
 define <4 x i64> @shuffle_v4i64_0001(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_0001:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -1244,13 +1244,13 @@
 ; AVX2-LABEL: shuffle_v4i64_1054:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_1054:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v4i64_1054:
@@ -1262,7 +1262,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v4i64_1054:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
   ret <4 x i64> %shuffle
@@ -1312,7 +1312,7 @@
 ; AVX2-LABEL: shuffle_v4i64_3276:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_3276:
@@ -1346,13 +1346,13 @@
 ; AVX2-LABEL: shuffle_v4i64_1076:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v4i64_1076:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v4i64_1076:
@@ -1507,7 +1507,7 @@
 define <4 x i64> @shuffle_v4i64_11uu(<4 x i64> %a, <4 x i64> %b) {
 ; ALL-LABEL: shuffle_v4i64_11uu:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 1, i32 undef, i32 undef>
   ret <4 x i64> %shuffle
@@ -1517,7 +1517,7 @@
 ; AVX1-LABEL: shuffle_v4i64_22uu:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_22uu:
@@ -1585,14 +1585,14 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i64_0044_v2i64:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; AVX512VL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX512VL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX512VL-NEXT:    retq
   %1 = shufflevector <2 x i64> %a, <2 x i64> undef, <2 x i32> <i32 0, i32 0>
   %2 = shufflevector <2 x i64> %b, <2 x i64> undef, <2 x i32> <i32 0, i32 0>
@@ -1605,7 +1605,7 @@
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; ALL-NEXT:    retq
   %1 = shufflevector <2 x i64> %a, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
   %2 = shufflevector <2 x i64> %b, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
@@ -1767,7 +1767,7 @@
 define <4 x double> @broadcast_v4f64_0000_from_v2i64(<2 x i64> %a0) {
 ; AVX1-LABEL: broadcast_v4f64_0000_from_v2i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -10,7 +10,7 @@
 define <8 x float> @shuffle_v8f32_00000000(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00000000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -25,14 +25,14 @@
 define <8 x float> @shuffle_v8f32_00000010(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00000010:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_00000010:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -44,13 +44,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_00000010:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_00000010:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -62,7 +62,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_00000010:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
@@ -72,14 +72,14 @@
 define <8 x float> @shuffle_v8f32_00000200(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00000200:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_00000200:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -91,13 +91,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_00000200:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_00000200:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -109,7 +109,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_00000200:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
@@ -119,14 +119,14 @@
 define <8 x float> @shuffle_v8f32_00003000(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00003000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_00003000:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -138,13 +138,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_00003000:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_00003000:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -156,7 +156,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_00003000:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
@@ -244,8 +244,8 @@
 define <8 x float> @shuffle_v8f32_00112233(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00112233:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -261,14 +261,14 @@
 define <8 x float> @shuffle_v8f32_00001111(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00001111:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_00001111:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -280,13 +280,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_00001111:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_00001111:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -298,7 +298,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_00001111:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
@@ -318,7 +318,7 @@
 ; AVX1-LABEL: shuffle_v8f32_08080808:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -353,13 +353,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_08084c4c:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_08084c4c:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8f32_08084c4c:
@@ -439,7 +439,7 @@
 ; AVX1-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -447,7 +447,7 @@
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    retq
@@ -465,7 +465,7 @@
 ; AVX2-FAST-PERLANE:       # %bb.0:
 ; AVX2-FAST-PERLANE-NEXT:    vmovaps {{.*#+}} ymm2 = <u,0,1,1,u,2,3,3>
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    retq
@@ -483,8 +483,8 @@
 define <8 x float> @shuffle_v8f32_091b2d3f(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_091b2d3f:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,1,1,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-NEXT:    retq
@@ -526,7 +526,7 @@
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_09ab1def:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-SLOW-NEXT:    retq
@@ -540,7 +540,7 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_09ab1def:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    retq
@@ -558,7 +558,7 @@
 define <8 x float> @shuffle_v8f32_00014445(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_00014445:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
   ret <8 x float> %shuffle
@@ -567,7 +567,7 @@
 define <8 x float> @shuffle_v8f32_00204464(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_00204464:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
   ret <8 x float> %shuffle
@@ -576,7 +576,7 @@
 define <8 x float> @shuffle_v8f32_03004744(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_03004744:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
   ret <8 x float> %shuffle
@@ -585,7 +585,7 @@
 define <8 x float> @shuffle_v8f32_10005444(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_10005444:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
   ret <8 x float> %shuffle
@@ -594,7 +594,7 @@
 define <8 x float> @shuffle_v8f32_22006644(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_22006644:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
   ret <8 x float> %shuffle
@@ -603,7 +603,7 @@
 define <8 x float> @shuffle_v8f32_33307774(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_33307774:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
   ret <8 x float> %shuffle
@@ -612,7 +612,7 @@
 define <8 x float> @shuffle_v8f32_32107654(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_32107654:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shuffle
@@ -621,7 +621,7 @@
 define <8 x float> @shuffle_v8f32_00234467(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_00234467:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
   ret <8 x float> %shuffle
@@ -654,7 +654,7 @@
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; ALL-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
   ret <8 x float> %1
@@ -663,7 +663,7 @@
 define <8 x float> @shuffle_v8f32_10325476(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_10325476:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
   ret <8 x float> %shuffle
@@ -694,7 +694,7 @@
 define <8 x float> @shuffle_v8f32_10235467(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_10235467:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
   ret <8 x float> %shuffle
@@ -703,7 +703,7 @@
 define <8 x float> @shuffle_v8f32_10225466(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_10225466:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
   ret <8 x float> %shuffle
@@ -896,7 +896,7 @@
 ; AVX2-SLOW-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4,5,2,0,4,5,2,0]
 ; AVX2-SLOW-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,2,0,4,7,6,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3,2,0,4,7,6,4]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,1]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
 ; AVX2-SLOW-NEXT:    retq
@@ -916,7 +916,7 @@
 ; AVX2-FAST-PERLANE-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = [4,5,2,0,4,5,2,0]
 ; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,2,0,4,7,6,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3,2,0,4,7,6,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,1]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    retq
@@ -935,18 +935,18 @@
 ; AVX1-LABEL: shuffle_v8f32_f511235a:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,1,1,5,5,5,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1,1,1,5,5,5,5]
 ; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[3],ymm0[3]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_f511235a:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,2,3,7,6,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,2,2,3,7,6,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,0]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,5,5,6,7]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1,2,3,5,5,6,7]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,0,1,2]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-SLOW-NEXT:    retq
@@ -962,9 +962,9 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_f511235a:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,2,3,7,6,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,2,2,3,7,6,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,0]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,5,5,6,7]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1,2,3,5,5,6,7]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,0,1,2]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX2-FAST-PERLANE-NEXT:    retq
@@ -981,13 +981,13 @@
 define <8 x float> @shuffle_v8f32_32103210(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_32103210:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_32103210:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1000,13 +1000,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_32103210:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_32103210:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1019,7 +1019,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_32103210:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
@@ -1030,12 +1030,12 @@
 ; AVX1-LABEL: shuffle_v8f32_76547654:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_76547654:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1048,13 +1048,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_76547654:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_76547654:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1067,7 +1067,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_76547654:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 7, i32 6, i32 5, i32 4>
@@ -1078,12 +1078,12 @@
 ; AVX1-LABEL: shuffle_v8f32_76543210:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_76543210:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1095,13 +1095,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8f32_76543210:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_76543210:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1113,7 +1113,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_76543210:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
@@ -1124,13 +1124,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_3210ba98:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_3210ba98:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8f32_3210ba98:
@@ -1142,7 +1142,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_3210ba98:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x float> %shuffle
@@ -1152,13 +1152,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_3210fedc:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_3210fedc:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8f32_3210fedc:
@@ -1174,13 +1174,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_7654fedc:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_7654fedc:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8f32_7654fedc:
@@ -1192,7 +1192,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_7654fedc:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 15, i32 14, i32 13, i32 12>
   ret <8 x float> %shuffle
@@ -1202,13 +1202,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_fedc7654:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_fedc7654:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8f32_fedc7654:
@@ -1221,7 +1221,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_fedc7654:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 7, i32 6, i32 5, i32 4>
   ret <8 x float> %shuffle
@@ -1266,13 +1266,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_ba987654:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_ba987654:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8f32_ba987654:
@@ -1289,13 +1289,13 @@
 ; AVX1OR2-LABEL: shuffle_v8f32_ba983210:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8f32_ba983210:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8f32_ba983210:
@@ -1308,7 +1308,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8f32_ba983210:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 11, i32 10, i32 9, i32 8, i32 3, i32 2, i32 1, i32 0>
   ret <8 x float> %shuffle
@@ -1336,12 +1336,12 @@
 ; AVX1-LABEL: shuffle_v8f32_084c195d:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5],ymm1[6,7]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm2 = ymm2[1,1,0,3,5,5,4,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,0,3,5,5,4,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,5,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5],ymm0[6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-NEXT:    retq
@@ -1474,7 +1474,7 @@
 define <8 x float> @shuffle_v8f32_uuuu1111(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_uuuu1111:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 1, i32 1, i32 1>
@@ -1485,7 +1485,7 @@
 ; AVX1-LABEL: shuffle_v8f32_44444444:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_44444444:
@@ -1539,7 +1539,7 @@
 define <8 x float> @shuffle_v8f32_uuuu3210(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_uuuu3210:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 3, i32 2, i32 1, i32 0>
@@ -1559,7 +1559,7 @@
 define <8 x float> @shuffle_v8f32_1111uuuu(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_1111uuuu:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x float> %shuffle
@@ -1569,7 +1569,7 @@
 ; ALL-LABEL: shuffle_v8f32_5555uuuu:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x float> %shuffle
@@ -1580,7 +1580,7 @@
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; ALL-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %2 = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
@@ -1614,7 +1614,7 @@
 define <8 x i32> @shuffle_v8i32_00000000(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00000000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -1629,14 +1629,14 @@
 define <8 x i32> @shuffle_v8i32_00000010(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00000010:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_00000010:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1648,13 +1648,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_00000010:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_00000010:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1666,7 +1666,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_00000010:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
@@ -1676,14 +1676,14 @@
 define <8 x i32> @shuffle_v8i32_00000200(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00000200:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_00000200:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1695,13 +1695,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_00000200:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_00000200:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1713,7 +1713,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_00000200:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,2]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,2]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
@@ -1723,14 +1723,14 @@
 define <8 x i32> @shuffle_v8i32_00003000(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00003000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_00003000:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1742,13 +1742,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_00003000:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_00003000:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1760,7 +1760,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_00003000:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,3,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
@@ -1844,7 +1844,7 @@
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_01014545:
 ; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2OR512VL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX2OR512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
   ret <8 x i32> %shuffle
@@ -1853,15 +1853,15 @@
 define <8 x i32> @shuffle_v8i32_00112233(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00112233:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_00112233:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-ALL-LABEL: shuffle_v8i32_00112233:
@@ -1873,13 +1873,13 @@
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_00112233:
 ; AVX2-FAST-PERLANE:       # %bb.0:
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_00112233:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8i32_00112233:
@@ -1891,7 +1891,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_00112233:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3>
   ret <8 x i32> %shuffle
@@ -1900,14 +1900,14 @@
 define <8 x i32> @shuffle_v8i32_00001111(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00001111:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_00001111:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1919,13 +1919,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_00001111:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_00001111:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -1937,7 +1937,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_00001111:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1>
@@ -1957,7 +1957,7 @@
 ; AVX1-LABEL: shuffle_v8i32_08080808:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -1992,19 +1992,19 @@
 ; AVX1-LABEL: shuffle_v8i32_08084c4c:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_08084c4c:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_08084c4c:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8i32_08084c4c:
@@ -2084,7 +2084,7 @@
 ; AVX1-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -2164,7 +2164,7 @@
 define <8 x i32> @shuffle_v8i32_09ab1def(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_09ab1def:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX1-NEXT:    retq
@@ -2203,7 +2203,7 @@
 define <8 x i32> @shuffle_v8i32_00014445(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_00014445:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 1, i32 4, i32 4, i32 4, i32 5>
   ret <8 x i32> %shuffle
@@ -2212,7 +2212,7 @@
 define <8 x i32> @shuffle_v8i32_00204464(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_00204464:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,0,4,4,6,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4>
   ret <8 x i32> %shuffle
@@ -2221,7 +2221,7 @@
 define <8 x i32> @shuffle_v8i32_03004744(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_03004744:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3,0,0,4,7,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 3, i32 0, i32 0, i32 4, i32 7, i32 4, i32 4>
   ret <8 x i32> %shuffle
@@ -2230,7 +2230,7 @@
 define <8 x i32> @shuffle_v8i32_10005444(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_10005444:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,0,0,5,4,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4>
   ret <8 x i32> %shuffle
@@ -2239,7 +2239,7 @@
 define <8 x i32> @shuffle_v8i32_22006644(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_22006644:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,0,0,6,6,4,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 2, i32 0, i32 0, i32 6, i32 6, i32 4, i32 4>
   ret <8 x i32> %shuffle
@@ -2248,7 +2248,7 @@
 define <8 x i32> @shuffle_v8i32_33307774(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_33307774:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3,3,0,7,7,7,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 3, i32 3, i32 0, i32 7, i32 7, i32 7, i32 4>
   ret <8 x i32> %shuffle
@@ -2257,7 +2257,7 @@
 define <8 x i32> @shuffle_v8i32_32107654(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_32107654:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   ret <8 x i32> %shuffle
@@ -2266,7 +2266,7 @@
 define <8 x i32> @shuffle_v8i32_00234467(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_00234467:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,3,4,4,6,7]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
   ret <8 x i32> %shuffle
@@ -2280,7 +2280,7 @@
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_00224466:
 ; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
+; AVX2OR512VL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
 ; AVX2OR512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
   ret <8 x i32> %shuffle
@@ -2289,7 +2289,7 @@
 define <8 x i32> @shuffle_v8i32_10325476(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_10325476:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
   ret <8 x i32> %shuffle
@@ -2303,7 +2303,7 @@
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_11335577:
 ; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
+; AVX2OR512VL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
 ; AVX2OR512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
   ret <8 x i32> %shuffle
@@ -2312,7 +2312,7 @@
 define <8 x i32> @shuffle_v8i32_10235467(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_10235467:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
   ret <8 x i32> %shuffle
@@ -2321,7 +2321,7 @@
 define <8 x i32> @shuffle_v8i32_10225466(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_10225466:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0,2,2,5,4,6,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
   ret <8 x i32> %shuffle
@@ -2625,7 +2625,7 @@
 ; AVX2-SLOW-LABEL: shuffle_v8i32_6caa87e5:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,1,3,2]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,2,4,4,6,6]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,2,4,4,6,6]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,0,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
 ; AVX2-SLOW-NEXT:    retq
@@ -2641,7 +2641,7 @@
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_6caa87e5:
 ; AVX2-FAST-PERLANE:       # %bb.0:
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,1,3,2]
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,2,4,4,6,6]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0,2,2,4,4,6,6]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,0,3]
 ; AVX2-FAST-PERLANE-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
 ; AVX2-FAST-PERLANE-NEXT:    retq
@@ -2659,13 +2659,13 @@
 define <8 x i32> @shuffle_v8i32_32103210(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_32103210:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_32103210:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -2678,13 +2678,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_32103210:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_32103210:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -2697,7 +2697,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_32103210:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
@@ -2708,12 +2708,12 @@
 ; AVX1-LABEL: shuffle_v8i32_76547654:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_76547654:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -2726,13 +2726,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_76547654:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_76547654:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -2745,7 +2745,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_76547654:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 7, i32 6, i32 5, i32 4>
@@ -2756,12 +2756,12 @@
 ; AVX1-LABEL: shuffle_v8i32_76543210:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_76543210:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -2773,13 +2773,13 @@
 ;
 ; AVX2-FAST-PERLANE-LABEL: shuffle_v8i32_76543210:
 ; AVX2-FAST-PERLANE:       # %bb.0:
-; AVX2-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX2-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX2-FAST-PERLANE-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_76543210:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
@@ -2791,7 +2791,7 @@
 ;
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_76543210:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
@@ -2802,13 +2802,13 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_3210ba98:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_3210ba98:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-ALL-LABEL: shuffle_v8i32_3210ba98:
@@ -2820,7 +2820,7 @@
 ; AVX512VL-FAST-PERLANE-LABEL: shuffle_v8i32_3210ba98:
 ; AVX512VL-FAST-PERLANE:       # %bb.0:
 ; AVX512VL-FAST-PERLANE-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-FAST-PERLANE-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-FAST-PERLANE-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-FAST-PERLANE-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 11, i32 10, i32 9, i32 8>
   ret <8 x i32> %shuffle
@@ -2830,13 +2830,13 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_3210fedc:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_3210fedc:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8i32_3210fedc:
@@ -2852,7 +2852,7 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_7654fedc:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_7654fedc:
@@ -2880,7 +2880,7 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_fedc7654:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_fedc7654:
@@ -2909,13 +2909,13 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_ba987654:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_ba987654:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8i32_ba987654:
@@ -2932,13 +2932,13 @@
 ; AVX1OR2-LABEL: shuffle_v8i32_ba983210:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1OR2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i32_ba983210:
 ; AVX512VL-SLOW:       # %bb.0:
 ; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8i32_ba983210:
@@ -3105,7 +3105,7 @@
 define <8 x i32> @shuffle_v8i32_uuuu1111(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_uuuu1111:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 1, i32 1, i32 1>
@@ -3115,7 +3115,7 @@
 define <8 x i32> @shuffle_v8i32_2222uuuu(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_2222uuuu:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i32> %shuffle
@@ -3134,7 +3134,7 @@
 ; AVX1-LABEL: shuffle_v8i32_44444444:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_44444444:
@@ -3180,7 +3180,7 @@
 ; AVX1-LABEL: shuffle_v8i32_44444444_bc:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_44444444_bc:
@@ -3228,7 +3228,7 @@
 ; ALL-LABEL: shuffle_v8i32_5555uuuu:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; ALL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; ALL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i32> %shuffle
@@ -3239,7 +3239,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm0[1,2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[3,1,1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,1,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
@@ -3273,7 +3273,7 @@
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_uuuuuu7u:
 ; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,3,3,3,7,7,7,7]
+; AVX2OR512VL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,3,3,3,7,7,7,7]
 ; AVX2OR512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 7, i32 undef>
   ret <8 x i32> %shuffle
@@ -3284,7 +3284,7 @@
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; ALL-NEXT:    retq
   %1 = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %2 = shufflevector <4 x i32> %b, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
@@ -3297,7 +3297,7 @@
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; ALL-NEXT:    retq
   %1 = shufflevector <4 x i32> %a, <4 x i32> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
   ret <8 x i32> %1
@@ -3317,7 +3317,7 @@
 define <8 x float> @splat_v8f32(<4 x float> %r) {
 ; AVX1-LABEL: splat_v8f32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -3338,7 +3338,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm1[0,2],ymm0[4,6],ymm1[4,6]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_z0U2zUz6:
@@ -3354,7 +3354,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_1U3z5zUU:
@@ -3428,7 +3428,7 @@
 define <8 x i32> @shuffle_v8i32_30127456(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_30127456:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,0,1,2,7,4,5,6]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 3, i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6>
   ret <8 x i32> %shuffle
@@ -3437,7 +3437,7 @@
 define <8 x i32> @shuffle_v8i32_12305674(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_12305674:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
+; ALL-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,2,3,0,5,6,7,4]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 0, i32 5, i32 6, i32 7, i32 4>
   ret <8 x i32> %shuffle
@@ -3690,7 +3690,7 @@
 define <8 x float> @broadcast_concat_crash(<4 x float> %x, <4 x float> %y, float %z) {
 ; AVX1-LABEL: broadcast_concat_crash:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[3,3,3,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
@@ -3699,7 +3699,7 @@
 ; AVX2:       # %bb.0: # %entry
 ; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
 ; AVX2-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
@@ -3708,7 +3708,7 @@
 ; AVX512VL-SLOW:       # %bb.0: # %entry
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
-; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX512VL-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX512VL-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
 ; AVX512VL-SLOW-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512VL-SLOW-NEXT:    retq
@@ -3812,7 +3812,7 @@
 ; AVX1-LABEL: lowhalf_v8i32:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; AVX1-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    retq
 ;
@@ -3839,7 +3839,7 @@
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[2,2]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: lowhalf_v8f32:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
@@ -141,7 +141,7 @@
 define <16 x float> @shuffle_v16f32_00_00_02_00_04_04_06_04_08_08_10_08_12_12_14_12(<16 x float> %a, <16 x float> %b) {
 ; ALL-LABEL: shuffle_v16f32_00_00_02_00_04_04_06_04_08_08_10_08_12_12_14_12:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[0,0,2,0,4,4,6,4,8,8,10,8,12,12,14,12]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[0,0,2,0,4,4,6,4,8,8,10,8,12,12,14,12]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 0, i32 0, i32 2, i32 0, i32 4, i32 4, i32 6, i32 4, i32 8, i32 8, i32 10, i32 8, i32 12, i32 12, i32 14, i32 12>
   ret <16 x float> %shuffle
@@ -150,7 +150,7 @@
 define <16 x float> @shuffle_v16f32_03_uu_uu_uu_uu_04_uu_uu_uu_uu_11_uu_uu_uu_uu_12(<16 x float> %a, <16 x float> %b) {
 ; ALL-LABEL: shuffle_v16f32_03_uu_uu_uu_uu_04_uu_uu_uu_uu_11_uu_uu_uu_uu_12:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,0,3,0,7,4,7,4,11,8,11,8,15,12,15,12]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,0,3,0,7,4,7,4,11,8,11,8,15,12,15,12]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 4, i32 undef, i32 undef, i32 undef, i32 undef, i32 11, i32 undef, i32 undef, i32 undef, i32 undef, i32 12>
   ret <16 x float> %shuffle
@@ -284,7 +284,7 @@
 define <16 x float> @shuffle_v16f32_0b_0a_09_08_0f_0e_0d_0c_03_02_01_00_07_06_05_04(<16 x float> %a) {
 ; ALL-LABEL: shuffle_v16f32_0b_0a_09_08_0f_0e_0d_0c_03_02_01_00_07_06_05_04:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; ALL-NEXT:    vshuff64x2 {{.*#+}} zmm0 = zmm0[4,5,6,7,0,1,2,3]
 ; ALL-NEXT:    retq
   %1 = shufflevector <16 x float> %a, <16 x float> undef, <16 x i32> <i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12, i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
@@ -411,7 +411,7 @@
 define <16 x i32> @shuffle_v16i16_1_0_0_0_5_4_4_4_9_8_8_8_13_12_12_12(<16 x i32> %a, <16 x i32> %b)  {
 ; ALL-LABEL: shuffle_v16i16_1_0_0_0_5_4_4_4_9_8_8_8_13_12_12_12:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[1,0,0,0,5,4,4,4,9,8,8,8,13,12,12,12]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[1,0,0,0,5,4,4,4,9,8,8,8,13,12,12,12]
 ; ALL-NEXT:    retq
   %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 1, i32 0, i32 0, i32 0, i32 5, i32 4, i32 4, i32 4, i32 9, i32 8, i32 8, i32 8, i32 13, i32 12, i32 12, i32 12>
   ret <16 x i32> %c
@@ -420,7 +420,7 @@
 define <16 x i32> @shuffle_v16i16_3_3_0_0_7_7_4_4_11_11_8_8_15_15_12_12(<16 x i32> %a, <16 x i32> %b)  {
 ; ALL-LABEL: shuffle_v16i16_3_3_0_0_7_7_4_4_11_11_8_8_15_15_12_12:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
 ; ALL-NEXT:    retq
   %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9, i32 14, i32 15, i32 12, i32 13>
   ret <16 x i32> %c
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -1300,7 +1300,7 @@
 define <8 x i64> @shuffle_v8i64_00224466(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00224466:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
 ; ALL-NEXT:    ret{{[l|q]}}
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
   ret <8 x i64> %shuffle
@@ -1309,7 +1309,7 @@
 define <8 x i64> @shuffle_v8i64_10325476(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10325476:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13]
 ; ALL-NEXT:    ret{{[l|q]}}
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
   ret <8 x i64> %shuffle
@@ -1318,7 +1318,7 @@
 define <8 x i64> @shuffle_v8i64_11335577(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_11335577:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[2,3,2,3,6,7,6,7,10,11,10,11,14,15,14,15]
+; ALL-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[2,3,2,3,6,7,6,7,10,11,10,11,14,15,14,15]
 ; ALL-NEXT:    ret{{[l|q]}}
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
   ret <8 x i64> %shuffle
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-avx512.ll b/llvm/test/CodeGen/X86/vector-shuffle-avx512.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-avx512.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-avx512.ll
@@ -16,7 +16,7 @@
 ;
 ; KNL-LABEL: expand:
 ; KNL:       # %bb.0:
-; KNL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; KNL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; KNL-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3,4,5,6,7]
 ; KNL-NEXT:    ret{{[l|q]}}
@@ -291,7 +291,7 @@
 ;
 ; KNL-LABEL: expand15:
 ; KNL:       # %bb.0:
-; KNL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; KNL-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; KNL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,3]
 ; KNL-NEXT:    vblendps {{.*#+}} ymm0 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6,7]
 ; KNL-NEXT:    ret{{[l|q]}}
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
@@ -76,7 +76,7 @@
 define <4 x float> @combine_vpermilvar_4f32_unpckh(<4 x float> %a0) {
 ; CHECK-LABEL: combine_vpermilvar_4f32_unpckh:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = tail call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %a0, <4 x i32> <i32 2, i32 2, i32 3, i32 3>)
   ret <4 x float> %1
@@ -85,7 +85,7 @@
 define <4 x float> @combine_vpermilvar_4f32_unpckl(<4 x float> %a0) {
 ; CHECK-LABEL: combine_vpermilvar_4f32_unpckl:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = tail call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %a0, <4 x i32> <i32 0, i32 0, i32 1, i32 1>)
   ret <4 x float> %1
@@ -143,7 +143,7 @@
 ; AVX512-NEXT:    vmovaps {{.*#+}} ymm1 = [16,17,18,19,3,2,1,0]
 ; AVX512-NEXT:    vxorps %xmm2, %xmm2, %xmm2
 ; AVX512-NEXT:    vpermt2ps %zmm2, %zmm1, %zmm0
-; AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX512-NEXT:    ret{{[l|q]}}
   %1 = tail call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>)
   %2 = shufflevector <8 x float> %1, <8 x float> zeroinitializer, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 0, i32 1, i32 2, i32 3>
@@ -274,7 +274,7 @@
 define <4 x float> @combine_vpermilvar_4f32_4stage(<4 x float> %a0) {
 ; CHECK-LABEL: combine_vpermilvar_4f32_4stage:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,3,1]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0,3,1]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = tail call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %a0, <4 x i32> <i32 3, i32 2, i32 1, i32 0>)
   %2 = tail call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float>  %1, <4 x i32> <i32 2, i32 3, i32 0, i32 1>)
@@ -286,7 +286,7 @@
 define <8 x float> @combine_vpermilvar_8f32_4stage(<8 x float> %a0) {
 ; CHECK-LABEL: combine_vpermilvar_8f32_4stage:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,3,1,6,4,7,5]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = tail call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>)
   %2 = tail call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float>  %1, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll
@@ -755,7 +755,7 @@
 define <32 x i8> @combine_pshufb_pshufb_or_pshufb(<32 x i8> %a0) {
 ; CHECK-LABEL: combine_pshufb_pshufb_or_pshufb:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; CHECK-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %1 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1>)
   %2 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3, i8 -1, i8 -1, i8 -1, i8 -1, i8 0, i8 1, i8 2, i8 3>)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
@@ -519,7 +519,7 @@
 define <16 x float> @combine_vpermt2var_16f32_vpermilps(<16 x float> %x0, <16 x float> %x1) {
 ; CHECK-LABEL: combine_vpermt2var_16f32_vpermilps:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
+; CHECK-NEXT:    vshufps {{.*#+}} zmm0 = zmm0[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>, <16 x float> %x0, <16 x float> %x1, i16 -1)
   ret <16 x float> %res0
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-ssse3.ll
@@ -204,7 +204,7 @@
 ;
 ; AVX-LABEL: combine_pshufb_palignr:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <16 x i8> %a0, <16 x i8> %a1, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
   %2 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %1, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>)
@@ -723,7 +723,7 @@
 ;
 ; AVX1-LABEL: combine_pshufb_pshufb_or_pshufb:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_pshufb_pshufb_or_pshufb:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -100,7 +100,7 @@
 ;
 ; AVX1-LABEL: combine_pshufd6:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_pshufd6:
@@ -178,7 +178,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test1:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
   %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
@@ -196,7 +196,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test2:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
   %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
@@ -214,7 +214,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test3:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
   %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
@@ -232,7 +232,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test4:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
   %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
@@ -250,7 +250,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test5:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vorps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
   %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
@@ -268,7 +268,7 @@
 ; AVX-LABEL: combine_bitwise_ops_test6:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vxorps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
   %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
   %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
@@ -628,7 +628,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test1:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1,0,1]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 4, i32 3, i32 1>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 4, i32 0, i32 3>
@@ -643,7 +643,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test2:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 5, i32 2, i32 3>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 4, i32 0, i32 3>
@@ -658,7 +658,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test3:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 6, i32 2, i32 3>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 4, i32 0, i32 3>
@@ -673,7 +673,7 @@
 ;
 ; AVX1-LABEL: combine_nested_undef_test4:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test4:
@@ -693,7 +693,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test5:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 5, i32 5, i32 2, i32 3>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 4, i32 4, i32 3>
@@ -708,7 +708,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test6:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 6, i32 2, i32 4>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 4, i32 0, i32 4>
@@ -723,7 +723,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test7:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,0,2]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 0, i32 2>
@@ -738,7 +738,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test8:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 1, i32 6, i32 3>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 1, i32 4, i32 3, i32 4>
@@ -753,7 +753,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test9:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,2,2]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 1, i32 3, i32 2, i32 5>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 1, i32 4, i32 2>
@@ -768,7 +768,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test10:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 1, i32 1, i32 5, i32 5>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 4>
@@ -783,7 +783,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test11:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1,2,1]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 1, i32 2, i32 5, i32 4>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 0>
@@ -798,7 +798,7 @@
 ;
 ; AVX1-LABEL: combine_nested_undef_test12:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test12:
@@ -867,15 +867,15 @@
 ;
 ; AVX1-LABEL: combine_nested_undef_test15:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1,0,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test15:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastss %xmm1, %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1,0,1]
 ; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 4, i32 3, i32 1>
@@ -906,7 +906,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test16:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
@@ -936,7 +936,7 @@
 ; AVX-LABEL: combine_nested_undef_test17:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1,0,1]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 1, i32 3, i32 1>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 1, i32 0, i32 3>
@@ -951,7 +951,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test18:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[1,1,0,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1,0,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 5, i32 2, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 0, i32 3>
@@ -980,7 +980,7 @@
 ; AVX-LABEL: combine_nested_undef_test19:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,0,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,0,0]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 4, i32 5, i32 6>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 0, i32 0, i32 0>
@@ -1011,7 +1011,7 @@
 ; AVX-LABEL: combine_nested_undef_test20:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,3,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,3,0]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 3, i32 2, i32 4, i32 4>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 1, i32 0, i32 3>
@@ -1040,7 +1040,7 @@
 ; AVX1-LABEL: combine_nested_undef_test21:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test21:
@@ -1065,7 +1065,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test22:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[1,1,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1,1,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 5, i32 2, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 3>
@@ -1080,7 +1080,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test23:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[0,1,0,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[0,1,0,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 5, i32 2, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 3>
@@ -1095,7 +1095,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test24:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[0,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[0,3,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 4, i32 1, i32 6, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 0, i32 3, i32 2, i32 4>
@@ -1110,7 +1110,7 @@
 ;
 ; AVX1-LABEL: combine_nested_undef_test25:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test25:
@@ -1130,7 +1130,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test26:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %B, <4 x i32> %A, <4 x i32> <i32 1, i32 2, i32 6, i32 7>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 2, i32 3>
@@ -1145,7 +1145,7 @@
 ;
 ; AVX1-LABEL: combine_nested_undef_test27:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_nested_undef_test27:
@@ -1165,7 +1165,7 @@
 ;
 ; AVX-LABEL: combine_nested_undef_test28:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1,1,0]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x i32> %B, <4 x i32> %A, <4 x i32> <i32 1, i32 2, i32 4, i32 5>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 3, i32 2>
@@ -1640,7 +1640,7 @@
 ;
 ; AVX-LABEL: combine_test1b:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[0,1,2,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[0,1,2,0]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 4, i32 1, i32 6, i32 3>
   %2 = shufflevector <4 x float> %1, <4 x float> %b, <4 x i32> <i32 0, i32 5, i32 2, i32 0>
@@ -1695,7 +1695,7 @@
 ; AVX-LABEL: combine_test3b:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 6, i32 3>
   %2 = shufflevector <4 x float> %1, <4 x float> %b, <4 x i32> <i32 0, i32 7, i32 2, i32 7>
@@ -1711,7 +1711,7 @@
 ;
 ; AVX-LABEL: combine_test4b:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[1,1,2,3]
 ; AVX-NEXT:    retq
   %1 = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 4, i32 1, i32 6, i32 3>
   %2 = shufflevector <4 x float> %1, <4 x float> %b, <4 x i32> <i32 5, i32 5, i32 2, i32 7>
@@ -2457,7 +2457,7 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_unneeded_subvector1:
@@ -2500,7 +2500,7 @@
 ; AVX1-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: combine_unneeded_subvector2:
@@ -2742,7 +2742,7 @@
 ; AVX:       # %bb.0: # %entry
 ; AVX-NEXT:    vhaddps %xmm0, %xmm0, %xmm1
 ; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,1]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; AVX-NEXT:    retq
 entry:
   %s1 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 1, i32 3>
@@ -2778,7 +2778,7 @@
 ;
 ; AVX-LABEL: PR22390:
 ; AVX:       # %bb.0: # %entry
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,0,1,2]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,0,1,2]
 ; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm0[1,2,3]
 ; AVX-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll b/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
--- a/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
@@ -22,7 +22,7 @@
 ; AVX-LABEL: concat_a_to_shuf_of_a:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps (%rdi), %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX-NEXT:    vmovaps %ymm0, (%rsi)
 ; AVX-NEXT:    vzeroupper
@@ -69,7 +69,7 @@
 ; AVX-LABEL: concat_shuf_of_a_to_a:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps (%rdi), %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX-NEXT:    vmovaps %ymm0, (%rdx)
 ; AVX-NEXT:    vzeroupper
@@ -119,7 +119,7 @@
 ; AVX-LABEL: concat_a_to_shuf_of_a_extrause_of_shuf:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps (%rdi), %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vmovaps %xmm1, (%rdx)
 ; AVX-NEXT:    vmovaps %xmm0, 16(%rsi)
 ; AVX-NEXT:    vmovaps %xmm1, (%rsi)
@@ -128,7 +128,7 @@
 ; AVX2-LABEL: concat_a_to_shuf_of_a_extrause_of_shuf:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovaps (%rdi), %xmm0
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX2-NEXT:    vmovaps %xmm1, (%rdx)
 ; AVX2-NEXT:    vmovaps %xmm0, 16(%rsi)
 ; AVX2-NEXT:    vmovaps %xmm1, (%rsi)
@@ -137,7 +137,7 @@
 ; AVX512F-LABEL: concat_a_to_shuf_of_a_extrause_of_shuf:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vmovaps (%rdi), %xmm0
-; AVX512F-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX512F-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX512F-NEXT:    vmovaps %xmm1, (%rdx)
 ; AVX512F-NEXT:    vmovaps %xmm0, 16(%rsi)
 ; AVX512F-NEXT:    vmovaps %xmm1, (%rsi)
@@ -146,7 +146,7 @@
 ; AVX512BW-LABEL: concat_a_to_shuf_of_a_extrause_of_shuf:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vmovaps (%rdi), %xmm0
-; AVX512BW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX512BW-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX512BW-NEXT:    vmovaps %xmm1, (%rdx)
 ; AVX512BW-NEXT:    vmovaps %xmm0, 16(%rsi)
 ; AVX512BW-NEXT:    vmovaps %xmm1, (%rsi)
@@ -618,7 +618,7 @@
 ; AVX-LABEL: concat_aaa_to_shuf_of_a:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
 ; AVX-NEXT:    vmovaps %ymm0, 32(%rsi)
 ; AVX-NEXT:    vmovaps %ymm1, (%rsi)
@@ -675,7 +675,7 @@
 ; AVX-LABEL: concat_shuf_of_a_to_aaa:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; AVX-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX-NEXT:    vmovaps %ymm0, (%rsi)
 ; AVX-NEXT:    vmovaps %ymm1, 32(%rsi)
diff --git a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
--- a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
@@ -119,7 +119,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744071562067968,18446744071562067968]
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: trunc_ssat_v2i64_v2i32:
@@ -263,7 +263,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744071562067968,18446744071562067968]
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    vmovlpd %xmm0, (%rdi)
 ; AVX-NEXT:    retq
 ;
@@ -1119,7 +1119,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709518848,18446744073709518848]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX1-NEXT:    retq
 ;
@@ -1131,7 +1131,7 @@
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709518848,18446744073709518848]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-SLOW-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -1288,7 +1288,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709518848,18446744073709518848]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX1-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
@@ -1301,7 +1301,7 @@
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709518848,18446744073709518848]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm2
 ; AVX2-SLOW-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX2-SLOW-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-trunc-usat.ll b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
--- a/llvm/test/CodeGen/X86/vector-trunc-usat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
@@ -80,7 +80,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372041149743103,9223372041149743103]
 ; AVX-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    retq
 ;
 ; AVX512F-LABEL: trunc_usat_v2i64_v2i32:
@@ -183,7 +183,7 @@
 ; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372041149743103,9223372041149743103]
 ; AVX-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX-NEXT:    vmovlpd %xmm0, (%rdi)
 ; AVX-NEXT:    retq
 ;
@@ -781,7 +781,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372036854841343,9223372036854841343]
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX1-NEXT:    retq
 ;
@@ -792,7 +792,7 @@
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372036854841343,9223372036854841343]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX2-SLOW-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    retq
 ;
@@ -907,7 +907,7 @@
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372036854841343,9223372036854841343]
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX1-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
@@ -919,7 +919,7 @@
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm3 = [9223372036854841343,9223372036854841343]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX2-SLOW-NEXT:    vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; AVX2-SLOW-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX2-SLOW-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vselect-avx.ll b/llvm/test/CodeGen/X86/vselect-avx.ll
--- a/llvm/test/CodeGen/X86/vselect-avx.ll
+++ b/llvm/test/CodeGen/X86/vselect-avx.ll
@@ -288,16 +288,16 @@
 ; AVX1-LABEL: vselect_concat_splat:
 ; AVX1:       ## %bb.0: ## %entry
 ; AVX1-NEXT:    vmovups (%rax), %xmm0
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,3,2,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,3,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm0[0,3,2,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,3,2]
 ; AVX1-NEXT:    vmovups 16, %xmm2
 ; AVX1-NEXT:    vmovups 32, %xmm3
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm4 = mem[0],xmm3[1],mem[2,3]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2],xmm4[3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[0,3,2,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,3,2,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm3 = mem[0,1],xmm3[2,3]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[1,0,3,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,0,3,2]
 ; AVX1-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX1-NEXT:    vcmpneqps %xmm3, %xmm1, %xmm3
 ; AVX1-NEXT:    vblendvps %xmm3, %xmm4, %xmm1, %xmm1
diff --git a/llvm/test/CodeGen/X86/x86-interleaved-access.ll b/llvm/test/CodeGen/X86/x86-interleaved-access.ll
--- a/llvm/test/CodeGen/X86/x86-interleaved-access.ll
+++ b/llvm/test/CodeGen/X86/x86-interleaved-access.ll
@@ -1720,13 +1720,13 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovups (%rdi), %ymm0
 ; AVX2-NEXT:    vmovups (%rdi), %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,1,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,2,3,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,3,3]
 ; AVX2-NEXT:    vmovups %ymm0, 96(%rsi)
 ; AVX2-NEXT:    vmovups %ymm3, 64(%rsi)
@@ -1780,13 +1780,13 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovups (%rdi), %ymm0
 ; AVX2-NEXT:    vmovups (%rdi), %xmm1
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,1,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,2,3,3]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,1]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,2,3,3]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,3,3]
 ; AVX2-NEXT:    vmovups %ymm0, 96(%rsi)
 ; AVX2-NEXT:    vmovups %ymm3, 64(%rsi)
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -1942,7 +1942,7 @@
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[1,3],ymm0[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 16(%rdx), %xmm1, %xmm1
 ; AVX-NEXT:    vpaddb (%rdx), %xmm0, %xmm0
@@ -4645,7 +4645,7 @@
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm2
 ; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[1,3],ymm2[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,2,1,3]
 ; AVX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
 ; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
@@ -4852,7 +4852,7 @@
 ; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6],ymm3[7]
 ; AVX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; AVX-NEXT:    vpaddb (%rdx), %xmm1, %xmm1
 ; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX-NEXT:    vpaddb 16(%rdx), %xmm2, %xmm2
@@ -6609,7 +6609,7 @@
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[1,3],ymm0[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 48(%rdx), %xmm1, %xmm2
 ; AVX-NEXT:    vpaddb 32(%rdx), %xmm0, %xmm3
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -1562,7 +1562,7 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],mem[1,3],ymm0[4,4],mem[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 16(%rsi), %xmm1, %xmm1
 ; AVX-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
@@ -3743,7 +3743,7 @@
 ; AVX-NEXT:    vmovaps 48(%rdi), %xmm1
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm2 = mem[0,1,0,1]
 ; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[1,3],ymm2[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[0,2,1,3]
+; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm1[0,2,1,3]
 ; AVX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX-NEXT:    vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
 ; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm1
@@ -3896,13 +3896,13 @@
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0],mem[1,2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX-NEXT:    vbroadcastss (%rdi), %ymm2
 ; AVX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6],ymm3[7]
 ; AVX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
 ; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; AVX-NEXT:    vpaddb 32(%rsi), %xmm0, %xmm0
 ; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm2
 ; AVX-NEXT:    vpaddb 16(%rsi), %xmm2, %xmm2
@@ -5369,7 +5369,7 @@
 ; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1]
 ; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[1,3],ymm0[4,4],ymm1[5,7]
-; AVX-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
 ; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm2
 ; AVX-NEXT:    vpaddb 32(%rsi), %xmm0, %xmm3
diff --git a/llvm/utils/gn/secondary/llvm/lib/Target/X86/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/Target/X86/BUILD.gn
--- a/llvm/utils/gn/secondary/llvm/lib/Target/X86/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/Target/X86/BUILD.gn
@@ -94,6 +94,7 @@
     "X86FastTileConfig.cpp",
     "X86FixupBWInsts.cpp",
     "X86FixupLEAs.cpp",
+    "X86FixupInstTuning.cpp",
     "X86FixupSetCC.cpp",
     "X86FlagsCopyLowering.cpp",
     "X86FloatingPoint.cpp",