[X86][SSE] Add cpu feature for aggressive combining to variable shuffles

[RKSimon]

Summary

As mentioned in D38318 and D40865, modern Intel processors prefer to combine multiple shuffles to a variable shuffle mask (PSHUFB/VPERMPS etc.) instead of having multiple stage 'fixed' shuffles which put more pressure on Port 5 (at the expense of extra shuffle mask loads).

As discussed, this patch provides a FeatureFastVariableShuffle target flag for Haswell+ CPUs that prefers combining 2 or more fixed shuffles to a single variable shuffle (default is 3 shuffles).

If everybody is happy with this approach I will refactor some of the vector-shuffle-* tests to run with -fast-variable-shuffle enabled to compare shuffles.

The long term aim is to drive more of this from schedule data (probably via the MC) but we're not close to being ready for that yet.

[spatel]

LGTM.

The only question is whether we prefer a 'Fast' feature or a 'Slow' feature to distinguish these cases. I don't have a preference, so you might want to wait for more feedback before committing.

[zvi]

LGTM. Thanks

[zvi]

On second thought, i think we need to update the tests with a -mattr=+fast-variable-shuffle configuration, right?

[RKSimon]

In D41323#958329, @zvi wrote:

On second thought, i think we need to update the tests with a -mattr=+fast-variable-shuffle configuration, right?

Yup, that's what I'm working on now.

[RKSimon]

Added slow/fast passes for AVX2/AVX512 shuffle lowering tests - did this for v8i16/v16i8 and all 256-bit types as these have useful tests to show the diffs. I can add them for the others if you wish but these will mostly not cause any codegen diffs.

[zvi]

Here's the full list of tests that are affected by setting AllowVariableMask for Depth=2. I think that we should have the full list covered with the new configuration.
I would be happy to assist with the work involved.

CodeGen/X86/avx-intrinsics-fast-isel.ll
CodeGen/X86/avx-splat.ll
CodeGen/X86/avx2-conversions.ll
CodeGen/X86/avx2-vector-shifts.ll
CodeGen/X86/avx512-shuffles/broadcast-vector-int.ll
CodeGen/X86/avx512-shuffles/partial_permute.ll
CodeGen/X86/avx512-trunc.ll
CodeGen/X86/bitcast-and-setcc-512.ll
CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
CodeGen/X86/bitcast-int-to-vector-bool.ll
CodeGen/X86/cast-vsel.ll
CodeGen/X86/combine-or.ll
CodeGen/X86/combine-shl.ll
CodeGen/X86/combine-sra.ll
CodeGen/X86/combine-srl.ll
CodeGen/X86/insertelement-duplicates.ll
CodeGen/X86/insertelement-zero.ll
CodeGen/X86/known-signbits-vector.ll
CodeGen/X86/merge-consecutive-loads-128.ll
CodeGen/X86/oddshuffles.ll
CodeGen/X86/psubus.ll
CodeGen/X86/reduce-trunc-shl.ll
CodeGen/X86/select.ll
CodeGen/X86/shuffle-of-splat-multiuses.ll
CodeGen/X86/shuffle-strided-with-offset-128.ll
CodeGen/X86/shuffle-strided-with-offset-256.ll
CodeGen/X86/shuffle-strided-with-offset-512.ll
CodeGen/X86/shuffle-vs-trunc-128.ll
CodeGen/X86/shuffle-vs-trunc-256.ll
CodeGen/X86/shuffle-vs-trunc-512.ll
CodeGen/X86/sse41.ll
CodeGen/X86/swizzle-2.ll
CodeGen/X86/trunc-ext-ld-st.ll
CodeGen/X86/vec_cast2.ll
CodeGen/X86/vec_insert-5.ll
CodeGen/X86/vec_insert-mmx.ll
CodeGen/X86/vec_set-3.ll
CodeGen/X86/vec_trunc_sext.ll
CodeGen/X86/vector-compare-results.ll
CodeGen/X86/vector-half-conversions.ll
CodeGen/X86/vector-rotate-128.ll
CodeGen/X86/vector-rotate-256.ll
CodeGen/X86/vector-sext.ll
CodeGen/X86/vector-shuffle-128-v16.ll
CodeGen/X86/vector-shuffle-128-v4.ll
CodeGen/X86/vector-shuffle-128-v8.ll

[RKSimon]

In D41323#958558, @zvi wrote:

Here's the full list of tests that are affected by setting AllowVariableMask for Depth=2. I think that we should have the full list covered with the new configuration.
I would be happy to assist with the work involved.

Sure - as long as we're testing with the fast and slow cases for them all. Are you happy with this patch with its tests changes as it is and you just update the remaining tests as followups?

[zvi]

In D41323#958561, @RKSimon wrote:

In D41323#958558, @zvi wrote:

Here's the full list of tests that are affected by setting AllowVariableMask for Depth=2. I think that we should have the full list covered with the new configuration.
I would be happy to assist with the work involved.

Sure - as long as we're testing with the fast and slow cases for them all. Are you happy with this patch with its tests changes as it is and you just update the remaining tests as followups?

Ok, let's do that.

[zvi]

Since all known processors with AVX512 will prefer this new feature turned on, can we make AVX512 imply Fast-var-shuffles?

[RKSimon]

In D41323#960589, @zvi wrote:

Since all known processors with AVX512 will prefer this new feature turned on, can we make AVX512 imply Fast-var-shuffles?

I've created D41436 - the main issue is whether KNL prefers variable shuffles the same as SkylakeServer

[zvi]

In D41323#960715, @RKSimon wrote:

I've created D41436 - the main issue is whether KNL prefers variable shuffles the same as SkylakeServer

The Intel Optimization Reference Manual is the only source of information i have for KNL. It says that of the two core's VPUs, one can execute shuffles. It says that single-source shuffles have a reciprocal throughput of 1 cycle, dual-source shuffles have a reciprocal throughput of 2 cycles. So i don't have any information indicating that KNL differs from SKX in preference of variable-shuffles.

llvm/trunk/lib/Target/X86/X86.td

def FeatureSlowLEA : SubtargetFeature<"slow-lea", "SlowLEA", "true",
                                    "LEA instruction with certain arguments is slow">;
 def FeatureSlow3OpsLEA : SubtargetFeature<"slow-3ops-lea", "Slow3OpsLEA", "true",
                                    "LEA instruction with 3 ops or certain registers is slow">;
 def FeatureSlowIncDec : SubtargetFeature<"slow-incdec", "SlowIncDec", "true",
                                    "INC and DEC instructions are slower than ADD and SUB">;
 def FeatureSoftFloat
     : SubtargetFeature<"soft-float", "UseSoftFloat", "true",
                        "Use software floating point features.">;
+// On recent X86 (port bound) processors, its preferable to combine to a single shuffle
+// using a variable mask over multiple fixed shuffles.
+def FeatureFastVariableShuffle
+    : SubtargetFeature<"fast-variable-shuffle",
+                       "HasFastVariableShuffle",
+                       "true", "Shuffles with variable masks are fast">;
 // On some X86 processors, there is no performance hazard to writing only the
 // lower parts of a YMM or ZMM register without clearing the upper part.
 def FeatureFastPartialYMMorZMMWrite
     : SubtargetFeature<"fast-partial-ymm-or-zmm-write",
                        "HasFastPartialYMMorZMMWrite",
                        "true", "Partial writes to YMM/ZMM registers are fast">;
 // FeatureFastScalarFSQRT should be enabled if scalar FSQRT has shorter latency
 // than the corresponding NR code. FeatureFastVectorFSQRT should be enabled if

 
 def HSWFeatures : ProcessorFeatures<IVBFeatures.Value, [
   FeatureAVX2,
   FeatureBMI,
   FeatureBMI2,
   FeatureERMSB,
   FeatureFMA,
   FeatureLZCNT,
-  FeatureMOVBE
+  FeatureMOVBE,
+  FeatureFastVariableShuffle
 ]>;
 
 class HaswellProc<string Name> : ProcModel<Name, HaswellModel,
                                            HSWFeatures.Value, [
   ProcIntelHSW
 ]>;
 def : HaswellProc<"haswell">;
 def : HaswellProc<"core-avx2">; // Legacy alias.

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

static SDValue combineX86ShuffleChain(ArrayRef<SDValue> Inputs, SDValue Root,
   }
 
   // Don't try to re-form single instruction chains under any circumstances now
   // that we've done encoding canonicalization for them.
   if (Depth < 2)
     return SDValue();
 
   // Depth threshold above which we can efficiently use variable mask shuffles.
-  // TODO This should probably be target specific.
-  bool AllowVariableMask = (Depth >= 3) || HasVariableMask;
+  int VariableShuffleDepth = Subtarget.hasFastVariableShuffle() ? 2 : 3;
+  bool AllowVariableMask = (Depth >= VariableShuffleDepth) || HasVariableMask;
 
   bool MaskContainsZeros =
       any_of(Mask, [](int M) { return M == SM_SentinelZero; });
 
   if (is128BitLaneCrossingShuffleMask(MaskVT, Mask)) {
     // If we have a single input lane-crossing shuffle then lower to VPERMV.
     if (UnaryShuffle && AllowVariableMask && !MaskContainsZeros &&
         ((Subtarget.hasAVX2() &&

llvm/trunk/lib/Target/X86/X86Subtarget.h

protected:
   /// True if this processor has the CMPXCHG16B instruction;
   /// this is true for most x86-64 chips, but not the first AMD chips.
   bool HasCmpxchg16b;
 
   /// True if the LEA instruction should be used for adjusting
   /// the stack pointer. This is an optimization for Intel Atom processors.
   bool UseLeaForSP;
 
+  /// True if its preferable to combine to a single shuffle using a variable
+  /// mask over multiple fixed shuffles.
+  bool HasFastVariableShuffle;
+
   /// True if there is no performance penalty to writing only the lower parts
   /// of a YMM or ZMM register without clearing the upper part.
   bool HasFastPartialYMMorZMMWrite;
 
   /// True if gather is reasonably fast. This is true for Skylake client and
   /// all AVX-512 CPUs.
   bool HasFastGather;
 
public:
   bool isPMULLDSlow() const { return IsPMULLDSlow; }
   bool isUnalignedMem16Slow() const { return IsUAMem16Slow; }
   bool isUnalignedMem32Slow() const { return IsUAMem32Slow; }
   int getGatherOverhead() const { return GatherOverhead; }
   int getScatterOverhead() const { return ScatterOverhead; }
   bool hasSSEUnalignedMem() const { return HasSSEUnalignedMem; }
   bool hasCmpxchg16b() const { return HasCmpxchg16b; }
   bool useLeaForSP() const { return UseLeaForSP; }
+  bool hasFastVariableShuffle() const {
+    return HasFastVariableShuffle;
+  }
   bool hasFastPartialYMMorZMMWrite() const {
     return HasFastPartialYMMorZMMWrite;
   }
   bool hasFastGather() const { return HasFastGather; }
   bool hasFastScalarFSQRT() const { return HasFastScalarFSQRT; }
   bool hasFastVectorFSQRT() const { return HasFastVectorFSQRT; }
   bool hasFastLZCNT() const { return HasFastLZCNT; }
   bool hasFastSHLDRotate() const { return HasFastSHLDRotate; }

llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v16.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq

 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03(<16 x i8> %a, <16 x i8> %b) {
 ; SSE-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i8_00_00_00_00_01_01_01_01_02_02_02_02_03_03_03_03:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VL-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 2, i32 3, i32 3, i32 3, i32 3>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07(<16 x i8> %a, <16 x i8> %b) {
 ; SSE-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; AVX-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,4,4,4,5,5,5,5,6,6,6,6,7,7,7,7]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i8_04_04_04_04_05_05_05_05_06_06_06_06_07_07_07_07:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; AVX512VL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 4, i32 4, i32 4, i32 4, i32 5, i32 5, i32 5, i32 5, i32 6, i32 6, i32 6, i32 6, i32 7, i32 7, i32 7, i32 7>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-LABEL: shuffle_v16i8_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]

llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
 
 define <8 x i16> @shuffle_v8i16_01012323(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_01012323:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v8i16_01012323:

 }
 define <8 x i16> @shuffle_v8i16_00004444(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_00004444:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_00004444:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_00004444:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_00004444:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_00004444:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_00004444:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_00004444:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
   ret <8 x i16> %shuffle
 }
 define <8 x i16> @shuffle_v8i16_u0u1u2u3(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_u0u1u2u3:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSE-NEXT:    retq

 }
 define <8 x i16> @shuffle_v8i16_31206745(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_31206745:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_31206745:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_31206745:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_31206745:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_31206745:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[6,7,2,3,4,5,0,1,12,13,14,15,8,9,10,11]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i16_31206745:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 3, i32 1, i32 2, i32 0, i32 6, i32 7, i32 4, i32 5>
   ret <8 x i16> %shuffle
 }
 define <8 x i16> @shuffle_v8i16_44440000(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_44440000:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]

 }
 define <8 x i16> @shuffle_v8i16_23026745(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_23026745:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_23026745:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_23026745:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_23026745:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_23026745:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,6,7,0,1,4,5,12,13,14,15,8,9,10,11]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i16_23026745:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 2, i32 3, i32 0, i32 2, i32 6, i32 7, i32 4, i32 5>
   ret <8 x i16> %shuffle
 }
 define <8 x i16> @shuffle_v8i16_23016747(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_23016747:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,7]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_23016747:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
-; AVX-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,7]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_23016747:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,7]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_23016747:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_23016747:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,6,7,0,1,2,3,12,13,14,15,8,9,14,15]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_23016747:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,4,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_23016747:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,6,7,0,1,2,3,12,13,14,15,8,9,14,15]
+; AVX512VL-FAST-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 7>
   ret <8 x i16> %shuffle
 }
 define <8 x i16> @shuffle_v8i16_75643120(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_75643120:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]

 }
 define <8 x i16> @shuffle_v8i16_04404567(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_04404567:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_04404567:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_04404567:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_04404567:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_04404567:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,0,1,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_04404567:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_04404567:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,0,1,8,9,10,11,12,13,14,15]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 4, i32 4, i32 0, i32 4, i32 5, i32 6, i32 7>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_0X444444(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_0X444444:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]

 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,14,15,4,5,14,15,12,13,14,15]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v8i16_0127XXXX:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,14,15,4,5,14,15,12,13,14,15]
 ; AVX1OR2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_0127XXXX:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX512VL-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_0127XXXX:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_0127XXXX:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,8,9,10,11,4,5,14,15,12,13,14,15]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_XXXX4563(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_XXXX4563:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]

 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[12,13,6,7,4,5,6,7,8,9,10,11,12,13,6,7]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v8i16_XXXX4563:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,6,7,4,5,6,7,8,9,10,11,12,13,6,7]
 ; AVX1OR2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_XXXX4563:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_XXXX4563:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_XXXX4563:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,6,7,4,5,6,7,8,9,10,11,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 4, i32 5, i32 6, i32 3>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_4563XXXX(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_4563XXXX:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]

 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,6,7,8,9,10,11,0,1,2,3]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v8i16_4563XXXX:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,6,7,8,9,10,11,0,1,2,3]
 ; AVX1OR2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_4563XXXX:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,2,3]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_4563XXXX:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,2,3]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_4563XXXX:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,6,7,4,5,6,7,8,9,10,11,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,2,3]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_01274563(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_01274563:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]

 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,14,15,8,9,10,11,12,13,6,7]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v8i16_01274563:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,14,15,8,9,10,11,12,13,6,7]
 ; AVX1OR2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_01274563:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX512VL-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,1,2]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_01274563:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,1,2]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_01274563:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,8,9,10,11,12,13,6,7,4,5,14,15]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,1,2]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 7, i32 4, i32 5, i32 6, i32 3>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_45630127(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_45630127:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]

 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,6,7,0,1,2,3,4,5,14,15]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v8i16_45630127:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,6,7,0,1,2,3,4,5,14,15]
 ; AVX1OR2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v8i16_45630127:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,3,1]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_45630127:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,3,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_45630127:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,6,7,4,5,14,15,8,9,10,11,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,0,3,1]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 3, i32 0, i32 1, i32 2, i32 7>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_37102735(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_37102735:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]

 define <8 x i16> @shuffle_v8i16_109832ba(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_109832ba:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_109832ba:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
-; AVX-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_109832ba:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
+; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_109832ba:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_109832ba:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,0,1,6,7,2,3,12,13,8,9,14,15,10,11]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_109832ba:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_109832ba:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,0,1,6,7,2,3,12,13,8,9,14,15,10,11]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 1, i32 0, i32 9, i32 8, i32 3, i32 2, i32 11, i32 10>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_8091a2b3(<8 x i16> %a, <8 x i16> %b) {
 ; SSE-LABEL: shuffle_v8i16_8091a2b3:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]

 ; SSE-LABEL: shuffle_v8i16_0213cedf:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_0213cedf:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
-; AVX-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_0213cedf:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_0213cedf:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_0213cedf:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[8,9,12,13,10,11,14,15,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_0213cedf:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_0213cedf:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[8,9,12,13,10,11,14,15,8,9,10,11,12,13,14,15]
+; AVX512VL-FAST-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX512VL-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 2, i32 1, i32 3, i32 12, i32 14, i32 13, i32 15>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_443aXXXX(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_443aXXXX:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,65535,65535,65535]

 ;
 ; SSE41-LABEL: shuffle_v8i16_443aXXXX:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE41-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
 ; SSE41-NEXT:    retq
 ;
-; AVX-LABEL: shuffle_v8i16_443aXXXX:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
-; AVX-NEXT:    retq
+; AVX1-LABEL: shuffle_v8i16_443aXXXX:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v8i16_443aXXXX:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_443aXXXX:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,8,9,6,7,4,5,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_443aXXXX:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_443aXXXX:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,8,9,6,7,4,5,8,9,10,11,12,13,14,15]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 3, i32 10, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_032dXXXX(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_032dXXXX:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]

 ; AVX1-LABEL: shuffle_v8i16_XXX1X579:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i16_XXX1X579:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpbroadcastd %xmm1, %xmm1
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
-; AVX2OR512VL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8i16_XXX1X579:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpbroadcastd %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i16_XXX1X579:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpbroadcastd %xmm1, %xmm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,2,3,8,9,10,11,14,15,14,15]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i16_XXX1X579:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpbroadcastd %xmm1, %xmm1
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
+; AVX512VL-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i16_XXX1X579:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpbroadcastd %xmm1, %xmm1
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,2,3,8,9,10,11,14,15,14,15]
+; AVX512VL-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 1, i32 undef, i32 5, i32 7, i32 9>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_XX4X8acX(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_XX4X8acX:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]

llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpbroadcastw %xmm0, %xmm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpbroadcastw %xmm0, %xmm1
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,8,0,0,0,0,0,0,0,0]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,10,11,0,1,0,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,10,0,0,0,0,0,0,0,0,0,0]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 10, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,14,15,0,1,0,1,0,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,11,0,0,0,0,0,0,0,0,0,0,0]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 11, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2OR512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,7,7,7,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
-; AVX2OR512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,14,15,14,15,14,15,14,15,14,15,14,15,14,15,30,31,30,31,30,31,30,31,30,31,30,31,30,31,30,31]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[3,3,3,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,6,7,6,7,6,7,14,15,14,15,14,15,14,15,22,23,22,23,22,23,22,23,30,31,30,31,30,31,30,31]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,6,7,6,7,6,7,14,15,14,15,14,15,14,15,22,23,22,23,22,23,22,23,30,31,30,31,30,31,30,31]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 7, i32 7, i32 7, i32 7, i32 11, i32 11, i32 11, i32 11, i32 15, i32 15, i32 15, i32 15>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,2,2,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,6,6]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,2,2,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,6,6]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,2,4,5,6,7,8,8,10,10,12,13,14,15]
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,6,6,8,9,10,11,12,12,14,14]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,2,4,5,6,7,8,8,10,10,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,6,6,8,9,10,11,12,12,14,14]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13,16,17,16,17,20,21,20,21,24,25,24,25,28,29,28,29]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,2,4,5,6,7,8,8,10,10,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,6,6,8,9,10,11,12,12,14,14]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13,16,17,16,17,20,21,20,21,24,25,24,25,28,29,28,29]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6, i32 8, i32 8, i32 10, i32 10, i32 12, i32 12, i32 14, i32 14>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[1,1,3,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,7,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,3,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,7,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7,9,9,11,11,12,13,14,15]
-; AVX2OR512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,5,7,7,8,9,10,11,13,13,15,15]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7,9,9,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,5,7,7,8,9,10,11,13,13,15,15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7,10,11,10,11,14,15,14,15,18,19,18,19,22,23,22,23,26,27,26,27,30,31,30,31]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7,9,9,11,11,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,5,7,7,8,9,10,11,13,13,15,15]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7,10,11,10,11,14,15,14,15,18,19,18,19,22,23,22,23,26,27,26,27,30,31,30,31]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7, i32 9, i32 9, i32 11, i32 11, i32 13, i32 13, i32 15, i32 15>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_01_00_00_00_00_00_00_00_01_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_00_00_00_00_00_00_01_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1]

 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,16,0,16,0,16,0,16,8,24,8,24,8,24,8,24]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,4,5,2,3,0,1,14,15,12,13,10,11,8,9,22,23,20,21,18,19,16,17,30,31,28,29,26,27,24,25]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,2,1,0,23,22,21,20,11,10,9,8,31,30,29,28]
 ; AVX512VL-NEXT:    vpermi2w %ymm0, %ymm1, %ymm2
 ; AVX512VL-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 19, i32 18, i32 17, i32 16, i32 7, i32 6, i32 5, i32 4, i32 27, i32 26, i32 25, i32 24, i32 15, i32 14, i32 13, i32 12>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [12,13,8,9,4,5,0,1,14,15,10,11,6,7,2,3]
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,7,6,7,4,5,2,3,0,1,16,17,18,19,20,21,22,23,22,23,20,21,18,19,16,17]
+; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,2,1,0,19,18,17,16,11,10,9,8,27,26,25,24]
 ; AVX512VL-NEXT:    vpermi2w %ymm0, %ymm1, %ymm2
 ; AVX512VL-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 19, i32 18, i32 17, i32 16, i32 3, i32 2, i32 1, i32 0, i32 27, i32 26, i32 25, i32 24, i32 11, i32 10, i32 9, i32 8>

 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,4,4,4,4,16,16,16,16,20,20,20,20]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,12,12,12,12,16,16,16,16,20,20,20,20]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,12,12,12,12,24,24,24,24,28,28,28,28]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 24, i32 24, i32 24, i32 24, i32 28, i32 28, i32 28, i32 28>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,4,4,4,4,24,24,24,24,28,28,28,28]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 24, i32 24, i32 24, i32 24, i32 28, i32 28, i32 28, i32 28>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpbroadcastw %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpbroadcastw %xmm1, %xmm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,0,0,0,8,8,8,8,8,8,8,8,8]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpsllq $48, %xmm1, %xmm2
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm2[7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpsllq $48, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpsllq $48, %xmm1, %xmm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,0,0,0,4,4,4,12,8,8,8,8,12,12,12,12]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 12, i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,1,2,0,4,5,6,7,11,9,10,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,1,2,0,4,5,6,7,11,9,10,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,2,3,4,5,0,1,12,13,14,15,8,9,10,11,22,23,18,19,20,21,16,17,28,29,30,31,24,25,26,27]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,1,2,0,6,7,4,13,11,9,10,8,14,15,12,13]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 3, i32 1, i32 2, i32 0, i32 6, i32 7, i32 4, i32 13, i32 11, i32 9, i32 10, i32 8, i32 14, i32 15, i32 12, i32 13>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[2,3,0,2,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[2,3,0,2,4,5,6,7,10,11,8,10,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[2,3,0,2,4,5,6,7,10,11,8,10,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,6,7,0,1,4,5,12,13,14,15,8,9,10,11,20,21,22,23,16,17,20,21,28,29,30,31,24,25,26,27]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,3,0,2,6,7,4,13,10,11,8,10,14,15,12,13]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 2, i32 3, i32 0, i32 2, i32 6, i32 7, i32 4, i32 13, i32 10, i32 11, i32 8, i32 10, i32 14, i32 15, i32 12, i32 13>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,0,2,3]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,7,4,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,7,4,7,8,9,10,11,14,15,12,15]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,7,4,7,8,9,10,11,14,15,12,15]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[4,5,6,7,0,1,2,3,12,13,14,15,8,9,14,15,20,21,22,23,16,17,18,19,28,29,30,31,24,25,30,31]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,3,0,1,6,7,4,15,10,11,8,9,14,15,12,15]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 15, i32 10, i32 11, i32 8, i32 9, i32 14, i32 15, i32 12, i32 15>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7]
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm3
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,2,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,7,5,1,6,4,11,14,14,15,13,9,14,12,11]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 6, i32 6, i32 7, i32 5, i32 1, i32 6, i32 4, i32 11, i32 14, i32 14, i32 15, i32 13, i32 9, i32 14, i32 12, i32 11>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,0,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,2,0,4,5,6,7,8,10,10,8,12,13,14,15]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,2,0,4,5,6,7,8,10,10,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,8,9,10,11,12,13,14,15,16,17,24,25,24,25,16,17,24,25,26,27,28,29,30,31]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,4,4,0,4,5,6,15,8,12,12,8,12,13,14,15]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 4, i32 4, i32 0, i32 4, i32 5, i32 6, i32 15, i32 8, i32 12, i32 12, i32 8, i32 12, i32 13, i32 14, i32 15>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_uu:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,4,5,14,15,12,13,14,15,16,17,18,19,20,21,30,31,20,21,30,31,28,29,30,31]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_uu:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
-; AVX512VL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,7,6,7,8,9,10,11,12,15,14,15]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_uu:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX512VL-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,7,6,7,8,9,10,11,12,15,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_uu:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,8,9,10,11,4,5,14,15,12,13,14,15,16,17,18,19,24,25,26,27,20,21,30,31,28,29,30,31]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 8, i32 9, i32 10, i32 15, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_uu_uu_uu_uu_04_05_06_11_uu_uu_uu_uu_12_13_14_11(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_uu_uu_uu_uu_04_05_06_11_uu_uu_uu_uu_12_13_14_11:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1

 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_uu:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,6,7,8,9,10,11,0,1,2,3,24,25,26,27,28,29,22,23,24,25,26,27,16,17,18,19]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_uu:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,1,2,0,7,5,6,4]
-; AVX512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,3,2,3,4,5,6,7,8,11,10,11,12,13,14,15]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_uu:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,1,2,0,7,5,6,4]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,3,2,3,4,5,6,7,8,11,10,11,12,13,14,15]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_uu:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,6,7,4,5,6,7,8,9,10,11,0,1,2,3,28,29,22,23,20,21,22,23,24,25,26,27,16,17,18,19]
+; AVX512VL-FAST-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 5, i32 6, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 12, i32 13, i32 14, i32 11, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_01_zz_02_zz_04_uu_06_07_08_09_10_11_12_13_14_15(<16 x i16> %a) {
 ; AVX1-LABEL: shuffle_v16i16_01_zz_02_zz_04_uu_06_07_08_09_10_11_12_13_14_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[1,1,2,3,4,5,6,7]

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11]
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm3
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5],xmm0[6,7]
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,6,3,4,5,6,7]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,7,1,0,2,7,3,13,11,15,9,8,10,15,11,13]
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm1, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 3, i32 7, i32 1, i32 0, i32 2, i32 7, i32 3, i32 13, i32 11, i32 15, i32 9, i32 8, i32 10, i32 15, i32 11, i32 13>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,2,3,8,9,12,13,12,13,14,15,16,17,16,17,20,21,18,19,24,25,28,29,28,29,30,31]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,2,3,6,7,12,13,10,11,14,15,14,15,16,17,18,19,18,19,22,23,28,29,26,27,30,31,30,31]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,16,1,17,6,22,7,31,8,24,9,25,14,30,15,31]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 6, i32 22, i32 7, i32 31, i32 8, i32 24, i32 9, i32 25, i32 14, i32 30, i32 15, i32 31>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[8,9,10,11,0,1,2,3,2,3,0,1,12,13,2,3]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u>
-; AVX2-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[8,9,8,9,4,5,10,11,0,1,0,1,12,13,2,3,24,25,24,25,20,21,26,27,16,17,16,17,28,29,18,19]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u>
+; AVX2-SLOW-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[8,9,8,9,4,5,10,11,0,1,0,1,12,13,2,3,24,25,24,25,20,21,26,27,16,17,16,17,28,29,18,19]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u>
+; AVX2-FAST-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[8,9,8,9,4,5,10,11,0,1,0,1,12,13,2,3,24,25,24,25,20,21,26,27,16,17,16,17,28,29,18,19]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,2,3,6,7,12,13,10,11,14,15,14,15,16,17,18,19,18,19,22,23,28,29,26,27,30,31,30,31]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,20,1,21,6,16,7,25,8,28,9,29,14,24,15,25]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 20, i32 1, i32 21, i32 6, i32 16, i32 7, i32 25, i32 8, i32 28, i32 9, i32 29, i32 14, i32 24, i32 15, i32 25>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm4[0]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,8,9,4,5,14,15,0,1,4,5,4,5,6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,1,3,4,5,6,7,8,10,9,11,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,5,7,8,9,10,11,12,14,13,15]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,1,3,4,5,6,7,8,10,9,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,5,7,8,9,10,11,12,14,13,15]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,4,5,2,3,6,7,8,9,12,13,10,11,14,15,16,17,20,21,18,19,22,23,24,25,28,29,26,27,30,31]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,1,3,20,22,21,31,8,10,9,11,28,30,29,31]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 2, i32 1, i32 3, i32 20, i32 22, i32 21, i32 31, i32 8, i32 10, i32 9, i32 11, i32 28, i32 30, i32 29, i32 31>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,3,2,4,5,6,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7,8,9],ymm1[10],ymm0[11,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,1,2,3,6,5,6,7]
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,3,2,4,5,6,7,8,8,11,10,12,13,14,15]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7,8,9],ymm1[10],ymm0[11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,3,2,4,5,6,7,8,8,11,10,12,13,14,15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7,8,9],ymm1[10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,8,9,6,7,4,5,8,9,10,11,12,13,14,15,24,25,24,25,22,23,20,21,24,25,26,27,28,29,30,31]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = <4,4,3,18,u,u,u,u,12,12,11,26,u,u,u,u>
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 4, i32 3, i32 18, i32 undef, i32 undef, i32 undef, i32 undef, i32 12, i32 12, i32 11, i32 26, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle

 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm1[7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
-; AVX2-NEXT:    vpbroadcastd %xmm1, %ymm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,2,1,4,5,6,7,8,9,10,9,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,7,7,8,9,10,11,12,13,15,15]
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-SLOW-NEXT:    vpbroadcastd %xmm1, %ymm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,2,1,4,5,6,7,8,9,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,7,7,8,9,10,11,12,13,15,15]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,2,3,8,9,10,11,14,15,14,15,16,17,18,19,20,21,18,19,24,25,26,27,30,31,30,31]
+; AVX2-FAST-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-FAST-NEXT:    vpbroadcastd %xmm1, %ymm1
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,1,u,5,7,25,u,u,u,9,u,13,15,25>
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 1, i32 undef, i32 5, i32 7, i32 25, i32 undef, i32 undef, i32 undef, i32 9, i32 undef, i32 13, i32 15, i32 25>
   ret <16 x i16> %shuffle

 define <16 x i16> @shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[6,7,6,7,6,7,6,7,6,7,6,7,6,7,6,7]
+; AVX2-FAST-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0

 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u(<16 x i16> %a, <16 x i16> %b) {
-; ALL-LABEL: shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u:
-; ALL:       # %bb.0:
-; ALL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; ALL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; ALL-NEXT:    retq
+; AVX1-LABEL: shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX1-NEXT:    retq
+;
+; AVX2-SLOW-LABEL: shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[6,7,6,7,6,7,6,7,6,7,6,7,6,7,6,7]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i16_3_3_3_3_3_3_3_3_u_u_u_u_u_u_u_u:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2OR512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25(<16 x i16> %a0, <16 x i16> %a1) {
 ; AVX1-LABEL: shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]

 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3,4,5],xmm1[6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[6,7,4,5,4,5,10,11,4,5,14,15,12,13,0,1]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4,5],xmm0[6],xmm1[7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: PR24935:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm2 = ymm1[8,9,10,11,4,5,8,9,0,1,14,15,12,13,0,1,24,25,26,27,20,21,24,25,16,17,30,31,28,29,16,17]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[6,7,4,5,0,1,10,11,4,5,10,11,4,5,6,7,22,23,20,21,16,17,26,27,20,21,26,27,20,21,22,23]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,u,u,255,255,255,255,0,0,u,u,0,0,u,u,u,u,255,255,0,0,u,u,u,u,u,u,0,0>
-; AVX2-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
-; AVX2-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,5,5,6,7,8,9,10,11,13,13,14,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,6,7,u,u,18,19,u,u,u,u,u,u,u,u,24,25,16,17,u,u]
-; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5,6,7,8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13,14,15]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,0,0,255,255,0,0,0,0,255,255,255,255,0,0,0,0,0,0,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: PR24935:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm2 = ymm1[8,9,10,11,4,5,8,9,0,1,14,15,12,13,0,1,24,25,26,27,20,21,24,25,16,17,30,31,28,29,16,17]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[6,7,4,5,0,1,10,11,4,5,10,11,4,5,6,7,22,23,20,21,16,17,26,27,20,21,26,27,20,21,22,23]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,u,u,255,255,255,255,0,0,u,u,0,0,u,u,u,u,255,255,0,0,u,u,u,u,u,u,0,0>
+; AVX2-SLOW-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,5,5,6,7,8,9,10,11,13,13,14,15]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,6,7,u,u,18,19,u,u,u,u,u,u,u,u,24,25,16,17,u,u]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5,6,7,8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13,14,15]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,0,0,255,255,0,0,0,0,255,255,255,255,0,0,0,0,0,0,255,255]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: PR24935:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm2 = ymm1[8,9,10,11,4,5,8,9,0,1,14,15,12,13,0,1,24,25,26,27,20,21,24,25,16,17,30,31,28,29,16,17]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[6,7,4,5,0,1,10,11,4,5,10,11,4,5,6,7,22,23,20,21,16,17,26,27,20,21,26,27,20,21,22,23]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,u,u,255,255,255,255,0,0,u,u,0,0,u,u,u,u,255,255,0,0,u,u,u,u,u,u,0,0>
+; AVX2-FAST-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm2 = ymm0[0,1,2,3,2,3,6,7,10,11,10,11,12,13,14,15,16,17,18,19,18,19,22,23,26,27,26,27,28,29,30,31]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,6,7,u,u,18,19,u,u,u,u,u,u,u,u,24,25,16,17,u,u]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5,6,7,8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13,14,15]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,0,0,255,255,0,0,0,0,255,255,255,255,0,0,0,0,0,0,255,255]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: PR24935:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [11,10,17,13,10,7,27,0,17,25,0,12,29,20,16,8]
 ; AVX512VL-NEXT:    vpermi2w %ymm0, %ymm1, %ymm2
 ; AVX512VL-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 27, i32 26, i32 1, i32 29, i32 26, i32 23, i32 11, i32 16, i32 1, i32 9, i32 16, i32 28, i32 13, i32 4, i32 0, i32 24>

llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v32.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
 
 define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq

 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,0,9,0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_20_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2OR512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_20_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_20_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_20_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_20_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,0,4,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 20, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,0,11,0,0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2OR512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_21_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,0,5,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 21, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,0,13,0,0,0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2OR512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_22_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,0,6,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 22, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,15,0,0,0,0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2OR512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
-; AVX2OR512VL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_23_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,6,7,0,1,2,3]
+; AVX512VL-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,7,0,0,0,0,0,0,0,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 23, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_00_00_00_00_00_00_24_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_24_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; 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:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-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:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-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:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-SLOW-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
+; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
+; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-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:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-FAST-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
+; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-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:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
 ; AVX512VL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX512VL-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX512VL-NEXT:    movl $-1431655766, %eax # imm = 0xAAAAAAAA
 ; AVX512VL-NEXT:    kmovd %eax, %k1

 ;
 ; AVX2-LABEL: shuffle_v32i8_15_15_15_15_15_15_15_15_32_32_32_32_32_32_32_32_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastb %xmm1, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[15,15,15,15,15,15,15,15,12,12,13,13,14,14,15,15]
 ; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v32i8_15_15_15_15_15_15_15_15_32_32_32_32_32_32_32_32_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpbroadcastb %xmm1, %xmm1
-; AVX512VL-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX512VL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; AVX512VL-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v32i8_15_15_15_15_15_15_15_15_32_32_32_32_32_32_32_32_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpbroadcastb %xmm1, %xmm1
+; AVX512VL-SLOW-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX512VL-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX512VL-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v32i8_15_15_15_15_15_15_15_15_32_32_32_32_32_32_32_32_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpbroadcastb %xmm1, %xmm1
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[15,15,15,15,15,15,15,15,12,12,13,13,14,14,15,15]
+; AVX512VL-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1OR2-LABEL: shuffle_v32i8_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu_uu:
 ; AVX1OR2:       # %bb.0:
 ; AVX1OR2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]

llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v4.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
 
 define <4 x double> @shuffle_v4f64_0000(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: shuffle_v4f64_0000:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;

 define <4 x double> @shuffle_v4f64_0z3z(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: shuffle_v4f64_0z3z:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v4f64_0z3z:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
-; AVX2-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v4f64_0z3z:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
-; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
-; AVX512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v4f64_0z3z:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
+; AVX2-SLOW-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v4f64_0z3z:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v4f64_0z3z:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
+; AVX512VL-SLOW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-SLOW-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4f64_0z3z:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> <double 0.000000e+00, double undef, double undef, double undef>, <4 x i32> <i32 0, i32 4, i32 3, i32 4>
   ret <4 x double> %shuffle
 }
 
 define <4 x double> @shuffle_v4f64_1z2z(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: shuffle_v4f64_1z2z:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v4f64_1z2z:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v4f64_1z2z:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
-; AVX512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v4f64_1z2z:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v4f64_1z2z:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v4f64_1z2z:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-SLOW-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4f64_1z2z:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-FAST-NEXT:    retq
   %1 = shufflevector <4 x double> %a, <4 x double> <double 0.000000e+00, double undef, double undef, double undef>, <4 x i32> <i32 1, i32 4, i32 2, i32 4>
   ret <4 x double> %1
 }
 
 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:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]

 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_2u35:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
 ; AVX2-NEXT:    retq
 ;
-; AVX512VL-LABEL: shuffle_v4i64_2u35:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
-; AVX512VL-NEXT:    retq
+; AVX512VL-SLOW-LABEL: shuffle_v4i64_2u35:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,1]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4i64_2u35:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,5,3,5]
+; AVX512VL-FAST-NEXT:    vpermt2q %ymm1, %ymm2, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 2, i32 undef, i32 3, i32 5>
   ret <4 x i64> %shuffle
 }
 
 define <4 x i64> @shuffle_v4i64_1251(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_1251:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]

 define <4 x i64> @shuffle_v4i64_z0z3(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_z0z3:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v4i64_z0z3:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
-; AVX2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v4i64_z0z3:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
+; AVX2-SLOW-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v4i64_z0z3:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[24,25,26,27,28,29,30,31]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i64_z0z3:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,3]
 ; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX512VL-NEXT:    retq
   %1 = shufflevector <4 x i64> %a, <4 x i64> <i64 0, i64 undef, i64 undef, i64 undef>, <4 x i32> <i32 4, i32 0, i32 4, i32 3>
   ret <4 x i64> %1
 }
 
 define <4 x i64> @shuffle_v4i64_1z2z(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_1z2z:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v4i64_1z2z:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v4i64_1z2z:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,2,0]
-; AVX512VL-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v4i64_1z2z:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,2,0]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v4i64_1z2z:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v4i64_1z2z:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,2,0]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v4i64_1z2z:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,18,19,20,21,22,23],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512VL-FAST-NEXT:    retq
   %1 = shufflevector <4 x i64> %a, <4 x i64> <i64 0, i64 undef, i64 undef, i64 undef>, <4 x i32> <i32 1, i32 4, i32 2, i32 4>
   ret <4 x i64> %1
 }

llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v8.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX1OR2 --check-prefix=AVX2OR512VL --check-prefix=AVX2 --check-prefix=AVX2-FAST
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512dq,+fast-variable-shuffle | FileCheck %s --check-prefix=ALL --check-prefix=AVX2OR512VL --check-prefix=AVX512VL --check-prefix=AVX512VL-FAST
 
 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:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;

 define <8 x float> @shuffle_v8f32_09ab1def(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_09ab1def:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovshdup {{.*#+}} 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
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_09ab1def:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2OR512VL-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v8f32_09ab1def:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8f32_09ab1def:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,3]
+; AVX512VL-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX512VL-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8f32_09ab1def:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
+; AVX512VL-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [8,1,2,3,10,5,6,7]
+; AVX512VL-FAST-NEXT:    vpermi2ps %ymm2, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
   ret <8 x float> %shuffle
 }
 
 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]

 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX1-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8f32_c348cda0:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <4,u,u,0,4,5,2,u>
-; AVX2-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,2,0,4,7,6,4]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,1]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: shuffle_v8f32_c348cda0:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm2 = <4,u,u,0,4,5,2,u>
+; AVX2-SLOW-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} 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
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_c348cda0:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [0,3,4,7,4,7,2,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <4,u,u,0,4,5,2,u>
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v8f32_c348cda0:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovaps {{.*#+}} ymm2 = [4,11,12,0,4,5,2,8]
 ; AVX512VL-NEXT:    vpermi2ps %ymm0, %ymm1, %ymm2
 ; AVX512VL-NEXT:    vmovaps %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 12, i32 3, i32 4, i32 8, i32 12, i32 13, i32 10, i32 0>
   ret <8 x float> %shuffle
 }
 
 define <8 x float> @shuffle_v8f32_f511235a(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_f511235a:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
 ; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[3],ymm0[3]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8f32_f511235a:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,2,2,3,7,6,6,7]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,0]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,5,5,6,7]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,0,1,2]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
-; AVX2-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:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,0]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} 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
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_f511235a:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [7,6,2,3,7,6,3,2]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [5,5,1,1,2,3,5,5]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v8f32_f511235a:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovaps {{.*#+}} ymm2 = [15,5,1,1,2,3,5,10]
 ; AVX512VL-NEXT:    vpermt2ps %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 15, i32 5, i32 1, i32 1, i32 2, i32 3, i32 5, i32 10>
   ret <8 x float> %shuffle

 
 define <8 x float> @shuffle_v8f32_76547654(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_76547654:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_76547654:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX2OR512VL-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:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_76547654:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [7,6,5,4,7,6,5,4]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8f32_76547654:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX512VL-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>
   ret <8 x float> %shuffle
 }
 
 define <8 x float> @shuffle_v8f32_76543210(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_76543210:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8f32_76543210:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2OR512VL-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:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8f32_76543210:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [7,6,5,4,3,2,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8f32_76543210:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX512VL-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>
   ret <8 x float> %shuffle
 }
 
 define <8 x float> @shuffle_v8f32_3210ba98(<8 x float> %a, <8 x float> %b) {
 ; ALL-LABEL: shuffle_v8f32_3210ba98:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0

 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:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_09ab1def:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
-; AVX2OR512VL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
-; AVX2OR512VL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2OR512VL-NEXT:    retq
+; AVX2-LABEL: shuffle_v8i32_09ab1def:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-NEXT:    retq
+;
+; AVX512VL-SLOW-LABEL: shuffle_v8i32_09ab1def:
+; AVX512VL-SLOW:       # %bb.0:
+; AVX512VL-SLOW-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX512VL-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX512VL-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX512VL-SLOW-NEXT:    retq
+;
+; AVX512VL-FAST-LABEL: shuffle_v8i32_09ab1def:
+; AVX512VL-FAST:       # %bb.0:
+; AVX512VL-FAST-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero
+; AVX512VL-FAST-NEXT:    vmovdqa {{.*#+}} ymm0 = [8,1,2,3,10,5,6,7]
+; AVX512VL-FAST-NEXT:    vpermi2d %ymm2, %ymm1, %ymm0
+; AVX512VL-FAST-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 1, i32 13, i32 14, i32 15>
   ret <8 x i32> %shuffle
 }
 
 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]

 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,2],ymm2[4,4],ymm1[6,6]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8i32_6caa87e5:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,1,3,2]
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,0,2,2,4,4,6,6]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,1,0,3]
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
-; AVX2-NEXT:    retq
+; 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:    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
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_6caa87e5:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [4,4,2,2,0,0,6,6]
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,1,3,2]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v8i32_6caa87e5:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [14,4,2,2,0,15,6,13]
 ; AVX512VL-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
 ; AVX512VL-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 6, i32 12, i32 10, i32 10, i32 8, i32 7, i32 14, i32 5>

 
 define <8 x i32> @shuffle_v8i32_76547654(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_76547654:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_76547654:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX2OR512VL-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:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_76547654:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [7,6,5,4,7,6,5,4]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i32_76547654:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX512VL-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>
   ret <8 x i32> %shuffle
 }
 
 define <8 x i32> @shuffle_v8i32_76543210(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_76543210:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    retq
 ;
-; AVX2OR512VL-LABEL: shuffle_v8i32_76543210:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2OR512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2OR512VL-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:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: shuffle_v8i32_76543210:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [7,6,5,4,3,2,1,0]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i32_76543210:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; AVX512VL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX512VL-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>
   ret <8 x i32> %shuffle
 }
 
 define <8 x i32> @shuffle_v8i32_3210ba98(<8 x i32> %a, <8 x i32> %b) {
 ; ALL-LABEL: shuffle_v8i32_3210ba98:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0

llvm/trunk/test/CodeGen/X86/vector-shuffle-512-v32.ll

 ; KNL-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,0,0,4,5,6,7,9,9,8,8,12,13,14,15]
 ; KNL-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,5,4,4,8,9,10,11,13,13,12,12]
 ; KNL-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,1,0,0,4,5,6,7,9,9,8,8,12,13,14,15]
 ; KNL-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,5,5,4,4,8,9,10,11,13,13,12,12]
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: shuffle_v32i16_1_1_0_0_5_5_4_4_9_9_11_11_13_13_12_12_17_17_19_19_21_21_20_20_25_25_27_27_29_29_28_28:
 ; SKX:       ## %bb.0:
-; SKX-NEXT:    vpshuflw {{.*#+}} zmm0 = zmm0[1,1,0,0,4,5,6,7,9,9,8,8,12,13,14,15,17,17,16,16,20,21,22,23,25,25,24,24,28,29,30,31]
-; SKX-NEXT:    vpshufhw {{.*#+}} zmm0 = zmm0[0,1,2,3,5,5,4,4,8,9,10,11,13,13,12,12,16,17,18,19,21,21,20,20,24,25,26,27,29,29,28,28]
+; SKX-NEXT:    vpshufb {{.*#+}} zmm0 = zmm0[2,3,2,3,0,1,0,1,10,11,10,11,8,9,8,9,18,19,18,19,16,17,16,17,26,27,26,27,24,25,24,25,34,35,34,35,32,33,32,33,42,43,42,43,40,41,40,41,50,51,50,51,48,49,48,49,58,59,58,59,56,57,56,57]
 ; SKX-NEXT:    retq
   %c = shufflevector <32 x i16> %a, <32 x i16> zeroinitializer, <32 x i32> <i32 1, i32 1, i32 0, i32 0, i32 5, i32 5, i32 4, i32 4, i32 9, i32 9, i32 8, i32 8, i32 13, i32 13, i32 12, i32 12, i32 17, i32 17, i32 16, i32 16, i32 21, i32 21, i32 20, i32 20, i32 25, i32 25, i32 24, i32 24, i32 29, i32 29, i32 28, i32 28>
   ret <32 x i16> %c
 }
 
 define <32 x i16> @shuffle_v32i16_0zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz(<32 x i16> %a) {
 ; KNL-LABEL: shuffle_v32i16_0zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz:
 ; KNL:       ## %bb.0: