AVX512: support AVX512BW Intrinsic in 32bit mode.

[igorb]

Summary

AVX512: support AVX512BW Intrinsic in 32bit mode.

[delena]

lib/Target/X86/X86ISelLowering.cpp

16009 ↗

(On Diff #41393)

You can put assert:
if (Mask.getSimpleValueType() == MVT::i64 && Subtarget->is32Bit()) {
assert(MaskVT == MVT::v64i1 &&
assert(Subtarget->hasBWI() & ..

but it is not necessary

16010 ↗

(On Diff #41393)

extend / split

16012 ↗

(On Diff #41393)

Lo, Hi

16033 ↗

(On Diff #41393)

remove empty line

[igorb]

Thanks for the review!

[delena]

lib/Target/X86/X86ISelLowering.cpp

16004 ↗

(On Diff #41402)

Can newMaskVT be i64?

[igorb]

Thanks for the review!
Update patch according to you comments.

lib/Target/X86/X86ISelLowering.cpp

16004 ↗

(On Diff #41402)

we don't have such intrinsic meanwhile, but theoretically it is possible.

[delena]

LGTM

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

if (!Subtarget->useSoftFloat() && Subtarget->hasVLX()) {
     setOperationAction(ISD::UMIN,               MVT::v2i64, Legal);
     setOperationAction(ISD::UMIN,               MVT::v4i64, Legal);
   }
 
   // We want to custom lower some of our intrinsics.
   setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom);
   setOperationAction(ISD::INTRINSIC_W_CHAIN, MVT::Other, Custom);
   setOperationAction(ISD::INTRINSIC_VOID, MVT::Other, Custom);
-  if (!Subtarget->is64Bit())
+  if (!Subtarget->is64Bit()) {
     setOperationAction(ISD::INTRINSIC_W_CHAIN, MVT::i64, Custom);
+    setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::i64, Custom);
+  }
 
   // Only custom-lower 64-bit SADDO and friends on 64-bit because we don't
   // handle type legalization for these operations here.
   //
   // FIXME: We really should do custom legalization for addition and
   // subtraction on x86-32 once PR3203 is fixed.  We really can't do much better
   // than generic legalization for 64-bit multiplication-with-overflow, though.
   for (auto VT : { MVT::i8, MVT::i16, MVT::i32, MVT::i64 }) {
static SDValue getTargetVShiftNode(unsigned Opc, SDLoc dl, MVT VT,
   // type as the input type.
   MVT EltVT = VT.getVectorElementType();
   MVT ShVT = MVT::getVectorVT(EltVT, 128/EltVT.getSizeInBits());
 
   ShAmt = DAG.getBitcast(ShVT, ShAmt);
   return DAG.getNode(Opc, dl, VT, SrcOp, ShAmt);
 }
 
+/// \brief Return Mask with the necessary casting or extending
+/// for \p Mask according to \p MaskVT when lowering masking intrinsics
+static SDValue getMaskNode(SDValue Mask, MVT MaskVT,
+                           const X86Subtarget *Subtarget,
+                           SelectionDAG &DAG, SDLoc dl) {
+
+  if (MaskVT.bitsGT(Mask.getSimpleValueType())) {
+    // Mask should be extended
+    Mask = DAG.getNode(ISD::ANY_EXTEND, dl,
+                       MVT::getIntegerVT(MaskVT.getSizeInBits()), Mask);
+  }
+
+  if (Mask.getSimpleValueType() == MVT::i64 && Subtarget->is32Bit()) {
+    assert(MaskVT == MVT::v64i1 && "Unexpected mask VT!");
+    assert(Subtarget->hasBWI() && "Expected AVX512BW target!");
+    // In case 32bit mode, bitcast i64 is illegal, extend/split it.
+    SDValue Lo, Hi;
+    Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, Mask,
+                        DAG.getConstant(0, dl, MVT::i32));
+    Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, Mask,
+                        DAG.getConstant(1, dl, MVT::i32));
+
+    Lo = DAG.getNode(ISD::BITCAST, dl, MVT::v32i1, Lo);
+    Hi = DAG.getNode(ISD::BITCAST, dl, MVT::v32i1, Hi);
+
+    return DAG.getNode(ISD::CONCAT_VECTORS, dl, MVT::v64i1, Hi, Lo);
+
+  } else {
+    MVT BitcastVT = MVT::getVectorVT(MVT::i1,
+                                     Mask.getSimpleValueType().getSizeInBits());
+    // In case when MaskVT equals v2i1 or v4i1, low 2 or 4 elements
+    // are extracted by EXTRACT_SUBVECTOR.
+    return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MaskVT,
+                       DAG.getBitcast(BitcastVT, Mask),
+                       DAG.getIntPtrConstant(0, dl));
+  }
+}
+
 /// \brief Return (and \p Op, \p Mask) for compare instructions or
 /// (vselect \p Mask, \p Op, \p PreservedSrc) for others along with the
 /// necessary casting or extending for \p Mask when lowering masking intrinsics
 static SDValue getVectorMaskingNode(SDValue Op, SDValue Mask,
-                                    SDValue PreservedSrc,
-                                    const X86Subtarget *Subtarget,
-                                    SelectionDAG &DAG) {
-    MVT VT = Op.getSimpleValueType();
-    MVT MaskVT = MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
-    SDValue VMask;
-    unsigned OpcodeSelect = ISD::VSELECT;
-    SDLoc dl(Op);
 
-    if (isAllOnesConstant(Mask))
-      return Op;
 
-    if (MaskVT.bitsGT(Mask.getSimpleValueType())) {
-      MVT newMaskVT = MVT::getIntegerVT(MaskVT.getSizeInBits());
-      VMask = DAG.getBitcast(MaskVT,
-                             DAG.getNode(ISD::ANY_EXTEND, dl, newMaskVT, Mask));
-    } else {
-      MVT BitcastVT = MVT::getVectorVT(MVT::i1,
-                                       Mask.getSimpleValueType().getSizeInBits());
-      // In case when MaskVT equals v2i1 or v4i1, low 2 or 4 elements
-      // are extracted by EXTRACT_SUBVECTOR.
-      VMask = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MaskVT,
-                          DAG.getBitcast(BitcastVT, Mask),
-                          DAG.getIntPtrConstant(0, dl));
-    }
+  SDValue VMask = getMaskNode(Mask, MaskVT, Subtarget, DAG, dl);
 
-    switch (Op.getOpcode()) {
-    default: break;
-    case X86ISD::PCMPEQM:
-    case X86ISD::PCMPGTM:
-    case X86ISD::CMPM:
-    case X86ISD::CMPMU:
-      return DAG.getNode(ISD::AND, dl, VT, Op, VMask);
-    case X86ISD::VFPCLASS:
     case X86ISD::VFPCLASSS:
-      return DAG.getNode(ISD::OR, dl, VT, Op, VMask);
-    case X86ISD::VTRUNC:
-    case X86ISD::VTRUNCS:
-    case X86ISD::VTRUNCUS:
-      // We can't use ISD::VSELECT here because it is not always "Legal"
-      // for the destination type. For example vpmovqb require only AVX512
-      // and vselect that can operate on byte element type require BWI
-      OpcodeSelect = X86ISD::SELECT;
-      break;
-    }
-    if (PreservedSrc.getOpcode() == ISD::UNDEF)
-      PreservedSrc = getZeroVector(VT, Subtarget, DAG, dl);
-    return DAG.getNode(OpcodeSelect, dl, VT, VMask, Op, PreservedSrc);
 }
 
 /// \brief Creates an SDNode for a predicated scalar operation.
 /// \returns (X86vselect \p Mask, \p Op, \p PreservedSrc).
 /// The mask is coming as MVT::i8 and it should be truncated
 /// to MVT::i1 while lowering masking intrinsics.
 /// The main difference between ScalarMaskingNode and VectorMaskingNode is using
 /// "X86select" instead of "vselect". We just can't create the "vselect" node
case BROADCASTM: {
       MVT MaskVT = MVT::getVectorVT(MVT::i1, Mask.getSimpleValueType().getSizeInBits());
       Mask = DAG.getBitcast(MaskVT, Mask);
       return DAG.getNode(IntrData->Opc0, dl, Op.getValueType(), Mask);
     }
     case BLEND: {
       SDValue Mask = Op.getOperand(3);
       MVT VT = Op.getSimpleValueType();
       MVT MaskVT = MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
-      MVT BitcastVT = MVT::getVectorVT(MVT::i1,
-                                       Mask.getSimpleValueType().getSizeInBits());
-      SDLoc dl(Op);
-      SDValue VMask = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MaskVT,
-                                  DAG.getBitcast(BitcastVT, Mask),
-                                  DAG.getIntPtrConstant(0, dl));
+      SDValue VMask = getMaskNode(Mask, MaskVT, Subtarget, DAG, dl);
       return DAG.getNode(IntrData->Opc0, dl, VT, VMask, Op.getOperand(1),
                          Op.getOperand(2));
     }
     default:
       break;
     }
   }
 
case Intrinsic::x86_rdtsc:
                                      Results);
     case Intrinsic::x86_rdtscp:
       return getReadTimeStampCounter(N, dl, X86ISD::RDTSCP_DAG, DAG, Subtarget,
                                      Results);
     case Intrinsic::x86_rdpmc:
       return getReadPerformanceCounter(N, dl, DAG, Subtarget, Results);
     }
   }
+  case ISD::INTRINSIC_WO_CHAIN: {
+    Results.push_back(LowerINTRINSIC_WO_CHAIN(SDValue(N, 0), Subtarget, DAG));
+    return;
+  }
   case ISD::READCYCLECOUNTER: {
     return getReadTimeStampCounter(N, dl, X86ISD::RDTSC_DAG, DAG, Subtarget,
                                    Results);
   }
   case ISD::ATOMIC_CMP_SWAP_WITH_SUCCESS: {
     EVT T = N->getValueType(0);
     assert((T == MVT::i64 || T == MVT::i128) && "can only expand cmpxchg pair");
     bool Regs64bit = T == MVT::i128;

llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=AVX512BW
-
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX512BW
+; RUN: llc < %s -mtriple=i386-unknown-linux-gnu -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX512F-32
 
 define i64 @test_pcmpeq_b(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: test_pcmpeq_b:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_pcmpeq_b:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $12, %esp
+; AVX512F-32-NEXT:  .Ltmp0:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $12, %esp
+; AVX512F-32-NEXT:    retl
   %res = call i64 @llvm.x86.avx512.mask.pcmpeq.b.512(<64 x i8> %a, <64 x i8> %b, i64 -1)
   ret i64 %res
 }
 
 define i64 @test_mask_pcmpeq_b(<64 x i8> %a, <64 x i8> %b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_pcmpeq_b:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_pcmpeq_b:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $12, %esp
+; AVX512F-32-NEXT:  .Ltmp1:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $12, %esp
+; AVX512F-32-NEXT:    retl
   %res = call i64 @llvm.x86.avx512.mask.pcmpeq.b.512(<64 x i8> %a, <64 x i8> %b, i64 %mask)
   ret i64 %res
 }
 
 declare i64 @llvm.x86.avx512.mask.pcmpeq.b.512(<64 x i8>, <64 x i8>, i64)
 
 define i32 @test_pcmpeq_w(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_pcmpeq_w:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_pcmpeq_w:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    retl
   %res = call i32 @llvm.x86.avx512.mask.pcmpeq.w.512(<32 x i16> %a, <32 x i16> %b, i32 -1)
   ret i32 %res
 }
 
 define i32 @test_mask_pcmpeq_w(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_pcmpeq_w:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_pcmpeq_w:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    retl
   %res = call i32 @llvm.x86.avx512.mask.pcmpeq.w.512(<32 x i16> %a, <32 x i16> %b, i32 %mask)
   ret i32 %res
 }
 
 declare i32 @llvm.x86.avx512.mask.pcmpeq.w.512(<32 x i16>, <32 x i16>, i32)
 
 define i64 @test_pcmpgt_b(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: test_pcmpgt_b:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_pcmpgt_b:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $12, %esp
+; AVX512F-32-NEXT:  .Ltmp2:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-32-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $12, %esp
+; AVX512F-32-NEXT:    retl
   %res = call i64 @llvm.x86.avx512.mask.pcmpgt.b.512(<64 x i8> %a, <64 x i8> %b, i64 -1)
   ret i64 %res
 }
 
 define i64 @test_mask_pcmpgt_b(<64 x i8> %a, <64 x i8> %b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_pcmpgt_b:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_pcmpgt_b:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $12, %esp
+; AVX512F-32-NEXT:  .Ltmp3:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $12, %esp
+; AVX512F-32-NEXT:    retl
   %res = call i64 @llvm.x86.avx512.mask.pcmpgt.b.512(<64 x i8> %a, <64 x i8> %b, i64 %mask)
   ret i64 %res
 }
 
 declare i64 @llvm.x86.avx512.mask.pcmpgt.b.512(<64 x i8>, <64 x i8>, i64)
 
 define i32 @test_pcmpgt_w(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_pcmpgt_w:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_pcmpgt_w:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    retl
   %res = call i32 @llvm.x86.avx512.mask.pcmpgt.w.512(<32 x i16> %a, <32 x i16> %b, i32 -1)
   ret i32 %res
 }
 
 define i32 @test_mask_pcmpgt_w(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_pcmpgt_w:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_pcmpgt_w:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    retl
   %res = call i32 @llvm.x86.avx512.mask.pcmpgt.w.512(<32 x i16> %a, <32 x i16> %b, i32 %mask)
   ret i32 %res
 }
 
 declare i32 @llvm.x86.avx512.mask.pcmpgt.w.512(<32 x i16>, <32 x i16>, i32)
 
 define i64 @test_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
 ; AVX512BW-LABEL: test_cmp_b_512:

 ; AVX512BW-NEXT:    addq %rax, %rcx
 ; AVX512BW-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rdx
 ; AVX512BW-NEXT:    addq %rcx, %rdx
 ; AVX512BW-NEXT:    vpcmpordb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rdx, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_cmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $68, %esp
+; AVX512F-32-NEXT:  .Ltmp4:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpltb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpordb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    addl (%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $68, %esp
+; AVX512F-32-NEXT:    retl
   %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
   %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
   %ret1 = add i64 %res0, %res1
   %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
   %ret2 = add i64 %ret1, %res2
   %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
   %ret3 = add i64 %ret2, %res3
   %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)

 ; AVX512BW-NEXT:    addq %rax, %rcx
 ; AVX512BW-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rdx
 ; AVX512BW-NEXT:    addq %rcx, %rdx
 ; AVX512BW-NEXT:    vpcmpordb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rdx, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_cmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $68, %esp
+; AVX512F-32-NEXT:  .Ltmp5:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpltb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpordb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $68, %esp
+; AVX512F-32-NEXT:    retl
   %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
   %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
   %ret1 = add i64 %res0, %res1
   %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
   %ret2 = add i64 %ret1, %res2
   %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
   %ret3 = add i64 %ret2, %res3
   %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)

 ; AVX512BW-NEXT:    addq %rax, %rcx
 ; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rdx
 ; AVX512BW-NEXT:    addq %rcx, %rdx
 ; AVX512BW-NEXT:    vpcmpordub %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rdx, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_ucmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $68, %esp
+; AVX512F-32-NEXT:  .Ltmp6:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:    vpcmpequb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpltub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpordub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    addl (%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $68, %esp
+; AVX512F-32-NEXT:    retl
   %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
   %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
   %ret1 = add i64 %res0, %res1
   %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
   %ret2 = add i64 %ret1, %res2
   %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
   %ret3 = add i64 %ret2, %res3
   %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)

 ; AVX512BW-NEXT:    addq %rax, %rcx
 ; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rdx
 ; AVX512BW-NEXT:    addq %rcx, %rdx
 ; AVX512BW-NEXT:    vpcmpordub %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rdx, %rax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_x86_avx512_ucmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $68, %esp
+; AVX512F-32-NEXT:  .Ltmp7:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpcmpequb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpltub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpordub %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $68, %esp
+; AVX512F-32-NEXT:    retl
   %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
   %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
   %ret1 = add i64 %res0, %res1
   %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
   %ret2 = add i64 %ret1, %res2
   %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
   %ret3 = add i64 %ret2, %res3
   %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)

 ; AVX512BW-NEXT:    addl %eax, %ecx
 ; AVX512BW-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %edx
 ; AVX512BW-NEXT:    addl %ecx, %edx
 ; AVX512BW-NEXT:    vpcmpordw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    addl %edx, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_cmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %ecx, %edx
+; AVX512F-32-NEXT:    vpcmpordw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
   %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
   %ret1 = add i32 %res0, %res1
   %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
   %ret2 = add i32 %ret1, %res2
   %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
   %ret3 = add i32 %ret2, %res3
   %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)

 ; AVX512BW-NEXT:    addl %eax, %ecx
 ; AVX512BW-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %edx
 ; AVX512BW-NEXT:    addl %ecx, %edx
 ; AVX512BW-NEXT:    vpcmpordw %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    addl %edx, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_cmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %ecx, %edx
+; AVX512F-32-NEXT:    vpcmpordw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
   %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
   %ret1 = add i32 %res0, %res1
   %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
   %ret2 = add i32 %ret1, %res2
   %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
   %ret3 = add i32 %ret2, %res3
   %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)

 ; AVX512BW-NEXT:    addl %eax, %ecx
 ; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %edx
 ; AVX512BW-NEXT:    addl %ecx, %edx
 ; AVX512BW-NEXT:    vpcmporduw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    addl %edx, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_ucmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpequw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %ecx, %edx
+; AVX512F-32-NEXT:    vpcmporduw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
   %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
   %ret1 = add i32 %res0, %res1
   %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
   %ret2 = add i32 %ret1, %res2
   %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
   %ret3 = add i32 %ret2, %res3
   %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)

 ; AVX512BW-NEXT:    addl %eax, %ecx
 ; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %edx
 ; AVX512BW-NEXT:    addl %ecx, %edx
 ; AVX512BW-NEXT:    vpcmporduw %zmm1, %zmm0, %k0 {%k1}
 ; AVX512BW-NEXT:    kmovd %k0, %eax
 ; AVX512BW-NEXT:    addl %edx, %eax
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_ucmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpcmpequw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %ecx, %edx
+; AVX512F-32-NEXT:    vpcmporduw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
   %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
   %ret1 = add i32 %res0, %res1
   %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
   %ret2 = add i32 %ret1, %res2
   %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
   %ret3 = add i32 %ret2, %res3
   %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)

 declare <64 x i8> @llvm.x86.avx512.mask.blend.b.512(<64 x i8>, <64 x i8>, i64) nounwind readonly
 
 define <32 x i16> @test_x86_mask_blend_w_512(i32 %mask, <32 x i16> %a1, <32 x i16> %a2) {
 ; AVX512BW-LABEL: test_x86_mask_blend_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpblendmw %zmm1, %zmm0, %zmm0 {%k1}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_x86_mask_blend_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpblendmw %zmm1, %zmm0, %zmm0 {%k1}
+; AVX512F-32-NEXT:    retl
     %res = call <32 x i16> @llvm.x86.avx512.mask.blend.w.512(<32 x i16> %a1, <32 x i16> %a2, i32 %mask) ; <<32 x i16>> [#uses=1]
   ret <32 x i16> %res
 }
 declare <32 x i16> @llvm.x86.avx512.mask.blend.w.512(<32 x i16>, <32 x i16>, i32) nounwind readonly
 
 define <64 x i8> @test_x86_mask_blend_b_512(i64 %a0, <64 x i8> %a1, <64 x i8> %a2) {
 ; AVX512BW-LABEL: test_x86_mask_blend_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpblendmb %zmm1, %zmm0, %zmm0 {%k1}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_x86_mask_blend_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpblendmb %zmm1, %zmm0, %zmm0 {%k1}
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.blend.b.512(<64 x i8> %a1, <64 x i8> %a2, i64 %a0) ; <<64 x i8>> [#uses=1]
   ret <64 x i8> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackssdw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rmb_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rmbk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmbk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packs_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rmbkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackssdw (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi32_rmbkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackssdw (%eax){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.packssdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
 
 define <64 x i8> @test_mask_packs_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packs_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packs_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpacksswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packs_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packs_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rsi, %k1
 ; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packs_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packs_epi16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rsi, %k1
 ; AVX512BW-NEXT:    vpacksswb (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packs_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpacksswb (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
   ret <64 x i8> %res
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.packsswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
 
 
 define <32 x i16> @test_mask_packus_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rrk_512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rrkz_512(<16 x i32> %a, <16 x i32> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rm_512(<16 x i32> %a, <16 x i32>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmk_512(<16 x i32> %a, <16 x i32>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmkz_512(<16 x i32> %a, <16 x i32>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackusdw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <16 x i32>, <16 x i32>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmb_512(<16 x i32> %a, i32* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rmb_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmbk_512(<16 x i32> %a, i32* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rmbk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmbk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_packus_epi32_rmbkz_512(<16 x i32> %a, i32* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi32_rmbkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpackusdw (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi32_rmbkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpackusdw (%eax){1to16}, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %q = load i32, i32* %ptr_b
   %vecinit.i = insertelement <16 x i32> undef, i32 %q, i32 0
   %b = shufflevector <16 x i32> %vecinit.i, <16 x i32> undef, <16 x i32> zeroinitializer
   %res = call <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32> %a, <16 x i32> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.packusdw.512(<16 x i32>, <16 x i32>, <32 x i16>, i32)
 
 define <64 x i8> @test_mask_packus_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packus_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packus_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packus_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 -1)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packus_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <64 x i8> %passThru, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rsi, %k1
 ; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> %passThru, i64 %mask)
   ret <64 x i8> %res
 }
 
 define <64 x i8> @test_mask_packus_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i64 %mask) {
 ; AVX512BW-LABEL: test_mask_packus_epi16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rsi, %k1
 ; AVX512BW-NEXT:    vpackuswb (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_packus_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
   ret <64 x i8> %res
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
 
 define <32 x i16> @test_mask_adds_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpaddsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpaddsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpaddsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpaddsw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddsw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpaddsw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpaddsw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpaddsw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddsw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epi16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpaddsw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddsw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.padds.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16> @test_mask_subs_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsubsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpsubsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpsubsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpsubsw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubsw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsubsw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpsubsw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpsubsw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubsw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epi16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpsubsw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epi16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubsw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.psubs.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16> @test_mask_adds_epu16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpaddusw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpaddusw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epu16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpaddusw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddusw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epu16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpaddusw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddusw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epu16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpaddusw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpaddusw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epu16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpaddusw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddusw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_adds_epu16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_adds_epu16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpaddusw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_adds_epu16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpaddusw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.paddus.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16> @test_mask_subs_epu16_rr_512(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsubusw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpsubusw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epu16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rrk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpsubusw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rrk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubusw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epu16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rrkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpsubusw %zmm1, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rrkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubusw %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epu16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rm_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsubusw (%rdi), %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rm_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpsubusw (%eax), %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epu16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rmk_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpsubusw (%rdi), %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rmk_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubusw (%eax), %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
   ret <32 x i16> %res
 }
 
 define <32 x i16> @test_mask_subs_epu16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
 ; AVX512BW-LABEL: test_mask_subs_epu16_rmkz_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpsubusw (%rdi), %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_subs_epu16_rmkz_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpsubusw (%eax), %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
   ret <32 x i16> %res
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.psubus.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 declare <64 x i8> @llvm.x86.avx512.mask.pmaxs.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pmaxs_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaxs_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpmaxsb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaxsb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaxs_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpmaxsb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaxsb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pmaxs.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pmaxs.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmaxs.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmaxs_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaxs_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmaxsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaxsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaxs_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmaxsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaxsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmaxs.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmaxs.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pmaxu.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pmaxu_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaxu_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpmaxub %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaxub %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaxu_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpmaxub %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaxub %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pmaxu.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pmaxu.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmaxu.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmaxu_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaxu_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmaxuw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaxuw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaxu_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmaxuw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaxuw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmaxu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmaxu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pmins.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pmins_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmins_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpminsb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpminsb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmins_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpminsb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpminsb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pmins.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pmins.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmins.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmins_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmins_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpminsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpminsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmins_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpminsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpminsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmins.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmins.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pminu.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pminu_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pminu_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpminub %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpminub %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pminu_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpminub %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpminub %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pminu.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pminu.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pminu.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pminu_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pminu_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpminuw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpminuw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pminu_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpminuw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpminuw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pminu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pminu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_vpermt2var_hi_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm3
 ; AVX512BW-NEXT:    vpermt2w %zmm2, %zmm0, %zmm3 {%k1}
 ; AVX512BW-NEXT:    vpermt2w %zmm2, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm3
+; AVX512F-32-NEXT:    vpermt2w %zmm2, %zmm0, %zmm3 {%k1}
+; AVX512F-32-NEXT:    vpermt2w %zmm2, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_maskz_vpermt2var_hi_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm3
 ; AVX512BW-NEXT:    vpermt2w %zmm2, %zmm0, %zmm3 {%k1} {z}
 ; AVX512BW-NEXT:    vpermt2w %zmm2, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm3
+; AVX512F-32-NEXT:    vpermt2w %zmm2, %zmm0, %zmm3 {%k1} {z}
+; AVX512F-32-NEXT:    vpermt2w %zmm2, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_vpermi2var_hi_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vmovaps %zmm1, %zmm3
 ; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm0, %zmm3 {%k1}
 ; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vmovaps %zmm1, %zmm3
+; AVX512F-32-NEXT:    vpermi2w %zmm2, %zmm0, %zmm3 {%k1}
+; AVX512F-32-NEXT:    vpermi2w %zmm2, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpaddw %zmm1, %zmm3, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pavg.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pavg_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pavg_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpavgb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpavgb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pavg_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpavgb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpavgb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pavg.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pavg.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pavg.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pavg_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pavg_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpavgw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpavgw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pavg_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpavgw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpavgw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pavg.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pavg.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pshuf_b_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pshuf_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpshufb %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpshufb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pshuf_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpshufb %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpshufb %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pshuf.b.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pabs_w_512(<32 x i16> %x0, <32 x i16> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pabs_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpabsw %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vpabsw %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pabs_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpabsw %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vpabsw %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 %x2)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pabs.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_pabs_b_512(<64 x i8> %x0, <64 x i8> %x1, i64 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pabs_b_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpabsb %zmm0, %zmm1 {%k1}
 ; AVX512BW-NEXT:    vpabsb %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pabs_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpabsb %zmm0, %zmm1 {%k1}
+; AVX512F-32-NEXT:    vpabsb %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8> %x0, <64 x i8> %x1, i64 %x2)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.pabs.b.512(<64 x i8> %x0, <64 x i8> %x1, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmulhu.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmulhu_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmulhu_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmulhu_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmulhu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmulhu.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmulh.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmulh_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmulh_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmulhw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmulhw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmulh_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmulhw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmulhw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmulh.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmulh.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmulhr_sw_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmulhr_sw_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmulhrsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmulhrsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmulhr_sw_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmulhrsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmulhrsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16>, <32 x i8>, i32)
 
 define <32 x i8>@test_int_x86_avx512_mask_pmov_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmov_wb_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm1 {%k1}
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm2 {%k1} {z}
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmov_wb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmovwb %zmm0, %ymm1 {%k1}
+; AVX512F-32-NEXT:    vpmovwb %zmm0, %ymm2 {%k1} {z}
+; AVX512F-32-NEXT:    vpmovwb %zmm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512F-32-NEXT:    retl
     %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
     %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
     %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
     %res3 = add <32 x i8> %res0, %res1
     %res4 = add <32 x i8> %res3, %res2
     ret <32 x i8> %res4
 }
 
 declare void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16>, i32)
 
 define void @test_int_x86_avx512_mask_pmov_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpmovwb %zmm0, (%rdi)
 ; AVX512BW-NEXT:    vpmovwb %zmm0, (%rdi) {%k1}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    vpmovwb %zmm0, (%eax)
+; AVX512F-32-NEXT:    vpmovwb %zmm0, (%eax) {%k1}
+; AVX512F-32-NEXT:    retl
     call void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
     call void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
     ret void
 }
 
 declare <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16>, <32 x i8>, i32)
 
 define <32 x i8>@test_int_x86_avx512_mask_pmovs_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmovs_wb_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmovswb %zmm0, %ymm1 {%k1}
 ; AVX512BW-NEXT:    vpmovswb %zmm0, %ymm2 {%k1} {z}
 ; AVX512BW-NEXT:    vpmovswb %zmm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmovs_wb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmovswb %zmm0, %ymm1 {%k1}
+; AVX512F-32-NEXT:    vpmovswb %zmm0, %ymm2 {%k1} {z}
+; AVX512F-32-NEXT:    vpmovswb %zmm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512F-32-NEXT:    retl
     %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
     %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
     %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
     %res3 = add <32 x i8> %res0, %res1
     %res4 = add <32 x i8> %res3, %res2
     ret <32 x i8> %res4
 }
 
 declare void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16>, i32)
 
 define void @test_int_x86_avx512_mask_pmovs_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpmovswb %zmm0, (%rdi)
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpmovswb %zmm0, (%rdi) {%k1}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    vpmovswb %zmm0, (%ecx)
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovswb %zmm0, (%ecx) {%k1}
+; AVX512F-32-NEXT:    retl
     call void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
     call void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
     ret void
 }
 
 declare <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16>, <32 x i8>, i32)
 
 define <32 x i8>@test_int_x86_avx512_mask_pmovus_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmovus_wb_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmovuswb %zmm0, %ymm1 {%k1}
 ; AVX512BW-NEXT:    vpmovuswb %zmm0, %ymm2 {%k1} {z}
 ; AVX512BW-NEXT:    vpmovuswb %zmm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmovus_wb_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmovuswb %zmm0, %ymm1 {%k1}
+; AVX512F-32-NEXT:    vpmovuswb %zmm0, %ymm2 {%k1} {z}
+; AVX512F-32-NEXT:    vpmovuswb %zmm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512F-32-NEXT:    retl
     %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
     %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
     %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
     %res3 = add <32 x i8> %res0, %res1
     %res4 = add <32 x i8> %res3, %res2
     ret <32 x i8> %res4
 }
 
 declare void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16>, i32)
 
 define void @test_int_x86_avx512_mask_pmovus_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpmovuswb %zmm0, (%rdi)
 ; AVX512BW-NEXT:    kmovd %esi, %k1
 ; AVX512BW-NEXT:    vpmovuswb %zmm0, (%rdi) {%k1}
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    vpmovuswb %zmm0, (%ecx)
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovuswb %zmm0, (%ecx) {%k1}
+; AVX512F-32-NEXT:    retl
     call void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
     call void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
     ret void
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pmaddubs.w.512(<64 x i8>, <64 x i8>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_pmaddubs_w_512(<64 x i8> %x0, <64 x i8> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaddubs_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaddubs_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.pmaddubs.w.512(<64 x i8> %x0, <64 x i8> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.pmaddubs.w.512(<64 x i8> %x0, <64 x i8> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <16 x i32> @llvm.x86.avx512.mask.pmaddw.d.512(<32 x i16>, <32 x i16>, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_pmaddw_d_512(<32 x i16> %x0, <32 x i16> %x1, <16 x i32> %x2, i16 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_pmaddw_d_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovw %edi, %k1
 ; AVX512BW-NEXT:    vpmaddwd %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpmaddwd %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddd %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_pmaddw_d_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpmaddwd %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpmaddwd %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddd %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <16 x i32> @llvm.x86.avx512.mask.pmaddw.d.512(<32 x i16> %x0, <32 x i16> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.pmaddw.d.512(<32 x i16> %x0, <32 x i16> %x1, <16 x i32> %x2, i16 -1)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.punpckhb.w.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_punpckhb_w_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_punpckhb_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm2 = zmm2[8],k1[8],zmm2[9],k1[9],zmm2[10],k1[10],zmm2[11],k1[11],zmm2[12],k1[12],zmm2[13],k1[13],zmm2[14],k1[14],zmm2[15],k1[15],zmm2[24],k1[24],zmm2[25],k1[25],zmm2[26],k1[26],zmm2[27],k1[27],zmm2[28],k1[28],zmm2[29],k1[29],zmm2[30],k1[30],zmm2[31],k1[31],zmm2[40],k1[40],zmm2[41],k1[41],zmm2[42],k1[42],zmm2[43],k1[43],zmm2[44],k1[44],zmm2[45],k1[45],zmm2[46],k1[46],zmm2[47],k1[47],zmm2[56],k1[56],zmm2[57],k1[57],zmm2[58],k1[58],zmm2[59],k1[59],zmm2[60],k1[60],zmm2[61],k1[61],zmm2[62],k1[62],zmm2[63],k1[63]
 ; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm0 = zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[12],zmm1[12],zmm0[13],zmm1[13],zmm0[14],zmm1[14],zmm0[15],zmm1[15],zmm0[24],zmm1[24],zmm0[25],zmm1[25],zmm0[26],zmm1[26],zmm0[27],zmm1[27],zmm0[28],zmm1[28],zmm0[29],zmm1[29],zmm0[30],zmm1[30],zmm0[31],zmm1[31],zmm0[40],zmm1[40],zmm0[41],zmm1[41],zmm0[42],zmm1[42],zmm0[43],zmm1[43],zmm0[44],zmm1[44],zmm0[45],zmm1[45],zmm0[46],zmm1[46],zmm0[47],zmm1[47],zmm0[56],zmm1[56],zmm0[57],zmm1[57],zmm0[58],zmm1[58],zmm0[59],zmm1[59],zmm0[60],zmm1[60],zmm0[61],zmm1[61],zmm0[62],zmm1[62],zmm0[63],zmm1[63]
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_punpckhb_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpunpckhbw {{.*#+}} zmm2 = zmm2[8],k1[8],zmm2[9],k1[9],zmm2[10],k1[10],zmm2[11],k1[11],zmm2[12],k1[12],zmm2[13],k1[13],zmm2[14],k1[14],zmm2[15],k1[15],zmm2[24],k1[24],zmm2[25],k1[25],zmm2[26],k1[26],zmm2[27],k1[27],zmm2[28],k1[28],zmm2[29],k1[29],zmm2[30],k1[30],zmm2[31],k1[31],zmm2[40],k1[40],zmm2[41],k1[41],zmm2[42],k1[42],zmm2[43],k1[43],zmm2[44],k1[44],zmm2[45],k1[45],zmm2[46],k1[46],zmm2[47],k1[47],zmm2[56],k1[56],zmm2[57],k1[57],zmm2[58],k1[58],zmm2[59],k1[59],zmm2[60],k1[60],zmm2[61],k1[61],zmm2[62],k1[62],zmm2[63],k1[63]
+; AVX512F-32-NEXT:    vpunpckhbw {{.*#+}} zmm0 = zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[12],zmm1[12],zmm0[13],zmm1[13],zmm0[14],zmm1[14],zmm0[15],zmm1[15],zmm0[24],zmm1[24],zmm0[25],zmm1[25],zmm0[26],zmm1[26],zmm0[27],zmm1[27],zmm0[28],zmm1[28],zmm0[29],zmm1[29],zmm0[30],zmm1[30],zmm0[31],zmm1[31],zmm0[40],zmm1[40],zmm0[41],zmm1[41],zmm0[42],zmm1[42],zmm0[43],zmm1[43],zmm0[44],zmm1[44],zmm0[45],zmm1[45],zmm0[46],zmm1[46],zmm0[47],zmm1[47],zmm0[56],zmm1[56],zmm0[57],zmm1[57],zmm0[58],zmm1[58],zmm0[59],zmm1[59],zmm0[60],zmm1[60],zmm0[61],zmm1[61],zmm0[62],zmm1[62],zmm0[63],zmm1[63]
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.punpckhb.w.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.punpckhb.w.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.punpcklb.w.512(<64 x i8>, <64 x i8>, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_punpcklb_w_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_punpcklb_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm2 = zmm2[0],k1[0],zmm2[1],k1[1],zmm2[2],k1[2],zmm2[3],k1[3],zmm2[4],k1[4],zmm2[5],k1[5],zmm2[6],k1[6],zmm2[7],k1[7],zmm2[16],k1[16],zmm2[17],k1[17],zmm2[18],k1[18],zmm2[19],k1[19],zmm2[20],k1[20],zmm2[21],k1[21],zmm2[22],k1[22],zmm2[23],k1[23],zmm2[32],k1[32],zmm2[33],k1[33],zmm2[34],k1[34],zmm2[35],k1[35],zmm2[36],k1[36],zmm2[37],k1[37],zmm2[38],k1[38],zmm2[39],k1[39],zmm2[48],k1[48],zmm2[49],k1[49],zmm2[50],k1[50],zmm2[51],k1[51],zmm2[52],k1[52],zmm2[53],k1[53],zmm2[54],k1[54],zmm2[55],k1[55]
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[16],zmm1[16],zmm0[17],zmm1[17],zmm0[18],zmm1[18],zmm0[19],zmm1[19],zmm0[20],zmm1[20],zmm0[21],zmm1[21],zmm0[22],zmm1[22],zmm0[23],zmm1[23],zmm0[32],zmm1[32],zmm0[33],zmm1[33],zmm0[34],zmm1[34],zmm0[35],zmm1[35],zmm0[36],zmm1[36],zmm0[37],zmm1[37],zmm0[38],zmm1[38],zmm0[39],zmm1[39],zmm0[48],zmm1[48],zmm0[49],zmm1[49],zmm0[50],zmm1[50],zmm0[51],zmm1[51],zmm0[52],zmm1[52],zmm0[53],zmm1[53],zmm0[54],zmm1[54],zmm0[55],zmm1[55]
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_punpcklb_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpunpcklbw {{.*#+}} zmm2 = zmm2[0],k1[0],zmm2[1],k1[1],zmm2[2],k1[2],zmm2[3],k1[3],zmm2[4],k1[4],zmm2[5],k1[5],zmm2[6],k1[6],zmm2[7],k1[7],zmm2[16],k1[16],zmm2[17],k1[17],zmm2[18],k1[18],zmm2[19],k1[19],zmm2[20],k1[20],zmm2[21],k1[21],zmm2[22],k1[22],zmm2[23],k1[23],zmm2[32],k1[32],zmm2[33],k1[33],zmm2[34],k1[34],zmm2[35],k1[35],zmm2[36],k1[36],zmm2[37],k1[37],zmm2[38],k1[38],zmm2[39],k1[39],zmm2[48],k1[48],zmm2[49],k1[49],zmm2[50],k1[50],zmm2[51],k1[51],zmm2[52],k1[52],zmm2[53],k1[53],zmm2[54],k1[54],zmm2[55],k1[55]
+; AVX512F-32-NEXT:    vpunpcklbw {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[16],zmm1[16],zmm0[17],zmm1[17],zmm0[18],zmm1[18],zmm0[19],zmm1[19],zmm0[20],zmm1[20],zmm0[21],zmm1[21],zmm0[22],zmm1[22],zmm0[23],zmm1[23],zmm0[32],zmm1[32],zmm0[33],zmm1[33],zmm0[34],zmm1[34],zmm0[35],zmm1[35],zmm0[36],zmm1[36],zmm0[37],zmm1[37],zmm0[38],zmm1[38],zmm0[39],zmm1[39],zmm0[48],zmm1[48],zmm0[49],zmm1[49],zmm0[50],zmm1[50],zmm0[51],zmm1[51],zmm0[52],zmm1[52],zmm0[53],zmm1[53],zmm0[54],zmm1[54],zmm0[55],zmm1[55]
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.punpcklb.w.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 %x3)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.punpcklb.w.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2, i64 -1)
   %res2 = add <64 x i8> %res, %res1
   ret <64 x i8> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.punpckhw.d.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_punpckhw_d_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_punpckhw_d_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpunpckhwd {{.*#+}} zmm2 = zmm2[4],k1[4],zmm2[5],k1[5],zmm2[6],k1[6],zmm2[7],k1[7],zmm2[12],k1[12],zmm2[13],k1[13],zmm2[14],k1[14],zmm2[15],k1[15],zmm2[20],k1[20],zmm2[21],k1[21],zmm2[22],k1[22],zmm2[23],k1[23],zmm2[28],k1[28],zmm2[29],k1[29],zmm2[30],k1[30],zmm2[31],k1[31]
 ; AVX512BW-NEXT:    vpunpckhwd {{.*#+}} zmm0 = zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[12],zmm1[12],zmm0[13],zmm1[13],zmm0[14],zmm1[14],zmm0[15],zmm1[15],zmm0[20],zmm1[20],zmm0[21],zmm1[21],zmm0[22],zmm1[22],zmm0[23],zmm1[23],zmm0[28],zmm1[28],zmm0[29],zmm1[29],zmm0[30],zmm1[30],zmm0[31],zmm1[31]
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_punpckhw_d_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpunpckhwd {{.*#+}} zmm2 = zmm2[4],k1[4],zmm2[5],k1[5],zmm2[6],k1[6],zmm2[7],k1[7],zmm2[12],k1[12],zmm2[13],k1[13],zmm2[14],k1[14],zmm2[15],k1[15],zmm2[20],k1[20],zmm2[21],k1[21],zmm2[22],k1[22],zmm2[23],k1[23],zmm2[28],k1[28],zmm2[29],k1[29],zmm2[30],k1[30],zmm2[31],k1[31]
+; AVX512F-32-NEXT:    vpunpckhwd {{.*#+}} zmm0 = zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[12],zmm1[12],zmm0[13],zmm1[13],zmm0[14],zmm1[14],zmm0[15],zmm1[15],zmm0[20],zmm1[20],zmm0[21],zmm1[21],zmm0[22],zmm1[22],zmm0[23],zmm1[23],zmm0[28],zmm1[28],zmm0[29],zmm1[29],zmm0[30],zmm1[30],zmm0[31],zmm1[31]
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.punpckhw.d.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.punpckhw.d.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.punpcklw.d.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_punpcklw_d_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_punpcklw_d_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vpunpcklwd {{.*#+}} zmm2 = zmm2[0],k1[0],zmm2[1],k1[1],zmm2[2],k1[2],zmm2[3],k1[3],zmm2[8],k1[8],zmm2[9],k1[9],zmm2[10],k1[10],zmm2[11],k1[11],zmm2[16],k1[16],zmm2[17],k1[17],zmm2[18],k1[18],zmm2[19],k1[19],zmm2[24],k1[24],zmm2[25],k1[25],zmm2[26],k1[26],zmm2[27],k1[27]
 ; AVX512BW-NEXT:    vpunpcklwd {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[16],zmm1[16],zmm0[17],zmm1[17],zmm0[18],zmm1[18],zmm0[19],zmm1[19],zmm0[24],zmm1[24],zmm0[25],zmm1[25],zmm0[26],zmm1[26],zmm0[27],zmm1[27]
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_punpcklw_d_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vpunpcklwd {{.*#+}} zmm2 = zmm2[0],k1[0],zmm2[1],k1[1],zmm2[2],k1[2],zmm2[3],k1[3],zmm2[8],k1[8],zmm2[9],k1[9],zmm2[10],k1[10],zmm2[11],k1[11],zmm2[16],k1[16],zmm2[17],k1[17],zmm2[18],k1[18],zmm2[19],k1[19],zmm2[24],k1[24],zmm2[25],k1[25],zmm2[26],k1[26],zmm2[27],k1[27]
+; AVX512F-32-NEXT:    vpunpcklwd {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[16],zmm1[16],zmm0[17],zmm1[17],zmm0[18],zmm1[18],zmm0[19],zmm1[19],zmm0[24],zmm1[24],zmm0[25],zmm1[25],zmm0[26],zmm1[26],zmm0[27],zmm1[27]
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.punpcklw.d.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.punpcklw.d.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 -1)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.palignr.512(<64 x i8>, <64 x i8>, i32, <64 x i8>, i64)
 
 define <64 x i8>@test_int_x86_avx512_mask_palignr_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x3, i64 %x4) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_palignr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovq %rdi, %k1
 ; AVX512BW-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm3 {%k1} {z}
 ; AVX512BW-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddb %zmm3, %zmm2, %zmm1
 ; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_palignr_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm3
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
+; AVX512F-32-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vpalignr $2, %zmm1, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-32-NEXT:    vpaddb %zmm0, %zmm2, %zmm0
+; AVX512F-32-NEXT:    vpaddb %zmm3, %zmm0, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <64 x i8> @llvm.x86.avx512.mask.palignr.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <64 x i8> %x3, i64 %x4)
   %res1 = call <64 x i8> @llvm.x86.avx512.mask.palignr.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <64 x i8> zeroinitializer, i64 %x4)
   %res2 = call <64 x i8> @llvm.x86.avx512.mask.palignr.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <64 x i8> %x3, i64 -1)
   %res3 = add <64 x i8> %res, %res1
   %res4 = add <64 x i8> %res3, %res2
   ret <64 x i8> %res4
 }
 
 declare <32 x i16> @llvm.x86.avx512.mask.dbpsadbw.512(<64 x i8>, <64 x i8>, i32, <32 x i16>, i32)
 
 define <32 x i16>@test_int_x86_avx512_mask_dbpsadbw_512(<64 x i8> %x0, <64 x i8> %x1, <32 x i16> %x3, i32 %x4) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    kmovd %edi, %k1
 ; AVX512BW-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1}
 ; AVX512BW-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm3 {%k1} {z}
 ; AVX512BW-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddw %zmm3, %zmm2, %zmm1
 ; AVX512BW-NEXT:    vpaddw %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
+; AVX512F-32-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1}
+; AVX512F-32-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm3 {%k1} {z}
+; AVX512F-32-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddw %zmm3, %zmm2, %zmm1
+; AVX512F-32-NEXT:    vpaddw %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <32 x i16> @llvm.x86.avx512.mask.dbpsadbw.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <32 x i16> %x3, i32 %x4)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.dbpsadbw.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <32 x i16> zeroinitializer, i32 %x4)
   %res2 = call <32 x i16> @llvm.x86.avx512.mask.dbpsadbw.512(<64 x i8> %x0, <64 x i8> %x1, i32 2, <32 x i16> %x3, i32 -1)
   %res3 = add <32 x i16> %res, %res1
   %res4 = add <32 x i16> %res3, %res2
   ret <32 x i16> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.psll.dq.512(<8 x i64>, i32)
 
 define <8 x i64>@test_int_x86_avx512_mask_psll_dq_512(<8 x i64> %x0) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_psll_dq_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpslldq $8, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpslldq $4, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_psll_dq_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpslldq $8, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpslldq $4, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <8 x i64> @llvm.x86.avx512.psll.dq.512(<8 x i64> %x0, i32 8)
   %res1 = call <8 x i64> @llvm.x86.avx512.psll.dq.512(<8 x i64> %x0, i32 4)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 
 declare <8 x i64> @llvm.x86.avx512.psrl.dq.512(<8 x i64>, i32)
 
 define <8 x i64>@test_int_x86_avx512_mask_psrl_dq_512(<8 x i64> %x0) {
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_psrl_dq_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsrldq $8, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpsrldq $4, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_psrl_dq_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpsrldq $8, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpsrldq $4, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call <8 x i64> @llvm.x86.avx512.psrl.dq.512(<8 x i64> %x0, i32 8)
   %res1 = call <8 x i64> @llvm.x86.avx512.psrl.dq.512(<8 x i64> %x0, i32 4)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 declare  <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8>, <64 x i8>)
 
 define  <8 x i64>@test_int_x86_avx512_mask_psadb_w_512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2){
 ; AVX512BW-LABEL: test_int_x86_avx512_mask_psadb_w_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm1
 ; AVX512BW-NEXT:    vpsadbw %zmm2, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_int_x86_avx512_mask_psadb_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpsadbw %zmm1, %zmm0, %zmm1
+; AVX512F-32-NEXT:    vpsadbw %zmm2, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    retl
   %res = call  <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8> %x0, <64 x i8> %x1)
   %res1 = call  <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8> %x0, <64 x i8> %x2)
   %res2 = add  <8 x i64> %res, %res1
   ret  <8 x i64> %res2
 }