[SDAG] enable binop identity constant folds for add

[xbolva00]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

2148

You can use Constantexpr::getBinOpIdentity to check if identity constant.

[LuoYuanke]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

2148

Sorry, this is SDNode and it seems there is no getBinOpIdentity() API.
BTW, there is some regression on this patch.

1. When the select can be combined with its operands, we don't need to invert the select folding. See below example.

; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi2,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
define <16 x i16> @test_int_x86_avx512_mask_vpshldv_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16>* %x2p, <16 x i16> %x4, i16 %x3) {
  %x2 = load <16 x i16>, <16 x i16>* %x2p
  %1 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x0
  %4 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x4)
  %5 = bitcast i16 %x3 to <16 x i1>
  %6 = select <16 x i1> %5, <16 x i16> %4, <16 x i16> zeroinitializer
  %res3 = add <16 x i16> %3, %6
  ret <16 x i16> %res3
}

declare <16 x i16> @llvm.fshl.v16i16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x4)

2. The freeze node seems to prevent the sub combine for below case.

define <4 x i32> @test_srem_allones(<4 x i32> %X) nounwind {
  %srem = srem <4 x i32> %X, <i32 4294967295, i32 4294967295, i32 4294967295, i32 4294967295>
  %cmp = icmp eq <4 x i32> %srem, <i32 0, i32 0, i32 0, i32 0>
  %ret = zext <4 x i1> %cmp to <4 x i32>
  ret <4 x i32> %ret
}

[RKSimon]

llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll

695 ↗

(On Diff #408232)

Any chance you can track down the missing combine please?

The ISD::SUB should definitely fold away, not sure if the ISD::XOR is a zeroinitializer or not - but X86ISD::PCMPEQ will fold to all ones if the inputs are equal. And we should have constant folding for X86ISD::VSRLI

[RKSimon]

llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll

695 ↗

(On Diff #408232)

Sorry - missed your reply above!

[LuoYuanke]

Fix regression for sub(freeze(x), x).

[LuoYuanke]

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

4241

This vmovdqa64 is emitted because the function need to return value by zmm0. Not sure if it is a regression.

[xbolva00]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

3333

Some general solution? FREEZE should be dropped much sooner, no?

[LuoYuanke]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

3333

It is not dropped soon, because compiler can't guarantee it is NOT undef or poison value.

13469 SDValue DAGCombiner::visitFREEZE(SDNode *N) {
13470   SDValue N0 = N->getOperand(0);
13471
13472   if (DAG.isGuaranteedNotToBeUndefOrPoison(N0, /*PoisonOnly*/ false))
13473     return N0;
13474
13475   return SDValue();
13476 }

The freeze node live until instruction selection.

ISEL: Starting selection on root node: t40: v4i32 = freeze t2

[RKSimon]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

3336

Is there anyway we can reduce the scope of this initially, its likely that losing freeze like this might have other effects - if we're just after the (sub x, x) -> 0 fold, maybe create a peekThroughFreeze helper:

if (peekThroughFreeze(N0) == peekThroughFreeze(N1))

return tryFoldToZero(DL, TLI, VT, DAG, LegalOperations);

[RKSimon]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

2148

You can use Constantexpr::getBinOpIdentity to check if identity constant.

We already have SelectionDAG::getNeutralElement - I wonder if adding a SelectionDAG::isNeutralElement helper sibling would be useful?

[LuoYuanke]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

2148

I prefer to inverting the operation one by one, so that the patch can be small. After all the operators are inverted, we can refactor the code by using isNeutralElement() and getNeutralElement(). What do you think?

3336

Good suggestion. I'll apply your idea. Thanks.

[LuoYuanke]

Address Simon's comments.

[LuoYuanke]

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

4241

It seems fold select to its previous operands (psrl) is better, because the add operands is communitive so there is more chance to meet the hint (return register) of register allocator.

[LuoYuanke]

llvm/test/CodeGen/X86/avx512vnni-intrinsics-upgrade.ll

66

This can be improved in RA by evict the previous assigned physical register (zmm0) with below patch, but there is some risk on performance regression, because we change the general RA evicting rule. If anyone concern about this additional vmovdqa64, I can separate sub from add in the patch and we may submit sub patch first.

diff --git a/llvm/lib/CodeGen/RegAllocEvictionAdvisor.cpp b/llvm/lib/CodeGen/RegAllocEvictionAdvisor.cpp
index 718e12e5d602..863394fffeb6 100644
--- a/llvm/lib/CodeGen/RegAllocEvictionAdvisor.cpp
+++ b/llvm/lib/CodeGen/RegAllocEvictionAdvisor.cpp
@@ -168,6 +168,7 @@ bool DefaultEvictionAdvisor::canEvictHintInterference(
     const SmallVirtRegSet &FixedRegisters) const {
   EvictionCost MaxCost;
   MaxCost.setBrokenHints(1);
+  MaxCost.MaxWeight = VirtReg.weight();
   return canEvictInterferenceBasedOnCost(VirtReg, PhysReg, true, MaxCost,
                                          FixedRegisters);
 }

[RKSimon]

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

4241

These adds were just used for simplicity to make the result dependent on all 3 intrinsics.

We'd avoid all of the intrinsics-upgrade changes if we just changed these add ops to something else, preferably something that we're not going to add to foldSelectWithIdentityConstant in the future.

Alternatively we split these tests into the 3 normal / {k} / {k}{z} variants

[xbolva00]

llvm/test/CodeGen/X86/avx512vnni-intrinsics-upgrade.ll

66

In any case, Consider posting this patch for RA on Phabricator.

[RKSimon]

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

4241

@LuoYuanke Something that might work is to return a { <8 x i64>, <8 x i64>, <8 x i64> } structure : https://gcc.godbolt.org/z/39ahrqM7E

define { <8 x i64>, <8 x i64>, <8 x i64> } @test_int_x86_avx512_mask_psrl_qi_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
  %res = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 4, <8 x i64> %x2, i8 %x3)
  %res1 = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
  %res2 = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 6, <8 x i64> zeroinitializer, i8 %x3)

  %r0 = insertvalue { <8 x i64>, <8 x i64>, <8 x i64> } poison, <8 x i64> %res, 0
  %r1 = insertvalue { <8 x i64>, <8 x i64>, <8 x i64> } %r0, <8 x i64> %res1, 1
  %r2 = insertvalue { <8 x i64>, <8 x i64>, <8 x i64> } %r1, <8 x i64> %res2, 2
  ret { <8 x i64>, <8 x i64>, <8 x i64> } %r2
}
declare <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64>, i32, <8 x i64>, i8)

test_int_x86_avx512_mask_psrl_qi_512:   # @test_int_x86_avx512_mask_psrl_qi_512
        vmovdqa64       %zmm1, %zmm3            # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd9]
        kmovw   %esi, %k1                       # encoding: [0xc5,0xf8,0x92,0xce]
        vpsrlq  $4, %zmm0, %zmm3 {%k1}          # encoding: [0x62,0xf1,0xe5,0x49,0x73,0xd0,0x04]
        vpsrlq  $5, %zmm0, %zmm1                # encoding: [0x62,0xf1,0xf5,0x48,0x73,0xd0,0x05]
        vpsrlq  $6, %zmm0, %zmm2 {%k1} {z}      # encoding: [0x62,0xf1,0xed,0xc9,0x73,0xd0,0x06]
        vmovdqa64       %zmm3, %zmm0            # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
        retq                                    # encoding: [0xc3]

[RKSimon]

@LuoYuanke Please can you rebase and add test coverage for 'add+select' to vector-bo-select.ll? I've updated some of the intrinsic tests to avoid the issue so these shouldn't show up any more, I'll finish this cleanup when I have a free moment.

[LuoYuanke]

Rebase.

[LuoYuanke]

In D119654#3362458, @RKSimon wrote:

@LuoYuanke Please can you rebase and add test coverage for 'add+select' to vector-bo-select.ll? I've updated some of the intrinsic tests to avoid the issue so these shouldn't show up any more, I'll finish this cleanup when I have a free moment.

@RKSimon, it is very kind of you. Thanks! Let me follow your approach and update other test cases.

[LuoYuanke]

Rebase

[RKSimon]

@LuoYuanke please can you rebase this?

[LuoYuanke]

Rebase.

[LuoYuanke]

Rebase.

[RKSimon]

@LuoYuanke rebase? I think this might be ready now

[LuoYuanke]

Rebase.

[LuoYuanke]

In D119654#3394240, @RKSimon wrote:

@LuoYuanke rebase? I think this might be ready now

Thanks, Simon. Yes, I think it is ready now.

[RKSimon]

LGTM - cheers!

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

if (ConstantFPSDNode *C = isConstOrConstSplatFP(V)) {
         return C->isZero() && C->isNegative();
       case ISD::FSUB: // X - 0.0 --> X
         return C->isZero() && !C->isNegative();
       case ISD::FMUL: // X * 1.0 --> X
       case ISD::FDIV: // X / 1.0 --> X
         return C->isExactlyValue(1.0);
       }
     }
+    if (ConstantSDNode *C = isConstOrConstSplat(V)) {
+      switch (Opcode) {
+      case ISD::ADD: // X + 0 --> X
+      case ISD::SUB: // X - 0 --> X
+        return C->isZero();
+      }
+    }
     return false;
   };
 
   // This transform increases uses of N0, so freeze it to be safe.
   // binop N0, (vselect Cond, IDC, FVal) --> vselect Cond, N0, (binop N0, FVal)
   if (isIdentityConstantForOpcode(Opcode, TVal)) {
     SDValue F0 = DAG.getFreeze(N0);
+
     SDValue NewBO = DAG.getNode(Opcode, SDLoc(N), VT, F0, FVal, N->getFlags());
     return DAG.getSelect(SDLoc(N), VT, Cond, F0, NewBO);
   }
   // binop N0, (vselect Cond, TVal, IDC) --> vselect Cond, (binop N0, TVal), N0
   if (isIdentityConstantForOpcode(Opcode, FVal)) {
     SDValue F0 = DAG.getFreeze(N0);
+
     SDValue NewBO = DAG.getNode(Opcode, SDLoc(N), VT, F0, TVal, N->getFlags());
     return DAG.getSelect(SDLoc(N), VT, Cond, NewBO, F0);
   }
 
   return SDValue();
 }
 
 SDValue DAGCombiner::foldBinOpIntoSelect(SDNode *BO) {

 }
 
 SDValue DAGCombiner::visitSUB(SDNode *N) {
   SDValue N0 = N->getOperand(0);
   SDValue N1 = N->getOperand(1);
   EVT VT = N0.getValueType();
   SDLoc DL(N);
 
+  if (N0->getOpcode() == ISD::FREEZE && N0.hasOneUse())
+    N0 = N0->getOperand(0);
+  if (N1->getOpcode() == ISD::FREEZE && N0.hasOneUse())
+    N1 = N1->getOperand(0);
+
   // fold (sub x, x) -> 0
   // FIXME: Refactor this and xor and other similar operations together.
   if (N0 == N1)
     return tryFoldToZero(DL, TLI, VT, DAG, LegalOperations);
 
   // fold (sub c1, c2) -> c3
   if (SDValue C = DAG.FoldConstantArithmetic(ISD::SUB, DL, VT, {N0, N1}))
     return C;

llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

 }
 
 define <16 x i32>@test_int_x86_avx512_mask_pbroadcastd_gpr_512(i32 %x0, <16 x i32> %x1, i16 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcastd_gpr_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %zmm1 ## encoding: [0x62,0xf2,0x7d,0x48,0x58,0x4c,0x24,0x01]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0xc1]
-; X86-NEXT:    vmovdqa32 %zmm1, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0x6f,0xd1]
-; X86-NEXT:    vpaddd %zmm2, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfe,0xc2]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm1, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfe,0xc1]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcastd_gpr_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    vpbroadcastd %edi, %zmm1 ## encoding: [0x62,0xf2,0x7d,0x48,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastd %edi, %zmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x49,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastd %edi, %zmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xc9,0x7c,0xd7]
-; X64-NEXT:    vpaddd %zmm2, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfe,0xc2]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfe,0xc1]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.512(i32 %x0, <16 x i32> %x1, i16 -1)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.512(i32 %x0, <16 x i32> %x1, i16 %mask)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.512(i32 %x0, <16 x i32> zeroinitializer, i16 %mask)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res2, %res3
   ret <16 x i32> %res4
 }
 declare <16 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.512(i32, <16 x i32>, i16)
 
 
 define <8 x i64>@test_int_x86_avx512_mask_pbroadcastq_gpr_512(i64 %x0, <8 x i64> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcastq_gpr_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    vpbroadcastq {{[0-9]+}}(%esp), %zmm1 ## encoding: [0x62,0xf2,0xfd,0x48,0x59,0x8c,0x24,0x04,0x00,0x00,0x00]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x0c]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x6f,0xc1]
-; X86-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0x6f,0xd1]
-; X86-NEXT:    vpaddq %zmm2, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0xd4,0xc2]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm1, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xd4,0xc1]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcastq_gpr_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    vpbroadcastq %rdi, %zmm1 ## encoding: [0x62,0xf2,0xfd,0x48,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastq %rdi, %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastq %rdi, %zmm2 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xc9,0x7c,0xd7]
-; X64-NEXT:    vpaddq %zmm2, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0xd4,0xc2]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xd4,0xc1]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.512(i64 %x0, <8 x i64> %x1,i8 -1)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.512(i64 %x0, <8 x i64> %x1,i8 %mask)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.512(i64 %x0, <8 x i64> zeroinitializer,i8 %mask)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res2, %res3
   ret <8 x i64> %res4
 }

 
 declare <16 x i32> @llvm.x86.avx512.prol.d.512(<16 x i32>, i32)
 
 define <16 x i32>@test_int_x86_avx512_prol_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_d_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vprold $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_d_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %1 = call <16 x i32> @llvm.x86.avx512.prol.d.512(<16 x i32> %x0, i32 3)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x2
   %4 = call <16 x i32> @llvm.x86.avx512.prol.d.512(<16 x i32> %x0, i32 4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   %7 = call <16 x i32> @llvm.x86.avx512.prol.d.512(<16 x i32> %x0, i32 5)
   %res3 = add <16 x i32> %3, %6
   %res4 = add <16 x i32> %res3, %7
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.prol.q.512(<8 x i64>, i32)
 
 define <8 x i64>@test_int_x86_avx512_prol_q_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_q_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_q_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %1 = call <8 x i64> @llvm.x86.avx512.prol.q.512(<8 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x2
   %4 = call <8 x i64> @llvm.x86.avx512.prol.q.512(<8 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   %7 = call <8 x i64> @llvm.x86.avx512.prol.q.512(<8 x i64> %x0, i32 5)
   %res3 = add <8 x i64> %3, %6
   %res4 = add <8 x i64> %res3, %7
   ret <8 x i64> %res4
 }
 
 declare <16 x i32> @llvm.x86.avx512.pror.d.512(<16 x i32>, i32)
 
 define <16 x i32>@test_int_x86_avx512_pror_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_d_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vprord $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_d_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %1 = call <16 x i32> @llvm.x86.avx512.pror.d.512(<16 x i32> %x0, i32 3)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x2
   %4 = call <16 x i32> @llvm.x86.avx512.pror.d.512(<16 x i32> %x0, i32 4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   %7 = call <16 x i32> @llvm.x86.avx512.pror.d.512(<16 x i32> %x0, i32 5)
   %res3 = add <16 x i32> %3, %6
   %res4 = add <16 x i32> %res3, %7
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.pror.q.512(<8 x i64>, i32)
 
 define <8 x i64>@test_int_x86_avx512_pror_q_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_q_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_q_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %1 = call <8 x i64> @llvm.x86.avx512.pror.q.512(<8 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x2
   %4 = call <8 x i64> @llvm.x86.avx512.pror.q.512(<8 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   %7 = call <8 x i64> @llvm.x86.avx512.pror.q.512(<8 x i64> %x0, i32 5)
   %res3 = add <8 x i64> %3, %6
   %res4 = add <8 x i64> %res3, %7
   ret <8 x i64> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_psrl_qi_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_qi_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsrlq $4, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x73,0xd0,0x04]
 ; X86-NEXT:    vpsrlq $5, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x73,0xd0,0x05]
-; X86-NEXT:    vpsrlq $6, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0x73,0xd0,0x06]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
-; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xca]
+; X86-NEXT:    vpsrlq $6, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x73,0xd0,0x06]
+; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xc8]
+; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_qi_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrlq $4, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x73,0xd0,0x04]
 ; X64-NEXT:    vpsrlq $5, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x73,0xd0,0x05]
-; X64-NEXT:    vpsrlq $6, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0x73,0xd0,0x06]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
-; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xca]
+; X64-NEXT:    vpsrlq $6, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x73,0xd0,0x06]
+; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xc8]
+; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 4, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.psrl.qi.512(<8 x i64> %x0, i32 6, <8 x i64> zeroinitializer, i8 %x3)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 declare <16 x i32> @llvm.x86.avx512.mask.psrl.di.512(<16 x i32>, i32, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_psrl_di_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_di_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsrld $4, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xd0,0x04]
 ; X86-NEXT:    vpsrld $5, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xd0,0x05]
-; X86-NEXT:    vpsrld $6, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0x72,0xd0,0x06]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
-; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xca]
+; X86-NEXT:    vpsrld $6, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xd0,0x06]
+; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xc8]
+; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_di_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrld $4, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xd0,0x04]
 ; X64-NEXT:    vpsrld $5, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xd0,0x05]
-; X64-NEXT:    vpsrld $6, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0x72,0xd0,0x06]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
-; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xca]
+; X64-NEXT:    vpsrld $6, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xd0,0x06]
+; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xc8]
+; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.psrl.di.512(<16 x i32> %x0, i32 4, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.psrl.di.512(<16 x i32> %x0, i32 5, <16 x i32> %x2, i16 -1)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.psrl.di.512(<16 x i32> %x0, i32 6, <16 x i32> zeroinitializer, i16 %x3)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 declare <16 x i32> @llvm.x86.avx512.mask.psra.di.512(<16 x i32>, i32, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_psra_di_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_di_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsrad $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xe0,0x03]
-; X86-NEXT:    vpsrad $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xe0,0x04]
+; X86-NEXT:    vpsrad $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xe0,0x04]
 ; X86-NEXT:    vpsrad $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xe0,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_di_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrad $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xe0,0x03]
-; X64-NEXT:    vpsrad $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xe0,0x04]
+; X64-NEXT:    vpsrad $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xe0,0x04]
 ; X64-NEXT:    vpsrad $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xe0,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.psra.di.512(<16 x i32> %x0, i32 3, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.psra.di.512(<16 x i32> %x0, i32 4, <16 x i32> zeroinitializer, i16 %x3)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.psra.di.512(<16 x i32> %x0, i32 5, <16 x i32> %x2, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.psra.qi.512(<8 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_psra_qi_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_qi_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsraq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xe0,0x03]
-; X86-NEXT:    vpsraq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xe0,0x04]
+; X86-NEXT:    vpsraq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xe0,0x04]
 ; X86-NEXT:    vpsraq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xe0,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_qi_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsraq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xe0,0x03]
-; X64-NEXT:    vpsraq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xe0,0x04]
+; X64-NEXT:    vpsraq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xe0,0x04]
 ; X64-NEXT:    vpsraq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xe0,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.psra.qi.512(<8 x i64> %x0, i32 3, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.psra.qi.512(<8 x i64> %x0, i32 4, <8 x i64> zeroinitializer, i8 %x3)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.psra.qi.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 declare <16 x i32> @llvm.x86.avx512.mask.psll.di.512(<16 x i32>, i32, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_psll_di_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_di_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpslld $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xf0,0x03]
-; X86-NEXT:    vpslld $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xf0,0x04]
+; X86-NEXT:    vpslld $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xf0,0x04]
 ; X86-NEXT:    vpslld $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xf0,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_di_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpslld $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xf0,0x03]
-; X64-NEXT:    vpslld $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xf0,0x04]
+; X64-NEXT:    vpslld $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xf0,0x04]
 ; X64-NEXT:    vpslld $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xf0,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.psll.di.512(<16 x i32> %x0, i32 3, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.psll.di.512(<16 x i32> %x0, i32 4, <16 x i32> zeroinitializer, i16 %x3)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.psll.di.512(<16 x i32> %x0, i32 5, <16 x i32> %x2, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.psll.qi.512(<8 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_psll_qi_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_qi_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsllq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x73,0xf0,0x03]
-; X86-NEXT:    vpsllq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x73,0xf0,0x04]
+; X86-NEXT:    vpsllq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x73,0xf0,0x04]
 ; X86-NEXT:    vpsllq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x73,0xf0,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_qi_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsllq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x73,0xf0,0x03]
-; X64-NEXT:    vpsllq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x73,0xf0,0x04]
+; X64-NEXT:    vpsllq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x73,0xf0,0x04]
 ; X64-NEXT:    vpsllq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x73,0xf0,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.psll.qi.512(<8 x i64> %x0, i32 3, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.psll.qi.512(<8 x i64> %x0, i32 4, <8 x i64> zeroinitializer, i8 %x3)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.psll.qi.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4

 
 declare <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32>, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x4_512(<4 x i32> %x0, <16 x i32> %x2, i16 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_broadcasti32x4_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    ## kill: def $xmm0 killed $xmm0 def $ymm0
 ; X86-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
-; X86-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0 ## encoding: [0x62,0xf3,0xfd,0x48,0x3a,0xc0,0x01]
+; X86-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm2 ## encoding: [0x62,0xf3,0xfd,0x48,0x3a,0xd0,0x01]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
-; X86-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0xc8]
-; X86-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0x6f,0xd0]
-; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xca]
-; X86-NEXT:    vpaddd %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfe,0xc1]
+; X86-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0xca]
+; X86-NEXT:    vpaddd %zmm1, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc1]
+; X86-NEXT:    vpaddd %zmm2, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfe,0xc2]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_broadcasti32x4_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    ## kill: def $xmm0 killed $xmm0 def $ymm0
 ; X64-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0 ## encoding: [0xc4,0xe3,0x7d,0x38,0xc0,0x01]
-; X64-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0 ## encoding: [0x62,0xf3,0xfd,0x48,0x3a,0xc0,0x01]
+; X64-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm2 ## encoding: [0x62,0xf3,0xfd,0x48,0x3a,0xd0,0x01]
 ; X64-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0xc8]
-; X64-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0x6f,0xd0]
-; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xca]
-; X64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfe,0xc1]
+; X64-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0xca]
+; X64-NEXT:    vpaddd %zmm1, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc1]
+; X64-NEXT:    vpaddd %zmm2, %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfe,0xc2]
 ; X64-NEXT:    retq ## encoding: [0xc3]
 
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32> %x0, <16 x i32> %x2, i16 -1)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32> %x0, <16 x i32> %x2, i16 %mask)
   %res3 = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32> %x0, <16 x i32> zeroinitializer, i16 %mask)
   %res4 = add <16 x i32> %res1, %res2
   %res5 = add <16 x i32> %res3, %res4
   ret <16 x i32> %res5

 
 declare <16 x i32> @llvm.x86.avx512.mask.prol.d.512(<16 x i32>, i32, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_prol_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_d_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vprold $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_d_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.prol.d.512(<16 x i32> %x0, i32 3, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.prol.d.512(<16 x i32> %x0, i32 4, <16 x i32> zeroinitializer, i16 %x3)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.prol.d.512(<16 x i32> %x0, i32 5, <16 x i32> %x2, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.prol.q.512(<8 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_prol_q_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_q_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_q_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.prol.q.512(<8 x i64> %x0, i32 3, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.prol.q.512(<8 x i64> %x0, i32 4, <8 x i64> zeroinitializer, i8 %x3)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.prol.q.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 declare <16 x i32> @llvm.x86.avx512.mask.pror.d.512(<16 x i32>, i32, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_pror_d_512(<16 x i32> %x0, i32 %x1, <16 x i32> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_d_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vprord $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X86-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_d_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xc9,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc0]
+; X64-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x49,0xfe,0xca]
 ; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x75,0x48,0xfe,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <16 x i32> @llvm.x86.avx512.mask.pror.d.512(<16 x i32> %x0, i32 3, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.pror.d.512(<16 x i32> %x0, i32 4, <16 x i32> zeroinitializer, i16 %x3)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.pror.d.512(<16 x i32> %x0, i32 5, <16 x i32> %x2, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.pror.q.512(<8 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_pror_q_512(<8 x i64> %x0, i32 %x1, <8 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_q_512:
 ; X86:       ## %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X86-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X86-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_q_512:
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %zmm0, %zmm2 {%k1} {z} ## encoding: [0x62,0xf1,0xed,0xc9,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %zmm0, %zmm2 ## encoding: [0x62,0xf1,0xed,0x48,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0 ## encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc0]
+; X64-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xd4,0xca]
 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq ## encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.pror.q.512(<8 x i64> %x0, i32 3, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.pror.q.512(<8 x i64> %x0, i32 4, <8 x i64> zeroinitializer, i8 %x3)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.pror.q.512(<8 x i64> %x0, i32 5, <8 x i64> %x2, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4

llvm/test/CodeGen/X86/avx512-rotate.ll

 
 ; Tests showing replacement of variable rotates with immediate splat versions.
 
 define <16 x i32> @test_splat_rol_v16i32(<16 x i32> %x0, <16 x i32> %x1, i16 %x2) {
 ; KNL-LABEL: test_splat_rol_v16i32:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprold $5, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprold $6, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprold $6, %zmm0, %zmm2
 ; KNL-NEXT:    vprold $7, %zmm0, %zmm0
+; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_rol_v16i32:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprold $5, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprold $6, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprold $6, %zmm0, %zmm2
 ; SKX-NEXT:    vprold $7, %zmm0, %zmm0
+; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>, <16 x i32> %x1, i16 %x2)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>, <16 x i32> zeroinitializer, i16 %x2)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>, <16 x i32> %x1, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 define <8 x i64>@test_splat_rol_v8i64(<8 x i64> %x0, <8 x i64> %x1, i8 %x2) {
 ; KNL-LABEL: test_splat_rol_v8i64:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprolq $5, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprolq $6, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprolq $6, %zmm0, %zmm2
 ; KNL-NEXT:    vprolq $7, %zmm0, %zmm0
+; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_rol_v8i64:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprolq $5, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprolq $6, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprolq $6, %zmm0, %zmm2
 ; SKX-NEXT:    vprolq $7, %zmm0, %zmm0
+; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>, <8 x i64> %x1, i8 %x2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6>, <8 x i64> zeroinitializer, i8 %x2)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>, <8 x i64> %x1, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 define <16 x i32> @test_splat_ror_v16i32(<16 x i32> %x0, <16 x i32> %x1, i16 %x2) {
 ; KNL-LABEL: test_splat_ror_v16i32:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprord $5, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprord $6, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprord $6, %zmm0, %zmm2
 ; KNL-NEXT:    vprord $7, %zmm0, %zmm0
+; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_ror_v16i32:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprord $5, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprord $6, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprord $6, %zmm0, %zmm2
 ; SKX-NEXT:    vprord $7, %zmm0, %zmm0
+; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>, <16 x i32> %x1, i16 %x2)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>, <16 x i32> zeroinitializer, i16 %x2)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>, <16 x i32> %x1, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 define <8 x i64>@test_splat_ror_v8i64(<8 x i64> %x0, <8 x i64> %x1, i8 %x2) {
 ; KNL-LABEL: test_splat_ror_v8i64:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprorq $5, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprorq $6, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprorq $6, %zmm0, %zmm2
 ; KNL-NEXT:    vprorq $7, %zmm0, %zmm0
+; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_ror_v8i64:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprorq $5, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprorq $6, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprorq $6, %zmm0, %zmm2
 ; SKX-NEXT:    vprorq $7, %zmm0, %zmm0
+; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>, <8 x i64> %x1, i8 %x2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6>, <8 x i64> zeroinitializer, i8 %x2)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>, <8 x i64> %x1, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 ; Tests showing replacement of out-of-bounds variable rotates with in-bounds immediate splat versions.
 
 define <16 x i32> @test_splat_bounds_rol_v16i32(<16 x i32> %x0, <16 x i32> %x1, i16 %x2) {
 ; KNL-LABEL: test_splat_bounds_rol_v16i32:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprold $1, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprold $31, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprold $31, %zmm0, %zmm2
 ; KNL-NEXT:    vprold $30, %zmm0, %zmm0
+; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_bounds_rol_v16i32:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprold $1, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprold $31, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprold $31, %zmm0, %zmm2
 ; SKX-NEXT:    vprold $30, %zmm0, %zmm0
+; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33>, <16 x i32> %x1, i16 %x2)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <16 x i32> zeroinitializer, i16 %x2)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.prolv.d.512(<16 x i32> %x0, <16 x i32> <i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534>, <16 x i32> %x1, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 define <8 x i64>@test_splat_bounds_rol_v8i64(<8 x i64> %x0, <8 x i64> %x1, i8 %x2) {
 ; KNL-LABEL: test_splat_bounds_rol_v8i64:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprolq $62, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprolq $1, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprolq $1, %zmm0, %zmm2
 ; KNL-NEXT:    vprolq $63, %zmm0, %zmm0
+; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_bounds_rol_v8i64:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprolq $62, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprolq $1, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprolq $1, %zmm0, %zmm2
 ; SKX-NEXT:    vprolq $63, %zmm0, %zmm0
+; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534>, <8 x i64> %x1, i8 %x2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65>, <8 x i64> zeroinitializer, i8 %x2)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.prolv.q.512(<8 x i64> %x0, <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, <8 x i64> %x1, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4
 }
 
 define <16 x i32> @test_splat_bounds_ror_v16i32(<16 x i32> %x0, <16 x i32> %x1, i16 %x2) {
 ; KNL-LABEL: test_splat_bounds_ror_v16i32:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprord $1, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprord $31, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprord $31, %zmm0, %zmm2
 ; KNL-NEXT:    vprord $30, %zmm0, %zmm0
+; KNL-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_bounds_ror_v16i32:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprord $1, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprord $31, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprord $31, %zmm0, %zmm2
 ; SKX-NEXT:    vprord $30, %zmm0, %zmm0
+; SKX-NEXT:    vpaddd %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33>, <16 x i32> %x1, i16 %x2)
   %res1 = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <16 x i32> zeroinitializer, i16 %x2)
   %res2 = call <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32> %x0, <16 x i32> <i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534, i32 65534>, <16 x i32> %x1, i16 -1)
   %res3 = add <16 x i32> %res, %res1
   %res4 = add <16 x i32> %res3, %res2
   ret <16 x i32> %res4
 }
 
 define <8 x i64>@test_splat_bounds_ror_v8i64(<8 x i64> %x0, <8 x i64> %x1, i8 %x2) {
 ; KNL-LABEL: test_splat_bounds_ror_v8i64:
 ; KNL:       # %bb.0:
 ; KNL-NEXT:    kmovw %edi, %k1
 ; KNL-NEXT:    vprorq $62, %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vprorq $1, %zmm0, %zmm2 {%k1} {z}
-; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; KNL-NEXT:    vprorq $1, %zmm0, %zmm2
 ; KNL-NEXT:    vprorq $63, %zmm0, %zmm0
+; KNL-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; KNL-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test_splat_bounds_ror_v8i64:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovd %edi, %k1
 ; SKX-NEXT:    vprorq $62, %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vprorq $1, %zmm0, %zmm2 {%k1} {z}
-; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1
+; SKX-NEXT:    vprorq $1, %zmm0, %zmm2
 ; SKX-NEXT:    vprorq $63, %zmm0, %zmm0
+; SKX-NEXT:    vpaddq %zmm2, %zmm1, %zmm1 {%k1}
 ; SKX-NEXT:    vpaddq %zmm0, %zmm1, %zmm0
 ; SKX-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534, i64 65534>, <8 x i64> %x1, i8 %x2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65>, <8 x i64> zeroinitializer, i8 %x2)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.prorv.q.512(<8 x i64> %x0, <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, <8 x i64> %x1, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res3, %res2
   ret <8 x i64> %res4

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

 declare <64 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.512(i8, <64 x i8>, i64)
 
 define <64 x i8> @test_int_x86_avx512_mask_pbroadcast_b_gpr_512(i8 %x0, <64 x i8> %x1, i64 %mask) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpbroadcastb {{[0-9]+}}(%esp), %zmm1 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0x4c,0x24,0x04]
 ; X86-NEXT:    vmovdqu8 %zmm1, %zmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x49,0x6f,0xc1]
-; X86-NEXT:    vmovdqu8 %zmm1, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xc9,0x6f,0xd1]
-; X86-NEXT:    vpaddb %zmm2, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfc,0xc2]
 ; X86-NEXT:    vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
+; X86-NEXT:    vpaddb %zmm1, %zmm0, %zmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfc,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastb %edi, %zmm1 # encoding: [0x62,0xf2,0x7d,0x48,0x7a,0xcf]
 ; X64-NEXT:    kmovq %rsi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastb %edi, %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x7a,0xc7]
-; X64-NEXT:    vpbroadcastb %edi, %zmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x7a,0xd7]
-; X64-NEXT:    vpaddb %zmm2, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfc,0xc2]
 ; X64-NEXT:    vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
+; X64-NEXT:    vpaddb %zmm1, %zmm0, %zmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfc,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
     %res = call <64 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.512(i8 %x0, <64 x i8> %x1, i64 -1)
     %res1 = call <64 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.512(i8 %x0, <64 x i8> %x1, i64 %mask)
     %res2 = call <64 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.512(i8 %x0, <64 x i8> zeroinitializer, i64 %mask)
     %res3 = add <64 x i8> %res, %res1
     %res4 = add <64 x i8> %res2, %res3
     ret <64 x i8> %res4
   }
 
 declare <32 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.512(i16, <32 x i16>, i32)
 define <32 x i16> @test_int_x86_avx512_mask_pbroadcast_w_gpr_512(i16 %x0, <32 x i16> %x1, i32 %mask) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastw {{[0-9]+}}(%esp), %zmm1 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0x4c,0x24,0x02]
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqu16 %zmm1, %zmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x6f,0xc1]
-; X86-NEXT:    vmovdqu16 %zmm1, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xc9,0x6f,0xd1]
-; X86-NEXT:    vpaddw %zmm2, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0xc2]
 ; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastw %edi, %zmm1 # encoding: [0x62,0xf2,0x7d,0x48,0x7b,0xcf]
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastw %edi, %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x7b,0xc7]
-; X64-NEXT:    vpbroadcastw %edi, %zmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x7b,0xd7]
-; X64-NEXT:    vpaddw %zmm2, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0xc2]
 ; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
     %res = call <32 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.512(i16 %x0, <32 x i16> %x1, i32 -1)
     %res1 = call <32 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.512(i16 %x0, <32 x i16> %x1, i32 %mask)
    %res2 = call <32 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.512(i16 %x0, <32 x i16> zeroinitializer, i32 %mask)
     %res3 = add <32 x i16> %res, %res1
    %res4 = add <32 x i16> %res2, %res3
     ret <32 x i16> %res4
  }

 declare <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16>, i32, <32 x i16>, i32)
 
 define <32 x i16> @test_int_x86_avx512_mask_psrl_wi_512(<32 x i16> %x0, i32 %x1, <32 x i16> %x2, i32 %x3) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsrlw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xd0,0x03]
 ; X86-NEXT:    vpsrlw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xd0,0x04]
-; X86-NEXT:    vpsrlw $5, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0x71,0xd0,0x05]
-; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
-; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xca]
+; X86-NEXT:    vpsrlw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xd0,0x05]
+; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xc8]
+; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsrlw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xd0,0x03]
 ; X64-NEXT:    vpsrlw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xd0,0x04]
-; X64-NEXT:    vpsrlw $5, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0x71,0xd0,0x05]
-; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
-; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xca]
+; X64-NEXT:    vpsrlw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xd0,0x05]
+; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xc8]
+; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i32 3, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i32 4, <32 x i16> %x2, i32 -1)
   %res2 = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i32 5, <32 x i16> zeroinitializer, i32 %x3)
   %res3 = add <32 x i16> %res, %res1
   %res4 = add <32 x i16> %res3, %res2
   ret <32 x i16> %res4
 }

 
 declare <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16>, i32, <32 x i16>, i32)
 
 define <32 x i16> @test_int_x86_avx512_mask_psra_wi_512(<32 x i16> %x0, i32 %x1, <32 x i16> %x2, i32 %x3) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsraw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xe0,0x03]
-; X86-NEXT:    vpsraw $4, %zmm0, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xc9,0x71,0xe0,0x04]
+; X86-NEXT:    vpsraw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xe0,0x04]
 ; X86-NEXT:    vpsraw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xe0,0x05]
-; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xca]
 ; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsraw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xe0,0x03]
-; X64-NEXT:    vpsraw $4, %zmm0, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xc9,0x71,0xe0,0x04]
+; X64-NEXT:    vpsraw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xe0,0x04]
 ; X64-NEXT:    vpsraw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xe0,0x05]
-; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xca]
 ; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i32 3, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i32 4, <32 x i16> zeroinitializer, i32 %x3)
   %res2 = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i32 5, <32 x i16> %x2, i32 -1)
   %res3 = add <32 x i16> %res, %res1
   %res4 = add <32 x i16> %res3, %res2
   ret <32 x i16> %res4

 
 declare <32 x i16> @llvm.x86.avx512.mask.psll.wi.512(<32 x i16>, i32, <32 x i16>, i32)
 
 define <32 x i16> @test_int_x86_avx512_mask_psll_wi_512(<32 x i16> %x0, i32 %x1, <32 x i16> %x2, i32 %x3) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsllw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xf0,0x03]
-; X86-NEXT:    vpsllw $4, %zmm0, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xc9,0x71,0xf0,0x04]
+; X86-NEXT:    vpsllw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xf0,0x04]
 ; X86-NEXT:    vpsllw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xf0,0x05]
-; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xca]
 ; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsllw $3, %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0x71,0xf0,0x03]
-; X64-NEXT:    vpsllw $4, %zmm0, %zmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xc9,0x71,0xf0,0x04]
+; X64-NEXT:    vpsllw $4, %zmm0, %zmm2 # encoding: [0x62,0xf1,0x6d,0x48,0x71,0xf0,0x04]
 ; X64-NEXT:    vpsllw $5, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0x71,0xf0,0x05]
-; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm2, %zmm1, %zmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x49,0xfd,0xca]
 ; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <32 x i16> @llvm.x86.avx512.mask.psll.wi.512(<32 x i16> %x0, i32 3, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.mask.psll.wi.512(<32 x i16> %x0, i32 4, <32 x i16> zeroinitializer, i32 %x3)
   %res2 = call <32 x i16> @llvm.x86.avx512.mask.psll.wi.512(<32 x i16> %x0, i32 5, <32 x i16> %x2, i32 -1)
   %res3 = add <32 x i16> %res, %res1
   %res4 = add <32 x i16> %res3, %res2
   ret <32 x i16> %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) nounwind {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x49,0xfd,0xd3]
 ; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x49,0xfd,0xd3]
 ; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %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 3, <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 4, <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

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

 
 declare <32 x i16> @llvm.x86.avx512.dbpsadbw.512(<64 x i8>, <64 x i8>, 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) {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x49,0xfd,0xd3]
 ; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xc9,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf3,0x7d,0x48,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x49,0xfd,0xd3]
 ; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <32 x i16> @llvm.x86.avx512.dbpsadbw.512(<64 x i8> %x0, <64 x i8> %x1, i32 2)
   %2 = bitcast i32 %x4 to <32 x i1>
   %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %x3
   %4 = call <32 x i16> @llvm.x86.avx512.dbpsadbw.512(<64 x i8> %x0, <64 x i8> %x1, i32 3)
   %5 = bitcast i32 %x4 to <32 x i1>
   %6 = select <32 x i1> %5, <32 x i16> %4, <32 x i16> zeroinitializer

llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
 
 declare <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8, <16 x i8>, i16)
 
 define <16 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_128(i8 %x0, <16 x i8> %x1, i16 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastb {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0x4c,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqu8 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0xc1]
-; X86-NEXT:    vmovdqu8 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0xd1]
-; X86-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
 ; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
+; X86-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfc,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastb %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xcf]
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastb %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc7]
-; X64-NEXT:    vpbroadcastb %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd7]
-; X64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
 ; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
+; X64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfc,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 -1)
   %res1 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 %mask)
   %res2 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> zeroinitializer, i16 %mask)
   %res3 = add <16 x i8> %res, %res1
   %res4 = add <16 x i8> %res2, %res3
   ret <16 x i8> %res4
 }
 
 
 declare <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16, <8 x i16>, i8)
 
 define <8 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_128(i16 %x0, <8 x i16> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastw {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0x4c,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vmovdqu16 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0xc1]
-; X86-NEXT:    vmovdqu16 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0xd1]
-; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastw %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xcf]
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastw %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc7]
-; X64-NEXT:    vpbroadcastw %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd7]
-; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 -1)
   %res1 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 %mask)
   %res2 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> zeroinitializer, i8 %mask)
   %res3 = add <8 x i16> %res, %res1
   %res4 = add <8 x i16> %res2, %res3
   ret <8 x i16> %res4
 }
 
 
 declare <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8, <32 x i8>, i32)
 
 define <32 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_256(i8 %x0, <32 x i8> %x1, i32 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastb {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0x4c,0x24,0x04]
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqu8 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0xc1]
-; X86-NEXT:    vmovdqu8 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0xd1]
-; X86-NEXT:    vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
 ; X86-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
+; X86-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfc,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastb %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xcf]
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastb %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc7]
-; X64-NEXT:    vpbroadcastb %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd7]
-; X64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
 ; X64-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
+; X64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfc,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 -1)
   %res1 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 %mask)
   %res2 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> zeroinitializer, i32 %mask)
   %res3 = add <32 x i8> %res, %res1
   %res4 = add <32 x i8> %res2, %res3
   ret <32 x i8> %res4
 }
 
 
 
 declare <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16, <16 x i16>, i16)
 
 define <16 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_256(i16 %x0, <16 x i16> %x1, i16 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastw {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0x4c,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqu16 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0xc1]
-; X86-NEXT:    vmovdqu16 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0xd1]
-; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
 ; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastw %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xcf]
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpbroadcastw %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc7]
-; X64-NEXT:    vpbroadcastw %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd7]
-; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
 ; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 -1)
   %res1 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 %mask)
   %res2 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> zeroinitializer, i16 %mask)
   %res3 = add <16 x i16> %res, %res1
   %res4 = add <16 x i16> %res2, %res3
   ret <16 x i16> %res4
 }

 
 define <8 x i16>@test_int_x86_avx512_mask_psrl_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
 ; X86-NEXT:    vpsrlw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x04]
-; X86-NEXT:    vpsrlw $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x05]
-; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
-; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xca]
+; X86-NEXT:    vpsrlw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xd0,0x05]
+; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xc8]
+; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
 ; X64-NEXT:    vpsrlw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x04]
-; X64-NEXT:    vpsrlw $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x05]
-; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
-; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xca]
+; X64-NEXT:    vpsrlw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xd0,0x05]
+; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xc8]
+; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
   %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 4, <8 x i16> %x2, i8 -1)
   %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 5, <8 x i16> zeroinitializer, i8 %x3)
   %res3 = add <8 x i16> %res, %res1
   %res4 = add <8 x i16> %res2, %res3
   ret <8 x i16> %res4
 }
 
 declare <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16>, i32, <16 x i16>, i16)
 
 define <16 x i16>@test_int_x86_avx512_mask_psrl_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
 ; X86-NEXT:    vpsrlw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x04]
-; X86-NEXT:    vpsrlw $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x05]
-; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
-; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xca]
+; X86-NEXT:    vpsrlw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xd0,0x05]
+; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xc8]
+; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
 ; X64-NEXT:    vpsrlw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x04]
-; X64-NEXT:    vpsrlw $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x05]
-; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
-; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xca]
+; X64-NEXT:    vpsrlw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xd0,0x05]
+; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xc8]
+; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
   %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 4, <16 x i16> %x2, i16 -1)
   %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 5, <16 x i16> zeroinitializer, i16 %x3)
   %res3 = add <16 x i16> %res, %res1
   %res4 = add <16 x i16> %res3, %res2
   ret <16 x i16> %res4
 }
 
 declare <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16>, i32, <8 x i16>, i8)
 
 define <8 x i16>@test_int_x86_avx512_mask_psra_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
-; X86-NEXT:    vpsraw $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x71,0xe0,0x04]
+; X86-NEXT:    vpsraw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x04]
 ; X86-NEXT:    vpsraw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xe0,0x05]
-; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xca]
 ; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
-; X64-NEXT:    vpsraw $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x71,0xe0,0x04]
+; X64-NEXT:    vpsraw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x04]
 ; X64-NEXT:    vpsraw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xe0,0x05]
-; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xca]
 ; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
   %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 4, <8 x i16> zeroinitializer, i8 %x3)
   %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 5, <8 x i16> %x2, i8 -1)
   %res3 = add <8 x i16> %res, %res1
   %res4 = add <8 x i16> %res3, %res2
   ret <8 x i16> %res4
 }
 
 declare <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16>, i32, <16 x i16>, i16)
 
 define <16 x i16>@test_int_x86_avx512_mask_psra_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
-; X86-NEXT:    vpsraw $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xe0,0x04]
+; X86-NEXT:    vpsraw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x04]
 ; X86-NEXT:    vpsraw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xe0,0x05]
-; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xca]
 ; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
-; X64-NEXT:    vpsraw $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xe0,0x04]
+; X64-NEXT:    vpsraw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x04]
 ; X64-NEXT:    vpsraw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xe0,0x05]
-; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xca]
 ; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
   %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 4, <16 x i16> zeroinitializer, i16 %x3)
   %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 5, <16 x i16> %x2, i16 -1)
   %res3 = add <16 x i16> %res, %res1
   %res4 = add <16 x i16> %res3, %res2
   ret <16 x i16> %res4
 }
 
 declare <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16>, i32, <8 x i16>, i8)
 
 define <8 x i16>@test_int_x86_avx512_mask_psll_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
-; X86-NEXT:    vpsllw $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x71,0xf0,0x04]
+; X86-NEXT:    vpsllw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x04]
 ; X86-NEXT:    vpsllw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xf0,0x05]
-; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xca]
 ; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
-; X64-NEXT:    vpsllw $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x71,0xf0,0x04]
+; X64-NEXT:    vpsllw $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x04]
 ; X64-NEXT:    vpsllw $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xf0,0x05]
-; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfd,0xca]
 ; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
   %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 4, <8 x i16> zeroinitializer, i8 %x3)
   %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 5, <8 x i16> %x2, i8 -1)
   %res3 = add <8 x i16> %res, %res1
   %res4 = add <8 x i16> %res3, %res2
   ret <8 x i16> %res4
 }
 
 declare <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16>, i32, <16 x i16>, i16)
 
 define <16 x i16>@test_int_x86_avx512_mask_psll_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
-; X86-NEXT:    vpsllw $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xf0,0x04]
+; X86-NEXT:    vpsllw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x04]
 ; X86-NEXT:    vpsllw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xf0,0x05]
-; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xca]
 ; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
-; X64-NEXT:    vpsllw $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xf0,0x04]
+; X64-NEXT:    vpsllw $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x04]
 ; X64-NEXT:    vpsllw $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x71,0xf0,0x05]
-; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfd,0xca]
 ; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
   %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 4, <16 x i16> zeroinitializer, i16 %x3)
   %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 5, <16 x i16> %x2, i16 -1)
   %res3 = add <16 x i16> %res, %res1
   %res4 = add <16 x i16> %res3, %res2
   ret <16 x i16> %res4

 declare <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8>, <16 x i8>, i32, <8 x i16>, i8)
 
 define <8 x i16>@test_int_x86_avx512_mask_dbpsadbw_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
-; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfd,0xd3]
+; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
-; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfd,0xd3]
+; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> %x3, i8 %x4)
   %res1 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 3, <8 x i16> zeroinitializer, i8 %x4)
   %res2 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 4, <8 x i16> %x3, i8 -1)
   %res3 = add <8 x i16> %res, %res1
   %res4 = add <8 x i16> %res2, %res3
   ret <8 x i16> %res4
 }
 
 declare <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8>, <32 x i8>, i32, <16 x i16>, i16)
 
 define <16 x i16>@test_int_x86_avx512_mask_dbpsadbw_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x3, i16 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x29,0xfd,0xd3]
 ; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x29,0xfd,0xd3]
 ; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> %x3, i16 %x4)
   %res1 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 3, <16 x i16> zeroinitializer, i16 %x4)
   %res2 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 4, <16 x i16> %x3, i16 -1)
   %res3 = add <16 x i16> %res, %res1
   %res4 = add <16 x i16> %res3, %res2
   ret <16 x i16> %res4

llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll

 declare <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8>, <16 x i8>, i32)
 
 define <8 x i16>@test_int_x86_avx512_mask_dbpsadbw_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
-; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfd,0xd3]
+; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
-; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
+; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfd,0xd3]
+; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2)
   %2 = bitcast i8 %x4 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x3
   %4 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 3)
   %5 = bitcast i8 %x4 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i16> %4, <8 x i16> zeroinitializer
   %7 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 4)
   %res3 = add <8 x i16> %3, %6
   %res4 = add <8 x i16> %7, %res3
   ret <8 x i16> %res4
 }
 
 declare <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8>, <32 x i8>, i32)
 
 define <16 x i16>@test_int_x86_avx512_mask_dbpsadbw_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x3, i16 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
-; X86-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
+; X86-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x03]
 ; X86-NEXT:    vdbpsadbw $4, %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xc1,0x04]
-; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x29,0xfd,0xd3]
 ; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
-; X64-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
+; X64-NEXT:    vdbpsadbw $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x03]
 ; X64-NEXT:    vdbpsadbw $4, %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xc1,0x04]
-; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x29,0xfd,0xd3]
 ; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2)
   %2 = bitcast i16 %x4 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x3
   %4 = call <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 3)
   %5 = bitcast i16 %x4 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i16> %4, <16 x i16> zeroinitializer

llvm/test/CodeGen/X86/avx512vbmi2-intrinsics-upgrade.ll

 
 define <16 x i32>@test_int_x86_avx512_mask_vpshrdv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x18]
-; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpshrdv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpshrdv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

 define <8 x i64>@test_int_x86_avx512_mask_vpshrdv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x18]
-; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
-; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xc2]
+; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
-; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xc2]
+; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i64>, <8 x i64>* %x2p
   %res = call <8 x i64> @llvm.x86.avx512.mask.vpshrdv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.maskz.vpshrdv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8  %x3)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 

 
 define <16 x i32>@test_int_x86_avx512_mask_vpshldv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x18]
-; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x1f]
-; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpshldv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpshldv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

 define <8 x i64>@test_int_x86_avx512_mask_vpshldv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x18]
-; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
-; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xc2]
+; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x1f]
-; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
-; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xc2]
+; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i64>, <8 x i64>* %x2p
   %res = call <8 x i64> @llvm.x86.avx512.mask.vpshldv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3)
   %res1 = call <8 x i64> @llvm.x86.avx512.maskz.vpshldv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8  %x3)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 

 
 define <32 x i16>@test_int_x86_avx512_mask_vpshldv_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16>* %x2p, <32 x i16> %x4, i32 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvw (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x18]
-; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
-; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xc2]
+; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvw (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x1f]
-; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
-; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xc2]
+; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <32 x i16>, <32 x i16>* %x2p
   %res = call <32 x i16> @llvm.x86.avx512.mask.vpshldv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
   %res1 = call <32 x i16> @llvm.x86.avx512.maskz.vpshldv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4, i32  %x3)
   %res2 = add <32 x i16> %res, %res1
   ret <32 x i16> %res2
 }

llvm/test/CodeGen/X86/avx512vbmi2-intrinsics.ll

 
 define <16 x i32> @test_int_x86_avx512_mask_vpshrdv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x18]
-; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   %res3 = add <16 x i32> %3, %6
   ret <16 x i32> %res3
 }
 
 define <8 x i64> @test_int_x86_avx512_mask_vpshrdv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x18]
-; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
-; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xc2]
+; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
-; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xc2]
+; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i64>, <8 x i64>* %x2p
   %1 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x0
   %4 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   %res3 = add <8 x i64> %3, %6
   ret <8 x i64> %res3
 }
 
 define <32 x i16> @test_int_x86_avx512_mask_vpshrdv_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16>* %x2p, <32 x i16> %x4, i32 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_w_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvw (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x72,0x18]
-; X86-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x72,0xc2]
-; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X86-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x72,0xc2]
+; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_w_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvw (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x72,0x1f]
-; X64-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x72,0xc2]
-; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X64-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x72,0xc2]
+; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <32 x i16>, <32 x i16>* %x2p
   %1 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %x1, <32 x i16> %x0, <32 x i16> %x2)
   %2 = bitcast i32 %x3 to <32 x i1>
   %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %x0
   %4 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %x1, <32 x i16> %x0, <32 x i16> %x4)
   %5 = bitcast i32 %x3 to <32 x i1>
   %6 = select <32 x i1> %5, <32 x i16> %4, <32 x i16> zeroinitializer
   %res3 = add <32 x i16> %3, %6
   ret <32 x i16> %res3
 }
 
 define <16 x i32> @test_int_x86_avx512_mask_vpshldv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x18]
-; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x1f]
-; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   %res3 = add <16 x i32> %3, %6
   ret <16 x i32> %res3
 }
 
 define <8 x i64> @test_int_x86_avx512_mask_vpshldv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x18]
-; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
-; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xc2]
+; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x1f]
-; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
-; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xc2]
+; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x49,0xd4,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i64>, <8 x i64>* %x2p
   %1 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x0
   %4 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   %res3 = add <8 x i64> %3, %6
   ret <8 x i64> %res3
 }
 
 define <32 x i16> @test_int_x86_avx512_mask_vpshldv_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16>* %x2p, <32 x i16> %x4, i32 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpshldvw (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x18]
-; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
-; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xc2]
+; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpshldvw (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x1f]
-; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
-; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
+; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xc2]
+; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfd,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <32 x i16>, <32 x i16>* %x2p
   %1 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2)
   %2 = bitcast i32 %x3 to <32 x i1>
   %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %x0
   %4 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4)
   %5 = bitcast i32 %x3 to <32 x i1>
   %6 = select <32 x i1> %5, <32 x i16> %4, <32 x i16> zeroinitializer

llvm/test/CodeGen/X86/avx512vbmi2vl-intrinsics-upgrade.ll

 
 define <4 x i32>@test_int_x86_avx512_mask_vpshld_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshld_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X86-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshld_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X64-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22, <4 x i32> %x3, i8 %x4)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 23, <4 x i32> %x3, i8 -1)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 24, <4 x i32> zeroinitializer,i8 %x4)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res3, %res2
   ret <4 x i32> %res4
 }

 
 define <4 x i32>@test_int_x86_avx512_mask_vpshrd_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrd_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X86-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrd_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X64-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22, <4 x i32> %x3, i8 %x4)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 23, <4 x i32> %x3, i8 -1)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 24, <4 x i32> zeroinitializer,i8 %x4)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res3, %res2
   ret <4 x i32> %res4
 }

 
 define <4 x i32>@test_int_x86_avx512_mask_vpshld_d_128_2(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshld_d_128_2:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X86-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshld_d_128_2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X64-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.x86.avx512.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22)
   %2 = bitcast i8 %x4 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x3
   %4 = call <4 x i32> @llvm.x86.avx512.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 23)
   %5 = call <4 x i32> @llvm.x86.avx512.vpshld.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 24)
   %6 = bitcast i8 %x4 to <8 x i1>

 
 define <4 x i32>@test_int_x86_avx512_mask_vpshrd_d_128_2(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrd_d_128_2:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X86-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrd_d_128_2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X64-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.x86.avx512.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 22)
   %2 = bitcast i8 %x4 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x3
   %4 = call <4 x i32> @llvm.x86.avx512.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 23)
   %5 = call <4 x i32> @llvm.x86.avx512.vpshrd.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 24)
   %6 = bitcast i8 %x4 to <8 x i1>

 define <8 x i32>@test_int_x86_avx512_mask_vpshrdv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x73,0x18]
-; X86-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x73,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x73,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x73,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x73,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpshrdv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpshrdv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpshrdv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x73,0x18]
-; X86-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x73,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x73,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x73,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x73,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpshrdv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpshrdv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }
 

 define <4 x i64>@test_int_x86_avx512_mask_vpshrdv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64>* %x2p, <4 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x73,0x18]
-; X86-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x73,0xc2]
-; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x73,0xc2]
+; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x73,0xc2]
-; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x73,0xc2]
+; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i64>, <4 x i64>* %x2p
   %res = call <4 x i64> @llvm.x86.avx512.mask.vpshrdv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.maskz.vpshrdv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x4, i8  %x3)
   %res2 = add <4 x i64> %res, %res1
   ret <4 x i64> %res2
 }
 

 define <2 x i64>@test_int_x86_avx512_mask_vpshrdv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64>* %x2p, <2 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x73,0x18]
-; X86-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x73,0xc2]
-; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x73,0xc2]
+; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x73,0xc2]
-; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x73,0xc2]
+; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <2 x i64>, <2 x i64>* %x2p
   %res = call <2 x i64> @llvm.x86.avx512.mask.vpshrdv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.maskz.vpshrdv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x4, i8  %x3)
   %res2 = add <2 x i64> %res, %res1
   ret <2 x i64> %res2
 }
 

 
 define <16 x i16>@test_int_x86_avx512_mask_vpshrdv_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16>* %x2p, <16 x i16> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_w_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvw (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x72,0x18]
-; X86-NEXT:    vpshrdvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x72,0xc2]
-; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X86-NEXT:    vpshrdvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x72,0xc2]
+; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_w_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvw (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x72,0x1f]
-; X64-NEXT:    vpshrdvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x72,0xc2]
-; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X64-NEXT:    vpshrdvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x72,0xc2]
+; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i16>, <16 x i16>* %x2p
   %res = call <16 x i16> @llvm.x86.avx512.mask.vpshrdv.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
   %res1 = call <16 x i16> @llvm.x86.avx512.maskz.vpshrdv.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x4, i16  %x3)
   %res2 = add <16 x i16> %res, %res1
   ret <16 x i16> %res2
 }
 

 define <8 x i16>@test_int_x86_avx512_mask_vpshrdv_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16>* %x2p, <8 x i16> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_w_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvw (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x72,0x18]
-; X86-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x72,0xc2]
-; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X86-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x72,0xc2]
+; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_w_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvw (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x72,0x1f]
-; X64-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x72,0xc2]
-; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X64-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x72,0xc2]
+; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i16>, <8 x i16>* %x2p
   %res = call <8 x i16> @llvm.x86.avx512.mask.vpshrdv.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
   %res1 = call <8 x i16> @llvm.x86.avx512.maskz.vpshrdv.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x4, i8  %x3)
   %res2 = add <8 x i16> %res, %res1
   ret <8 x i16> %res2
 }
 

 define <8 x i32>@test_int_x86_avx512_mask_vpshldv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x71,0x18]
-; X86-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x71,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x71,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x71,0x1f]
-; X64-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x71,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x71,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpshldv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpshldv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpshldv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x71,0x18]
-; X86-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x71,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x71,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x71,0x1f]
-; X64-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x71,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x71,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpshldv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpshldv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }
 

 define <4 x i64>@test_int_x86_avx512_mask_vpshldv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64>* %x2p, <4 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x71,0x18]
-; X86-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x71,0xc2]
-; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x71,0xc2]
+; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x71,0x1f]
-; X64-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x71,0xc2]
-; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x71,0xc2]
+; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i64>, <4 x i64>* %x2p
   %res = call <4 x i64> @llvm.x86.avx512.mask.vpshldv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.maskz.vpshldv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x4, i8  %x3)
   %res2 = add <4 x i64> %res, %res1
   ret <4 x i64> %res2
 }
 

 define <2 x i64>@test_int_x86_avx512_mask_vpshldv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64>* %x2p, <2 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x71,0x18]
-; X86-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x71,0xc2]
-; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x71,0xc2]
+; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x71,0x1f]
-; X64-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x71,0xc2]
-; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x71,0xc2]
+; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <2 x i64>, <2 x i64>* %x2p
   %res = call <2 x i64> @llvm.x86.avx512.mask.vpshldv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.maskz.vpshldv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x4, i8  %x3)
   %res2 = add <2 x i64> %res, %res1
   ret <2 x i64> %res2
 }
 

 
 define <16 x i16>@test_int_x86_avx512_mask_vpshldv_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16>* %x2p, <16 x i16> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvw (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x70,0x18]
-; X86-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x70,0xc2]
-; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X86-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x70,0xc2]
+; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvw (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x70,0x1f]
-; X64-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x70,0xc2]
-; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X64-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x70,0xc2]
+; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i16>, <16 x i16>* %x2p
   %res = call <16 x i16> @llvm.x86.avx512.mask.vpshldv.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
   %res1 = call <16 x i16> @llvm.x86.avx512.maskz.vpshldv.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x4, i16  %x3)
   %res2 = add <16 x i16> %res, %res1
   ret <16 x i16> %res2
 }
 

 define <8 x i16>@test_int_x86_avx512_mask_vpshldv_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16>* %x2p, <8 x i16> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshldvw (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x70,0x18]
-; X86-NEXT:    vpshldvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x70,0xc2]
-; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X86-NEXT:    vpshldvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x70,0xc2]
+; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshldvw (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x70,0x1f]
-; X64-NEXT:    vpshldvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x70,0xc2]
-; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X64-NEXT:    vpshldvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x70,0xc2]
+; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i16>, <8 x i16>* %x2p
   %res = call <8 x i16> @llvm.x86.avx512.mask.vpshldv.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
   %res1 = call <8 x i16> @llvm.x86.avx512.maskz.vpshldv.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x4, i8  %x3)
   %res2 = add <8 x i16> %res, %res1
   ret <8 x i16> %res2
 }

llvm/test/CodeGen/X86/avx512vbmi2vl-intrinsics.ll

 
 define <4 x i32> @test_int_x86_avx512_mask_vpshld_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshld_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X86-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshld_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshldd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x71,0xd1,0x16]
 ; X64-NEXT:    vpshldd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xd9,0x17]
-; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x71,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshldd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x71,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> <i32 22, i32 22, i32 22, i32 22>)
   %2 = bitcast i8 %x4 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x3
   %4 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> <i32 23, i32 23, i32 23, i32 23>)
   %5 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> <i32 24, i32 24, i32 24, i32 24>)
   %6 = bitcast i8 %x4 to <8 x i1>

 
 define <4 x i32> @test_int_x86_avx512_mask_vpshrd_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrd_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
 ; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
 ; X86-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X86-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X86-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrd_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
 ; X64-NEXT:    vpshrdd $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x73,0xd1,0x16]
 ; X64-NEXT:    vpshrdd $23, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xd9,0x17]
-; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x73,0xc1,0x18]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xd3]
+; X64-NEXT:    vpshrdd $24, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x73,0xc1,0x18]
+; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm2 {%k1} # encoding: [0x62,0xf1,0x6d,0x09,0xfe,0xd0]
+; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> <i32 22, i32 22, i32 22, i32 22>)
   %2 = bitcast i8 %x4 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x3
   %4 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> <i32 23, i32 23, i32 23, i32 23>)
   %5 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> <i32 24, i32 24, i32 24, i32 24>)
   %6 = bitcast i8 %x4 to <8 x i1>

 define <8 x i32> @test_int_x86_avx512_mask_vpshrdv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x73,0x18]
-; X86-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x73,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x73,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x73,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x73,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
   %res3 = add <8 x i32> %3, %6
   ret <8 x i32> %res3
 }
 
 define <4 x i32> @test_int_x86_avx512_mask_vpshrdv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x73,0x18]
-; X86-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x73,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x73,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x73,0x1f]
-; X64-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x73,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpshrdvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x73,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract1, <4 x i32> %4, <4 x i32> zeroinitializer
   %res3 = add <4 x i32> %3, %6
   ret <4 x i32> %res3
 }
 
 define <4 x i64> @test_int_x86_avx512_mask_vpshrdv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64>* %x2p, <4 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x73,0x18]
-; X86-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x73,0xc2]
-; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x73,0xc2]
+; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x73,0xc2]
-; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x73,0xc2]
+; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i64>, <4 x i64>* %x2p
   %1 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i64> %1, <4 x i64> %x0
   %4 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract1, <4 x i64> %4, <4 x i64> zeroinitializer
   %res3 = add <4 x i64> %3, %6
   ret <4 x i64> %res3
 }
 
 define <2 x i64> @test_int_x86_avx512_mask_vpshrdv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64>* %x2p, <2 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvq (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x73,0x18]
-; X86-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x73,0xc2]
-; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X86-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x73,0xc2]
+; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvq (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x73,0x1f]
-; X64-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x73,0xc2]
-; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X64-NEXT:    vpshrdvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x73,0xc2]
+; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <2 x i64>, <2 x i64>* %x2p
   %1 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
   %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> %x0
   %4 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>

 define <8 x i16> @test_int_x86_avx512_mask_vpshrdv_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16>* %x2p, <8 x i16> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_w_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshrdvw (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x72,0x18]
-; X86-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x72,0xc2]
-; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X86-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x72,0xc2]
+; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_w_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshrdvw (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x72,0x1f]
-; X64-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x72,0xc2]
-; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
+; X64-NEXT:    vpshrdvw %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x72,0xc2]
+; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i16>, <8 x i16>* %x2p
   %1 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %x1, <8 x i16> %x0, <8 x i16> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x0
   %4 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %x1, <8 x i16> %x0, <8 x i16> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i16> %4, <8 x i16> zeroinitializer
   %res3 = add <8 x i16> %3, %6
   ret <8 x i16> %res3
 }
 
 define <8 x i32> @test_int_x86_avx512_mask_vpshldv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x71,0x18]
-; X86-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x71,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x71,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x71,0x1f]
-; X64-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x71,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x71,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
   %res3 = add <8 x i32> %3, %6
   ret <8 x i32> %res3
 }
 
 define <4 x i32> @test_int_x86_avx512_mask_vpshldv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshldvd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x71,0x18]
-; X86-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x71,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x71,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshldvd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x71,0x1f]
-; X64-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x71,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpshldvd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x71,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract1, <4 x i32> %4, <4 x i32> zeroinitializer
   %res3 = add <4 x i32> %3, %6
   ret <4 x i32> %res3
 }
 
 define <4 x i64> @test_int_x86_avx512_mask_vpshldv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64>* %x2p, <4 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x71,0x18]
-; X86-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x71,0xc2]
-; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x71,0xc2]
+; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x71,0x1f]
-; X64-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x71,0xc2]
-; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x71,0xc2]
+; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x29,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i64>, <4 x i64>* %x2p
   %1 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i64> %1, <4 x i64> %x0
   %4 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract1, <4 x i64> %4, <4 x i64> zeroinitializer
   %res3 = add <4 x i64> %3, %6
   ret <4 x i64> %res3
 }
 
 define <2 x i64> @test_int_x86_avx512_mask_vpshldv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64>* %x2p, <2 x i64> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpshldvq (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x71,0x18]
-; X86-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x71,0xc2]
-; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X86-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x71,0xc2]
+; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpshldvq (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x71,0x1f]
-; X64-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x71,0xc2]
-; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
+; X64-NEXT:    vpshldvq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x71,0xc2]
+; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0xe5,0x09,0xd4,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <2 x i64>, <2 x i64>* %x2p
   %1 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
   %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> %x0
   %4 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
   %6 = select <2 x i1> %extract1, <2 x i64> %4, <2 x i64> zeroinitializer
   %res3 = add <2 x i64> %3, %6
   ret <2 x i64> %res3
 }
 
 define <16 x i16> @test_int_x86_avx512_mask_vpshldv_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16>* %x2p, <16 x i16> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpshldvw (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x70,0x18]
-; X86-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x70,0xc2]
-; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X86-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x70,0xc2]
+; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpshldvw (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x70,0x1f]
-; X64-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x70,0xc2]
-; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
+; X64-NEXT:    vpshldvw %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x70,0xc2]
+; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfd,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i16>, <16 x i16>* %x2p
   %1 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x0
   %4 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i16> %4, <16 x i16> zeroinitializer

llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
 
 declare <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_pbroadcast_d_gpr_128(i32 %x0, <4 x i32> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x4c,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vmovdqa32 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6f,0xc1]
-; X86-NEXT:    vmovdqa32 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6f,0xd1]
-; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastd %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastd %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastd %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7c,0xd7]
-; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> %x1, i8 -1)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> %x1, i8 %mask)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> zeroinitializer, i8 %mask)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res2, %res3
   ret <4 x i32> %res4
 }
 
 
 declare <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_pbroadcast_q_gpr_128(i64 %x0, <2 x i64> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastq {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0x4c,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x0c]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vmovdqa64 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6f,0xc1]
-; X86-NEXT:    vmovdqa64 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x6f,0xd1]
-; X86-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xd4,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastq %rdi, %xmm1 # encoding: [0x62,0xf2,0xfd,0x08,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastq %rdi, %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastq %rdi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7c,0xd7]
-; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xd4,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> %x1,i8 -1)
   %res1 = call <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> %x1,i8 %mask)
   %res2 = call <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> zeroinitializer,i8 %mask)
   %res3 = add <2 x i64> %res, %res1
   %res4 = add <2 x i64> %res2, %res3
   ret <2 x i64> %res4
 }
 
 
  declare <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32, <8 x i32>, i8)
 
   define <8 x i32>@test_int_x86_avx512_mask_pbroadcast_d_gpr_256(i32 %x0, <8 x i32> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x4c,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vmovdqa32 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6f,0xc1]
-; X86-NEXT:    vmovdqa32 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6f,0xd1]
-; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastd %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastd %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastd %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7c,0xd7]
-; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> %x1, i8 -1)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> %x1, i8 %mask)
   %res2 = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> zeroinitializer, i8 %mask)
   %res3 = add <8 x i32> %res, %res1
   %res4 = add <8 x i32> %res2, %res3
   ret <8 x i32> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_pbroadcast_q_gpr_256(i64 %x0, <4 x i64> %x1, i8 %mask) {
 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vpbroadcastq {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0x4c,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x0c]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vmovdqa64 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6f,0xc1]
-; X86-NEXT:    vmovdqa64 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x6f,0xd1]
-; X86-NEXT:    vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xd4,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vpbroadcastq %rdi, %ymm1 # encoding: [0x62,0xf2,0xfd,0x28,0x7c,0xcf]
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpbroadcastq %rdi, %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7c,0xc7]
-; X64-NEXT:    vpbroadcastq %rdi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7c,0xd7]
-; X64-NEXT:    vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xd4,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> %x1,i8 -1)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> %x1,i8 %mask)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> zeroinitializer,i8 %mask)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res2, %res3
   ret <4 x i64> %res4
 }

 
 define <2 x i64>@test_int_x86_avx512_mask_psrl_qi_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_qi_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsrlq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x73,0xd0,0x03]
 ; X86-NEXT:    vpsrlq $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x73,0xd0,0x04]
-; X86-NEXT:    vpsrlq $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x73,0xd0,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
-; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xca]
+; X86-NEXT:    vpsrlq $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x73,0xd0,0x05]
+; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xc8]
+; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_qi_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrlq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x73,0xd0,0x03]
 ; X64-NEXT:    vpsrlq $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x73,0xd0,0x04]
-; X64-NEXT:    vpsrlq $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x73,0xd0,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
-; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xca]
+; X64-NEXT:    vpsrlq $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x73,0xd0,0x05]
+; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xc8]
+; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64> %x0, i32 4, <2 x i64> %x2, i8 -1)
   %res2 = call <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64> %x0, i32 5, <2 x i64> zeroinitializer, i8 %x3)
   %res3 = add <2 x i64> %res, %res1
   %res4 = add <2 x i64> %res2, %res3
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64>, i32, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_psrl_qi_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_qi_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsrlq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x73,0xd0,0x03]
 ; X86-NEXT:    vpsrlq $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x73,0xd0,0x04]
-; X86-NEXT:    vpsrlq $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x73,0xd0,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
-; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xca]
+; X86-NEXT:    vpsrlq $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x73,0xd0,0x05]
+; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xc8]
+; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_qi_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrlq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x73,0xd0,0x03]
 ; X64-NEXT:    vpsrlq $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x73,0xd0,0x04]
-; X64-NEXT:    vpsrlq $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x73,0xd0,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
-; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xca]
+; X64-NEXT:    vpsrlq $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x73,0xd0,0x05]
+; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xc8]
+; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64> %x0, i32 4, <4 x i64> %x2, i8 -1)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64> %x0, i32 5, <4 x i64> zeroinitializer, i8 %x3)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res2, %res3
   ret <4 x i64> %res4
 }
 
 declare <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32>, i32, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_psrl_di_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_di_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsrld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xd0,0x03]
 ; X86-NEXT:    vpsrld $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xd0,0x04]
-; X86-NEXT:    vpsrld $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xd0,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
-; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
+; X86-NEXT:    vpsrld $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xd0,0x05]
+; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xc8]
+; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_di_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xd0,0x03]
 ; X64-NEXT:    vpsrld $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xd0,0x04]
-; X64-NEXT:    vpsrld $5, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xd0,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
-; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
+; X64-NEXT:    vpsrld $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xd0,0x05]
+; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xc8]
+; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 4, <4 x i32> %x2, i8 -1)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 5, <4 x i32> zeroinitializer, i8 %x3)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res2, %res3
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32>, i32, <8 x i32>, i8)
 
 define <8 x i32>@test_int_x86_avx512_mask_psrl_di_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psrl_di_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsrld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xd0,0x03]
 ; X86-NEXT:    vpsrld $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xd0,0x04]
-; X86-NEXT:    vpsrld $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xd0,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
-; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xca]
+; X86-NEXT:    vpsrld $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x72,0xd0,0x05]
+; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xc8]
+; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psrl_di_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsrld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xd0,0x03]
 ; X64-NEXT:    vpsrld $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xd0,0x04]
-; X64-NEXT:    vpsrld $5, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xd0,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
-; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xca]
+; X64-NEXT:    vpsrld $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x72,0xd0,0x05]
+; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xc8]
+; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 4, <8 x i32> %x2, i8 -1)
   %res2 = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 5, <8 x i32> zeroinitializer, i8 %x3)
   %res3 = add <8 x i32> %res, %res1
   %res4 = add <8 x i32> %res2, %res3
   ret <8 x i32> %res4
 }
 
 declare <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32>, i32, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_psll_di_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_di_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpslld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xf0,0x03]
-; X86-NEXT:    vpslld $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xf0,0x04]
+; X86-NEXT:    vpslld $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xf0,0x04]
 ; X86-NEXT:    vpslld $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xf0,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_di_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpslld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xf0,0x03]
-; X64-NEXT:    vpslld $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xf0,0x04]
+; X64-NEXT:    vpslld $4, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xf0,0x04]
 ; X64-NEXT:    vpslld $5, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xf0,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 4, <4 x i32> zeroinitializer, i8 %x3)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 5, <4 x i32> %x2, i8 -1)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res3, %res2
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32>, i32, <8 x i32>, i8)
 
 define <8 x i32>@test_int_x86_avx512_mask_psll_di_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psll_di_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpslld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xf0,0x03]
-; X86-NEXT:    vpslld $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xf0,0x04]
+; X86-NEXT:    vpslld $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xf0,0x04]
 ; X86-NEXT:    vpslld $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x72,0xf0,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psll_di_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpslld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xf0,0x03]
-; X64-NEXT:    vpslld $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xf0,0x04]
+; X64-NEXT:    vpslld $4, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xf0,0x04]
 ; X64-NEXT:    vpslld $5, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x72,0xf0,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 4, <8 x i32> zeroinitializer, i8 %x3)
   %res2 = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 5, <8 x i32> %x2, i8 -1)
   %res3 = add <8 x i32> %res, %res1
   %res4 = add <8 x i32> %res3, %res2
   ret <8 x i32> %res4

 declare <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64>, i32, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_psra_qi_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_qi_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsraq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x03]
-; X86-NEXT:    vpsraq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xe0,0x04]
+; X86-NEXT:    vpsraq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xe0,0x04]
 ; X86-NEXT:    vpsraq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xe0,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_qi_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsraq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x03]
-; X64-NEXT:    vpsraq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xe0,0x04]
+; X64-NEXT:    vpsraq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xe0,0x04]
 ; X64-NEXT:    vpsraq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xe0,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64> %x0, i32 4, <2 x i64> zeroinitializer, i8 %x3)
   %res2 = call <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64> %x0, i32 5, <2 x i64> %x2, i8 -1)
   %res3 = add <2 x i64> %res, %res1
   %res4 = add <2 x i64> %res3, %res2
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64>, i32, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_psra_qi_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_psra_qi_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpsraq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x03]
-; X86-NEXT:    vpsraq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xe0,0x04]
+; X86-NEXT:    vpsraq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xe0,0x04]
 ; X86-NEXT:    vpsraq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xe0,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_psra_qi_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vpsraq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x03]
-; X64-NEXT:    vpsraq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xe0,0x04]
+; X64-NEXT:    vpsraq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xe0,0x04]
 ; X64-NEXT:    vpsraq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xe0,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64> %x0, i32 4, <4 x i64> zeroinitializer, i8 %x3)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64> %x0, i32 5, <4 x i64> %x2, i8 -1)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res3, %res2
   ret <4 x i64> %res4

 declare <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32>, i32, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_prol_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 4, <4 x i32> zeroinitializer, i8 %x3)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 5, <4 x i32> %x2, i8 -1)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res3, %res2
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32>, i32, <8 x i32>, i8)
 
 define <8 x i32>@test_int_x86_avx512_mask_prol_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 4, <8 x i32> zeroinitializer, i8 %x3)
   %res2 = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 5, <8 x i32> %x2, i8 -1)
   %res3 = add <8 x i32> %res, %res1
   %res4 = add <8 x i32> %res3, %res2
   ret <8 x i32> %res4
 }
 
 declare <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64>, i32, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_prol_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64> %x0, i32 4, <2 x i64> zeroinitializer, i8 %x3)
   %res2 = call <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64> %x0, i32 5, <2 x i64> %x2, i8 -1)
   %res3 = add <2 x i64> %res, %res1
   %res4 = add <2 x i64> %res3, %res2
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64>, i32, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_prol_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_prol_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_prol_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64> %x0, i32 4, <4 x i64> zeroinitializer, i8 %x3)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64> %x0, i32 5, <4 x i64> %x2, i8 -1)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res3, %res2
   ret <4 x i64> %res4

 declare <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32>, i32, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_pror_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 4, <4 x i32> zeroinitializer, i8 %x3)
   %res2 = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 5, <4 x i32> %x2, i8 -1)
   %res3 = add <4 x i32> %res, %res1
   %res4 = add <4 x i32> %res3, %res2
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32>, i32, <8 x i32>, i8)
 
 define <8 x i32>@test_int_x86_avx512_mask_pror_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 4, <8 x i32> zeroinitializer, i8 %x3)
   %res2 = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 5, <8 x i32> %x2, i8 -1)
   %res3 = add <8 x i32> %res, %res1
   %res4 = add <8 x i32> %res3, %res2
   ret <8 x i32> %res4
 }
 
 declare <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64>, i32, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_pror_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
   %res1 = call <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64> %x0, i32 4, <2 x i64> zeroinitializer, i8 %x3)
   %res2 = call <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64> %x0, i32 5, <2 x i64> %x2, i8 -1)
   %res3 = add <2 x i64> %res, %res1
   %res4 = add <2 x i64> %res3, %res2
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64>, i32, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_pror_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_pror_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_pror_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
   %res1 = call <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64> %x0, i32 4, <4 x i64> zeroinitializer, i8 %x3)
   %res2 = call <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64> %x0, i32 5, <4 x i64> %x2, i8 -1)
   %res3 = add <4 x i64> %res, %res1
   %res4 = add <4 x i64> %res3, %res2
   ret <4 x i64> %res4

 declare <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32>, i32)
 
 define <4 x i32>@test_int_x86_avx512_prol_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
   %4 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
   %7 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 5)
   %res3 = add <4 x i32> %3, %6
   %res4 = add <4 x i32> %res3, %7
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32>, i32)
 
 define <8 x i32>@test_int_x86_avx512_prol_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
-; X86-NEXT:    vprold $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc8,0x04]
+; X86-NEXT:    vprold $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x04]
 ; X86-NEXT:    vprold $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
-; X64-NEXT:    vprold $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc8,0x04]
+; X64-NEXT:    vprold $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x04]
 ; X64-NEXT:    vprold $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
   %4 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
   %7 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 5)
   %res3 = add <8 x i32> %3, %6
   %res4 = add <8 x i32> %res3, %7
   ret <8 x i32> %res4
 }
 
 declare <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64>, i32)
 
 define <2 x i64>@test_int_x86_avx512_prol_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
   %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
   %4 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
   %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
   %7 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 5)
   %res3 = add <2 x i64> %3, %6
   %res4 = add <2 x i64> %res3, %7
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64>, i32)
 
 define <4 x i64>@test_int_x86_avx512_prol_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_prol_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
-; X86-NEXT:    vprolq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc8,0x04]
+; X86-NEXT:    vprolq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x04]
 ; X86-NEXT:    vprolq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc8,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_prol_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
-; X64-NEXT:    vprolq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc8,0x04]
+; X64-NEXT:    vprolq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x04]
 ; X64-NEXT:    vprolq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc8,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
   %4 = call <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>

 declare <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32>, i32)
 
 define <4 x i32>@test_int_x86_avx512_pror_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_d_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_d_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0x89,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x08,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
+; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0xfe,0xca]
 ; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
   %4 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
   %7 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 5)
   %res3 = add <4 x i32> %3, %6
   %res4 = add <4 x i32> %res3, %7
   ret <4 x i32> %res4
 }
 
 declare <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32>, i32)
 
 define <8 x i32>@test_int_x86_avx512_pror_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_d_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
-; X86-NEXT:    vprord $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc0,0x04]
+; X86-NEXT:    vprord $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x04]
 ; X86-NEXT:    vprord $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_d_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
-; X64-NEXT:    vprord $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x6d,0xa9,0x72,0xc0,0x04]
+; X64-NEXT:    vprord $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x04]
 ; X64-NEXT:    vprord $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x28,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
+; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0xfe,0xca]
 ; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
   %4 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
   %7 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 5)
   %res3 = add <8 x i32> %3, %6
   %res4 = add <8 x i32> %res3, %7
   ret <8 x i32> %res4
 }
 
 declare <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64>, i32)
 
 define <2 x i64>@test_int_x86_avx512_pror_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_q_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_q_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0x89,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
+; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0xd4,0xca]
 ; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
   %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
   %4 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
   %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
   %7 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 5)
   %res3 = add <2 x i64> %3, %6
   %res4 = add <2 x i64> %res3, %7
   ret <2 x i64> %res4
 }
 
 declare <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64>, i32)
 
 define <4 x i64>@test_int_x86_avx512_pror_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_pror_q_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
-; X86-NEXT:    vprorq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc0,0x04]
+; X86-NEXT:    vprorq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x04]
 ; X86-NEXT:    vprorq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc0,0x05]
-; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_pror_q_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
-; X64-NEXT:    vprorq $4, %ymm0, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xed,0xa9,0x72,0xc0,0x04]
+; X64-NEXT:    vprorq $4, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x04]
 ; X64-NEXT:    vprorq $5, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xc0,0x05]
-; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
+; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0xd4,0xca]
 ; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %1 = call <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64> %x0, i32 3)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
   %4 = call <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64> %x0, i32 4)
   %5 = bitcast i8 %x3 to <8 x i1>

llvm/test/CodeGen/X86/avx512vl_vnni-intrinsics-upgrade.ll

 define <8 x i32>@test_int_x86_avx512_mask_vpdpbusd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x18]
-; X86-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x50,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x50,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpdpbusd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x18]
-; X86-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x50,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x50,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }
 

 define <8 x i32>@test_int_x86_avx512_mask_vpdpbusds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x18]
-; X86-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x51,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x51,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpdpbusds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x18]
-; X86-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x51,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x51,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }
 

 define <8 x i32>@test_int_x86_avx512_mask_vpdpwssd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x18]
-; X86-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x52,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x52,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpdpwssd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x18]
-; X86-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x52,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x52,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }
 

 define <8 x i32>@test_int_x86_avx512_mask_vpdpwssds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x18]
-; X86-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x53,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x53,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
   %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 

 define <4 x i32>@test_int_x86_avx512_mask_vpdpwssds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x18]
-; X86-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x53,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x53,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
   %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
   %res2 = add <4 x i32> %res, %res1
   ret <4 x i32> %res2
 }

llvm/test/CodeGen/X86/avx512vl_vnni-intrinsics.ll

 define <8 x i32>@test_int_x86_avx512_mask_vpdpbusd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x18]
-; X86-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x50,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x50,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.x86.avx512.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.x86.avx512.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer

 define <4 x i32>@test_int_x86_avx512_mask_vpdpbusd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x18]
-; X86-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x50,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x50,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.x86.avx512.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.x86.avx512.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>

 define <8 x i32>@test_int_x86_avx512_mask_vpdpbusds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x18]
-; X86-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x51,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x51,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.x86.avx512.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.x86.avx512.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer

 define <4 x i32>@test_int_x86_avx512_mask_vpdpbusds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x18]
-; X86-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x51,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x51,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.x86.avx512.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.x86.avx512.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>

 define <8 x i32>@test_int_x86_avx512_mask_vpdpwssd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x18]
-; X86-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x52,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x52,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.x86.avx512.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.x86.avx512.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer

 define <4 x i32>@test_int_x86_avx512_mask_vpdpwssd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x18]
-; X86-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x52,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x52,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.x86.avx512.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.x86.avx512.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>

 define <8 x i32>@test_int_x86_avx512_mask_vpdpwssds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32>* %x2p, <8 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x18]
-; X86-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xc2]
-; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x53,0xc2]
+; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xc2]
-; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x53,0xc2]
+; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm3 {%k1} # encoding: [0x62,0xf1,0x65,0x29,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <8 x i32>, <8 x i32>* %x2p
   %1 = call <8 x i32> @llvm.x86.avx512.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
   %4 = call <8 x i32> @llvm.x86.avx512.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer

 define <4 x i32>@test_int_x86_avx512_mask_vpdpwssds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32>* %x2p, <4 x i32> %x4, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
 ; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
 ; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x18]
-; X86-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xc2]
-; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x53,0xc2]
+; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X86-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xc2]
-; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x53,0xc2]
+; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x09,0xfe,0xd8]
+; X64-NEXT:    vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <4 x i32>, <4 x i32>* %x2p
   %1 = call <4 x i32> @llvm.x86.avx512.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
   %2 = bitcast i8 %x3 to <8 x i1>
   %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x0
   %4 = call <4 x i32> @llvm.x86.avx512.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4)
   %5 = bitcast i8 %x3 to <8 x i1>
   %extract1 = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = select <4 x i1> %extract1, <4 x i32> %4, <4 x i32> zeroinitializer
   %res3 = add <4 x i32> %3, %6
   ret <4 x i32> %res3
 }

llvm/test/CodeGen/X86/avx512vnni-intrinsics-upgrade.ll

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpbusd_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x50,0x18]
-; X86-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x50,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x50,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x50,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x50,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpdpbusd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpdpbusd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpbusds_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x51,0x18]
-; X86-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x51,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x51,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x51,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x51,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpdpbusds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpdpbusds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpwssd_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x52,0x18]
-; X86-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x52,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x52,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x52,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x52,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpdpwssd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpdpwssd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpwssds_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x53,0x18]
-; X86-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x53,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x53,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x53,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x53,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %res = call <16 x i32> @llvm.x86.avx512.mask.vpdpwssds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpdpwssds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
   %res2 = add <16 x i32> %res, %res1
   ret <16 x i32> %res2
 }
 

llvm/test/CodeGen/X86/avx512vnni-intrinsics.ll

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpbusd_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x50,0x18]
-; X86-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x50,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x50,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x50,0x1f]
-; X64-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x50,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpbusd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x50,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.x86.avx512.vpdpbusd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.x86.avx512.vpdpbusd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpbusds_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpbusds (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x51,0x18]
-; X86-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x51,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x51,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpbusds (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x51,0x1f]
-; X64-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x51,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpbusds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x51,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.x86.avx512.vpdpbusds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.x86.avx512.vpdpbusds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpwssd_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x52,0x18]
-; X86-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x52,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x52,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x52,0x1f]
-; X64-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x52,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpwssd %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x52,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer

 
 define <16 x i32>@test_int_x86_avx512_mask_vpdpwssds_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
 ; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
 ; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X86-NEXT:    vpdpwssds (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x53,0x18]
-; X86-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x53,0xc2]
-; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X86-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x53,0xc2]
+; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X86-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
 ; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
 ; X64-NEXT:    vpdpwssds (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x53,0x1f]
-; X64-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x53,0xc2]
-; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
+; X64-NEXT:    vpdpwssds %zmm2, %zmm1, %zmm0 # encoding: [0x62,0xf2,0x75,0x48,0x53,0xc2]
+; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm3 {%k1} # encoding: [0x62,0xf1,0x65,0x49,0xfe,0xd8]
+; X64-NEXT:    vmovdqa64 %zmm3, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc3]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %x2 = load <16 x i32>, <16 x i32>* %x2p
   %1 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   %2 = bitcast i16 %x3 to <16 x i1>
   %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
   %4 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
   %5 = bitcast i16 %x3 to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   %res3 = add <16 x i32> %3, %6
   ret <16 x i32> %res3
 }