diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 7eae08e54c29..d2a5f396d64e 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -32085,13 +32085,15 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetNode(
   }
 
   // Simplify target shuffles.
-  if (!isTargetShuffle(Opc))
+  if (!isTargetShuffle(Opc) || !VT.isSimple())
     return false;
 
   // Get target shuffle mask.
+  bool IsUnary;
   SmallVector<int, 64> OpMask;
   SmallVector<SDValue, 2> OpInputs;
-  if (!resolveTargetShuffleInputs(Op, OpInputs, OpMask, TLO.DAG))
+  if (!getTargetShuffleMask(Op.getNode(), VT.getSimpleVT(), true, OpInputs,
+                            OpMask, IsUnary))
     return false;
 
   // Shuffle inputs must be the same type as the result.
@@ -32101,9 +32103,13 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetNode(
 
   // Check if shuffle mask can be simplified to undef/zero/identity.
   int NumSrcs = OpInputs.size();
-  for (int i = 0; i != NumElts; ++i)
+  for (int i = 0; i != NumElts; ++i) {
+    int &M = OpMask[i];
     if (!DemandedElts[i])
-      OpMask[i] = SM_SentinelUndef;
+      M = SM_SentinelUndef;
+    else if (0 <= M && OpInputs[M / NumElts].isUndef())
+      M = SM_SentinelUndef;
+  }
 
   if (isUndefInRange(OpMask, 0, NumElts)) {
     KnownUndef.setAllBits();
diff --git a/llvm/test/CodeGen/X86/pmul.ll b/llvm/test/CodeGen/X86/pmul.ll
index 4874935f8a06..f5f7019043af 100644
--- a/llvm/test/CodeGen/X86/pmul.ll
+++ b/llvm/test/CodeGen/X86/pmul.ll
@@ -1318,76 +1318,74 @@ entry:
 define <8 x i64> @mul_v8i64_sext(<8 x i16> %val1, <8 x i32> %val2) {
 ; SSE2-LABEL: mul_v8i64_sext:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    movdqa %xmm0, %xmm8
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm0[4],xmm6[5],xmm0[5],xmm6[6],xmm0[6],xmm6[7],xmm0[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    psrad $31, %xmm3
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm0[4],xmm12[5],xmm0[5],xmm12[6],xmm0[6],xmm12[7],xmm0[7]
+; SSE2-NEXT:    movdqa %xmm12, %xmm9
+; SSE2-NEXT:    psrad $31, %xmm9
+; SSE2-NEXT:    psrad $16, %xmm12
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    movdqa %xmm0, %xmm4
+; SSE2-NEXT:    psrad $31, %xmm4
 ; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; SSE2-NEXT:    movdqa %xmm5, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE2-NEXT:    pxor %xmm7, %xmm7
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm5, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm0, %xmm4
-; SSE2-NEXT:    paddq %xmm3, %xmm4
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[0,2,2,3,4,5,6,7]
-; SSE2-NEXT:    pmuludq %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    psrad $31, %xmm5
-; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
-; SSE2-NEXT:    psllq $32, %xmm4
-; SSE2-NEXT:    paddq %xmm4, %xmm0
-; SSE2-NEXT:    movdqa %xmm1, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm5
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm4
-; SSE2-NEXT:    paddq %xmm5, %xmm4
-; SSE2-NEXT:    movdqa %xmm6, %xmm5
-; SSE2-NEXT:    psrad $31, %xmm5
-; SSE2-NEXT:    psrad $16, %xmm6
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,0,1]
-; SSE2-NEXT:    psllq $32, %xmm4
-; SSE2-NEXT:    paddq %xmm4, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm5
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm6, %xmm4
-; SSE2-NEXT:    paddq %xmm5, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
-; SSE2-NEXT:    pmuludq %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm5, %xmm6
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm15 = xmm5[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm15, %xmm8
+; SSE2-NEXT:    psrad $31, %xmm8
+; SSE2-NEXT:    psrad $16, %xmm15
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm7, %xmm13
+; SSE2-NEXT:    psrad $31, %xmm13
+; SSE2-NEXT:    psrad $16, %xmm7
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[2,3,0,1]
+; SSE2-NEXT:    movdqa %xmm11, %xmm10
+; SSE2-NEXT:    psrad $31, %xmm10
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; SSE2-NEXT:    movdqa %xmm2, %xmm14
+; SSE2-NEXT:    psrad $31, %xmm14
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,0,1]
+; SSE2-NEXT:    movdqa %xmm3, %xmm6
 ; SSE2-NEXT:    psrad $31, %xmm6
-; SSE2-NEXT:    psrad $16, %xmm5
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
+; SSE2-NEXT:    psrad $31, %xmm5
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm1, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm5
+; SSE2-NEXT:    paddq %xmm4, %xmm5
+; SSE2-NEXT:    psllq $32, %xmm5
+; SSE2-NEXT:    pmuludq %xmm1, %xmm0
+; SSE2-NEXT:    paddq %xmm5, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm7, %xmm1
+; SSE2-NEXT:    pmuludq %xmm3, %xmm7
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm4
+; SSE2-NEXT:    paddq %xmm4, %xmm1
+; SSE2-NEXT:    psllq $32, %xmm1
+; SSE2-NEXT:    paddq %xmm7, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm12, %xmm4
+; SSE2-NEXT:    paddq %xmm3, %xmm4
 ; SSE2-NEXT:    psllq $32, %xmm4
+; SSE2-NEXT:    pmuludq %xmm12, %xmm2
 ; SSE2-NEXT:    paddq %xmm4, %xmm2
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm6
-; SSE2-NEXT:    pmuludq %xmm5, %xmm4
-; SSE2-NEXT:    paddq %xmm6, %xmm4
-; SSE2-NEXT:    pmuludq %xmm5, %xmm3
-; SSE2-NEXT:    psllq $32, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm15, %xmm3
+; SSE2-NEXT:    pmuludq %xmm11, %xmm15
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm11, %xmm4
 ; SSE2-NEXT:    paddq %xmm4, %xmm3
+; SSE2-NEXT:    psllq $32, %xmm3
+; SSE2-NEXT:    paddq %xmm15, %xmm3
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mul_v8i64_sext:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
index 64dd50639d31..b8d3824f1caf 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
@@ -124,21 +124,21 @@ define <8 x double> @combine_vpermt2var_8f64_identity(<8 x double> %x0, <8 x dou
 define <8 x double> @combine_vpermt2var_8f64_identity_mask(<8 x double> %x0, <8 x double> %x1, i8 %m) {
 ; X32-LABEL: combine_vpermt2var_8f64_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovapd {{.*#+}} zmm2 = [7,0,6,0,5,0,4,0,3,0,2,0,1,0,0,0]
+; X32-NEXT:    vmovapd {{.*#+}} zmm1 = [7,0,6,0,5,0,4,0,3,0,2,0,1,0,0,0]
 ; X32-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    kmovd %eax, %k1
-; X32-NEXT:    vpermi2pd %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X32-NEXT:    vpermi2pd %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X32-NEXT:    vmovapd {{.*#+}} zmm0 = [7,0,14,0,5,0,12,0,3,0,10,0,1,0,8,0]
-; X32-NEXT:    vpermi2pd %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermi2pd %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_8f64_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovapd {{.*#+}} zmm2 = [7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovapd {{.*#+}} zmm1 = [7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2pd %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X64-NEXT:    vpermi2pd %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X64-NEXT:    vmovapd {{.*#+}} zmm0 = [7,14,5,12,3,10,1,8]
-; X64-NEXT:    vpermi2pd %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermi2pd %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <8 x double> @llvm.x86.avx512.maskz.vpermt2var.pd.512(<8 x i64> <i64 7, i64 6, i64 5, i64 4, i64 3, i64 2, i64 1, i64 0>, <8 x double> %x0, <8 x double> %x1, i8 %m)
   %res1 = call <8 x double> @llvm.x86.avx512.maskz.vpermt2var.pd.512(<8 x i64> <i64 7, i64 14, i64 5, i64 12, i64 3, i64 10, i64 1, i64 8>, <8 x double> %res0, <8 x double> %res0, i8 %m)
@@ -205,21 +205,21 @@ define <8 x i64> @combine_vpermt2var_8i64_identity(<8 x i64> %x0, <8 x i64> %x1)
 define <8 x i64> @combine_vpermt2var_8i64_identity_mask(<8 x i64> %x0, <8 x i64> %x1, i8 %m) {
 ; X32-LABEL: combine_vpermt2var_8i64_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [7,0,6,0,5,0,4,0,3,0,2,0,1,0,0,0]
+; X32-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [7,0,6,0,5,0,4,0,3,0,2,0,1,0,0,0]
 ; X32-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    kmovd %eax, %k1
-; X32-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X32-NEXT:    vpermi2q %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X32-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [7,0,14,0,5,0,12,0,3,0,10,0,1,0,8,0]
-; X32-NEXT:    vpermi2q %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermi2q %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_8i64_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X64-NEXT:    vpermi2q %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [7,14,5,12,3,10,1,8]
-; X64-NEXT:    vpermi2q %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermi2q %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <8 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.512(<8 x i64> <i64 7, i64 6, i64 5, i64 4, i64 3, i64 2, i64 1, i64 0>, <8 x i64> %x0, <8 x i64> %x1, i8 %m)
   %res1 = call <8 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.512(<8 x i64> <i64 7, i64 14, i64 5, i64 12, i64 3, i64 10, i64 1, i64 8>, <8 x i64> %res0, <8 x i64> %res0, i8 %m)
@@ -241,20 +241,20 @@ define <16 x float> @combine_vpermt2var_16f32_identity(<16 x float> %x0, <16 x f
 define <16 x float> @combine_vpermt2var_16f32_identity_mask(<16 x float> %x0, <16 x float> %x1, i16 %m) {
 ; X32-LABEL: combine_vpermt2var_16f32_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovaps {{.*#+}} zmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X32-NEXT:    vmovaps {{.*#+}} zmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2ps %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X32-NEXT:    vpermi2ps %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X32-NEXT:    vmovaps {{.*#+}} zmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X32-NEXT:    vpermi2ps %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermi2ps %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16f32_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovaps {{.*#+}} zmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovaps {{.*#+}} zmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2ps %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X64-NEXT:    vpermi2ps %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X64-NEXT:    vmovaps {{.*#+}} zmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X64-NEXT:    vpermi2ps %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermi2ps %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>, <16 x float> %x0, <16 x float> %x1, i16 %m)
   %res1 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 15, i32 30, i32 13, i32 28, i32 11, i32 26, i32 9, i32 24, i32 7, i32 22, i32 5, i32 20, i32 3, i32 18, i32 1, i32 16>, <16 x float> %res0, <16 x float> %res0, i16 %m)
@@ -264,14 +264,14 @@ define <16 x float> @combine_vpermt2var_16f32_identity_mask(<16 x float> %x0, <1
 define <16 x float> @combine_vpermt2var_16f32_vmovddup(<16 x float> %x0, <16 x float> %x1) {
 ; X32-LABEL: combine_vpermt2var_16f32_vmovddup:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovaps {{.*#+}} zmm2 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
-; X32-NEXT:    vpermt2ps %zmm1, %zmm2, %zmm0
+; X32-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X32-NEXT:    vpermt2ps %zmm0, %zmm1, %zmm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16f32_vmovddup:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovaps {{.*#+}} zmm2 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
-; X64-NEXT:    vpermt2ps %zmm1, %zmm2, %zmm0
+; X64-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X64-NEXT:    vpermt2ps %zmm0, %zmm1, %zmm0
 ; X64-NEXT:    retq
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5, i32 8, i32 9, i32 8, i32 9, i32 12, i32 13, i32 12, i32 13>, <16 x float> %x0, <16 x float> %x1, i16 -1)
   ret <16 x float> %res0
@@ -280,18 +280,16 @@ define <16 x float> @combine_vpermt2var_16f32_vmovddup_load(<16 x float> *%p0, <
 ; X32-LABEL: combine_vpermt2var_16f32_vmovddup_load:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    vmovaps (%eax), %zmm2
-; X32-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
-; X32-NEXT:    vpermi2ps %zmm0, %zmm2, %zmm1
-; X32-NEXT:    vmovaps %zmm1, %zmm0
+; X32-NEXT:    vmovaps (%eax), %zmm1
+; X32-NEXT:    vmovaps {{.*#+}} zmm0 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X32-NEXT:    vpermi2ps %zmm0, %zmm1, %zmm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16f32_vmovddup_load:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovaps (%rdi), %zmm2
-; X64-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
-; X64-NEXT:    vpermi2ps %zmm0, %zmm2, %zmm1
-; X64-NEXT:    vmovaps %zmm1, %zmm0
+; X64-NEXT:    vmovaps (%rdi), %zmm1
+; X64-NEXT:    vmovaps {{.*#+}} zmm0 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X64-NEXT:    vpermi2ps %zmm0, %zmm1, %zmm0
 ; X64-NEXT:    retq
   %x0 = load <16 x float>, <16 x float> *%p0
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5, i32 8, i32 9, i32 8, i32 9, i32 12, i32 13, i32 12, i32 13>, <16 x float> %x0, <16 x float> %x1, i16 -1)
@@ -300,16 +298,16 @@ define <16 x float> @combine_vpermt2var_16f32_vmovddup_load(<16 x float> *%p0, <
 define <16 x float> @combine_vpermt2var_16f32_vmovddup_mask(<16 x float> %x0, <16 x float> %x1, i16 %m) {
 ; X32-LABEL: combine_vpermt2var_16f32_vmovddup_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovaps {{.*#+}} zmm2 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X32-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermt2ps %zmm1, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermt2ps %zmm0, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16f32_vmovddup_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovaps {{.*#+}} zmm2 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X64-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermt2ps %zmm1, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermt2ps %zmm0, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5, i32 8, i32 9, i32 8, i32 9, i32 12, i32 13, i32 12, i32 13>, <16 x float> %x0, <16 x float> %x1, i16 %m)
   ret <16 x float> %res0
@@ -318,20 +316,18 @@ define <16 x float> @combine_vpermt2var_16f32_vmovddup_mask_load(<16 x float> *%
 ; X32-LABEL: combine_vpermt2var_16f32_vmovddup_mask_load:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    vmovaps (%eax), %zmm2
-; X32-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X32-NEXT:    vmovaps (%eax), %zmm1
+; X32-NEXT:    vmovaps {{.*#+}} zmm0 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2ps %zmm0, %zmm2, %zmm1 {%k1} {z}
-; X32-NEXT:    vmovaps %zmm1, %zmm0
+; X32-NEXT:    vpermi2ps %zmm0, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16f32_vmovddup_mask_load:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovaps (%rdi), %zmm2
-; X64-NEXT:    vmovaps {{.*#+}} zmm1 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
+; X64-NEXT:    vmovaps (%rdi), %zmm1
+; X64-NEXT:    vmovaps {{.*#+}} zmm0 = [0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13]
 ; X64-NEXT:    kmovd %esi, %k1
-; X64-NEXT:    vpermi2ps %zmm0, %zmm2, %zmm1 {%k1} {z}
-; X64-NEXT:    vmovaps %zmm1, %zmm0
+; X64-NEXT:    vpermi2ps %zmm0, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %x0 = load <16 x float>, <16 x float> *%p0
   %res0 = call <16 x float> @llvm.x86.avx512.maskz.vpermt2var.ps.512(<16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5, i32 8, i32 9, i32 8, i32 9, i32 12, i32 13, i32 12, i32 13>, <16 x float> %x0, <16 x float> %x1, i16 %m)
@@ -519,20 +515,20 @@ define <16 x i32> @combine_vpermt2var_16i32_identity(<16 x i32> %x0, <16 x i32>
 define <16 x i32> @combine_vpermt2var_16i32_identity_mask(<16 x i32> %x0, <16 x i32> %x1, i16 %m) {
 ; X32-LABEL: combine_vpermt2var_16i32_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X32-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X32-NEXT:    vpermi2d %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X32-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X32-NEXT:    vpermi2d %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermi2d %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16i32_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X64-NEXT:    vpermi2d %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X64-NEXT:    vpermi2d %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermi2d %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <16 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.512(<16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>, <16 x i32> %x0, <16 x i32> %x1, i16 %m)
   %res1 = call <16 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.512(<16 x i32> <i32 15, i32 30, i32 13, i32 28, i32 11, i32 26, i32 9, i32 24, i32 7, i32 22, i32 5, i32 20, i32 3, i32 18, i32 1, i32 16>, <16 x i32> %res0, <16 x i32> %res0, i16 %m)
@@ -554,20 +550,20 @@ define <32 x i16> @combine_vpermt2var_32i16_identity(<32 x i16> %x0, <32 x i16>
 define <32 x i16> @combine_vpermt2var_32i16_identity_mask(<32 x i16> %x0, <32 x i16> %x1, i32 %m) {
 ; X32-LABEL: combine_vpermt2var_32i16_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X32-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2w %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X32-NEXT:    vpermi2w %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X32-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [63,30,61,28,59,26,57,24,55,22,53,20,51,18,49,16,47,46,13,44,11,42,9,40,7,38,5,36,3,34,1,32]
-; X32-NEXT:    vpermi2w %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X32-NEXT:    vpermi2w %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_32i16_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2w %zmm1, %zmm0, %zmm2 {%k1} {z}
+; X64-NEXT:    vpermi2w %zmm0, %zmm0, %zmm1 {%k1} {z}
 ; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [63,30,61,28,59,26,57,24,55,22,53,20,51,18,49,16,47,46,13,44,11,42,9,40,7,38,5,36,3,34,1,32]
-; X64-NEXT:    vpermi2w %zmm2, %zmm2, %zmm0 {%k1} {z}
+; X64-NEXT:    vpermi2w %zmm1, %zmm1, %zmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <32 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.512(<32 x i16> <i16 31, i16 30, i16 29, i16 28, i16 27, i16 26, i16 25, i16 24, i16 23, i16 22, i16 21, i16 20, i16 19, i16 18, i16 17, i16 16, i16 15, i16 14, i16 13, i16 12, i16 11, i16 10, i16 9, i16 8, i16 7, i16 6, i16 5, i16 4, i16 3, i16 2, i16 1, i16 0>, <32 x i16> %x0, <32 x i16> %x1, i32 %m)
   %res1 = call <32 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.512(<32 x i16> <i16 63, i16 30, i16 61, i16 28, i16 59, i16 26, i16 57, i16 24, i16 55, i16 22, i16 53, i16 20, i16 51, i16 18, i16 49, i16 16, i16 47, i16 46, i16 13, i16 44, i16 11, i16 42, i16 9, i16 40, i16 7, i16 38, i16 5, i16 36, i16 3, i16 34, i16 1, i16 32>, <32 x i16> %res0, <32 x i16> %res0, i32 %m)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
index 0d56a24c5ff3..c4ab922d068c 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
@@ -20,20 +20,20 @@ define <16 x i16> @combine_vpermt2var_16i16_identity(<16 x i16> %x0, <16 x i16>
 define <16 x i16> @combine_vpermt2var_16i16_identity_mask(<16 x i16> %x0, <16 x i16> %x1, i16 %m) {
 ; X32-LABEL: combine_vpermt2var_16i16_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X32-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2w %ymm1, %ymm0, %ymm2 {%k1} {z}
+; X32-NEXT:    vpermi2w %ymm0, %ymm0, %ymm1 {%k1} {z}
 ; X32-NEXT:    vmovdqa {{.*#+}} ymm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X32-NEXT:    vpermi2w %ymm2, %ymm2, %ymm0 {%k1} {z}
+; X32-NEXT:    vpermi2w %ymm1, %ymm1, %ymm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16i16_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2w %ymm1, %ymm0, %ymm2 {%k1} {z}
+; X64-NEXT:    vpermi2w %ymm0, %ymm0, %ymm1 {%k1} {z}
 ; X64-NEXT:    vmovdqa {{.*#+}} ymm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X64-NEXT:    vpermi2w %ymm2, %ymm2, %ymm0 {%k1} {z}
+; X64-NEXT:    vpermi2w %ymm1, %ymm1, %ymm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> <i16 15, i16 14, i16 13, i16 12, i16 11, i16 10, i16 9, i16 8, i16 7, i16 6, i16 5, i16 4, i16 3, i16 2, i16 1, i16 0>, <16 x i16> %x0, <16 x i16> %x1, i16 %m)
   %res1 = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> <i16 15, i16 30, i16 13, i16 28, i16 11, i16 26, i16 9, i16 24, i16 7, i16 22, i16 5, i16 20, i16 3, i16 18, i16 1, i16 16>, <16 x i16> %res0, <16 x i16> %res0, i16 %m)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
index a9054102460c..5350ddab5ef7 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
@@ -33,20 +33,20 @@ define <16 x i8> @combine_vpermt2var_16i8_identity(<16 x i8> %x0, <16 x i8> %x1)
 define <16 x i8> @combine_vpermt2var_16i8_identity_mask(<16 x i8> %x0, <16 x i8> %x1, i16 %m) {
 ; X32-LABEL: combine_vpermt2var_16i8_identity_mask:
 ; X32:       # %bb.0:
-; X32-NEXT:    vmovdqa {{.*#+}} xmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X32-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
-; X32-NEXT:    vpermi2b %xmm1, %xmm0, %xmm2 {%k1} {z}
+; X32-NEXT:    vpermi2b %xmm0, %xmm0, %xmm1 {%k1} {z}
 ; X32-NEXT:    vmovdqa {{.*#+}} xmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X32-NEXT:    vpermi2b %xmm2, %xmm2, %xmm0 {%k1} {z}
+; X32-NEXT:    vpermi2b %xmm1, %xmm1, %xmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermt2var_16i8_identity_mask:
 ; X64:       # %bb.0:
-; X64-NEXT:    vmovdqa {{.*#+}} xmm2 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
+; X64-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0]
 ; X64-NEXT:    kmovd %edi, %k1
-; X64-NEXT:    vpermi2b %xmm1, %xmm0, %xmm2 {%k1} {z}
+; X64-NEXT:    vpermi2b %xmm0, %xmm0, %xmm1 {%k1} {z}
 ; X64-NEXT:    vmovdqa {{.*#+}} xmm0 = [15,30,13,28,11,26,9,24,7,22,5,20,3,18,1,16]
-; X64-NEXT:    vpermi2b %xmm2, %xmm2, %xmm0 {%k1} {z}
+; X64-NEXT:    vpermi2b %xmm1, %xmm1, %xmm0 {%k1} {z}
 ; X64-NEXT:    retq
   %res0 = call <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>, <16 x i8> %x0, <16 x i8> %x1, i16 %m)
   %res1 = call <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8> <i8 15, i8 30, i8 13, i8 28, i8 11, i8 26, i8 9, i8 24, i8 7, i8 22, i8 5, i8 20, i8 3, i8 18, i8 1, i8 16>, <16 x i8> %res0, <16 x i8> %res0, i16 %m)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
index b8f572f0c8d8..199a05eb2b8f 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -2827,16 +2827,13 @@ define <4 x float> @PR30264(<4 x float> %x) {
 define <8 x i16> @PR39549(<16 x i8> %x) {
 ; SSE-LABEL: PR39549:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
-; SSE-NEXT:    psraw $8, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    psraw $8, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: PR39549:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
+; AVX-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; AVX-NEXT:    vpsraw $8, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %a = shufflevector <16 x i8> %x, <16 x i8> undef, <16 x i32> <i32 8, i32 undef, i32 9, i32 undef, i32 10, i32 undef, i32 11, i32 undef, i32 12, i32 undef, i32 13, i32 undef, i32 14, i32 undef, i32 15, i32 undef>
diff --git a/llvm/test/CodeGen/X86/vector-trunc-math-widen.ll b/llvm/test/CodeGen/X86/vector-trunc-math-widen.ll
index c5095ae918ac..8f6a1fe57260 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-math-widen.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-math-widen.ll
@@ -5607,40 +5607,39 @@ define <4 x i32> @mul_add_const_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwi
 define <4 x i32> @mul_add_self_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwind {
 ; SSE-LABEL: mul_add_self_v4i64_v4i32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    psrad $31, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    psrad $31, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    psrad $31, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
 ; SSE-NEXT:    psrad $31, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,0,1]
+; SSE-NEXT:    movdqa %xmm6, %xmm7
 ; SSE-NEXT:    psrad $31, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT:    pxor %xmm8, %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm1, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm0, %xmm7
-; SSE-NEXT:    paddq %xmm6, %xmm7
-; SSE-NEXT:    psllq $32, %xmm7
-; SSE-NEXT:    pmuludq %xmm0, %xmm1
-; SSE-NEXT:    paddq %xmm7, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm4, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm2, %xmm5
-; SSE-NEXT:    paddq %xmm3, %xmm5
-; SSE-NEXT:    psllq $32, %xmm5
-; SSE-NEXT:    pmuludq %xmm2, %xmm4
-; SSE-NEXT:    paddq %xmm5, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
-; SSE-NEXT:    paddd %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    psrad $31, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm0, %xmm2
+; SSE-NEXT:    pmuludq %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm1, %xmm5
+; SSE-NEXT:    paddq %xmm5, %xmm2
+; SSE-NEXT:    psllq $32, %xmm2
+; SSE-NEXT:    paddq %xmm0, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm3, %xmm0
+; SSE-NEXT:    pmuludq %xmm6, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm6, %xmm1
+; SSE-NEXT:    paddq %xmm1, %xmm0
+; SSE-NEXT:    psllq $32, %xmm0
+; SSE-NEXT:    paddq %xmm3, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
+; SSE-NEXT:    paddd %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: mul_add_self_v4i64_v4i32:
diff --git a/llvm/test/CodeGen/X86/vector-trunc-math.ll b/llvm/test/CodeGen/X86/vector-trunc-math.ll
index b8aedf985a54..97c4f3da409f 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-math.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-math.ll
@@ -5607,40 +5607,39 @@ define <4 x i32> @mul_add_const_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwi
 define <4 x i32> @mul_add_self_v4i64_v4i32(<4 x i32> %a0, <4 x i32> %a1) nounwind {
 ; SSE-LABEL: mul_add_self_v4i64_v4i32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    psrad $31, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    psrad $31, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,0,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    psrad $31, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
 ; SSE-NEXT:    psrad $31, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,0,1]
+; SSE-NEXT:    movdqa %xmm6, %xmm7
 ; SSE-NEXT:    psrad $31, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT:    pxor %xmm8, %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm1, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm0, %xmm7
-; SSE-NEXT:    paddq %xmm6, %xmm7
-; SSE-NEXT:    psllq $32, %xmm7
-; SSE-NEXT:    pmuludq %xmm0, %xmm1
-; SSE-NEXT:    paddq %xmm7, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm4, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
-; SSE-NEXT:    pmuludq %xmm2, %xmm5
-; SSE-NEXT:    paddq %xmm3, %xmm5
-; SSE-NEXT:    psllq $32, %xmm5
-; SSE-NEXT:    pmuludq %xmm2, %xmm4
-; SSE-NEXT:    paddq %xmm5, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
-; SSE-NEXT:    paddd %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    psrad $31, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm0, %xmm2
+; SSE-NEXT:    pmuludq %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm1, %xmm5
+; SSE-NEXT:    paddq %xmm5, %xmm2
+; SSE-NEXT:    psllq $32, %xmm2
+; SSE-NEXT:    paddq %xmm0, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm3, %xmm0
+; SSE-NEXT:    pmuludq %xmm6, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
+; SSE-NEXT:    pmuludq %xmm6, %xmm1
+; SSE-NEXT:    paddq %xmm1, %xmm0
+; SSE-NEXT:    psllq $32, %xmm0
+; SSE-NEXT:    paddq %xmm3, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
+; SSE-NEXT:    paddd %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: mul_add_self_v4i64_v4i32: