From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 31 Mar 2020 13:37:48 +0000 (+0100)
Subject: [X86][SSE] lowerShuffleWithPACK - extend to use chained PACKs for larger truncations
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=efe59d6717dcdf7777acb9b7a734e1a520bdf22a;p=platform%2Fupstream%2Fllvm.git

[X86][SSE] lowerShuffleWithPACK - extend to use chained PACKs for larger truncations

If canLowerByDroppingEvenElements indicates that the shuffle is a N:1 compaction pattern and the inputs are suitably sign/zero extended then we can use a chain of PACKSS/PACKUS to compact.

This helps avoid PSHUFB (and its mask load) for short shuffle chains, shuffle combining will still replace with a PSHUFB if we have enough shuffles as getFauxShuffleMask can recognise PACKSS/PACKUS chains.
---

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index c7c599c..7f2046d 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -11358,6 +11358,72 @@ static SDValue lowerShuffleWithPACK(const SDLoc &DL, MVT VT, ArrayRef<int> Mask,
     return DAG.getNode(PackOpcode, DL, VT, DAG.getBitcast(PackVT, V1),
                        DAG.getBitcast(PackVT, V2));
 
+  // See if we can detect a compaction mask that we can lower with a chain of
+  // PACKUS/PACKSS.
+  // TODO: Handle 256/512-bit types.
+  // TODO: Can this be merged into matchShuffleWithPACK?
+  if ((VT == MVT::v8i16 || VT == MVT::v16i8) && !Subtarget.hasAVX512()) {
+    if (int NumEvenDrops = canLowerByDroppingEvenElements(Mask, V2.isUndef())) {
+      if (VT == MVT::v8i16 && NumEvenDrops == 3)
+        return SDValue();
+      unsigned NumElts = VT.getVectorNumElements();
+      unsigned NumDstEltBits = VT.getScalarSizeInBits();
+      unsigned NumSrcEltBits = NumDstEltBits << NumEvenDrops;
+      unsigned NumPackedEltBits = NumSrcEltBits - NumDstEltBits;
+      MVT SrcSVT = MVT::getIntegerVT(NumSrcEltBits);
+      MVT SrcVT = MVT::getVectorVT(SrcSVT, 128 / NumSrcEltBits);
+      SDValue Src1 = DAG.getBitcast(SrcVT, V1);
+      SDValue Src2 = DAG.getBitcast(SrcVT, V2.isUndef() ? V1 : V2);
+
+      // Try to use PACKUS - PACKUSDW (SSE41+ only).
+      unsigned PackOpcode = ISD::DELETED_NODE;
+      if (Subtarget.hasSSE41() || NumDstEltBits == 8) {
+        KnownBits KnownV1 = DAG.computeKnownBits(Src1);
+        KnownBits KnownV2 = DAG.computeKnownBits(Src2);
+        if (KnownV1.countMinLeadingZeros() >= NumPackedEltBits &&
+            KnownV2.countMinLeadingZeros() >= NumPackedEltBits) {
+          PackOpcode = X86ISD::PACKUS;
+        }
+      }
+
+      // Try to use PACKSS.
+      if (PackOpcode == ISD::DELETED_NODE &&
+          DAG.ComputeNumSignBits(Src1) > NumPackedEltBits &&
+          DAG.ComputeNumSignBits(Src2) > NumPackedEltBits) {
+        PackOpcode = X86ISD::PACKSS;
+      }
+
+      // Repeatedly pack until we're down to the target size.
+      if (PackOpcode != ISD::DELETED_NODE) {
+        SDValue Res;
+        unsigned NumPackEltBits = NumSrcEltBits;
+        for (int i = 0; i < NumEvenDrops; ++i, NumPackEltBits >>= 1) {
+          // Pack to the largest type possible:
+          // vXi64/vXi32 -> PACK*SDW and vXi16 -> PACK*SWB.
+          MVT InSVT = MVT::i16, OutSVT = MVT::i8;
+          if (NumPackEltBits > 16 &&
+              (PackOpcode == X86ISD::PACKSS || Subtarget.hasSSE41())) {
+            InSVT = MVT::i32;
+            OutSVT = MVT::i16;
+          }
+          MVT InVT = MVT::getVectorVT(InSVT, 128 / InSVT.getSizeInBits());
+          MVT OutVT = MVT::getVectorVT(OutSVT, 128 / OutSVT.getSizeInBits());
+          Src1 = DAG.getBitcast(InVT, i == 0 ? Src1 : Res);
+          Src2 = DAG.getBitcast(InVT, i == 0 ? Src2 : Res);
+          // If the mask doesn't need these packed elements, just use UNDEF.
+          unsigned UpperElts = NumElts >> (i + 1);
+          if (isUndefInRange(Mask, 0, UpperElts))
+            Src1 = DAG.getUNDEF(InVT);
+          if (isUndefInRange(Mask, UpperElts, NumElts - UpperElts))
+            Src2 = DAG.getUNDEF(InVT);
+          Res = DAG.getNode(PackOpcode, DL, OutVT, Src1, Src2);
+        }
+        assert(Res && Res.getValueType() == VT && "Failed to pack");
+        return Res;
+      }
+    }
+  }
+
   return SDValue();
 }
 
diff --git a/llvm/test/CodeGen/X86/avx-fp2int.ll b/llvm/test/CodeGen/X86/avx-fp2int.ll
index ac5fcfe..195b6f6 100644
--- a/llvm/test/CodeGen/X86/avx-fp2int.ll
+++ b/llvm/test/CodeGen/X86/avx-fp2int.ll
@@ -7,7 +7,8 @@ define <4 x i8> @test1(<4 x double> %d) {
 ; CHECK-LABEL: test1:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vcvttpd2dq %ymm0, %xmm0
-; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retl
   %c = fptoui <4 x double> %d to <4 x i8>
@@ -17,7 +18,8 @@ define <4 x i8> @test2(<4 x double> %d) {
 ; CHECK-LABEL: test2:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vcvttpd2dq %ymm0, %xmm0
-; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retl
   %c = fptosi <4 x double> %d to <4 x i8>
diff --git a/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll b/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
index 3852b0e..10d5975 100644
--- a/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
+++ b/llvm/test/CodeGen/X86/bitcast-and-setcc-512.ll
@@ -8,43 +8,25 @@
 define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
 ; SSE-LABEL: v8i64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm2
 ; SSE-NEXT:    pcmpgtq %xmm7, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    pcmpgtq %xmm6, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    pcmpgtq %xmm5, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    packssdw %xmm3, %xmm3
+; SSE-NEXT:    packssdw %xmm3, %xmm3
 ; SSE-NEXT:    pcmpgtq %xmm4, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm2[4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    packsswb %xmm0, %xmm3
-; SSE-NEXT:    pmovmskb %xmm3, %eax
+; SSE-NEXT:    packssdw %xmm0, %xmm0
+; SSE-NEXT:    packssdw %xmm0, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm2
+; SSE-NEXT:    packssdw %xmm2, %xmm2
+; SSE-NEXT:    packssdw %xmm2, %xmm2
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
 ; SSE-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE-NEXT:    retq
 ;
@@ -120,43 +102,25 @@ define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
 define i8 @v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double> %d) {
 ; SSE-LABEL: v8f64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm2
 ; SSE-NEXT:    cmpltpd %xmm3, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    cmpltpd %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    cmpltpd %xmm1, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    packssdw %xmm7, %xmm3
+; SSE-NEXT:    packssdw %xmm3, %xmm3
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm10[0,1,0,2,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm2[4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    packsswb %xmm0, %xmm3
-; SSE-NEXT:    pmovmskb %xmm3, %eax
+; SSE-NEXT:    packssdw %xmm0, %xmm4
+; SSE-NEXT:    packssdw %xmm0, %xmm4
+; SSE-NEXT:    pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm3[4,5,6,7]
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm2
+; SSE-NEXT:    packssdw %xmm2, %xmm0
+; SSE-NEXT:    packssdw %xmm0, %xmm0
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; SSE-NEXT:    pand %xmm4, %xmm1
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
 ; SSE-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE-NEXT:    retq
 ;
@@ -336,36 +300,24 @@ define i32 @v32i16(<32 x i16> %a, <32 x i16> %b, <32 x i16> %c, <32 x i16> %d) {
 define i16 @v16i32(<16 x i32> %a, <16 x i32> %b, <16 x i32> %c, <16 x i32> %d) {
 ; SSE-LABEL: v16i32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm2
 ; SSE-NEXT:    pcmpgtd %xmm7, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = <u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm7, %xmm3
-; SSE-NEXT:    pcmpgtd %xmm6, %xmm2
-; SSE-NEXT:    pshufb %xmm7, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    pcmpgtd %xmm5, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm3, %xmm1
+; SSE-NEXT:    packssdw %xmm3, %xmm3
+; SSE-NEXT:    packsswb %xmm3, %xmm3
 ; SSE-NEXT:    pcmpgtd %xmm4, %xmm0
-; SSE-NEXT:    pshufb %xmm3, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufb %xmm7, %xmm11
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufb %xmm7, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufb %xmm3, %xmm10
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufb %xmm3, %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm8 = xmm8[0,1,2,3],xmm9[4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pmovmskb %xmm8, %eax
+; SSE-NEXT:    packssdw %xmm0, %xmm0
+; SSE-NEXT:    packsswb %xmm0, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm2
+; SSE-NEXT:    packssdw %xmm2, %xmm2
+; SSE-NEXT:    packsswb %xmm2, %xmm2
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
 ; SSE-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE-NEXT:    retq
 ;
@@ -448,36 +400,24 @@ define i16 @v16i32(<16 x i32> %a, <16 x i32> %b, <16 x i32> %c, <16 x i32> %d) {
 define i16 @v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x float> %d) {
 ; SSE-LABEL: v16f32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm2
 ; SSE-NEXT:    cmpltps %xmm3, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = <u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm3, %xmm7
-; SSE-NEXT:    cmpltps %xmm2, %xmm6
-; SSE-NEXT:    pshufb %xmm3, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
-; SSE-NEXT:    cmpltps %xmm1, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm1, %xmm5
+; SSE-NEXT:    packssdw %xmm7, %xmm3
+; SSE-NEXT:    packsswb %xmm3, %xmm3
 ; SSE-NEXT:    cmpltps %xmm0, %xmm4
-; SSE-NEXT:    pshufb %xmm1, %xmm4
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm6[4,5,6,7]
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufb %xmm3, %xmm11
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufb %xmm3, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufb %xmm1, %xmm10
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufb %xmm1, %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
-; SSE-NEXT:    pblendw {{.*#+}} xmm8 = xmm8[0,1,2,3],xmm9[4,5,6,7]
-; SSE-NEXT:    pand %xmm4, %xmm8
-; SSE-NEXT:    pmovmskb %xmm8, %eax
+; SSE-NEXT:    packssdw %xmm0, %xmm4
+; SSE-NEXT:    packsswb %xmm0, %xmm4
+; SSE-NEXT:    pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm3[4,5,6,7]
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm2
+; SSE-NEXT:    packssdw %xmm2, %xmm0
+; SSE-NEXT:    packsswb %xmm0, %xmm0
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    packssdw %xmm0, %xmm1
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; SSE-NEXT:    pand %xmm4, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
 ; SSE-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
index b82d781..f2f9cc3 100644
--- a/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_ssat.ll
@@ -5263,7 +5263,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; SSE4-NEXT:    pxor %xmm2, %xmm2
 ; SSE4-NEXT:    pminsd {{.*}}(%rip), %xmm0
 ; SSE4-NEXT:    pmaxsd {{.*}}(%rip), %xmm0
-; SSE4-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE4-NEXT:    packssdw %xmm0, %xmm0
+; SSE4-NEXT:    packsswb %xmm0, %xmm0
 ; SSE4-NEXT:    pcmpeqd %xmm1, %xmm2
 ; SSE4-NEXT:    movmskps %xmm2, %eax
 ; SSE4-NEXT:    xorl $15, %eax
@@ -5301,7 +5302,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpminsd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpmaxsd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vmovmskps %xmm1, %eax
 ; AVX1-NEXT:    xorl $15, %eax
@@ -5341,7 +5343,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; AVX2-NEXT:    vpminsd %xmm3, %xmm0, %xmm0
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [4294967168,4294967168,4294967168,4294967168]
 ; AVX2-NEXT:    vpmaxsd %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpeqd %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vmovmskps %xmm1, %eax
 ; AVX2-NEXT:    xorl $15, %eax
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
index 84958fe..e1ee5d1 100644
--- a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
+++ b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll
@@ -4436,19 +4436,15 @@ define void @truncstore_v8i32_v8i8(<8 x i32> %x, <8 x i8>* %p, <8 x i32> %mask)
 ;
 ; SSE4-LABEL: truncstore_v8i32_v8i8:
 ; SSE4:       # %bb.0:
-; SSE4-NEXT:    pxor %xmm4, %xmm4
-; SSE4-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255]
-; SSE4-NEXT:    pminud %xmm5, %xmm0
-; SSE4-NEXT:    pminud %xmm5, %xmm1
-; SSE4-NEXT:    movdqa {{.*#+}} xmm5 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE4-NEXT:    pshufb %xmm5, %xmm1
-; SSE4-NEXT:    pshufb %xmm5, %xmm0
-; SSE4-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE4-NEXT:    pcmpeqd %xmm4, %xmm3
-; SSE4-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT:    pxor %xmm1, %xmm3
-; SSE4-NEXT:    pcmpeqd %xmm4, %xmm2
-; SSE4-NEXT:    pxor %xmm1, %xmm2
+; SSE4-NEXT:    pxor %xmm1, %xmm1
+; SSE4-NEXT:    pminud {{.*}}(%rip), %xmm0
+; SSE4-NEXT:    packusdw %xmm0, %xmm0
+; SSE4-NEXT:    packuswb %xmm0, %xmm0
+; SSE4-NEXT:    pcmpeqd %xmm1, %xmm3
+; SSE4-NEXT:    pcmpeqd %xmm4, %xmm4
+; SSE4-NEXT:    pxor %xmm4, %xmm3
+; SSE4-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE4-NEXT:    pxor %xmm4, %xmm2
 ; SSE4-NEXT:    packssdw %xmm3, %xmm2
 ; SSE4-NEXT:    packsswb %xmm0, %xmm2
 ; SSE4-NEXT:    pmovmskb %xmm2, %eax
@@ -4511,14 +4507,9 @@ define void @truncstore_v8i32_v8i8(<8 x i32> %x, <8 x i8>* %p, <8 x i32> %mask)
 ;
 ; AVX1-LABEL: truncstore_v8i32_v8i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [255,255,255,255]
-; AVX1-NEXT:    vpminud %xmm2, %xmm0, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpminud %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm2, %xmm2
@@ -4588,13 +4579,10 @@ define void @truncstore_v8i32_v8i8(<8 x i32> %x, <8 x i8>* %p, <8 x i32> %mask)
 ; AVX2-LABEL: truncstore_v8i32_v8i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpminud %ymm3, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
-; AVX2-NEXT:    vpshufb %xmm4, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [255,255,255,255]
+; AVX2-NEXT:    vpminud %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vmovmskps %ymm1, %eax
 ; AVX2-NEXT:    notl %eax
@@ -5030,7 +5018,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    pxor %xmm2, %xmm2
 ; SSE4-NEXT:    pminud {{.*}}(%rip), %xmm0
-; SSE4-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE4-NEXT:    packusdw %xmm0, %xmm0
+; SSE4-NEXT:    packuswb %xmm0, %xmm0
 ; SSE4-NEXT:    pcmpeqd %xmm1, %xmm2
 ; SSE4-NEXT:    movmskps %xmm2, %eax
 ; SSE4-NEXT:    xorl $15, %eax
@@ -5067,7 +5056,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vmovmskps %xmm1, %eax
 ; AVX1-NEXT:    xorl $15, %eax
@@ -5105,7 +5095,8 @@ define void @truncstore_v4i32_v4i8(<4 x i32> %x, <4 x i8>* %p, <4 x i32> %mask)
 ; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [255,255,255,255]
 ; AVX2-NEXT:    vpminud %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpeqd %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vmovmskps %xmm1, %eax
 ; AVX2-NEXT:    xorl $15, %eax
diff --git a/llvm/test/CodeGen/X86/psubus.ll b/llvm/test/CodeGen/X86/psubus.ll
index 2903cbd..d4ada24 100644
--- a/llvm/test/CodeGen/X86/psubus.ll
+++ b/llvm/test/CodeGen/X86/psubus.ll
@@ -607,49 +607,37 @@ define <16 x i8> @test14(<16 x i8> %x, <16 x i32> %y) nounwind {
 ; SSE2:       # %bb.0: # %vector.ph
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    movdqa %xmm5, %xmm6
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3],xmm6[4],xmm0[4],xmm6[5],xmm0[5],xmm6[6],xmm0[6],xmm6[7],xmm0[7]
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
+; SSE2-NEXT:    movdqa %xmm5, %xmm7
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
+; SSE2-NEXT:    movdqa %xmm7, %xmm8
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm0[4],xmm6[5],xmm0[5],xmm6[6],xmm0[6],xmm6[7],xmm0[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; SSE2-NEXT:    movdqa %xmm5, %xmm10
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3],xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE2-NEXT:    movdqa %xmm5, %xmm9
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm4, %xmm9
-; SSE2-NEXT:    pxor %xmm0, %xmm9
-; SSE2-NEXT:    psubd %xmm5, %xmm4
-; SSE2-NEXT:    por %xmm0, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm9, %xmm5
-; SSE2-NEXT:    movdqa {{.*#+}} xmm9 = [255,0,255,0,255,0,255,0]
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
+; SSE2-NEXT:    movdqa %xmm4, %xmm6
+; SSE2-NEXT:    pxor %xmm0, %xmm6
+; SSE2-NEXT:    psubd %xmm7, %xmm4
+; SSE2-NEXT:    por %xmm0, %xmm7
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm7
+; SSE2-NEXT:    packssdw %xmm7, %xmm6
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
 ; SSE2-NEXT:    pxor %xmm0, %xmm7
-; SSE2-NEXT:    psubd %xmm10, %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm10
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm10
-; SSE2-NEXT:    pand %xmm9, %xmm10
-; SSE2-NEXT:    packuswb %xmm5, %xmm10
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    pxor %xmm0, %xmm5
-; SSE2-NEXT:    psubd %xmm6, %xmm2
-; SSE2-NEXT:    por %xmm0, %xmm6
-; SSE2-NEXT:    pcmpgtd %xmm5, %xmm6
-; SSE2-NEXT:    pand %xmm9, %xmm6
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    pxor %xmm0, %xmm5
-; SSE2-NEXT:    por %xmm8, %xmm0
-; SSE2-NEXT:    pcmpgtd %xmm5, %xmm0
-; SSE2-NEXT:    pand %xmm9, %xmm0
-; SSE2-NEXT:    packuswb %xmm6, %xmm0
-; SSE2-NEXT:    packuswb %xmm10, %xmm0
-; SSE2-NEXT:    psubd %xmm8, %xmm1
-; SSE2-NEXT:    pand %xmm9, %xmm4
-; SSE2-NEXT:    pand %xmm9, %xmm3
+; SSE2-NEXT:    por %xmm5, %xmm0
+; SSE2-NEXT:    pcmpgtd %xmm7, %xmm0
+; SSE2-NEXT:    packssdw %xmm0, %xmm0
+; SSE2-NEXT:    packsswb %xmm6, %xmm0
+; SSE2-NEXT:    psubd %xmm5, %xmm1
+; SSE2-NEXT:    psubd %xmm9, %xmm2
+; SSE2-NEXT:    psubd %xmm8, %xmm3
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm3
 ; SSE2-NEXT:    packuswb %xmm4, %xmm3
-; SSE2-NEXT:    pand %xmm9, %xmm2
-; SSE2-NEXT:    pand %xmm9, %xmm1
+; SSE2-NEXT:    pand %xmm5, %xmm2
+; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    packuswb %xmm2, %xmm1
 ; SSE2-NEXT:    packuswb %xmm3, %xmm1
 ; SSE2-NEXT:    pandn %xmm1, %xmm0
@@ -657,47 +645,32 @@ define <16 x i8> @test14(<16 x i8> %x, <16 x i32> %y) nounwind {
 ;
 ; SSSE3-LABEL: test14:
 ; SSSE3:       # %bb.0: # %vector.ph
-; SSSE3-NEXT:    pxor %xmm7, %xmm7
-; SSSE3-NEXT:    movdqa %xmm0, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm7[8],xmm11[9],xmm7[9],xmm11[10],xmm7[10],xmm11[11],xmm7[11],xmm11[12],xmm7[12],xmm11[13],xmm7[13],xmm11[14],xmm7[14],xmm11[15],xmm7[15]
-; SSSE3-NEXT:    movdqa %xmm11, %xmm8
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm7[4],xmm11[5],xmm7[5],xmm11[6],xmm7[6],xmm11[7],xmm7[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3],xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm10
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm9
-; SSSE3-NEXT:    pxor %xmm7, %xmm9
-; SSSE3-NEXT:    psubd %xmm0, %xmm2
 ; SSSE3-NEXT:    movdqa %xmm0, %xmm5
-; SSSE3-NEXT:    por %xmm7, %xmm5
-; SSSE3-NEXT:    pcmpgtd %xmm9, %xmm5
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm9 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSSE3-NEXT:    pshufb %xmm9, %xmm5
-; SSSE3-NEXT:    movdqa %xmm1, %xmm6
-; SSSE3-NEXT:    pxor %xmm7, %xmm6
-; SSSE3-NEXT:    psubd %xmm10, %xmm1
-; SSSE3-NEXT:    movdqa %xmm10, %xmm0
-; SSSE3-NEXT:    por %xmm7, %xmm0
-; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm0
-; SSSE3-NEXT:    pshufb %xmm9, %xmm0
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSSE3-NEXT:    movdqa %xmm4, %xmm5
-; SSSE3-NEXT:    pxor %xmm7, %xmm5
-; SSSE3-NEXT:    psubd %xmm11, %xmm4
-; SSSE3-NEXT:    por %xmm7, %xmm11
-; SSSE3-NEXT:    pcmpgtd %xmm5, %xmm11
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm5 = <u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; SSSE3-NEXT:    pshufb %xmm5, %xmm11
-; SSSE3-NEXT:    movdqa %xmm3, %xmm6
-; SSSE3-NEXT:    pxor %xmm7, %xmm6
-; SSSE3-NEXT:    por %xmm8, %xmm7
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm5, %xmm7
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
+; SSSE3-NEXT:    movdqa %xmm7, %xmm8
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3],xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSSE3-NEXT:    movdqa %xmm5, %xmm9
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648]
+; SSSE3-NEXT:    movdqa %xmm4, %xmm6
+; SSSE3-NEXT:    pxor %xmm0, %xmm6
+; SSSE3-NEXT:    psubd %xmm7, %xmm4
+; SSSE3-NEXT:    por %xmm0, %xmm7
 ; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm7
-; SSSE3-NEXT:    pshufb %xmm5, %xmm7
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1]
-; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
+; SSSE3-NEXT:    packssdw %xmm7, %xmm6
+; SSSE3-NEXT:    movdqa %xmm1, %xmm7
+; SSSE3-NEXT:    pxor %xmm0, %xmm7
+; SSSE3-NEXT:    por %xmm5, %xmm0
+; SSSE3-NEXT:    pcmpgtd %xmm7, %xmm0
+; SSSE3-NEXT:    packssdw %xmm0, %xmm0
+; SSSE3-NEXT:    packsswb %xmm6, %xmm0
+; SSSE3-NEXT:    psubd %xmm5, %xmm1
+; SSSE3-NEXT:    psubd %xmm9, %xmm2
 ; SSSE3-NEXT:    psubd %xmm8, %xmm3
 ; SSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
 ; SSSE3-NEXT:    pand %xmm5, %xmm4
@@ -712,43 +685,31 @@ define <16 x i8> @test14(<16 x i8> %x, <16 x i32> %y) nounwind {
 ;
 ; SSE41-LABEL: test14:
 ; SSE41:       # %bb.0: # %vector.ph
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm11 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm8 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,3,0,1]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,3,0,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm8 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
+; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,1,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm9 = xmm6[0],zero,zero,zero,xmm6[1],zero,zero,zero,xmm6[2],zero,zero,zero,xmm6[3],zero,zero,zero
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm10 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
+; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[3,1,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,xmm7[1],zero,zero,zero,xmm7[2],zero,zero,zero,xmm7[3],zero,zero,zero
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm5 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm1, %xmm6
+; SSE41-NEXT:    pmaxud %xmm5, %xmm6
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm6
+; SSE41-NEXT:    pcmpeqd %xmm10, %xmm10
+; SSE41-NEXT:    pxor %xmm10, %xmm6
+; SSE41-NEXT:    packssdw %xmm0, %xmm6
+; SSE41-NEXT:    packsswb %xmm0, %xmm6
 ; SSE41-NEXT:    movdqa %xmm4, %xmm0
-; SSE41-NEXT:    pmaxud %xmm10, %xmm0
+; SSE41-NEXT:    pmaxud %xmm7, %xmm0
 ; SSE41-NEXT:    pcmpeqd %xmm4, %xmm0
-; SSE41-NEXT:    pcmpeqd %xmm6, %xmm6
-; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    movdqa {{.*#+}} xmm7 = <u,u,u,u,0,4,8,12,u,u,u,u,u,u,u,u>
-; SSE41-NEXT:    pshufb %xmm7, %xmm0
-; SSE41-NEXT:    movdqa %xmm3, %xmm5
-; SSE41-NEXT:    pmaxud %xmm9, %xmm5
-; SSE41-NEXT:    pcmpeqd %xmm3, %xmm5
-; SSE41-NEXT:    pxor %xmm6, %xmm5
-; SSE41-NEXT:    pshufb %xmm7, %xmm5
-; SSE41-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    pmaxud %xmm8, %xmm0
-; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
-; SSE41-NEXT:    pxor %xmm6, %xmm0
-; SSE41-NEXT:    movdqa {{.*#+}} xmm12 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE41-NEXT:    pshufb %xmm12, %xmm0
-; SSE41-NEXT:    movdqa %xmm2, %xmm7
-; SSE41-NEXT:    pmaxud %xmm11, %xmm7
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm7
-; SSE41-NEXT:    pxor %xmm6, %xmm7
-; SSE41-NEXT:    pshufb %xmm12, %xmm7
-; SSE41-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm5[4,5,6,7]
-; SSE41-NEXT:    psubd %xmm11, %xmm2
-; SSE41-NEXT:    psubd %xmm8, %xmm1
-; SSE41-NEXT:    psubd %xmm9, %xmm3
-; SSE41-NEXT:    psubd %xmm10, %xmm4
+; SSE41-NEXT:    pxor %xmm10, %xmm0
+; SSE41-NEXT:    packssdw %xmm0, %xmm0
+; SSE41-NEXT:    packsswb %xmm0, %xmm0
+; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm6[0,1,2,3],xmm0[4,5,6,7]
+; SSE41-NEXT:    psubd %xmm9, %xmm2
+; SSE41-NEXT:    psubd %xmm5, %xmm1
+; SSE41-NEXT:    psubd %xmm8, %xmm3
+; SSE41-NEXT:    psubd %xmm7, %xmm4
 ; SSE41-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
 ; SSE41-NEXT:    pand %xmm5, %xmm4
 ; SSE41-NEXT:    pand %xmm5, %xmm3
diff --git a/llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll b/llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll
index ed2db51..bd23c35 100644
--- a/llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll
@@ -2638,23 +2638,22 @@ define <4 x i8> @strict_vector_fptosi_v4f32_to_v4i8(<4 x float> %a) #0 {
 ; SSE-32-LABEL: strict_vector_fptosi_v4f32_to_v4i8:
 ; SSE-32:       # %bb.0:
 ; SSE-32-NEXT:    cvttps2dq %xmm0, %xmm0
-; SSE-32-NEXT:    pand {{\.LCPI.*}}, %xmm0
-; SSE-32-NEXT:    packuswb %xmm0, %xmm0
-; SSE-32-NEXT:    packuswb %xmm0, %xmm0
+; SSE-32-NEXT:    packssdw %xmm0, %xmm0
+; SSE-32-NEXT:    packsswb %xmm0, %xmm0
 ; SSE-32-NEXT:    retl
 ;
 ; SSE-64-LABEL: strict_vector_fptosi_v4f32_to_v4i8:
 ; SSE-64:       # %bb.0:
 ; SSE-64-NEXT:    cvttps2dq %xmm0, %xmm0
-; SSE-64-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE-64-NEXT:    packuswb %xmm0, %xmm0
-; SSE-64-NEXT:    packuswb %xmm0, %xmm0
+; SSE-64-NEXT:    packssdw %xmm0, %xmm0
+; SSE-64-NEXT:    packsswb %xmm0, %xmm0
 ; SSE-64-NEXT:    retq
 ;
 ; AVX-LABEL: strict_vector_fptosi_v4f32_to_v4i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcvttps2dq %xmm0, %xmm0
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    ret{{[l|q]}}
 ;
 ; AVX512F-LABEL: strict_vector_fptosi_v4f32_to_v4i8:
@@ -2689,7 +2688,6 @@ define <4 x i8> @strict_vector_fptoui_v4f32_to_v4i8(<4 x float> %a) #0 {
 ; SSE-32-LABEL: strict_vector_fptoui_v4f32_to_v4i8:
 ; SSE-32:       # %bb.0:
 ; SSE-32-NEXT:    cvttps2dq %xmm0, %xmm0
-; SSE-32-NEXT:    pand {{\.LCPI.*}}, %xmm0
 ; SSE-32-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-32-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-32-NEXT:    retl
@@ -2697,7 +2695,6 @@ define <4 x i8> @strict_vector_fptoui_v4f32_to_v4i8(<4 x float> %a) #0 {
 ; SSE-64-LABEL: strict_vector_fptoui_v4f32_to_v4i8:
 ; SSE-64:       # %bb.0:
 ; SSE-64-NEXT:    cvttps2dq %xmm0, %xmm0
-; SSE-64-NEXT:    pand {{.*}}(%rip), %xmm0
 ; SSE-64-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-64-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-64-NEXT:    retq
@@ -2705,7 +2702,8 @@ define <4 x i8> @strict_vector_fptoui_v4f32_to_v4i8(<4 x float> %a) #0 {
 ; AVX-LABEL: strict_vector_fptoui_v4f32_to_v4i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcvttps2dq %xmm0, %xmm0
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    ret{{[l|q]}}
 ;
 ; AVX512F-LABEL: strict_vector_fptoui_v4f32_to_v4i8:
diff --git a/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll b/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
index 0f9fc8d..8b69a9e 100644
--- a/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-fptoint-256.ll
@@ -1250,7 +1250,8 @@ define <4 x i8> @strict_vector_fptosi_v4f64_to_v4i8(<4 x double> %a) #0 {
 ; AVX-LABEL: strict_vector_fptosi_v4f64_to_v4i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcvttpd2dq %ymm0, %xmm0
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    ret{{[l|q]}}
 ;
@@ -1290,7 +1291,8 @@ define <4 x i8> @strict_vector_fptoui_v4f64_to_v4i8(<4 x double> %a) #0 {
 ; AVX-LABEL: strict_vector_fptoui_v4f64_to_v4i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcvttpd2dq %ymm0, %xmm0
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    ret{{[l|q]}}
 ;
diff --git a/llvm/test/CodeGen/X86/vec_cast2.ll b/llvm/test/CodeGen/X86/vec_cast2.ll
index e43216d..ed703f1 100644
--- a/llvm/test/CodeGen/X86/vec_cast2.ll
+++ b/llvm/test/CodeGen/X86/vec_cast2.ll
@@ -122,7 +122,8 @@ define <4 x i8> @cvt_v4f32_v4i8(<4 x float> %src) {
 ; CHECK-LABEL: cvt_v4f32_v4i8:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vcvttps2dq %xmm0, %xmm0
-; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    retl
   %res = fptosi <4 x float> %src to <4 x i8>
   ret <4 x i8> %res
@@ -167,7 +168,8 @@ define <4 x i8> @cvt_v4f32_v4u8(<4 x float> %src) {
 ; CHECK-LABEL: cvt_v4f32_v4u8:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vcvttps2dq %xmm0, %xmm0
-; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; CHECK-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    retl
   %res = fptoui <4 x float> %src to <4 x i8>
   ret <4 x i8> %res
diff --git a/llvm/test/CodeGen/X86/vector-trunc-packus.ll b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
index c74b879..9d044d5 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-packus.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
@@ -5212,7 +5212,8 @@ define <4 x i8> @trunc_packus_v4i32_v4i8(<4 x i32> %a0) "min-legal-vector-width"
 ; SSE41-NEXT:    pminsd {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pxor %xmm1, %xmm1
 ; SSE41-NEXT:    pmaxsd %xmm1, %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packusdw %xmm0, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_packus_v4i32_v4i8:
@@ -5220,7 +5221,8 @@ define <4 x i8> @trunc_packus_v4i32_v4i8(<4 x i32> %a0) "min-legal-vector-width"
 ; AVX1-NEXT:    vpminsd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_packus_v4i32_v4i8:
@@ -5229,7 +5231,8 @@ define <4 x i8> @trunc_packus_v4i32_v4i8(<4 x i32> %a0) "min-legal-vector-width"
 ; AVX2-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: trunc_packus_v4i32_v4i8:
@@ -5316,7 +5319,8 @@ define void @trunc_packus_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; SSE41-NEXT:    pminsd {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pxor %xmm1, %xmm1
 ; SSE41-NEXT:    pmaxsd %xmm0, %xmm1
-; SSE41-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packusdw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm0, %xmm1
 ; SSE41-NEXT:    movd %xmm1, (%rdi)
 ; SSE41-NEXT:    retq
 ;
@@ -5325,7 +5329,8 @@ define void @trunc_packus_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; AVX1-NEXT:    vpminsd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
 ;
@@ -5335,7 +5340,8 @@ define void @trunc_packus_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; AVX2-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX2-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
index 665e3fe..bed4553 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
@@ -5080,14 +5080,16 @@ define <4 x i8> @trunc_ssat_v4i32_v4i8(<4 x i32> %a0) {
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pminsd {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pmaxsd {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packssdw %xmm0, %xmm0
+; SSE41-NEXT:    packsswb %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_ssat_v4i32_v4i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpminsd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpmaxsd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_ssat_v4i32_v4i8:
@@ -5096,7 +5098,8 @@ define <4 x i8> @trunc_ssat_v4i32_v4i8(<4 x i32> %a0) {
 ; AVX2-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4294967168,4294967168,4294967168,4294967168]
 ; AVX2-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: trunc_ssat_v4i32_v4i8:
@@ -5178,7 +5181,8 @@ define void @trunc_ssat_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pminsd {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pmaxsd {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packssdw %xmm0, %xmm0
+; SSE41-NEXT:    packsswb %xmm0, %xmm0
 ; SSE41-NEXT:    movd %xmm0, (%rdi)
 ; SSE41-NEXT:    retq
 ;
@@ -5186,7 +5190,8 @@ define void @trunc_ssat_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpminsd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpmaxsd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
 ;
@@ -5196,7 +5201,8 @@ define void @trunc_ssat_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; AVX2-NEXT:    vpminsd %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4294967168,4294967168,4294967168,4294967168]
 ; AVX2-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX2-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-trunc-usat.ll b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
index 8434bac..586514e 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-usat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-usat.ll
@@ -3694,20 +3694,23 @@ define <4 x i8> @trunc_usat_v4i32_v4i8(<4 x i32> %a0) {
 ; SSE41-LABEL: trunc_usat_v4i32_v4i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pminud {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packusdw %xmm0, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_usat_v4i32_v4i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_usat_v4i32_v4i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255]
 ; AVX2-NEXT:    vpminud %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: trunc_usat_v4i32_v4i8:
@@ -3776,14 +3779,16 @@ define void @trunc_usat_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; SSE41-LABEL: trunc_usat_v4i32_v4i8_store:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pminud {{.*}}(%rip), %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE41-NEXT:    packusdw %xmm0, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm0
 ; SSE41-NEXT:    movd %xmm0, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_usat_v4i32_v4i8_store:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
 ;
@@ -3791,7 +3796,8 @@ define void @trunc_usat_v4i32_v4i8_store(<4 x i32> %a0, <4 x i8> *%p1) {
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255]
 ; AVX2-NEXT:    vpminud %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX2-NEXT:    retq
 ;
@@ -3880,37 +3886,25 @@ define <8 x i8> @trunc_usat_v8i32_v8i8(<8 x i32> %a0) {
 ;
 ; SSE41-LABEL: trunc_usat_v8i32_v8i8:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255]
-; SSE41-NEXT:    pminud %xmm2, %xmm0
-; SSE41-NEXT:    pminud %xmm2, %xmm1
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE41-NEXT:    pshufb %xmm2, %xmm1
-; SSE41-NEXT:    pshufb %xmm2, %xmm0
-; SSE41-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE41-NEXT:    pminud {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    packusdw %xmm0, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_usat_v8i32_v8i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255,255,255]
-; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_usat_v8i32_v8i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255]
+; AVX2-NEXT:    vpminud %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -4000,39 +3994,27 @@ define void @trunc_usat_v8i32_v8i8_store(<8 x i32> %a0, <8 x i8> *%p1) {
 ;
 ; SSE41-LABEL: trunc_usat_v8i32_v8i8_store:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255]
-; SSE41-NEXT:    pminud %xmm2, %xmm0
-; SSE41-NEXT:    pminud %xmm2, %xmm1
-; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; SSE41-NEXT:    pshufb %xmm2, %xmm1
-; SSE41-NEXT:    pshufb %xmm2, %xmm0
-; SSE41-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE41-NEXT:    pminud {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    packusdw %xmm0, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm0
 ; SSE41-NEXT:    movq %xmm0, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_usat_v8i32_v8i8_store:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255,255,255]
-; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-NEXT:    vpminud {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_usat_v8i32_v8i8_store:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255]
+; AVX2-NEXT:    vpminud %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq