[TargetLowering] SimplifyMultipleUseDemandedBits - add BITCAST pass through support.

This allows us to peek through BITCASTs and attempt simplify the source operand, and then bitcast back.

llvm-svn: 367091

3 files changed, 165 insertions, 109 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 8f2d45a1f40..6020877780f 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -565,12 +565,69 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, const APInt &DemandedBits,
 }
 
 // TODO: Can we merge SelectionDAG::GetDemandedBits into this?
-// TODO: Under what circumstances can we create nodes? BITCAST? Constant?
+// TODO: Under what circumstances can we create nodes? Constant folding?
 SDValue TargetLowering::SimplifyMultipleUseDemandedBits(
     SDValue Op, const APInt &DemandedBits, const APInt &DemandedElts,
     SelectionDAG &DAG, unsigned Depth) const {
+  unsigned NumElts = DemandedElts.getBitWidth();
   KnownBits LHSKnown, RHSKnown;
   switch (Op.getOpcode()) {
+  case ISD::BITCAST: {
+    SDValue Src = peekThroughBitcasts(Op.getOperand(0));
+    EVT SrcVT = Op.getOperand(0).getValueType();
+    EVT DstVT = Op.getValueType();
+    unsigned NumSrcEltBits = SrcVT.getScalarSizeInBits();
+    unsigned NumDstEltBits = DstVT.getScalarSizeInBits();
+
+    if (NumSrcEltBits == NumDstEltBits)
+      if (SDValue V = SimplifyMultipleUseDemandedBits(
+              Src, DemandedBits, DemandedElts, DAG, Depth + 1))
+        return DAG.getBitcast(DstVT, V);
+
+    // TODO - bigendian once we have test coverage.
+    if (SrcVT.isVector() && (NumDstEltBits % NumSrcEltBits) == 0 &&
+        DAG.getDataLayout().isLittleEndian()) {
+      unsigned Scale = NumDstEltBits / NumSrcEltBits;
+      unsigned NumSrcElts = SrcVT.getVectorNumElements();
+      APInt DemandedSrcBits = APInt::getNullValue(NumSrcEltBits);
+      APInt DemandedSrcElts = APInt::getNullValue(NumSrcElts);
+      for (unsigned i = 0; i != Scale; ++i) {
+        unsigned Offset = i * NumSrcEltBits;
+        APInt Sub = DemandedBits.extractBits(NumSrcEltBits, Offset);
+        if (!Sub.isNullValue()) {
+          DemandedSrcBits |= Sub;
+          for (unsigned j = 0; j != NumElts; ++j)
+            if (DemandedElts[j])
+              DemandedSrcElts.setBit((j * Scale) + i);
+        }
+      }
+
+      if (SDValue V = SimplifyMultipleUseDemandedBits(
+              Src, DemandedSrcBits, DemandedSrcElts, DAG, Depth + 1))
+        return DAG.getBitcast(DstVT, V);
+    }
+
+    // TODO - bigendian once we have test coverage.
+    if ((NumSrcEltBits % NumDstEltBits) == 0 &&
+        DAG.getDataLayout().isLittleEndian()) {
+      unsigned Scale = NumSrcEltBits / NumDstEltBits;
+      unsigned NumSrcElts = SrcVT.isVector() ? SrcVT.getVectorNumElements() : 1;
+      APInt DemandedSrcBits = APInt::getNullValue(NumSrcEltBits);
+      APInt DemandedSrcElts = APInt::getNullValue(NumSrcElts);
+      for (unsigned i = 0; i != NumElts; ++i)
+        if (DemandedElts[i]) {
+          unsigned Offset = (i % Scale) * NumDstEltBits;
+          DemandedSrcBits.insertBits(DemandedBits, Offset);
+          DemandedSrcElts.setBit(i / Scale);
+        }
+
+      if (SDValue V = SimplifyMultipleUseDemandedBits(
+              Src, DemandedSrcBits, DemandedSrcElts, DAG, Depth + 1))
+        return DAG.getBitcast(DstVT, V);
+    }
+
+    break;
+  }
   case ISD::AND: {
     LHSKnown = DAG.computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
     RHSKnown = DAG.computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
@@ -615,7 +672,7 @@ SDValue TargetLowering::SimplifyMultipleUseDemandedBits(
     // If all the demanded elts are from one operand and are inline,
     // then we can use the operand directly.
     bool AllUndef = true, IdentityLHS = true, IdentityRHS = true;
-    for (unsigned i = 0, NumElts = ShuffleMask.size(); i != NumElts; ++i) {
+    for (unsigned i = 0; i != NumElts; ++i) {
       int M = ShuffleMask[i];
       if (M < 0 || !DemandedElts[i])
         continue;
diff --git a/llvm/test/CodeGen/X86/vector-reduce-umax.ll b/llvm/test/CodeGen/X86/vector-reduce-umax.ll
index 075d219fd77..6efd3724d71 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-umax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-umax.ll
@@ -730,23 +730,24 @@ define i32 @test_v2i32(<2 x i32> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,0,4294967295,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm3
-; SSE2-NEXT:    pxor %xmm1, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    pxor %xmm2, %xmm4
+; SSE2-NEXT:    pxor %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm4, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm4, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v2i32:
@@ -1212,24 +1213,24 @@ define i16 @test_v2i16(<2 x i16> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,0,0,65535,0,0,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
 ; SSE2-NEXT:    pcmpgtd %xmm2, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE2-NEXT:    retq
 ;
@@ -1309,12 +1310,11 @@ define i16 @test_v4i16(<4 x i16> %a0) {
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pand %xmm2, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm2
-; SSE2-NEXT:    por %xmm1, %xmm2
-; SSE2-NEXT:    movd %xmm2, %eax
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm1
+; SSE2-NEXT:    pand %xmm1, %xmm3
+; SSE2-NEXT:    pandn %xmm0, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE2-NEXT:    retq
 ;
@@ -1765,25 +1765,25 @@ define i8 @test_v2i8(<2 x i8> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
 ; SSE2-NEXT:    pcmpgtd %xmm2, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
@@ -1861,12 +1861,11 @@ define i8 @test_v4i8(<4 x i8> %a0) {
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pand %xmm2, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm2
-; SSE2-NEXT:    por %xmm1, %xmm2
-; SSE2-NEXT:    movd %xmm2, %eax
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm1
+; SSE2-NEXT:    pand %xmm1, %xmm3
+; SSE2-NEXT:    pandn %xmm0, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-reduce-umin.ll b/llvm/test/CodeGen/X86/vector-reduce-umin.ll
index b75523562d7..184d5452863 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-umin.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-umin.ll
@@ -728,23 +728,23 @@ define i32 @test_v2i32(<2 x i32> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,0,4294967295,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm3
-; SSE2-NEXT:    pxor %xmm0, %xmm2
+; SSE2-NEXT:    pxor %xmm2, %xmm1
+; SSE2-NEXT:    pxor %xmm3, %xmm2
 ; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm4
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v2i32:
@@ -1210,24 +1210,24 @@ define i16 @test_v2i16(<2 x i16> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,0,0,65535,0,0,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm2
 ; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm4
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE2-NEXT:    retq
 ;
@@ -1309,9 +1309,9 @@ define i16 @test_v4i16(<4 x i16> %a0) {
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm3
 ; SSE2-NEXT:    pandn %xmm0, %xmm2
-; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    por %xmm3, %xmm2
 ; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE2-NEXT:    retq
@@ -1670,25 +1670,25 @@ define i8 @test_v2i8(<2 x i8> %a0) {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    por %xmm3, %xmm2
 ; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm4
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT:    por %xmm2, %xmm3
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    pandn %xmm0, %xmm3
-; SSE2-NEXT:    por %xmm1, %xmm3
-; SSE2-NEXT:    movd %xmm3, %eax
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pandn %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm0, %xmm2
+; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
@@ -1768,9 +1768,9 @@ define i8 @test_v4i8(<4 x i8> %a0) {
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm3
 ; SSE2-NEXT:    pandn %xmm0, %xmm2
-; SSE2-NEXT:    por %xmm1, %xmm2
+; SSE2-NEXT:    por %xmm3, %xmm2
 ; SSE2-NEXT:    movd %xmm2, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq