From 485b8789de8b327341072a2d0ef23709fb53a600 Mon Sep 17 00:00:00 2001
From: Vlad Tsyrklevich <vlad@tsyrklevich.net>
Date: Fri, 26 Jul 2019 18:14:21 +0000
Subject: [PATCH] Revert "[X86][SSE] Replace PMULDQ GetDemandedBits combine
 with SimplifyMultipleUseDemandedBits handler."

This reverts r367100, it appears to be causing test failures after
Nico's revert of r367091.

llvm-svn: 367141
---
 llvm/lib/Target/X86/X86ISelLowering.cpp          |  21 +-
 llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll | 396 ++++++++++++-----------
 llvm/test/CodeGen/X86/vector-reduce-mul-widen.ll | 103 +++---
 llvm/test/CodeGen/X86/vector-reduce-mul.ll       | 115 ++++---
 4 files changed, 357 insertions(+), 278 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 001862b..3086651 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -34381,18 +34381,6 @@ bool X86TargetLowering::SimplifyDemandedBitsForTargetNode(
     if (SimplifyDemandedBits(RHS, DemandedMask, OriginalDemandedElts, KnownOp,
                              TLO, Depth + 1))
       return true;
-
-    // Aggressively peek through ops to get at the demanded low bits.
-    SDValue DemandedLHS = SimplifyMultipleUseDemandedBits(
-        LHS, DemandedMask, OriginalDemandedElts, TLO.DAG, Depth + 1);
-    SDValue DemandedRHS = SimplifyMultipleUseDemandedBits(
-        RHS, DemandedMask, OriginalDemandedElts, TLO.DAG, Depth + 1);
-    if (DemandedLHS || DemandedRHS) {
-      DemandedLHS = DemandedLHS ? DemandedLHS : LHS;
-      DemandedRHS = DemandedRHS ? DemandedRHS : RHS;
-      return TLO.CombineTo(
-          Op, TLO.DAG.getNode(Opc, SDLoc(Op), VT, DemandedLHS, DemandedRHS));
-    }
     break;
   }
   case X86ISD::VSHLI: {
@@ -44232,6 +44220,15 @@ static SDValue combinePMULDQ(SDNode *N, SelectionDAG &DAG,
   if (ISD::isBuildVectorAllZeros(RHS.getNode()))
     return RHS;
 
+  // Aggressively peek through ops to get at the demanded low bits.
+  APInt DemandedMask = APInt::getLowBitsSet(64, 32);
+  SDValue DemandedLHS = DAG.GetDemandedBits(LHS, DemandedMask);
+  SDValue DemandedRHS = DAG.GetDemandedBits(RHS, DemandedMask);
+  if (DemandedLHS || DemandedRHS)
+    return DAG.getNode(N->getOpcode(), SDLoc(N), N->getValueType(0),
+                       DemandedLHS ? DemandedLHS : LHS,
+                       DemandedRHS ? DemandedRHS : RHS);
+
   // PMULDQ/PMULUDQ only uses lower 32 bits from each vector element.
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
   if (TLI.SimplifyDemandedBits(SDValue(N, 0), APInt::getAllOnesValue(64), DCI))
diff --git a/llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll b/llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
index 8717944e..c1669cf 100644
--- a/llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
+++ b/llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll
@@ -11,30 +11,32 @@
 define <4 x i32> @test_urem_odd_even(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2454267027,1374389535,1374389535]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    psrld $1, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE2-NEXT:    psrld $1, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,1374389535,1374389535]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $5, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    psrld $5, %xmm2
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm3
-; CHECK-SSE2-NEXT:    psrld $3, %xmm2
-; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,14,25,100]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    psrld $3, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,14,25,100]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
@@ -46,13 +48,14 @@ define <4 x i32> @test_urem_odd_even(<4 x i32> %X) nounwind {
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
-; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,1374389535,1374389535]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,1374389535,1374389535]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm3
@@ -71,13 +74,14 @@ define <4 x i32> @test_urem_odd_even(<4 x i32> %X) nounwind {
 ; CHECK-AVX1-LABEL: test_urem_odd_even:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,1374389535,1374389535]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [3435973837,2454267027,1374389535,1374389535]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
@@ -406,30 +410,32 @@ define <4 x i32> @test_urem_even_allones_ne(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_even_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even_allones_eq:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    psrld $1, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE2-NEXT:    psrld $1, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $5, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    psrld $5, %xmm2
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm3
-; CHECK-SSE2-NEXT:    psrld $31, %xmm2
-; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,14,4294967295,100]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    psrld $31, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,14,4294967295,100]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
@@ -441,13 +447,14 @@ define <4 x i32> @test_urem_odd_even_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
-; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm3
@@ -466,13 +473,14 @@ define <4 x i32> @test_urem_odd_even_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-AVX1-LABEL: test_urem_odd_even_allones_eq:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
@@ -520,30 +528,32 @@ define <4 x i32> @test_urem_odd_even_allones_eq(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_even_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even_allones_ne:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    psrld $1, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE2-NEXT:    psrld $1, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $5, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    psrld $5, %xmm2
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm3
-; CHECK-SSE2-NEXT:    psrld $31, %xmm2
-; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,14,4294967295,100]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    psrld $31, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,14,4294967295,100]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
@@ -555,13 +565,14 @@ define <4 x i32> @test_urem_odd_even_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
-; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm3
@@ -580,13 +591,14 @@ define <4 x i32> @test_urem_odd_even_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-AVX1-LABEL: test_urem_odd_even_allones_ne:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,2147483649,1374389535]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [3435973837,2454267027,2147483649,1374389535]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
@@ -828,29 +840,31 @@ define <4 x i32> @test_urem_even_poweroftwo(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even_poweroftwo:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2454267027,268435456,1374389535]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    psrld $1, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE2-NEXT:    psrld $1, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,268435456,1374389535]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $5, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    psrld $5, %xmm2
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm3
-; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,14,16,100]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,14,16,100]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
@@ -862,13 +876,14 @@ define <4 x i32> @test_urem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
-; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,268435456,1374389535]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,268435456,1374389535]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm3
@@ -886,13 +901,14 @@ define <4 x i32> @test_urem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-AVX1-LABEL: test_urem_odd_even_poweroftwo:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,268435456,1374389535]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [3435973837,2454267027,268435456,1374389535]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
@@ -1106,31 +1122,33 @@ define <4 x i32> @test_urem_even_one(<4 x i32> %X) nounwind {
 define <4 x i32> @test_urem_odd_even_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2454267027,0,1374389535]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    psrld $1, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE2-NEXT:    psrld $1, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,0,1374389535]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $5, %xmm1
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    psrld $5, %xmm2
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm3
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [5,14,1,100]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm5
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
 ; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm5[0,2]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm2[3,0]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[0,2]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
@@ -1142,13 +1160,14 @@ define <4 x i32> @test_urem_odd_even_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
-; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,0,1374389535]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [3435973837,2454267027,0,1374389535]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $2, %xmm1
@@ -1164,13 +1183,14 @@ define <4 x i32> @test_urem_odd_even_one(<4 x i32> %X) nounwind {
 ; CHECK-AVX1-LABEL: test_urem_odd_even_one:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,0,1374389535]
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [3435973837,2454267027,0,1374389535]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
@@ -2160,25 +2180,27 @@ define <4 x i32> @test_urem_odd_allones_and_poweroftwo_and_one(<4 x i32> %X) nou
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,0]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,3,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm2[2,1]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [5,4294967295,16,1]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    psrld $31, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm3, %xmm1
+; CHECK-SSE2-NEXT:    psrld $31, %xmm1
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [5,4294967295,16,1]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm5
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,2,2,3]
+; CHECK-SSE2-NEXT:    psrld $2, %xmm3
+; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,0],xmm2[3,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm5[2,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm3
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
@@ -2275,18 +2297,20 @@ define <4 x i32> @test_urem_even_allones_and_poweroftwo_and_one(<4 x i32> %X) no
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,1]
+; CHECK-SSE2-NEXT:    psrld $31, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm0[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [14,4294967295,16,1]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    psrld $31, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm5
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,0],xmm4[3,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[2,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-reduce-mul-widen.ll b/llvm/test/CodeGen/X86/vector-reduce-mul-widen.ll
index ce73ad5..1b81d04 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-mul-widen.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-mul-widen.ll
@@ -814,8 +814,12 @@ define i32 @test_v4i32(<4 x i32> %a0) {
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
@@ -852,16 +856,23 @@ define i32 @test_v4i32(<4 x i32> %a0) {
 define i32 @test_v8i32(<8 x i32> %a0) {
 ; SSE2-LABEL: test_v8i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm0
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v8i32:
@@ -916,22 +927,29 @@ define i32 @test_v8i32(<8 x i32> %a0) {
 define i32 @test_v16i32(<16 x i32> %a0) {
 ; SSE2-LABEL: test_v16i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm6
+; SSE2-NEXT:    pmuludq %xmm5, %xmm6
 ; SSE2-NEXT:    pmuludq %xmm3, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm5, %xmm2
-; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm0
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm6, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v16i32:
@@ -994,32 +1012,39 @@ define i32 @test_v16i32(<16 x i32> %a0) {
 define i32 @test_v32i32(<32 x i32> %a0) {
 ; SSE2-LABEL: test_v32i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm9
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm10
-; SSE2-NEXT:    pmuludq %xmm9, %xmm10
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm9
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm11
-; SSE2-NEXT:    pmuludq %xmm9, %xmm11
-; SSE2-NEXT:    pmuludq %xmm10, %xmm11
+; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm6, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm4, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm7, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm5, %xmm1
 ; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm11, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm8, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm9, %xmm3
+; SSE2-NEXT:    pmuludq %xmm0, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm10, %xmm2
+; SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
diff --git a/llvm/test/CodeGen/X86/vector-reduce-mul.ll b/llvm/test/CodeGen/X86/vector-reduce-mul.ll
index b474cc0..5f1455d 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-mul.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-mul.ll
@@ -807,8 +807,12 @@ define i32 @test_v4i32(<4 x i32> %a0) {
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
@@ -845,16 +849,23 @@ define i32 @test_v4i32(<4 x i32> %a0) {
 define i32 @test_v8i32(<8 x i32> %a0) {
 ; SSE2-LABEL: test_v8i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm0
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v8i32:
@@ -909,22 +920,29 @@ define i32 @test_v8i32(<8 x i32> %a0) {
 define i32 @test_v16i32(<16 x i32> %a0) {
 ; SSE2-LABEL: test_v16i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm6
+; SSE2-NEXT:    pmuludq %xmm5, %xmm6
 ; SSE2-NEXT:    pmuludq %xmm3, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm5, %xmm2
-; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm0
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm6, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v16i32:
@@ -987,32 +1005,39 @@ define i32 @test_v16i32(<16 x i32> %a0) {
 define i32 @test_v32i32(<32 x i32> %a0) {
 ; SSE2-LABEL: test_v32i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm9
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm10
-; SSE2-NEXT:    pmuludq %xmm9, %xmm10
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm9
-; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[1,1,3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm8, %xmm11
-; SSE2-NEXT:    pmuludq %xmm9, %xmm11
-; SSE2-NEXT:    pmuludq %xmm10, %xmm11
+; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm6, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm4, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm7, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm5, %xmm1
 ; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,0,0]
-; SSE2-NEXT:    pmuludq %xmm11, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm8, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm9, %xmm3
+; SSE2-NEXT:    pmuludq %xmm0, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm10, %xmm2
+; SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,0,0]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
@@ -1126,8 +1151,12 @@ define i16 @test_v4i16(<4 x i16> %a0) {
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    pextrw $0, %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; SSE2-NEXT:    retq
@@ -1574,8 +1603,12 @@ define i8 @test_v4i8(<4 x i8> %a0) {
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
-- 
2.7.4