Index: lib/CodeGen/SelectionDAG/SelectionDAG.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -2297,8 +2297,6 @@
     KnownOne &= KnownOne2;
     KnownZero &= KnownZero2;
     break;
-  case ISD::SADDO:
-  case ISD::UADDO:
   case ISD::SSUBO:
   case ISD::USUBO:
   case ISD::SMULO:
@@ -2505,6 +2503,17 @@
     }
     LLVM_FALLTHROUGH;
   }
+  case ISD::UADDO:
+  case ISD::SADDO:
+    if (Op.getResNo() == 1) {
+      // If we know the result of a setcc has the top bits zero, use this info.
+      if (TLI->getBooleanContents(Op.getOperand(0).getValueType()) ==
+              TargetLowering::ZeroOrOneBooleanContent &&
+          BitWidth > 1)
+        KnownZero |= APInt::getHighBitsSet(BitWidth, BitWidth - 1);
+      break;
+    }
+    LLVM_FALLTHROUGH;
   case ISD::ADD:
   case ISD::ADDC:
   case ISD::ADDE: {
@@ -2527,19 +2536,19 @@
     KnownZeroLow = std::min(KnownZeroLow,
                             KnownZero2.countTrailingOnes());
 
-    if (Opcode == ISD::ADD || Opcode == ISD::ADDC) {
-      KnownZero |= APInt::getLowBitsSet(BitWidth, KnownZeroLow);
-      if (KnownZeroHigh > 1)
-        KnownZero |= APInt::getHighBitsSet(BitWidth, KnownZeroHigh - 1);
+    if (Opcode == ISD::ADDE) {
+      // With ADDE, a carry bit may be added in, so we can only use this
+      // information if we know (at least) that the low two bits are clear.
+      // We then return to the caller that the low bit is unknown but that
+      // other bits are known zero.
+      if (KnownZeroLow >= 2)
+        KnownZero |= APInt::getBitsSet(BitWidth, 1, KnownZeroLow);
       break;
     }
 
-    // With ADDE, a carry bit may be added in, so we can only use this
-    // information if we know (at least) that the low two bits are clear.  We
-    // then return to the caller that the low bit is unknown but that other bits
-    // are known zero.
-    if (KnownZeroLow >= 2) // ADDE
-      KnownZero |= APInt::getBitsSet(BitWidth, 1, KnownZeroLow);
+    KnownZero |= APInt::getLowBitsSet(BitWidth, KnownZeroLow);
+    if (KnownZeroHigh > 1)
+      KnownZero |= APInt::getHighBitsSet(BitWidth, KnownZeroHigh - 1);
     break;
   }
   case ISD::SREM:
Index: test/CodeGen/X86/2008-02-14-BitMiscompile.ll
===================================================================
--- test/CodeGen/X86/2008-02-14-BitMiscompile.ll
+++ test/CodeGen/X86/2008-02-14-BitMiscompile.ll
@@ -7,7 +7,6 @@
 ; CHECK-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; CHECK-NEXT:    andl $1, %eax
 ; CHECK-NEXT:    negl %eax
-; CHECK-NEXT:    movzbl %al, %eax
 ; CHECK-NEXT:    retl
   %B = zext i1 %A to i32
   %C = sub i32 0, %B
Index: test/CodeGen/X86/known-bits.ll
===================================================================
--- test/CodeGen/X86/known-bits.ll
+++ test/CodeGen/X86/known-bits.ll
@@ -197,10 +197,10 @@
 ; X64-NEXT:    shlq $32, %rdi
 ; X64-NEXT:    shlq $32, %rsi
 ; X64-NEXT:    addq %rdi, %rsi
-; X64-NEXT:    setb %cl
+; X64-NEXT:    setb %al
 ; X64-NEXT:    seto %dl
-; X64-NEXT:    leal (%rsi,%rsi), %eax
-; X64-NEXT:    orb %cl, %dl
+; X64-NEXT:    orb %al, %dl
+; X64-NEXT:    xorl %eax, %eax
 ; X64-NEXT:    retq
   %1 = shl i64 %a0, 32
   %2 = shl i64 %a1, 32
Index: test/CodeGen/X86/sad.ll
===================================================================
--- test/CodeGen/X86/sad.ll
+++ test/CodeGen/X86/sad.ll
@@ -147,127 +147,99 @@
 define i32 @sad_32i8() nounwind {
 ; SSE2-LABEL: sad_32i8:
 ; SSE2:       # BB#0: # %entry
-; SSE2-NEXT:    pxor %xmm11, %xmm11
+; SSE2-NEXT:    pxor %xmm12, %xmm12
 ; SSE2-NEXT:    movq $-1024, %rax # imm = 0xFC00
 ; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pxor %xmm12, %xmm12
+; SSE2-NEXT:    pxor %xmm6, %xmm6
+; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    pxor %xmm4, %xmm4
+; SSE2-NEXT:    pxor %xmm7, %xmm7
 ; SSE2-NEXT:    pxor %xmm15, %xmm15
-; SSE2-NEXT:    pxor %xmm13, %xmm13
+; SSE2-NEXT:    pxor %xmm1, %xmm1
 ; SSE2-NEXT:    pxor %xmm14, %xmm14
 ; SSE2-NEXT:    .p2align 4, 0x90
 ; SSE2-NEXT:  .LBB1_1: # %vector.body
 ; SSE2-NEXT:    # =>This Inner Loop Header: Depth=1
-; SSE2-NEXT:    movdqa %xmm3, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm6, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa a+1040(%rax), %xmm6
 ; SSE2-NEXT:    movdqa a+1024(%rax), %xmm3
-; SSE2-NEXT:    movdqa %xmm3, %xmm8
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1],xmm8[2],xmm11[2],xmm8[3],xmm11[3],xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm11[8],xmm3[9],xmm11[9],xmm3[10],xmm11[10],xmm3[11],xmm11[11],xmm3[12],xmm11[12],xmm3[13],xmm11[13],xmm3[14],xmm11[14],xmm3[15],xmm11[15]
+; SSE2-NEXT:    movdqa a+1040(%rax), %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm1
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
 ; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm11[4],xmm3[5],xmm11[5],xmm3[6],xmm11[6],xmm3[7],xmm11[7]
-; SSE2-NEXT:    movdqa %xmm6, %xmm1
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3],xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm11[8],xmm6[9],xmm11[9],xmm6[10],xmm11[10],xmm6[11],xmm11[11],xmm6[12],xmm11[12],xmm6[13],xmm11[13],xmm6[14],xmm11[14],xmm6[15],xmm11[15]
-; SSE2-NEXT:    movdqa %xmm6, %xmm7
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1],xmm7[2],xmm11[2],xmm7[3],xmm11[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm12[8],xmm5[9],xmm12[9],xmm5[10],xmm12[10],xmm5[11],xmm12[11],xmm5[12],xmm12[12],xmm5[13],xmm12[13],xmm5[14],xmm12[14],xmm5[15],xmm12[15]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3],xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm0
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm12[8],xmm2[9],xmm12[9],xmm2[10],xmm12[10],xmm2[11],xmm12[11],xmm2[12],xmm12[12],xmm2[13],xmm12[13],xmm2[14],xmm12[14],xmm2[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm2, %xmm13
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE2-NEXT:    movdqa b+1024(%rax), %xmm8
 ; SSE2-NEXT:    movdqa b+1040(%rax), %xmm9
-; SSE2-NEXT:    movdqa %xmm9, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm11[8],xmm9[9],xmm11[9],xmm9[10],xmm11[10],xmm9[11],xmm11[11],xmm9[12],xmm11[12],xmm9[13],xmm11[13],xmm9[14],xmm11[14],xmm9[15],xmm11[15]
 ; SSE2-NEXT:    movdqa %xmm9, %xmm10
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
-; SSE2-NEXT:    psubd %xmm9, %xmm6
-; SSE2-NEXT:    movdqa b+1024(%rax), %xmm4
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3],xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
-; SSE2-NEXT:    psubd %xmm10, %xmm7
-; SSE2-NEXT:    movdqa %xmm2, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
-; SSE2-NEXT:    psubd %xmm2, %xmm1
-; SSE2-NEXT:    movdqa %xmm4, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm11[8],xmm4[9],xmm11[9],xmm4[10],xmm11[10],xmm4[11],xmm11[11],xmm4[12],xmm11[12],xmm4[13],xmm11[13],xmm4[14],xmm11[14],xmm4[15],xmm11[15]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1],xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; SSE2-NEXT:    psubd %xmm9, %xmm0
-; SSE2-NEXT:    movdqa %xmm4, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
-; SSE2-NEXT:    psubd %xmm4, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3],xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm12[8],xmm9[9],xmm12[9],xmm9[10],xmm12[10],xmm9[11],xmm12[11],xmm9[12],xmm12[12],xmm9[13],xmm12[13],xmm9[14],xmm12[14],xmm9[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm9, %xmm11
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm12[4],xmm9[5],xmm12[5],xmm9[6],xmm12[6],xmm9[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm9, %xmm2
+; SSE2-NEXT:    movdqa %xmm10, %xmm9
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm11, %xmm13
+; SSE2-NEXT:    movdqa %xmm4, %xmm11
+; SSE2-NEXT:    movdqa %xmm8, %xmm4
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3],xmm8[4],xmm12[4],xmm8[5],xmm12[5],xmm8[6],xmm12[6],xmm8[7],xmm12[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm10, %xmm1
 ; SSE2-NEXT:    movdqa %xmm8, %xmm10
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3],xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1],xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; SSE2-NEXT:    psubd %xmm9, %xmm5
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
-; SSE2-NEXT:    psubd %xmm2, %xmm8
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1],xmm4[2],xmm11[2],xmm4[3],xmm11[3]
-; SSE2-NEXT:    psubd %xmm4, %xmm10
-; SSE2-NEXT:    movdqa %xmm10, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm10
-; SSE2-NEXT:    pxor %xmm2, %xmm10
-; SSE2-NEXT:    movdqa %xmm8, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm8
-; SSE2-NEXT:    pxor %xmm2, %xmm8
-; SSE2-NEXT:    movdqa %xmm5, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm5
-; SSE2-NEXT:    pxor %xmm2, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm3
-; SSE2-NEXT:    pxor %xmm2, %xmm3
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm0
-; SSE2-NEXT:    pxor %xmm2, %xmm0
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm1
-; SSE2-NEXT:    movdqa %xmm7, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm7
-; SSE2-NEXT:    pxor %xmm2, %xmm7
-; SSE2-NEXT:    movdqa %xmm6, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    paddd %xmm2, %xmm6
-; SSE2-NEXT:    pxor %xmm2, %xmm6
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm6, %xmm14
-; SSE2-NEXT:    paddd %xmm7, %xmm13
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm8, %xmm3
+; SSE2-NEXT:    movdqa %xmm5, %xmm8
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm12[8],xmm4[9],xmm12[9],xmm4[10],xmm12[10],xmm4[11],xmm12[11],xmm4[12],xmm12[12],xmm4[13],xmm12[13],xmm4[14],xmm12[14],xmm4[15],xmm12[15]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm10, %xmm0
+; SSE2-NEXT:    movdqa %xmm4, %xmm10
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm4, %xmm5
+; SSE2-NEXT:    movdqa %xmm11, %xmm4
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm12[4],xmm8[5],xmm12[5],xmm8[6],xmm12[6],xmm8[7],xmm12[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm10, %xmm8
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1],xmm9[2],xmm12[2],xmm9[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm9, %xmm6
+; SSE2-NEXT:    paddd %xmm6, %xmm7
+; SSE2-NEXT:    paddd %xmm8, %xmm4
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm6 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm6 # 16-byte Reload
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm0, %xmm6
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm3, %xmm0
 ; SSE2-NEXT:    paddd %xmm1, %xmm15
 ; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm1 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm0, %xmm12
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm3, %xmm1
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm5, %xmm2
-; SSE2-NEXT:    paddd %xmm8, %xmm3
-; SSE2-NEXT:    paddd %xmm10, %xmm0
+; SSE2-NEXT:    paddd %xmm13, %xmm1
+; SSE2-NEXT:    paddd %xmm2, %xmm14
 ; SSE2-NEXT:    addq $4, %rax
 ; SSE2-NEXT:    jne .LBB1_1
 ; SSE2-NEXT:  # BB#2: # %middle.block
-; SSE2-NEXT:    paddd %xmm15, %xmm3
-; SSE2-NEXT:    paddd %xmm14, %xmm1
-; SSE2-NEXT:    paddd %xmm12, %xmm0
-; SSE2-NEXT:    paddd %xmm13, %xmm2
-; SSE2-NEXT:    paddd %xmm3, %xmm1
-; SSE2-NEXT:    paddd %xmm2, %xmm1
-; SSE2-NEXT:    paddd %xmm0, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-NEXT:    paddd %xmm1, %xmm0
+; SSE2-NEXT:    paddd %xmm15, %xmm6
+; SSE2-NEXT:    paddd %xmm14, %xmm4
+; SSE2-NEXT:    paddd %xmm7, %xmm0
+; SSE2-NEXT:    paddd %xmm1, %xmm5
+; SSE2-NEXT:    paddd %xmm6, %xmm4
+; SSE2-NEXT:    paddd %xmm5, %xmm4
+; SSE2-NEXT:    paddd %xmm0, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,0,1]
+; SSE2-NEXT:    paddd %xmm4, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    paddd %xmm0, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
@@ -394,288 +366,219 @@
 define i32 @sad_avx64i8() nounwind {
 ; SSE2-LABEL: sad_avx64i8:
 ; SSE2:       # BB#0: # %entry
-; SSE2-NEXT:    subq $184, %rsp
-; SSE2-NEXT:    pxor %xmm15, %xmm15
-; SSE2-NEXT:    movq $-1024, %rax # imm = 0xFC00
+; SSE2-NEXT:    subq $264, %rsp # imm = 0x108
 ; SSE2-NEXT:    pxor %xmm12, %xmm12
+; SSE2-NEXT:    movq $-1024, %rax # imm = 0xFC00
 ; SSE2-NEXT:    pxor %xmm8, %xmm8
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    pxor %xmm14, %xmm14
 ; SSE2-NEXT:    pxor %xmm0, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    pxor %xmm6, %xmm6
 ; SSE2-NEXT:    pxor %xmm0, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    pxor %xmm11, %xmm11
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pxor %xmm7, %xmm7
 ; SSE2-NEXT:    pxor %xmm13, %xmm13
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    pxor %xmm10, %xmm10
+; SSE2-NEXT:    pxor %xmm7, %xmm7
+; SSE2-NEXT:    pxor %xmm11, %xmm11
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pxor %xmm9, %xmm9
+; SSE2-NEXT:    pxor %xmm6, %xmm6
+; SSE2-NEXT:    pxor %xmm0, %xmm0
 ; SSE2-NEXT:    .p2align 4, 0x90
 ; SSE2-NEXT:  .LBB2_1: # %vector.body
 ; SSE2-NEXT:    # =>This Inner Loop Header: Depth=1
-; SSE2-NEXT:    movdqa %xmm3, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm2, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm13, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm6, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm7, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm1, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm5, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm11, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm9, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm10, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm14, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm4, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm6, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm13, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm8, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm12, {{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa a+1040(%rax), %xmm6
-; SSE2-NEXT:    movdqa a+1024(%rax), %xmm4
-; SSE2-NEXT:    movdqa a+1056(%rax), %xmm11
-; SSE2-NEXT:    movdqa a+1072(%rax), %xmm3
-; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3],xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm11, %xmm1
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm15[8],xmm1[9],xmm15[9],xmm1[10],xmm15[10],xmm1[11],xmm15[11],xmm1[12],xmm15[12],xmm1[13],xmm15[13],xmm1[14],xmm15[14],xmm1[15],xmm15[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1],xmm11[2],xmm15[2],xmm11[3],xmm15[3],xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm11, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1],xmm11[2],xmm15[2],xmm11[3],xmm15[3]
-; SSE2-NEXT:    movdqa %xmm4, %xmm12
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1],xmm12[2],xmm15[2],xmm12[3],xmm15[3],xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm12, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm15[8],xmm4[9],xmm15[9],xmm4[10],xmm15[10],xmm4[11],xmm15[11],xmm4[12],xmm15[12],xmm4[13],xmm15[13],xmm4[14],xmm15[14],xmm4[15],xmm15[15]
-; SSE2-NEXT:    movdqa %xmm4, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm15[4],xmm4[5],xmm15[5],xmm4[6],xmm15[6],xmm4[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm6, %xmm14
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1],xmm14[2],xmm15[2],xmm14[3],xmm15[3],xmm14[4],xmm15[4],xmm14[5],xmm15[5],xmm14[6],xmm15[6],xmm14[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm14, %xmm7
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm15[0],xmm7[1],xmm15[1],xmm7[2],xmm15[2],xmm7[3],xmm15[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm15[4],xmm14[5],xmm15[5],xmm14[6],xmm15[6],xmm14[7],xmm15[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm15[8],xmm6[9],xmm15[9],xmm6[10],xmm15[10],xmm6[11],xmm15[11],xmm6[12],xmm15[12],xmm6[13],xmm15[13],xmm6[14],xmm15[14],xmm6[15],xmm15[15]
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1],xmm8[2],xmm15[2],xmm8[3],xmm15[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm15[4],xmm6[5],xmm15[5],xmm6[6],xmm15[6],xmm6[7],xmm15[7]
-; SSE2-NEXT:    movdqa b+1040(%rax), %xmm9
-; SSE2-NEXT:    movdqa %xmm9, %xmm13
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm15[8],xmm9[9],xmm15[9],xmm9[10],xmm15[10],xmm9[11],xmm15[11],xmm9[12],xmm15[12],xmm9[13],xmm15[13],xmm9[14],xmm15[14],xmm9[15],xmm15[15]
-; SSE2-NEXT:    movdqa %xmm9, %xmm10
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm9, %xmm6
-; SSE2-NEXT:    movdqa b+1024(%rax), %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3],xmm13[4],xmm15[4],xmm13[5],xmm15[5],xmm13[6],xmm15[6],xmm13[7],xmm15[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1],xmm10[2],xmm15[2],xmm10[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm10, %xmm8
-; SSE2-NEXT:    movdqa %xmm13, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm15[4],xmm13[5],xmm15[5],xmm13[6],xmm15[6],xmm13[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm13, %xmm14
-; SSE2-NEXT:    movdqa %xmm2, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm15[8],xmm2[9],xmm15[9],xmm2[10],xmm15[10],xmm2[11],xmm15[11],xmm2[12],xmm15[12],xmm2[13],xmm15[13],xmm2[14],xmm15[14],xmm2[15],xmm15[15]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1],xmm9[2],xmm15[2],xmm9[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm9, %xmm7
-; SSE2-NEXT:    movdqa %xmm2, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
+; SSE2-NEXT:    movdqa a+1072(%rax), %xmm0
+; SSE2-NEXT:    movdqa a+1056(%rax), %xmm1
+; SSE2-NEXT:    movdqa a+1024(%rax), %xmm6
+; SSE2-NEXT:    movdqa a+1040(%rax), %xmm11
+; SSE2-NEXT:    movdqa %xmm11, %xmm13
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm13, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm12[8],xmm11[9],xmm12[9],xmm11[10],xmm12[10],xmm11[11],xmm12[11],xmm11[12],xmm12[12],xmm11[13],xmm12[13],xmm11[14],xmm12[14],xmm11[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm11, %xmm14
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm1, %xmm15
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm12[0],xmm15[1],xmm12[1],xmm15[2],xmm12[2],xmm15[3],xmm12[3],xmm15[4],xmm12[4],xmm15[5],xmm12[5],xmm15[6],xmm12[6],xmm15[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm15, %xmm10
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm12[4],xmm15[5],xmm12[5],xmm15[6],xmm12[6],xmm15[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
+; SSE2-NEXT:    movdqa %xmm0, %xmm9
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1],xmm9[2],xmm12[2],xmm9[3],xmm12[3],xmm9[4],xmm12[4],xmm9[5],xmm12[5],xmm9[6],xmm12[6],xmm9[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm9, %xmm4
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm12[4],xmm9[5],xmm12[5],xmm9[6],xmm12[6],xmm9[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    movdqa b+1072(%rax), %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm0, %xmm8
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm0, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa b+1056(%rax), %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm8, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm1, %xmm9
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
 ; SSE2-NEXT:    psubd %xmm2, %xmm4
-; SSE2-NEXT:    movdqa b+1056(%rax), %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1],xmm10[2],xmm15[2],xmm10[3],xmm15[3],xmm10[4],xmm15[4],xmm10[5],xmm15[5],xmm10[6],xmm15[6],xmm10[7],xmm15[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1],xmm9[2],xmm15[2],xmm9[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm9, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm10, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm15[4],xmm10[5],xmm15[5],xmm10[6],xmm15[6],xmm10[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm10, %xmm12
-; SSE2-NEXT:    movdqa %xmm2, %xmm10
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1],xmm9[2],xmm15[2],xmm9[3],xmm15[3]
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    psubd %xmm9, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm2, %xmm9
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm2, %xmm11
-; SSE2-NEXT:    movdqa %xmm1, %xmm13
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3]
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm15[4],xmm0[5],xmm15[5],xmm0[6],xmm15[6],xmm0[7],xmm15[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm15[8],xmm10[9],xmm15[9],xmm10[10],xmm15[10],xmm10[11],xmm15[11],xmm10[12],xmm15[12],xmm10[13],xmm15[13],xmm10[14],xmm15[14],xmm10[15],xmm15[15]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm9, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm10, %xmm2
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1],xmm10[2],xmm15[2],xmm10[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm10, %xmm1
-; SSE2-NEXT:    movdqa %xmm3, %xmm10
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm15[4],xmm13[5],xmm15[5],xmm13[6],xmm15[6],xmm13[7],xmm15[7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm2, %xmm13
-; SSE2-NEXT:    movdqa b+1072(%rax), %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
-; SSE2-NEXT:    movdqa %xmm2, %xmm9
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm2, %xmm3
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm15[4],xmm10[5],xmm15[5],xmm10[6],xmm15[6],xmm10[7],xmm15[7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm9, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm15[8],xmm5[9],xmm15[9],xmm5[10],xmm15[10],xmm5[11],xmm15[11],xmm5[12],xmm15[12],xmm5[13],xmm15[13],xmm5[14],xmm15[14],xmm5[15],xmm15[15]
-; SSE2-NEXT:    movdqa %xmm5, %xmm9
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm15[8],xmm0[9],xmm15[9],xmm0[10],xmm15[10],xmm0[11],xmm15[11],xmm0[12],xmm15[12],xmm0[13],xmm15[13],xmm0[14],xmm15[14],xmm0[15],xmm15[15]
+; SSE2-NEXT:    movdqa %xmm4, {{[0-9]+}}(%rsp) # 16-byte Spill
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE2-NEXT:    psubd %xmm0, %xmm5
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
-; SSE2-NEXT:    psubd %xmm2, %xmm9
-; SSE2-NEXT:    movdqa %xmm9, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm9
-; SSE2-NEXT:    pxor %xmm0, %xmm9
-; SSE2-NEXT:    movdqa %xmm5, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm5
-; SSE2-NEXT:    pxor %xmm0, %xmm5
-; SSE2-NEXT:    movdqa %xmm10, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm10
-; SSE2-NEXT:    pxor %xmm0, %xmm10
-; SSE2-NEXT:    movdqa %xmm3, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm3
-; SSE2-NEXT:    pxor %xmm0, %xmm3
-; SSE2-NEXT:    movdqa %xmm13, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm13
-; SSE2-NEXT:    pxor %xmm0, %xmm13
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm1
-; SSE2-NEXT:    pxor %xmm0, %xmm1
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm2
-; SSE2-NEXT:    pxor %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm11, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm11
-; SSE2-NEXT:    pxor %xmm0, %xmm11
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm2
-; SSE2-NEXT:    pxor %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm12, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm12
-; SSE2-NEXT:    pxor %xmm0, %xmm12
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm2
-; SSE2-NEXT:    pxor %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm4, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm4
-; SSE2-NEXT:    pxor %xmm0, %xmm4
-; SSE2-NEXT:    movdqa %xmm7, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm7
-; SSE2-NEXT:    pxor %xmm0, %xmm7
-; SSE2-NEXT:    movdqa %xmm14, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm14
-; SSE2-NEXT:    pxor %xmm0, %xmm14
-; SSE2-NEXT:    movdqa %xmm8, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm8
-; SSE2-NEXT:    pxor %xmm0, %xmm8
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm0, %xmm7
+; SSE2-NEXT:    movdqa %xmm7, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa b+1040(%rax), %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm2, %xmm5
+; SSE2-NEXT:    movdqa %xmm5, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm1, %xmm15
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm2, %xmm10
+; SSE2-NEXT:    movdqa %xmm10, (%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm0, %xmm11
 ; SSE2-NEXT:    movdqa %xmm6, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    paddd %xmm0, %xmm6
-; SSE2-NEXT:    pxor %xmm0, %xmm6
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm6 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm8, %xmm6
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8 # 16-byte Reload
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm14, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, {{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm2, %xmm14
+; SSE2-NEXT:    movdqa %xmm14, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm1, %xmm8
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm1, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa %xmm6, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3]
+; SSE2-NEXT:    movdqa b+1024(%rax), %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm14
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3],xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm2, %xmm6
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm5, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm12[8],xmm14[9],xmm12[9],xmm14[10],xmm12[10],xmm14[11],xmm12[11],xmm14[12],xmm12[12],xmm14[13],xmm12[13],xmm14[14],xmm12[14],xmm14[15],xmm12[15]
+; SSE2-NEXT:    movdqa %xmm14, %xmm5
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm14, %xmm0
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE2-NEXT:    psubd %xmm5, %xmm2
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm4 # 16-byte Reload
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm7 # 16-byte Reload
 ; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm14 # 16-byte Reload
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3]
+; SSE2-NEXT:    psubd %xmm8, %xmm13
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm13, %xmm14
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm2, %xmm5
+; SSE2-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    paddd %xmm0, %xmm7
+; SSE2-NEXT:    movdqa %xmm7, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8 # 16-byte Reload
 ; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm4, %xmm2
+; SSE2-NEXT:    paddd %xmm1, %xmm2
 ; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm4 # 16-byte Reload
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm13 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm6, %xmm8
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm6 # 16-byte Reload
 ; SSE2-NEXT:    paddd -{{[0-9]+}}(%rsp), %xmm4 # 16-byte Folded Reload
-; SSE2-NEXT:    paddd %xmm12, %xmm8
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
-; SSE2-NEXT:    paddd -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Folded Reload
-; SSE2-NEXT:    movdqa %xmm0, %xmm12
-; SSE2-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm11, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa (%rsp), %xmm11 # 16-byte Reload
+; SSE2-NEXT:    movdqa %xmm4, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    paddd -{{[0-9]+}}(%rsp), %xmm3 # 16-byte Folded Reload
+; SSE2-NEXT:    movdqa %xmm3, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    paddd %xmm11, %xmm13
 ; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm7 # 16-byte Reload
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    paddd -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Folded Reload
-; SSE2-NEXT:    paddd %xmm1, %xmm2
-; SSE2-NEXT:    paddd %xmm13, %xmm7
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm1 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm3, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm13 # 16-byte Reload
 ; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm1 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm10, %xmm1
-; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm5, %xmm3
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm9, %xmm5
-; SSE2-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
+; SSE2-NEXT:    paddd (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE2-NEXT:    paddd %xmm15, %xmm10
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
+; SSE2-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11 # 16-byte Reload
+; SSE2-NEXT:    paddd {{[0-9]+}}(%rsp), %xmm7 # 16-byte Folded Reload
+; SSE2-NEXT:    paddd {{[0-9]+}}(%rsp), %xmm11 # 16-byte Folded Reload
+; SSE2-NEXT:    paddd {{[0-9]+}}(%rsp), %xmm1 # 16-byte Folded Reload
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm9, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm9 # 16-byte Reload
+; SSE2-NEXT:    paddd {{[0-9]+}}(%rsp), %xmm6 # 16-byte Folded Reload
+; SSE2-NEXT:    paddd {{[0-9]+}}(%rsp), %xmm0 # 16-byte Folded Reload
 ; SSE2-NEXT:    addq $4, %rax
 ; SSE2-NEXT:    jne .LBB2_1
 ; SSE2-NEXT:  # BB#2: # %middle.block
-; SSE2-NEXT:    paddd %xmm2, %xmm4
-; SSE2-NEXT:    paddd %xmm3, %xmm6
-; SSE2-NEXT:    movdqa %xmm12, %xmm2
-; SSE2-NEXT:    paddd %xmm11, %xmm2
-; SSE2-NEXT:    paddd %xmm13, %xmm14
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm7, %xmm2
 ; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm3 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm7, %xmm3
+; SSE2-NEXT:    paddd %xmm6, %xmm3
+; SSE2-NEXT:    paddd %xmm5, %xmm8
+; SSE2-NEXT:    paddd %xmm1, %xmm14
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm5 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm11, %xmm5
+; SSE2-NEXT:    paddd %xmm0, %xmm13
 ; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm7 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm5, %xmm7
-; SSE2-NEXT:    paddd %xmm0, %xmm8
-; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
-; SSE2-NEXT:    paddd %xmm1, %xmm0
-; SSE2-NEXT:    paddd %xmm3, %xmm7
-; SSE2-NEXT:    paddd %xmm4, %xmm6
-; SSE2-NEXT:    paddd %xmm14, %xmm6
-; SSE2-NEXT:    paddd %xmm0, %xmm7
-; SSE2-NEXT:    paddd %xmm8, %xmm7
-; SSE2-NEXT:    paddd %xmm6, %xmm7
-; SSE2-NEXT:    paddd %xmm2, %xmm7
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,0,1]
-; SSE2-NEXT:    paddd %xmm7, %xmm0
+; SSE2-NEXT:    paddd %xmm10, %xmm7
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm4 # 16-byte Reload
+; SSE2-NEXT:    paddd %xmm9, %xmm4
+; SSE2-NEXT:    paddd %xmm5, %xmm13
+; SSE2-NEXT:    paddd %xmm2, %xmm3
+; SSE2-NEXT:    paddd %xmm14, %xmm3
+; SSE2-NEXT:    paddd %xmm4, %xmm13
+; SSE2-NEXT:    paddd %xmm7, %xmm13
+; SSE2-NEXT:    paddd %xmm3, %xmm13
+; SSE2-NEXT:    paddd %xmm8, %xmm13
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,0,1]
+; SSE2-NEXT:    paddd %xmm13, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
 ; SSE2-NEXT:    paddd %xmm0, %xmm1
 ; SSE2-NEXT:    movd %xmm1, %eax
-; SSE2-NEXT:    addq $184, %rsp
+; SSE2-NEXT:    addq $264, %rsp # imm = 0x108
 ; SSE2-NEXT:    retq
 ;
 ; AVX2-LABEL: sad_avx64i8:
@@ -703,6 +606,7 @@
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm15 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm8 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; AVX2-NEXT:    vpsubd %ymm8, %ymm15, %ymm8
+; AVX2-NEXT:    vmovdqu %ymm8, -{{[0-9]+}}(%rsp) # 32-byte Spill
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm15 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; AVX2-NEXT:    vpsubd %ymm15, %ymm14, %ymm14
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm15 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
@@ -715,26 +619,17 @@
 ; AVX2-NEXT:    vpsubd %ymm15, %ymm10, %ymm10
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm15 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; AVX2-NEXT:    vpsubd %ymm15, %ymm9, %ymm9
-; AVX2-NEXT:    vmovdqu %ymm9, -{{[0-9]+}}(%rsp) # 32-byte Spill
 ; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm15 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
-; AVX2-NEXT:    vmovdqu -{{[0-9]+}}(%rsp), %ymm9 # 32-byte Reload
-; AVX2-NEXT:    vpsubd %ymm15, %ymm9, %ymm15
-; AVX2-NEXT:    vpabsd %ymm8, %ymm8
-; AVX2-NEXT:    vpaddd %ymm3, %ymm8, %ymm3
-; AVX2-NEXT:    vpabsd %ymm14, %ymm8
-; AVX2-NEXT:    vpaddd %ymm1, %ymm8, %ymm1
-; AVX2-NEXT:    vpabsd %ymm13, %ymm8
-; AVX2-NEXT:    vpaddd %ymm2, %ymm8, %ymm2
-; AVX2-NEXT:    vpabsd %ymm12, %ymm8
-; AVX2-NEXT:    vpaddd %ymm0, %ymm8, %ymm0
-; AVX2-NEXT:    vpabsd %ymm11, %ymm8
-; AVX2-NEXT:    vpaddd %ymm4, %ymm8, %ymm4
-; AVX2-NEXT:    vpabsd %ymm10, %ymm8
-; AVX2-NEXT:    vpaddd %ymm6, %ymm8, %ymm6
-; AVX2-NEXT:    vpabsd -{{[0-9]+}}(%rsp), %ymm8 # 32-byte Folded Reload
-; AVX2-NEXT:    vpaddd %ymm5, %ymm8, %ymm5
-; AVX2-NEXT:    vpabsd %ymm15, %ymm8
-; AVX2-NEXT:    vpaddd %ymm7, %ymm8, %ymm7
+; AVX2-NEXT:    vmovdqu -{{[0-9]+}}(%rsp), %ymm8 # 32-byte Reload
+; AVX2-NEXT:    vpsubd %ymm15, %ymm8, %ymm15
+; AVX2-NEXT:    vpaddd %ymm1, %ymm15, %ymm1
+; AVX2-NEXT:    vpaddd %ymm2, %ymm9, %ymm2
+; AVX2-NEXT:    vpaddd %ymm0, %ymm10, %ymm0
+; AVX2-NEXT:    vpaddd %ymm3, %ymm11, %ymm3
+; AVX2-NEXT:    vpaddd %ymm4, %ymm12, %ymm4
+; AVX2-NEXT:    vpaddd %ymm6, %ymm13, %ymm6
+; AVX2-NEXT:    vpaddd %ymm5, %ymm14, %ymm5
+; AVX2-NEXT:    vpaddd -{{[0-9]+}}(%rsp), %ymm7, %ymm7 # 32-byte Folded Reload
 ; AVX2-NEXT:    addq $4, %rax
 ; AVX2-NEXT:    jne .LBB2_1
 ; AVX2-NEXT:  # BB#2: # %middle.block
@@ -776,14 +671,10 @@
 ; AVX512F-NEXT:    vpsubd %zmm10, %zmm6, %zmm6
 ; AVX512F-NEXT:    vpsubd %zmm9, %zmm5, %zmm5
 ; AVX512F-NEXT:    vpsubd %zmm8, %zmm4, %zmm4
-; AVX512F-NEXT:    vpabsd %zmm4, %zmm4
-; AVX512F-NEXT:    vpabsd %zmm5, %zmm5
-; AVX512F-NEXT:    vpabsd %zmm6, %zmm6
-; AVX512F-NEXT:    vpabsd %zmm7, %zmm7
-; AVX512F-NEXT:    vpaddd %zmm3, %zmm7, %zmm3
-; AVX512F-NEXT:    vpaddd %zmm2, %zmm6, %zmm2
+; AVX512F-NEXT:    vpaddd %zmm2, %zmm4, %zmm2
 ; AVX512F-NEXT:    vpaddd %zmm1, %zmm5, %zmm1
-; AVX512F-NEXT:    vpaddd %zmm0, %zmm4, %zmm0
+; AVX512F-NEXT:    vpaddd %zmm0, %zmm6, %zmm0
+; AVX512F-NEXT:    vpaddd %zmm3, %zmm7, %zmm3
 ; AVX512F-NEXT:    addq $4, %rax
 ; AVX512F-NEXT:    jne .LBB2_1
 ; AVX512F-NEXT:  # BB#2: # %middle.block
@@ -1147,95 +1038,64 @@
 define i32 @sad_nonloop_32i8(<32 x i8>* nocapture readonly %p, i64, <32 x i8>* nocapture readonly %q) local_unnamed_addr #0 {
 ; SSE2-LABEL: sad_nonloop_32i8:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movdqu (%rdi), %xmm0
+; SSE2-NEXT:    movdqu (%rdi), %xmm12
 ; SSE2-NEXT:    movdqu 16(%rdi), %xmm3
 ; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    movdqa %xmm3, %xmm12
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3],xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
-; SSE2-NEXT:    movdqa %xmm12, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
-; SSE2-NEXT:    movdqa %xmm0, %xmm13
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1],xmm13[2],xmm4[2],xmm13[3],xmm4[3],xmm13[4],xmm4[4],xmm13[5],xmm4[5],xmm13[6],xmm4[6],xmm13[7],xmm4[7]
-; SSE2-NEXT:    movdqa %xmm13, %xmm10
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm11
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm6
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1],xmm13[2],xmm4[2],xmm13[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm12, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE2-NEXT:    movdqu (%rdx), %xmm5
-; SSE2-NEXT:    movdqu 16(%rdx), %xmm7
-; SSE2-NEXT:    movdqa %xmm7, %xmm1
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm13
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm4[8],xmm13[9],xmm4[9],xmm13[10],xmm4[10],xmm13[11],xmm4[11],xmm13[12],xmm4[12],xmm13[13],xmm4[13],xmm13[14],xmm4[14],xmm13[15],xmm4[15]
+; SSE2-NEXT:    movdqa %xmm13, %xmm11
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3],xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
+; SSE2-NEXT:    movdqa %xmm12, %xmm2
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
 ; SSE2-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
-; SSE2-NEXT:    movdqa %xmm5, %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; SSE2-NEXT:    movdqa %xmm2, %xmm14
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm4[4],xmm14[5],xmm4[5],xmm14[6],xmm4[6],xmm14[7],xmm4[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm4[8],xmm7[9],xmm4[9],xmm7[10],xmm4[10],xmm7[11],xmm4[11],xmm7[12],xmm4[12],xmm7[13],xmm4[13],xmm7[14],xmm4[14],xmm7[15],xmm4[15]
-; SSE2-NEXT:    movdqa %xmm7, %xmm15
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm4[4],xmm15[5],xmm4[5],xmm15[6],xmm4[6],xmm15[7],xmm4[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm10
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm4[4],xmm13[5],xmm4[5],xmm13[6],xmm4[6],xmm13[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE2-NEXT:    movdqu (%rdx), %xmm6
+; SSE2-NEXT:    movdqu 16(%rdx), %xmm7
+; SSE2-NEXT:    movdqa %xmm6, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE2-NEXT:    movdqa %xmm2, %xmm9
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm7, %xmm5
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
-; SSE2-NEXT:    movdqa %xmm5, %xmm8
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; SSE2-NEXT:    psubd %xmm5, %xmm0
+; SSE2-NEXT:    movdqa %xmm5, %xmm14
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1],xmm14[2],xmm4[2],xmm14[3],xmm4[3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3],xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE2-NEXT:    movdqa %xmm6, %xmm15
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE2-NEXT:    movdqa %xmm7, %xmm8
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
 ; SSE2-NEXT:    psubd %xmm7, %xmm3
-; SSE2-NEXT:    psubd %xmm2, %xmm13
-; SSE2-NEXT:    psubd %xmm1, %xmm12
-; SSE2-NEXT:    psubd %xmm8, %xmm6
-; SSE2-NEXT:    psubd %xmm15, %xmm11
-; SSE2-NEXT:    psubd %xmm14, %xmm10
-; SSE2-NEXT:    psubd -{{[0-9]+}}(%rsp), %xmm9 # 16-byte Folded Reload
-; SSE2-NEXT:    movdqa %xmm9, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm9
-; SSE2-NEXT:    pxor %xmm1, %xmm9
-; SSE2-NEXT:    movdqa %xmm10, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm10
-; SSE2-NEXT:    pxor %xmm1, %xmm10
-; SSE2-NEXT:    movdqa %xmm11, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm11
-; SSE2-NEXT:    pxor %xmm1, %xmm11
-; SSE2-NEXT:    movdqa %xmm6, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm6
-; SSE2-NEXT:    pxor %xmm1, %xmm6
-; SSE2-NEXT:    movdqa %xmm12, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm12
-; SSE2-NEXT:    pxor %xmm1, %xmm12
-; SSE2-NEXT:    movdqa %xmm13, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm13
-; SSE2-NEXT:    pxor %xmm1, %xmm13
-; SSE2-NEXT:    movdqa %xmm3, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    paddd %xmm1, %xmm3
-; SSE2-NEXT:    pxor %xmm1, %xmm3
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrad $31, %xmm1
+; SSE2-NEXT:    psubd %xmm6, %xmm12
+; SSE2-NEXT:    psubd %xmm5, %xmm13
+; SSE2-NEXT:    psubd %xmm2, %xmm1
+; SSE2-NEXT:    psubd %xmm8, %xmm10
+; SSE2-NEXT:    movdqa -{{[0-9]+}}(%rsp), %xmm2 # 16-byte Reload
+; SSE2-NEXT:    psubd %xmm15, %xmm2
+; SSE2-NEXT:    psubd %xmm14, %xmm11
+; SSE2-NEXT:    psubd %xmm9, %xmm0
+; SSE2-NEXT:    paddd %xmm11, %xmm0
+; SSE2-NEXT:    paddd %xmm13, %xmm1
+; SSE2-NEXT:    paddd %xmm3, %xmm1
+; SSE2-NEXT:    paddd %xmm12, %xmm1
+; SSE2-NEXT:    paddd %xmm10, %xmm0
 ; SSE2-NEXT:    paddd %xmm1, %xmm0
-; SSE2-NEXT:    pxor %xmm1, %xmm0
-; SSE2-NEXT:    paddd %xmm3, %xmm0
-; SSE2-NEXT:    paddd %xmm11, %xmm6
-; SSE2-NEXT:    paddd %xmm9, %xmm6
-; SSE2-NEXT:    paddd %xmm10, %xmm6
-; SSE2-NEXT:    paddd %xmm12, %xmm0
-; SSE2-NEXT:    paddd %xmm6, %xmm0
-; SSE2-NEXT:    paddd %xmm13, %xmm0
+; SSE2-NEXT:    paddd %xmm2, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    paddd %xmm0, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
Index: test/CodeGen/X86/select.ll
===================================================================
--- test/CodeGen/X86/select.ll
+++ test/CodeGen/X86/select.ll
@@ -1082,7 +1082,6 @@
 ; MCU:       # BB#0: # %entry
 ; MCU-NEXT:    andl $1, %edx
 ; MCU-NEXT:    negl %edx
-; MCU-NEXT:    andl $43, %edx
 ; MCU-NEXT:    xorl %edx, %eax
 ; MCU-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
 ; MCU-NEXT:    retl