diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -970,6 +970,17 @@
     }
     if (SimplifyDemandedBits(Src, DemandedBits, SrcElts, Known, TLO, Depth + 1))
       return true;
+
+    // Attempt to avoid multi-use src if we don't need anything from it.
+    if (!DemandedBits.isAllOnesValue() || !SrcElts.isAllOnesValue()) {
+      SDValue DemandedSrc = SimplifyMultipleUseDemandedBits(
+          Src, DemandedBits, SrcElts, TLO.DAG, Depth + 1);
+      if (DemandedSrc) {
+        SDValue NewOp = TLO.DAG.getNode(Op.getOpcode(), dl, VT, DemandedSrc,
+                                        Op.getOperand(1));
+        return TLO.CombineTo(Op, NewOp);
+      }
+    }
     break;
   }
   case ISD::CONCAT_VECTORS: {
diff --git a/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll b/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
--- a/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
+++ b/llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll
@@ -129,8 +129,6 @@
 define i32 @test_v3i32(<3 x i32> %a) nounwind {
 ; CHECK-LABEL: test_v3i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    mov v0.s[3], w8
 ; CHECK-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
 ; CHECK-NEXT:    and v1.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    mov w8, v0.s[1]
diff --git a/llvm/test/CodeGen/X86/avx-vperm2x128.ll b/llvm/test/CodeGen/X86/avx-vperm2x128.ll
--- a/llvm/test/CodeGen/X86/avx-vperm2x128.ll
+++ b/llvm/test/CodeGen/X86/avx-vperm2x128.ll
@@ -627,11 +627,11 @@
 define <4 x i64> @ld1_hi0_hi1_4i64(<4 x i64> %a, <4 x i64> * %pb) nounwind uwtable readnone ssp {
 ; AVX1-LABEL: ld1_hi0_hi1_4i64:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[2,3]
-; AVX1-NEXT:    vpaddq {{.*}}(%rip), %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpaddq {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
+; AVX1-NEXT:    vpaddq {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: ld1_hi0_hi1_4i64:
@@ -672,12 +672,11 @@
 define <8 x i32> @ld1_hi0_hi1_8i32(<8 x i32> %a, <8 x i32> * %pb) nounwind uwtable readnone ssp {
 ; AVX1-LABEL: ld1_hi0_hi1_8i32:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[2,3]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [1,2,3,4]
-; AVX1-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,2,3,4]
+; AVX1-NEXT:    vpaddd 16(%rdi), %xmm1, %xmm2
+; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: ld1_hi0_hi1_8i32:
diff --git a/llvm/test/CodeGen/X86/bitcast-setcc-256.ll b/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
--- a/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
+++ b/llvm/test/CodeGen/X86/bitcast-setcc-256.ll
@@ -375,8 +375,6 @@
 ;
 ; AVX2-LABEL: bitcast_16i16_store:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
diff --git a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
--- a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
+++ b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
@@ -296,8 +296,6 @@
 ;
 ; AVX2-LABEL: bitcast_v16i16_to_v2i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %ecx
diff --git a/llvm/test/CodeGen/X86/masked_load.ll b/llvm/test/CodeGen/X86/masked_load.ll
--- a/llvm/test/CodeGen/X86/masked_load.ll
+++ b/llvm/test/CodeGen/X86/masked_load.ll
@@ -3198,8 +3198,6 @@
 ;
 ; AVX2-LABEL: load_v16i16_v16i16:
 ; AVX2:       ## %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
 ; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
--- a/llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -418,8 +418,6 @@
 ;
 ; AVX2-LABEL: allones_v16i16_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
@@ -473,8 +471,6 @@
 ;
 ; AVX2-LABEL: allzeros_v16i16_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
diff --git a/llvm/test/CodeGen/X86/pr31956.ll b/llvm/test/CodeGen/X86/pr31956.ll
--- a/llvm/test/CodeGen/X86/pr31956.ll
+++ b/llvm/test/CodeGen/X86/pr31956.ll
@@ -11,9 +11,8 @@
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; CHECK-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3,4,5,6,7]
-; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[0,2],xmm0[2,0]
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],mem[0,2]
+; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,0,3,1]
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
 entry:
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
@@ -1068,29 +1068,27 @@
 ;
 ; AVX2-64-LABEL: uitofp_v4i64_v4f32:
 ; AVX2-64:       # %bb.0:
-; AVX2-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-64-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm1
-; AVX2-64-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-64-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
-; AVX2-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [1,1,1,1]
-; AVX2-64-NEXT:    vpand %ymm2, %ymm0, %ymm2
-; AVX2-64-NEXT:    vpsrlq $1, %ymm0, %ymm3
-; AVX2-64-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-64-NEXT:    vblendvpd %ymm0, %ymm2, %ymm0, %ymm0
-; AVX2-64-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm2
-; AVX2-64-NEXT:    vmovq %xmm0, %rax
-; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
+; AVX2-64-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
+; AVX2-64-NEXT:    vpand %ymm1, %ymm0, %ymm1
+; AVX2-64-NEXT:    vpsrlq $1, %ymm0, %ymm2
+; AVX2-64-NEXT:    vpor %ymm1, %ymm2, %ymm1
+; AVX2-64-NEXT:    vblendvpd %ymm0, %ymm1, %ymm0, %ymm1
+; AVX2-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm2
+; AVX2-64-NEXT:    vmovq %xmm1, %rax
+; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm3
 ; AVX2-64-NEXT:    vinsertps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[2,3]
-; AVX2-64-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-64-NEXT:    vmovq %xmm0, %rax
+; AVX2-64-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-64-NEXT:    vmovq %xmm1, %rax
 ; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
 ; AVX2-64-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0],xmm2[3]
-; AVX2-64-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm0
-; AVX2-64-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
-; AVX2-64-NEXT:    vaddps %xmm0, %xmm0, %xmm2
-; AVX2-64-NEXT:    vblendvps %xmm1, %xmm2, %xmm0, %xmm0
+; AVX2-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-64-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm1
+; AVX2-64-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
+; AVX2-64-NEXT:    vaddps %xmm1, %xmm1, %xmm2
+; AVX2-64-NEXT:    vextracti128 $1, %ymm0, %xmm3
+; AVX2-64-NEXT:    vpackssdw %xmm3, %xmm0, %xmm0
+; AVX2-64-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
 ; AVX2-64-NEXT:    vzeroupper
 ; AVX2-64-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vec_int_to_fp.ll b/llvm/test/CodeGen/X86/vec_int_to_fp.ll
--- a/llvm/test/CodeGen/X86/vec_int_to_fp.ll
+++ b/llvm/test/CodeGen/X86/vec_int_to_fp.ll
@@ -2194,10 +2194,7 @@
 ; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm1
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
 ; AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm2
-; AVX2-NEXT:    vxorps %xmm3, %xmm3, %xmm3
-; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm3, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX2-NEXT:    vpackssdw %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vpackssdw %xmm0, %xmm0, %xmm0
 ; AVX2-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -2593,29 +2590,27 @@
 ;
 ; AVX2-LABEL: uitofp_4i64_to_4f32:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [1,1,1,1]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm2
-; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm3
-; AVX2-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vblendvpd %ymm0, %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm2
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm2
+; AVX2-NEXT:    vpor %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vblendvpd %ymm0, %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm2
+; AVX2-NEXT:    vmovq %xmm1, %rax
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm3
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[2,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vmovq %xmm0, %rax
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-NEXT:    vmovq %xmm1, %rax
 ; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
 ; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0],xmm2[3]
-; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm0
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
-; AVX2-NEXT:    vaddps %xmm0, %xmm0, %xmm2
-; AVX2-NEXT:    vblendvps %xmm1, %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm1
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
+; AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm2
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
+; AVX2-NEXT:    vpackssdw %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -4512,29 +4507,27 @@
 ; AVX2-LABEL: uitofp_load_4i64_to_4f32:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [1,1,1,1]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm2
-; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm3
-; AVX2-NEXT:    vpor %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vblendvpd %ymm0, %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm2
+; AVX2-NEXT:    vpor %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vblendvpd %ymm0, %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm2
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm1
 ; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[2,3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm2
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[2,3]
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm3
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0],xmm2[3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm2
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0],xmm1[3]
 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm4, %xmm0
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
-; AVX2-NEXT:    vaddps %xmm0, %xmm0, %xmm2
-; AVX2-NEXT:    vblendvps %xmm1, %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm0
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
+; AVX2-NEXT:    vaddps %xmm0, %xmm0, %xmm1
+; AVX2-NEXT:    vmovdqa (%rdi), %xmm2
+; AVX2-NEXT:    vpackssdw 16(%rdi), %xmm2, %xmm2
+; AVX2-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -4993,50 +4986,47 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovaps (%rdi), %ymm0
 ; AVX2-NEXT:    vmovdqa 32(%rdi), %ymm1
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3
-; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm4
-; AVX2-NEXT:    vpackssdw %xmm4, %xmm3, %xmm3
-; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm4 = [1,1,1,1]
-; AVX2-NEXT:    vpand %ymm4, %ymm1, %ymm5
-; AVX2-NEXT:    vpsrlq $1, %ymm1, %ymm6
-; AVX2-NEXT:    vpor %ymm5, %ymm6, %ymm5
-; AVX2-NEXT:    vblendvpd %ymm1, %ymm5, %ymm1, %ymm1
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [1,1,1,1]
+; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX2-NEXT:    vpsrlq $1, %ymm1, %ymm4
+; AVX2-NEXT:    vpor %ymm3, %ymm4, %ymm3
+; AVX2-NEXT:    vblendvpd %ymm1, %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    vpextrq $1, %xmm1, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm5
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm5, %xmm3
 ; AVX2-NEXT:    vmovq %xmm1, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm6
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[2,3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm5, %xmm4
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[2,3]
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
 ; AVX2-NEXT:    vmovq %xmm1, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm6
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm5 = xmm5[0,1],xmm6[0],xmm5[3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm5, %xmm4
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0],xmm3[3]
 ; AVX2-NEXT:    vpextrq $1, %xmm1, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm1
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm5[0,1,2],xmm1[0]
-; AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm5
-; AVX2-NEXT:    vblendvps %xmm3, %xmm5, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpackssdw %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpand %ymm4, %ymm0, %ymm3
-; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm4
-; AVX2-NEXT:    vpor %ymm3, %ymm4, %ymm3
-; AVX2-NEXT:    vblendvpd %ymm0, %ymm3, %ymm0, %ymm0
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm5, %xmm1
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[0]
+; AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm3
+; AVX2-NEXT:    vmovdqa (%rdi), %xmm4
+; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm5
+; AVX2-NEXT:    vpackssdw 48(%rdi), %xmm5, %xmm5
+; AVX2-NEXT:    vblendvps %xmm5, %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vandps %ymm2, %ymm0, %ymm2
+; AVX2-NEXT:    vpsrlq $1, %ymm0, %ymm3
+; AVX2-NEXT:    vpor %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vblendvpd %ymm0, %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm3
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm6, %xmm2
 ; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm4
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[2,3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm6, %xmm3
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[2,3]
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm4
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0],xmm3[3]
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm6, %xmm3
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0],xmm2[3]
 ; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    vcvtsi2ss %rax, %xmm7, %xmm0
-; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[0]
-; AVX2-NEXT:    vaddps %xmm0, %xmm0, %xmm3
-; AVX2-NEXT:    vblendvps %xmm2, %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vcvtsi2ss %rax, %xmm6, %xmm0
+; AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[0]
+; AVX2-NEXT:    vaddps %xmm0, %xmm0, %xmm2
+; AVX2-NEXT:    vpackssdw 16(%rdi), %xmm4, %xmm3
+; AVX2-NEXT:    vblendvps %xmm3, %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;