Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -28953,6 +28953,40 @@
       Other = LHS;
     }
 
+    // This optimization requires minimum target sse4.1, because it requires
+    // umin instruction.
+    if (Subtarget.hasSSE41() && Other.getNode() &&
+        Other.getOpcode() == ISD::TRUNCATE &&
+        Other.getOperand(0).getOpcode() == ISD::SUB && isUnsignedIntSetCC(CC)) {
+      //  x >= y ? x - y : 0, where y is wider then x. We can truncate y
+      //  to x, because it is only used if it is less than x.
+      SDValue SubXY = Other.getOperand(0);
+      EVT ExtType = SubXY.getValueType();
+
+      SDValue SubLHS = SubXY.getOperand(0), SubRHS = SubXY.getOperand(1);
+      SDValue CondLHS = Cond.getOperand(0), CondRHS = Cond.getOperand(1);
+      ISD::CondCode NewCC = CC;
+      // Else can occur, if the sub arguments are swapped
+      // x < y ? y - x : 0 -> y > x ? y - x : 0
+      if (CC == ISD::SETULE || CC == ISD::SETULT) {
+        std::swap(CondLHS, CondRHS);
+        NewCC = CC == ISD::SETULE ? ISD::SETUGE : ISD::SETUGT;
+      }
+      if (CondLHS.getOpcode() == ISD::ZERO_EXTEND &&
+          DAG.isEqualTo(SubRHS, CondRHS) && DAG.isEqualTo(SubLHS, CondLHS)) {
+        SDValue SaturationConst = DAG.getConstant(
+            APInt::getMaxValue(VT.getScalarSizeInBits()).getLimitedValue(),
+            SDLoc(SubRHS), ExtType);
+        SDValue UMin = DAG.getNode(ISD::UMIN, SDLoc(SubRHS), ExtType, SubRHS,
+                                   SaturationConst);
+        Other = DAG.getNode(ISD::SUB, SDLoc(SubXY), VT, CondLHS.getOperand(0),
+                            DAG.getZExtOrTrunc(UMin, SDLoc(SubRHS), VT));
+        Cond = DAG.getNode(NewCC, SDLoc(Cond), VT, Other.getOperand(0),
+                           Other.getOperand(1));
+        CC = NewCC;
+      }
+    }
+
     if (Other.getNode() && Other->getNumOperands() == 2 &&
         DAG.isEqualTo(Other->getOperand(0), Cond.getOperand(0))) {
       SDValue OpLHS = Other->getOperand(0), OpRHS = Other->getOperand(1);
Index: test/CodeGen/X86/psubus.ll
===================================================================
--- test/CodeGen/X86/psubus.ll
+++ test/CodeGen/X86/psubus.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mattr=+ssse3 | FileCheck %s --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE --check-prefix=SSE41
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
 
@@ -168,6 +169,16 @@
 ; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test6:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movd %esi, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    movdqu (%rdi), %xmm1
+; SSE41-NEXT:    psubusb %xmm0, %xmm1
+; SSE41-NEXT:    movdqu %xmm1, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test6:
 ; AVX1:       ## BB#0: ## %vector.ph
 ; AVX1-NEXT:    vmovd %esi, %xmm0
@@ -473,6 +484,19 @@
 ; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test12:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movd %esi, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    movdqu (%rdi), %xmm1
+; SSE41-NEXT:    movdqu 16(%rdi), %xmm2
+; SSE41-NEXT:    psubusb %xmm0, %xmm1
+; SSE41-NEXT:    psubusb %xmm0, %xmm2
+; SSE41-NEXT:    movdqu %xmm2, 16(%rdi)
+; SSE41-NEXT:    movdqu %xmm1, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test12:
 ; AVX1:       ## BB#0: ## %vector.ph
 ; AVX1-NEXT:    vmovups (%rdi), %ymm0
@@ -587,45 +611,47 @@
 ; SSSE3-NEXT:    movdqu %xmm6, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test13:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movdqu (%rdi), %xmm0
+; SSE41-NEXT:    movdqu (%rsi), %xmm1
+; SSE41-NEXT:    movdqu 16(%rsi), %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; SSE41-NEXT:    pminud %xmm3, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm4, %xmm2
+; SSE41-NEXT:    pminud %xmm3, %xmm1
+; SSE41-NEXT:    pshufb %xmm4, %xmm1
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE41-NEXT:    psubusw %xmm1, %xmm0
+; SSE41-NEXT:    movdqu %xmm0, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test13:
 ; AVX1:       ## BB#0: ## %vector.ph
-; AVX1-NEXT:    vmovups (%rsi), %ymm0
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
-; AVX1-NEXT:    vpxor %xmm3, %xmm2, %xmm4
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT:    vpxor %xmm3, %xmm5, %xmm6
-; AVX1-NEXT:    vpcmpgtd %xmm4, %xmm6, %xmm4
-; AVX1-NEXT:    vpxor %xmm3, %xmm1, %xmm6
-; AVX1-NEXT:    vxorps %xmm3, %xmm0, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm6, %xmm3, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm4, %xmm3, %xmm3
-; AVX1-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpsubd %xmm5, %xmm2, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-NEXT:    vpandn %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX1-NEXT:    vmovups (%rsi), %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; AVX1-NEXT:    vpminud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpminud %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test13:
 ; AVX2:       ## BB#0: ## %vector.ph
-; AVX2-NEXT:    vmovdqu (%rsi), %ymm0
-; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm2
-; AVX2-NEXT:    vpxor %ymm2, %ymm1, %ymm3
-; AVX2-NEXT:    vpxor %ymm2, %ymm0, %ymm2
-; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpsubd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpandn %xmm0, %xmm2, %xmm0
+; AVX2-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; AVX2-NEXT:    vpminud (%rsi), %ymm1, %ymm1
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -789,90 +815,69 @@
 ; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test14:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movdqu (%rdi), %xmm0
+; SSE41-NEXT:    movdqu (%rsi), %xmm1
+; SSE41-NEXT:    movdqu 16(%rsi), %xmm2
+; SSE41-NEXT:    movdqu 32(%rsi), %xmm3
+; SSE41-NEXT:    movdqu 48(%rsi), %xmm4
+; SSE41-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255]
+; SSE41-NEXT:    pminud %xmm5, %xmm4
+; SSE41-NEXT:    pand %xmm5, %xmm4
+; SSE41-NEXT:    pminud %xmm5, %xmm3
+; SSE41-NEXT:    pand %xmm5, %xmm3
+; SSE41-NEXT:    packuswb %xmm4, %xmm3
+; SSE41-NEXT:    pminud %xmm5, %xmm2
+; SSE41-NEXT:    pand %xmm5, %xmm2
+; SSE41-NEXT:    pminud %xmm5, %xmm1
+; SSE41-NEXT:    pand %xmm5, %xmm1
+; SSE41-NEXT:    packuswb %xmm2, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    psubusb %xmm1, %xmm0
+; SSE41-NEXT:    movdqu %xmm0, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test14:
 ; AVX1:       ## BB#0: ## %vector.ph
-; AVX1-NEXT:    vmovups (%rsi), %ymm0
-; AVX1-NEXT:    vmovups 32(%rsi), %ymm1
-; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm8 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero
-; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm9 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero
-; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm10 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero
-; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm5 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648]
-; AVX1-NEXT:    vpxor %xmm6, %xmm5, %xmm7
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpxor %xmm6, %xmm2, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm3, %xmm3
-; AVX1-NEXT:    vpxor %xmm6, %xmm10, %xmm7
-; AVX1-NEXT:    vxorps %xmm6, %xmm1, %xmm4
-; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpacksswb %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm11 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; AVX1-NEXT:    vpshufb %xmm11, %xmm3, %xmm12
-; AVX1-NEXT:    vpxor %xmm6, %xmm9, %xmm7
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
-; AVX1-NEXT:    vxorps %xmm6, %xmm4, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm3, %xmm3
-; AVX1-NEXT:    vpxor %xmm6, %xmm8, %xmm7
-; AVX1-NEXT:    vxorps %xmm6, %xmm0, %xmm6
-; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm6, %xmm6
-; AVX1-NEXT:    vpacksswb %xmm3, %xmm6, %xmm3
-; AVX1-NEXT:    vpshufb %xmm11, %xmm3, %xmm3
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm12[0]
-; AVX1-NEXT:    vpsllw $7, %xmm3, %xmm3
-; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpsubd %xmm0, %xmm8, %xmm0
-; AVX1-NEXT:    vpsubd %xmm4, %xmm9, %xmm4
-; AVX1-NEXT:    vpsubd %xmm1, %xmm10, %xmm1
-; AVX1-NEXT:    vpsubd %xmm2, %xmm5, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm5 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
-; AVX1-NEXT:    vpand %xmm5, %xmm2, %xmm2
-; AVX1-NEXT:    vpand %xmm5, %xmm1, %xmm1
+; AVX1-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX1-NEXT:    vmovups (%rsi), %ymm1
+; AVX1-NEXT:    vmovups 32(%rsi), %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255]
+; AVX1-NEXT:    vpminud %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vpand %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vpminud %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpand %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpminud %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vpand %xmm4, %xmm3, %xmm3
+; AVX1-NEXT:    vpminud %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpand %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpand %xmm5, %xmm4, %xmm2
-; AVX1-NEXT:    vpand %xmm5, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendvb %xmm3, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test14:
 ; AVX2:       ## BB#0: ## %vector.ph
-; AVX2-NEXT:    vmovdqu (%rsi), %ymm0
-; AVX2-NEXT:    vmovdqu 32(%rsi), %ymm1
-; AVX2-NEXT:    vpmovzxbd {{.*#+}} ymm2 = 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 {{.*#+}} ymm3 = 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:    vpbroadcastd {{.*}}(%rip), %ymm4
-; AVX2-NEXT:    vpxor %ymm4, %ymm3, %ymm5
-; AVX2-NEXT:    vpxor %ymm4, %ymm1, %ymm6
-; AVX2-NEXT:    vpcmpgtd %ymm5, %ymm6, %ymm5
-; AVX2-NEXT:    vextracti128 $1, %ymm5, %xmm6
-; AVX2-NEXT:    vpacksswb %xmm6, %xmm5, %xmm5
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm6 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX2-NEXT:    vpxor %ymm4, %ymm2, %ymm7
-; AVX2-NEXT:    vpxor %ymm4, %ymm0, %ymm4
-; AVX2-NEXT:    vpcmpgtd %ymm7, %ymm4, %ymm4
-; AVX2-NEXT:    vextracti128 $1, %ymm4, %xmm7
-; AVX2-NEXT:    vpacksswb %xmm7, %xmm4, %xmm4
-; AVX2-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
-; AVX2-NEXT:    vpsllw $7, %xmm4, %xmm4
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm4, %xmm4
-; AVX2-NEXT:    vpsubd %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpsubd %ymm1, %ymm3, %ymm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,4,5,8,9,12,13,128,128,128,128,128,128,128,128,0,1,4,5,8,9,12,13,128,128,128,128,128,128,128,128]
-; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; AVX2-NEXT:    vpminud 32(%rsi), %ymm1, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,128,128,128,128,128,128,128,128,0,1,4,5,8,9,12,13,128,128,128,128,128,128,128,128]
+; AVX2-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
+; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; AVX2-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX2-NEXT:    vpminud (%rsi), %ymm1, %ymm1
+; AVX2-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpshufb %xmm6, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpblendvb %xmm4, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -964,45 +969,47 @@
 ; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test15:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movdqu (%rdi), %xmm0
+; SSE41-NEXT:    movdqu (%rsi), %xmm1
+; SSE41-NEXT:    movdqu 16(%rsi), %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; SSE41-NEXT:    pminud %xmm3, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm4, %xmm2
+; SSE41-NEXT:    pminud %xmm3, %xmm1
+; SSE41-NEXT:    pshufb %xmm4, %xmm1
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE41-NEXT:    psubusw %xmm1, %xmm0
+; SSE41-NEXT:    movdqu %xmm0, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test15:
 ; AVX1:       ## BB#0: ## %vector.ph
-; AVX1-NEXT:    vmovups (%rsi), %ymm0
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
-; AVX1-NEXT:    vpxor %xmm3, %xmm2, %xmm4
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT:    vpxor %xmm3, %xmm5, %xmm6
-; AVX1-NEXT:    vpcmpgtd %xmm6, %xmm4, %xmm4
-; AVX1-NEXT:    vpxor %xmm3, %xmm1, %xmm6
-; AVX1-NEXT:    vxorps %xmm3, %xmm0, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm6, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm4, %xmm3, %xmm3
-; AVX1-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpsubd %xmm5, %xmm2, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-NEXT:    vpand %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX1-NEXT:    vmovups (%rsi), %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; AVX1-NEXT:    vpminud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpminud %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test15:
 ; AVX2:       ## BB#0: ## %vector.ph
-; AVX2-NEXT:    vmovdqu (%rsi), %ymm0
-; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm2
-; AVX2-NEXT:    vpxor %ymm2, %ymm0, %ymm3
-; AVX2-NEXT:    vpxor %ymm2, %ymm1, %ymm2
-; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpsubd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpand %xmm0, %xmm2, %xmm0
+; AVX2-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; AVX2-NEXT:    vpminud (%rsi), %ymm1, %ymm1
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -1094,45 +1101,47 @@
 ; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
+; SSE41-LABEL: test16:
+; SSE41:       ## BB#0: ## %vector.ph
+; SSE41-NEXT:    movdqu (%rdi), %xmm0
+; SSE41-NEXT:    movdqu (%rsi), %xmm1
+; SSE41-NEXT:    movdqu 16(%rsi), %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; SSE41-NEXT:    pminud %xmm3, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm4, %xmm2
+; SSE41-NEXT:    pminud %xmm3, %xmm1
+; SSE41-NEXT:    pshufb %xmm4, %xmm1
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE41-NEXT:    psubusw %xmm1, %xmm0
+; SSE41-NEXT:    movdqu %xmm0, (%rdi)
+; SSE41-NEXT:    retq
+;
 ; AVX1-LABEL: test16:
 ; AVX1:       ## BB#0: ## %vector.ph
-; AVX1-NEXT:    vmovups (%rsi), %ymm0
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
-; AVX1-NEXT:    vpxor %xmm3, %xmm2, %xmm4
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT:    vpxor %xmm3, %xmm5, %xmm6
-; AVX1-NEXT:    vpcmpgtd %xmm6, %xmm4, %xmm4
-; AVX1-NEXT:    vpxor %xmm3, %xmm1, %xmm6
-; AVX1-NEXT:    vxorps %xmm3, %xmm0, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm6, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm4, %xmm3, %xmm3
-; AVX1-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpsubd %xmm5, %xmm2, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-NEXT:    vpand %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX1-NEXT:    vmovups (%rsi), %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; AVX1-NEXT:    vpminud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpminud %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test16:
 ; AVX2:       ## BB#0: ## %vector.ph
-; AVX2-NEXT:    vmovdqu (%rsi), %ymm0
-; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm2
-; AVX2-NEXT:    vpxor %ymm2, %ymm0, %ymm3
-; AVX2-NEXT:    vpxor %ymm2, %ymm1, %ymm2
-; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpsubd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpand %xmm0, %xmm2, %xmm0
+; AVX2-NEXT:    vmovdqu (%rdi), %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; AVX2-NEXT:    vpminud (%rsi), %ymm1, %ymm1
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17,20,21,24,25,28,29],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq