Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -27356,6 +27356,37 @@
       Other = LHS;
     }
 
+    if (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
@@ -1,3 +1,4 @@
+; 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=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
@@ -513,3 +514,657 @@
   store <32 x i8> %6, <32 x i8>* %2, align 1
   ret void
 }
+
+define void @test13(i16* nocapture %head, i32* nocapture %w) nounwind {
+; SSE2-LABEL: test13:
+; SSE2:       ## BB#0: ## %vector.ph
+; SSE2-NEXT:    movdqu (%rdi), %xmm0
+; SSE2-NEXT:    movdqu (%rsi), %xmm1
+; SSE2-NEXT:    movdqu 16(%rsi), %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSE2-NEXT:    movd %xmm2, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm2
+; SSE2-NEXT:    psrad $16, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovbl %ecx, %eax
+; SSE2-NEXT:    movd %eax, %xmm1
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm3
+; SSE2-NEXT:    psrad $16, %xmm3
+; SSE2-NEXT:    packssdw %xmm2, %xmm3
+; SSE2-NEXT:    psubusw %xmm3, %xmm0
+; SSE2-NEXT:    movdqu %xmm0, (%rdi)
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: test13:
+; SSSE3:       ## BB#0: ## %vector.ph
+; SSSE3-NEXT:    movdqu (%rdi), %xmm0
+; SSSE3-NEXT:    movdqu (%rsi), %xmm1
+; SSSE3-NEXT:    movdqu 16(%rsi), %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm2, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm3
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm5, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm5
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovbl %ecx, %eax
+; SSSE3-NEXT:    movd %eax, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSSE3-NEXT:    psubusw %xmm4, %xmm0
+; SSSE3-NEXT:    movdqu %xmm0, (%rdi)
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: test13:
+; AVX1:       ## BB#0: ## %vector.ph
+; 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 (%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
+vector.ph:
+  %0 = getelementptr inbounds i16, i16* %head, i64 0
+  %1 = bitcast i16* %0 to <8 x i16>*
+  %2 = load <8 x i16>, <8 x i16>* %1, align 2
+  %3 = getelementptr inbounds i32, i32* %w, i64 0
+  %4 = bitcast i32* %3 to <8 x i32>*
+  %5 = load <8 x i32>, <8 x i32>* %4, align 2
+  %6 = zext <8 x i16> %2 to <8 x i32>
+  %7 = icmp ult <8 x i32> %6, %5
+  %8 = sub <8 x i32> %6, %5
+  %9 = trunc <8 x i32> %8 to <8 x i16>
+  %10 = select <8 x i1> %7, <8 x i16> zeroinitializer, <8 x i16> %9
+  store <8 x i16> %10, <8 x i16>* %1, align 1
+  ret void
+}
+
+define void @test14(i8* nocapture %head, i32* nocapture %w) nounwind {
+; SSE-LABEL: test14:
+; SSE:       ## BB#0: ## %vector.ph
+; SSE-NEXT:    movdqu (%rdi), %xmm0
+; SSE-NEXT:    movdqu (%rsi), %xmm1
+; SSE-NEXT:    movdqu 16(%rsi), %xmm3
+; SSE-NEXT:    movdqu 32(%rsi), %xmm5
+; SSE-NEXT:    movdqu 48(%rsi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[3,1,2,3]
+; SSE-NEXT:    movd %xmm4, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    movl $255, %eax
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[1,1,2,3]
+; SSE-NEXT:    movd %xmm6, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
+; SSE-NEXT:    movd %xmm2, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
+; SSE-NEXT:    movd %xmm2, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[3,1,2,3]
+; SSE-NEXT:    movd %xmm4, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[1,1,2,3]
+; SSE-NEXT:    movd %xmm7, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
+; SSE-NEXT:    movd %xmm5, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
+; SSE-NEXT:    movd %xmm5, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    packuswb %xmm6, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[3,1,2,3]
+; SSE-NEXT:    movd %xmm5, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[1,1,2,3]
+; SSE-NEXT:    movd %xmm6, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm6
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; SSE-NEXT:    movd %xmm3, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE-NEXT:    movd %xmm3, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE-NEXT:    movd %xmm3, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[1,1,2,3]
+; SSE-NEXT:    movd %xmm6, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm6
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
+; SSE-NEXT:    movd %xmm1, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovael %eax, %ecx
+; SSE-NEXT:    movd %ecx, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE-NEXT:    movd %xmm1, %ecx
+; SSE-NEXT:    cmpl $255, %ecx
+; SSE-NEXT:    cmovbl %ecx, %eax
+; SSE-NEXT:    movd %eax, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    packuswb %xmm5, %xmm3
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    psubusb %xmm3, %xmm0
+; SSE-NEXT:    movdqu %xmm0, (%rdi)
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: test14:
+; AVX1:       ## BB#0: ## %vector.ph
+; 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:    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 (%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 %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
+vector.ph:
+  %0 = getelementptr inbounds i8, i8* %head, i64 0
+  %1 = bitcast i8* %0 to <16 x i8>*
+  %2 = load <16 x i8>, <16 x i8>* %1, align 2
+  %3 = getelementptr inbounds i32, i32* %w, i64 0
+  %4 = bitcast i32* %3 to <16 x i32>*
+  %5 = load <16 x i32>, <16 x i32>* %4, align 2
+  %6 = zext <16 x i8> %2 to <16 x i32>
+  %7 = icmp ult <16 x i32> %6, %5
+  %8 = sub <16 x i32> %6, %5
+  %9 = trunc <16 x i32> %8 to <16 x i8>
+  %10 = select <16 x i1> %7, <16 x i8> zeroinitializer, <16 x i8> %9
+  store <16 x i8> %10, <16 x i8>* %1, align 1
+  ret void
+}
+
+define void @test15(i16* nocapture %head, i32* nocapture %w) nounwind {
+; SSE2-LABEL: test15:
+; SSE2:       ## BB#0: ## %vector.ph
+; SSE2-NEXT:    movdqu (%rdi), %xmm0
+; SSE2-NEXT:    movdqu (%rsi), %xmm1
+; SSE2-NEXT:    movdqu 16(%rsi), %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSE2-NEXT:    movd %xmm2, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm2
+; SSE2-NEXT:    psrad $16, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovbl %ecx, %eax
+; SSE2-NEXT:    movd %eax, %xmm1
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm3
+; SSE2-NEXT:    psrad $16, %xmm3
+; SSE2-NEXT:    packssdw %xmm2, %xmm3
+; SSE2-NEXT:    psubusw %xmm3, %xmm0
+; SSE2-NEXT:    movdqu %xmm0, (%rdi)
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: test15:
+; SSSE3:       ## BB#0: ## %vector.ph
+; SSSE3-NEXT:    movdqu (%rdi), %xmm0
+; SSSE3-NEXT:    movdqu (%rsi), %xmm1
+; SSSE3-NEXT:    movdqu 16(%rsi), %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm2, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm3
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm5, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm5
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovbl %ecx, %eax
+; SSSE3-NEXT:    movd %eax, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSSE3-NEXT:    psubusw %xmm4, %xmm0
+; SSSE3-NEXT:    movdqu %xmm0, (%rdi)
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: test15:
+; AVX1:       ## BB#0: ## %vector.ph
+; 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 (%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
+vector.ph:
+  %0 = getelementptr inbounds i16, i16* %head, i64 0
+  %1 = bitcast i16* %0 to <8 x i16>*
+  %2 = load <8 x i16>, <8 x i16>* %1, align 2
+  %3 = getelementptr inbounds i32, i32* %w, i64 0
+  %4 = bitcast i32* %3 to <8 x i32>*
+  %5 = load <8 x i32>, <8 x i32>* %4, align 2
+  %6 = zext <8 x i16> %2 to <8 x i32>
+  %7 = icmp ugt <8 x i32> %6, %5
+  %8 = sub <8 x i32> %6, %5
+  %9 = trunc <8 x i32> %8 to <8 x i16>
+  %10 = select <8 x i1> %7, <8 x i16> %9, <8 x i16> zeroinitializer
+  store <8 x i16> %10, <8 x i16>* %1, align 1
+  ret void
+}
+
+define void @test16(i16* nocapture %head, i32* nocapture %w) nounwind {
+; SSE2-LABEL: test16:
+; SSE2:       ## BB#0: ## %vector.ph
+; SSE2-NEXT:    movdqu (%rdi), %xmm0
+; SSE2-NEXT:    movdqu (%rsi), %xmm1
+; SSE2-NEXT:    movdqu 16(%rsi), %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSE2-NEXT:    movd %xmm2, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm2
+; SSE2-NEXT:    psrad $16, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE2-NEXT:    movd %xmm3, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,2,3]
+; SSE2-NEXT:    movd %xmm4, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm4
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovael %eax, %ecx
+; SSE2-NEXT:    movd %ecx, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    movd %xmm1, %ecx
+; SSE2-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSE2-NEXT:    cmovbl %ecx, %eax
+; SSE2-NEXT:    movd %eax, %xmm1
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE2-NEXT:    pslld $16, %xmm3
+; SSE2-NEXT:    psrad $16, %xmm3
+; SSE2-NEXT:    packssdw %xmm2, %xmm3
+; SSE2-NEXT:    psubusw %xmm3, %xmm0
+; SSE2-NEXT:    movdqu %xmm0, (%rdi)
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: test16:
+; SSSE3:       ## BB#0: ## %vector.ph
+; SSSE3-NEXT:    movdqu (%rdi), %xmm0
+; SSSE3-NEXT:    movdqu (%rsi), %xmm1
+; SSSE3-NEXT:    movdqu 16(%rsi), %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm2, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    movl $65535, %eax ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm3, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm3
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[3,1,2,3]
+; SSSE3-NEXT:    movd %xmm4, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,2,3]
+; SSSE3-NEXT:    movd %xmm5, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm5
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovael %eax, %ecx
+; SSSE3-NEXT:    movd %ecx, %xmm4
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSSE3-NEXT:    movd %xmm1, %ecx
+; SSSE3-NEXT:    cmpl $65535, %ecx ## imm = 0xFFFF
+; SSSE3-NEXT:    cmovbl %ecx, %eax
+; SSSE3-NEXT:    movd %eax, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSSE3-NEXT:    psubusw %xmm4, %xmm0
+; SSSE3-NEXT:    movdqu %xmm0, (%rdi)
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: test16:
+; AVX1:       ## BB#0: ## %vector.ph
+; 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 (%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
+vector.ph:
+  %0 = getelementptr inbounds i16, i16* %head, i64 0
+  %1 = bitcast i16* %0 to <8 x i16>*
+  %2 = load <8 x i16>, <8 x i16>* %1, align 2
+  %3 = getelementptr inbounds i32, i32* %w, i64 0
+  %4 = bitcast i32* %3 to <8 x i32>*
+  %5 = load <8 x i32>, <8 x i32>* %4, align 2
+  %6 = zext <8 x i16> %2 to <8 x i32>
+  %7 = icmp ult <8 x i32> %5, %6
+  %8 = sub <8 x i32> %6, %5
+  %9 = trunc <8 x i32> %8 to <8 x i16>
+  %10 = select <8 x i1> %7, <8 x i16> %9, <8 x i16> zeroinitializer
+  store <8 x i16> %10, <8 x i16>* %1, align 1
+  ret void
+}