Index: test/CodeGen/X86/psubus.ll
===================================================================
--- test/CodeGen/X86/psubus.ll
+++ test/CodeGen/X86/psubus.ll
@@ -1044,3 +1044,547 @@
   %res = select <8 x i1> %cond, <8 x i16> %truncsub, <8 x i16> zeroinitializer
   ret <8 x i16> %res
 }
+
+; Test converting max(x,y) - y pattern to psubus, where x and y are 8 x i16
+define <8 x i16> @psubus_8i16_max(<8 x i16> %x, <8 x i16> %y) nounwind {
+; SSE2-LABEL: psubus_8i16_max:
+; SSE2:       # BB#0: # %vector.ph
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768]
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pxor %xmm2, %xmm3
+; SSE2-NEXT:    pxor %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm3
+; SSE2-NEXT:    pand %xmm2, %xmm3
+; SSE2-NEXT:    pandn %xmm0, %xmm2
+; SSE2-NEXT:    por %xmm3, %xmm2
+; SSE2-NEXT:    psubw %xmm1, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: psubus_8i16_max:
+; SSSE3:       # BB#0: # %vector.ph
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pxor %xmm2, %xmm3
+; SSSE3-NEXT:    pxor %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSSE3-NEXT:    movdqa %xmm1, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    pandn %xmm0, %xmm2
+; SSSE3-NEXT:    por %xmm3, %xmm2
+; SSSE3-NEXT:    psubw %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa %xmm2, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: psubus_8i16_max:
+; SSE41:       # BB#0: # %vector.ph
+; SSE41-NEXT:    pmaxuw %xmm1, %xmm0
+; SSE41-NEXT:    psubw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX-LABEL: psubus_8i16_max:
+; AVX:       # BB#0: # %vector.ph
+; AVX-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpsubw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+vector.ph:
+  %cmp = icmp ult <8 x i16> %x, %y
+  %max = select <8 x i1> %cmp, <8 x i16> %y, <8 x i16> %x
+  %res = sub <8 x i16> %max, %y
+  ret <8 x i16> %res
+}
+
+; Test converting max(x,y) - y pattern to psubus, where x and y are 16 x i8
+define <16 x i8> @psubus_16i8_max(<16 x i8> %x, <16 x i8> %y) nounwind {
+; SSE-LABEL: psubus_16i8_max:
+; SSE:       # BB#0: # %vector.ph
+; SSE-NEXT:    pmaxub %xmm1, %xmm0
+; SSE-NEXT:    psubb %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: psubus_16i8_max:
+; AVX:       # BB#0: # %vector.ph
+; AVX-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+vector.ph:
+  %cmp = icmp ult <16 x i8> %x, %y
+  %max = select <16 x i1> %cmp, <16 x i8> %y, <16 x i8> %x
+  %res = sub <16 x i8> %max, %y
+  ret <16 x i8> %res
+}
+
+; Test converting max(x,y) - y pattern to psubus, where x and y are 16 x i16
+; Tests don't convert to psubus on sse2, because ISD umin/umax SDNode is not
+; generated there.
+define <16 x i16> @psubus_16i16_max(<16 x i16> %x, <16 x i16> %y) nounwind {
+; SSE2-LABEL: psubus_16i16_max:
+; SSE2:       # BB#0: # %vector.ph
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [32768,32768,32768,32768,32768,32768,32768,32768]
+; SSE2-NEXT:    movdqa %xmm0, %xmm6
+; SSE2-NEXT:    pxor %xmm4, %xmm6
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    pxor %xmm4, %xmm5
+; SSE2-NEXT:    pcmpgtw %xmm6, %xmm5
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    pxor %xmm4, %xmm6
+; SSE2-NEXT:    pxor %xmm3, %xmm4
+; SSE2-NEXT:    pcmpgtw %xmm6, %xmm4
+; SSE2-NEXT:    movdqa %xmm3, %xmm6
+; SSE2-NEXT:    pand %xmm4, %xmm6
+; SSE2-NEXT:    pandn %xmm1, %xmm4
+; SSE2-NEXT:    por %xmm6, %xmm4
+; SSE2-NEXT:    movdqa %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pandn %xmm0, %xmm5
+; SSE2-NEXT:    por %xmm1, %xmm5
+; SSE2-NEXT:    psubw %xmm2, %xmm5
+; SSE2-NEXT:    psubw %xmm3, %xmm4
+; SSE2-NEXT:    movdqa %xmm5, %xmm0
+; SSE2-NEXT:    movdqa %xmm4, %xmm1
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: psubus_16i16_max:
+; SSSE3:       # BB#0: # %vector.ph
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [32768,32768,32768,32768,32768,32768,32768,32768]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm6
+; SSSE3-NEXT:    pxor %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa %xmm2, %xmm5
+; SSSE3-NEXT:    pxor %xmm4, %xmm5
+; SSSE3-NEXT:    pcmpgtw %xmm6, %xmm5
+; SSSE3-NEXT:    movdqa %xmm1, %xmm6
+; SSSE3-NEXT:    pxor %xmm4, %xmm6
+; SSSE3-NEXT:    pxor %xmm3, %xmm4
+; SSSE3-NEXT:    pcmpgtw %xmm6, %xmm4
+; SSSE3-NEXT:    movdqa %xmm3, %xmm6
+; SSSE3-NEXT:    pand %xmm4, %xmm6
+; SSSE3-NEXT:    pandn %xmm1, %xmm4
+; SSSE3-NEXT:    por %xmm6, %xmm4
+; SSSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSSE3-NEXT:    pand %xmm5, %xmm1
+; SSSE3-NEXT:    pandn %xmm0, %xmm5
+; SSSE3-NEXT:    por %xmm1, %xmm5
+; SSSE3-NEXT:    psubw %xmm2, %xmm5
+; SSSE3-NEXT:    psubw %xmm3, %xmm4
+; SSSE3-NEXT:    movdqa %xmm5, %xmm0
+; SSSE3-NEXT:    movdqa %xmm4, %xmm1
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: psubus_16i16_max:
+; SSE41:       # BB#0: # %vector.ph
+; SSE41-NEXT:    pmaxuw %xmm3, %xmm1
+; SSE41-NEXT:    pmaxuw %xmm2, %xmm0
+; SSE41-NEXT:    psubw %xmm2, %xmm0
+; SSE41-NEXT:    psubw %xmm3, %xmm1
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: psubus_16i16_max:
+; AVX1:       # BB#0: # %vector.ph
+; AVX1-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpmaxuw %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsubw %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsubw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: psubus_16i16_max:
+; AVX2:       # BB#0: # %vector.ph
+; AVX2-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpsubw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+vector.ph:
+  %cmp = icmp ult <16 x i16> %x, %y
+  %max = select <16 x i1> %cmp, <16 x i16> %y, <16 x i16> %x
+  %res = sub <16 x i16> %max, %y
+  ret <16 x i16> %res
+}
+
+; Test converting max(x,y) - y pattern to psubus, where x and y are 32 x i8
+define <32 x i8> @psubus_32i8_max(<32 x i8> %x, <32 x i8> %y) nounwind {
+; SSE-LABEL: psubus_32i8_max:
+; SSE:       # BB#0: # %vector.ph
+; SSE-NEXT:    pmaxub %xmm3, %xmm1
+; SSE-NEXT:    pmaxub %xmm2, %xmm0
+; SSE-NEXT:    psubb %xmm2, %xmm0
+; SSE-NEXT:    psubb %xmm3, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: psubus_32i8_max:
+; AVX1:       # BB#0: # %vector.ph
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsubb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: psubus_32i8_max:
+; AVX2:       # BB#0: # %vector.ph
+; AVX2-NEXT:    vpmaxub %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpsubb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+vector.ph:
+  %cmp = icmp ult <32 x i8> %x, %y
+  %max = select <32 x i1> %cmp, <32 x i8> %y, <32 x i8> %x
+  %res = sub <32 x i8> %max, %y
+  ret <32 x i8> %res
+}
+
+; Test converting max(lhs,y) - y pattern to psubus(lhs,min(y, 0xFFFF)), where y is i32 and lhs is zero extended i16
+define <8 x i16> @psubus_i16_i32_max(<8 x i16> %x, <8 x i32> %y) nounwind {
+; SSE2-LABEL: psubus_i16_i32_max:
+; SSE2:       # BB#0: # %vector.ph
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648]
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    pxor %xmm5, %xmm0
+; SSE2-NEXT:    movdqa %xmm3, %xmm6
+; SSE2-NEXT:    por %xmm5, %xmm6
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm0
+; SSE2-NEXT:    movdqa %xmm2, %xmm6
+; SSE2-NEXT:    pxor %xmm5, %xmm6
+; SSE2-NEXT:    por %xmm4, %xmm5
+; SSE2-NEXT:    pcmpgtd %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    pandn %xmm4, %xmm6
+; SSE2-NEXT:    por %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    pand %xmm0, %xmm4
+; SSE2-NEXT:    pandn %xmm3, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    psubd %xmm1, %xmm0
+; SSE2-NEXT:    psubd %xmm2, %xmm6
+; SSE2-NEXT:    pslld $16, %xmm6
+; SSE2-NEXT:    psrad $16, %xmm6
+; SSE2-NEXT:    pslld $16, %xmm0
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    packssdw %xmm6, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: psubus_i16_i32_max:
+; SSSE3:       # BB#0: # %vector.ph
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    pxor %xmm5, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm6
+; SSSE3-NEXT:    por %xmm5, %xmm6
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm0
+; SSSE3-NEXT:    movdqa %xmm2, %xmm6
+; SSSE3-NEXT:    pxor %xmm5, %xmm6
+; SSSE3-NEXT:    por %xmm4, %xmm5
+; SSSE3-NEXT:    pcmpgtd %xmm5, %xmm6
+; SSSE3-NEXT:    movdqa %xmm2, %xmm5
+; SSSE3-NEXT:    pand %xmm6, %xmm5
+; SSSE3-NEXT:    pandn %xmm4, %xmm6
+; SSSE3-NEXT:    por %xmm5, %xmm6
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pand %xmm0, %xmm4
+; SSSE3-NEXT:    pandn %xmm3, %xmm0
+; SSSE3-NEXT:    por %xmm4, %xmm0
+; SSSE3-NEXT:    psubd %xmm1, %xmm0
+; SSSE3-NEXT:    psubd %xmm2, %xmm6
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm1, %xmm6
+; SSSE3-NEXT:    pshufb %xmm1, %xmm0
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm6[0]
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: psubus_i16_i32_max:
+; SSE41:       # BB#0: # %vector.ph
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SSE41-NEXT:    pmaxud %xmm1, %xmm0
+; SSE41-NEXT:    pmaxud %xmm2, %xmm3
+; SSE41-NEXT:    psubd %xmm2, %xmm3
+; SSE41-NEXT:    psubd %xmm1, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm1, %xmm3
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: psubus_i16_i32_max:
+; AVX1:       # BB#0: # %vector.ph
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX1-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpmaxud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpsubd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: psubus_i16_i32_max:
+; AVX2:       # BB#0: # %vector.ph
+; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT:    vpmaxud %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpsubd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    # kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+vector.ph:
+  %lhs = zext <8 x i16> %x to <8 x i32>
+  %cond = icmp ult <8 x i32> %lhs, %y
+  %max = select <8 x i1> %cond, <8 x i32> %y, <8 x i32> %lhs
+  %sub = sub <8 x i32> %max, %y
+  %res = trunc <8 x i32> %sub to <8 x i16>
+  ret <8 x i16> %res
+}
+
+; Test converting trunc(max(y,lhs) - y) pattern to psubus(lhs,min(y, 0xFFFF)), where y is i32 and lhs is zero extended i16
+define <8 x i16> @psubus_i16_i32_max_swapped(<8 x i16> %x, <8 x i32> %y) nounwind {
+; SSE2-LABEL: psubus_i16_i32_max_swapped:
+; SSE2:       # BB#0: # %vector.ph
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    pxor %xmm4, %xmm6
+; SSE2-NEXT:    movdqa %xmm3, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm0
+; SSE2-NEXT:    movdqa %xmm2, %xmm6
+; SSE2-NEXT:    pxor %xmm4, %xmm6
+; SSE2-NEXT:    por %xmm5, %xmm4
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm4, %xmm5
+; SSE2-NEXT:    pandn %xmm2, %xmm4
+; SSE2-NEXT:    por %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm0, %xmm3
+; SSE2-NEXT:    pandn %xmm1, %xmm0
+; SSE2-NEXT:    por %xmm3, %xmm0
+; SSE2-NEXT:    psubd %xmm1, %xmm0
+; SSE2-NEXT:    psubd %xmm2, %xmm4
+; SSE2-NEXT:    pslld $16, %xmm4
+; SSE2-NEXT:    psrad $16, %xmm4
+; SSE2-NEXT:    pslld $16, %xmm0
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    packssdw %xmm4, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: psubus_i16_i32_max_swapped:
+; SSSE3:       # BB#0: # %vector.ph
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm5
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm6
+; SSSE3-NEXT:    pxor %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    por %xmm4, %xmm0
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm0
+; SSSE3-NEXT:    movdqa %xmm2, %xmm6
+; SSSE3-NEXT:    pxor %xmm4, %xmm6
+; SSSE3-NEXT:    por %xmm5, %xmm4
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm4
+; SSSE3-NEXT:    pand %xmm4, %xmm5
+; SSSE3-NEXT:    pandn %xmm2, %xmm4
+; SSSE3-NEXT:    por %xmm5, %xmm4
+; SSSE3-NEXT:    pand %xmm0, %xmm3
+; SSSE3-NEXT:    pandn %xmm1, %xmm0
+; SSSE3-NEXT:    por %xmm3, %xmm0
+; SSSE3-NEXT:    psubd %xmm1, %xmm0
+; SSSE3-NEXT:    psubd %xmm2, %xmm4
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm1, %xmm0
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: psubus_i16_i32_max_swapped:
+; SSE41:       # BB#0: # %vector.ph
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SSE41-NEXT:    pmaxud %xmm1, %xmm0
+; SSE41-NEXT:    pmaxud %xmm2, %xmm3
+; SSE41-NEXT:    psubd %xmm2, %xmm3
+; SSE41-NEXT:    psubd %xmm1, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm1, %xmm3
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: psubus_i16_i32_max_swapped:
+; AVX1:       # BB#0: # %vector.ph
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX1-NEXT:    vpmaxud %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpmaxud %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsubd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: psubus_i16_i32_max_swapped:
+; AVX2:       # BB#0: # %vector.ph
+; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT:    vpmaxud %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsubd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    # kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+vector.ph:
+  %lhs = zext <8 x i16> %x to <8 x i32>
+  %cond = icmp ult <8 x i32> %y, %lhs
+  %max = select <8 x i1> %cond, <8 x i32> %lhs, <8 x i32> %y
+  %sub = sub <8 x i32> %max, %y
+  %res = trunc <8 x i32> %sub to <8 x i16>
+  ret <8 x i16> %res
+}
+
+; Test converting trunc(lhs - min(lhs,y)) pattern to psubus(lhs,min(y, 0xFFFF)), where y is i32 and lhs is a zero extended i16
+define <8 x i16> @psubus_i16_i32_min(<8 x i16> %x, <8 x i32> %y) nounwind {
+; SSE2-LABEL: psubus_i16_i32_min:
+; SSE2:       # BB#0: # %vector.ph
+; SSE2-NEXT:    pxor %xmm4, %xmm4
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
+; SSE2-NEXT:    pxor %xmm4, %xmm5
+; SSE2-NEXT:    movdqa %xmm0, %xmm6
+; SSE2-NEXT:    por %xmm4, %xmm6
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm5
+; SSE2-NEXT:    movdqa %xmm2, %xmm6
+; SSE2-NEXT:    pxor %xmm4, %xmm6
+; SSE2-NEXT:    por %xmm3, %xmm4
+; SSE2-NEXT:    pcmpgtd %xmm4, %xmm6
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pandn %xmm2, %xmm6
+; SSE2-NEXT:    por %xmm4, %xmm6
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm5, %xmm2
+; SSE2-NEXT:    pandn %xmm1, %xmm5
+; SSE2-NEXT:    por %xmm2, %xmm5
+; SSE2-NEXT:    psubd %xmm5, %xmm0
+; SSE2-NEXT:    psubd %xmm6, %xmm3
+; SSE2-NEXT:    pslld $16, %xmm3
+; SSE2-NEXT:    psrad $16, %xmm3
+; SSE2-NEXT:    pslld $16, %xmm0
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    packssdw %xmm3, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: psubus_i16_i32_min:
+; SSSE3:       # BB#0: # %vector.ph
+; SSSE3-NEXT:    pxor %xmm4, %xmm4
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm5
+; SSSE3-NEXT:    pxor %xmm4, %xmm5
+; SSSE3-NEXT:    movdqa %xmm0, %xmm6
+; SSSE3-NEXT:    por %xmm4, %xmm6
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm5
+; SSSE3-NEXT:    movdqa %xmm2, %xmm6
+; SSSE3-NEXT:    pxor %xmm4, %xmm6
+; SSSE3-NEXT:    por %xmm3, %xmm4
+; SSSE3-NEXT:    pcmpgtd %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pand %xmm6, %xmm4
+; SSSE3-NEXT:    pandn %xmm2, %xmm6
+; SSSE3-NEXT:    por %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm5, %xmm2
+; SSSE3-NEXT:    pandn %xmm1, %xmm5
+; SSSE3-NEXT:    por %xmm2, %xmm5
+; SSSE3-NEXT:    psubd %xmm5, %xmm0
+; SSSE3-NEXT:    psubd %xmm6, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm1, %xmm3
+; SSSE3-NEXT:    pshufb %xmm1, %xmm0
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: psubus_i16_i32_min:
+; SSE41:       # BB#0: # %vector.ph
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,0,1]
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SSE41-NEXT:    pminud %xmm0, %xmm1
+; SSE41-NEXT:    pminud %xmm3, %xmm2
+; SSE41-NEXT:    psubd %xmm2, %xmm3
+; SSE41-NEXT:    psubd %xmm1, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm1, %xmm3
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: psubus_i16_i32_min:
+; AVX1:       # BB#0: # %vector.ph
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vpminud %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpsubd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpsubd %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: psubus_i16_i32_min:
+; AVX2:       # BB#0: # %vector.ph
+; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpsubd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    # kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+vector.ph:
+  %lhs = zext <8 x i16> %x to <8 x i32>
+  %cond = icmp ult <8 x i32> %lhs, %y
+  %min = select <8 x i1> %cond, <8 x i32> %lhs, <8 x i32> %y
+  %sub = sub <8 x i32> %lhs, %min
+  %res = trunc <8 x i32> %sub to <8 x i16>
+  ret <8 x i16> %res
+}