Index: test/CodeGen/X86/vector-mul.ll
===================================================================
--- test/CodeGen/X86/vector-mul.ll
+++ test/CodeGen/X86/vector-mul.ll
@@ -245,24 +245,34 @@
 define <2 x i64> @mul_v2i64_17(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_17:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $4, %xmm1
-; X86-NEXT:    paddq %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [17,0,17,0]
+; X86-NEXT:    movdqa %xmm0, %xmm2
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_17:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $4, %xmm1
-; X64-NEXT:    paddq %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [17,17]
+; X64-NEXT:    movdqa %xmm0, %xmm2
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_v2i64_17:
 ; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpsllq $4, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
 ; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 17, i64 17>
   ret <2 x i64> %1
@@ -271,25 +281,24 @@
 define <4 x i32> @mul_v4i32_17(<4 x i32> %a0) nounwind {
 ; X86-LABEL: mul_v4i32_17:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    pslld $4, %xmm1
-; X86-NEXT:    paddd %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmulld {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v4i32_17:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    pslld $4, %xmm1
-; X64-NEXT:    paddd %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmulld {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-AVX-LABEL: mul_v4i32_17:
-; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpslld $4, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
-; X64-AVX-NEXT:    retq
+; X64-XOP-LABEL: mul_v4i32_17:
+; X64-XOP:       # BB#0:
+; X64-XOP-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; X64-XOP-NEXT:    retq
+;
+; X64-AVX2-LABEL: mul_v4i32_17:
+; X64-AVX2:       # BB#0:
+; X64-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; X64-AVX2-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
+; X64-AVX2-NEXT:    retq
   %1 = mul <4 x i32> %a0, <i32 17, i32 17, i32 17, i32 17>
   ret <4 x i32> %1
 }
@@ -297,24 +306,17 @@
 define <8 x i16> @mul_v8i16_17(<8 x i16> %a0) nounwind {
 ; X86-LABEL: mul_v8i16_17:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllw $4, %xmm1
-; X86-NEXT:    paddw %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmullw {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v8i16_17:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllw $4, %xmm1
-; X64-NEXT:    paddw %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_v8i16_17:
 ; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpsllw $4, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
+; X64-AVX-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; X64-AVX-NEXT:    retq
   %1 = mul <8 x i16> %a0, <i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17>
   ret <8 x i16> %1
@@ -323,33 +325,58 @@
 define <16 x i8> @mul_v16i8_17(<16 x i8> %a0) nounwind {
 ; X86-LABEL: mul_v16i8_17:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllw $4, %xmm1
-; X86-NEXT:    pand {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    paddb %xmm0, %xmm1
+; X86-NEXT:    pmovsxbw %xmm0, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm2 = [17,17,17,17,17,17,17,17]
+; X86-NEXT:    pmullw %xmm2, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X86-NEXT:    pand %xmm3, %xmm1
+; X86-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X86-NEXT:    pmovsxbw %xmm0, %xmm0
+; X86-NEXT:    pmullw %xmm2, %xmm0
+; X86-NEXT:    pand %xmm3, %xmm0
+; X86-NEXT:    packuswb %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v16i8_17:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllw $4, %xmm1
-; X64-NEXT:    pand {{.*}}(%rip), %xmm1
-; X64-NEXT:    paddb %xmm0, %xmm1
+; X64-NEXT:    pmovsxbw %xmm0, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm2 = [17,17,17,17,17,17,17,17]
+; X64-NEXT:    pmullw %xmm2, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X64-NEXT:    pand %xmm3, %xmm1
+; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-NEXT:    pmovsxbw %xmm0, %xmm0
+; X64-NEXT:    pmullw %xmm2, %xmm0
+; X64-NEXT:    pand %xmm3, %xmm0
+; X64-NEXT:    packuswb %xmm0, %xmm1
 ; X64-NEXT:    movdqa %xmm1, %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-XOP-LABEL: mul_v16i8_17:
 ; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlb {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; X64-XOP-NEXT:    vpmovsxbw %xmm0, %xmm1
+; X64-XOP-NEXT:    vmovdqa {{.*#+}} xmm2 = [17,17,17,17,17,17,17,17]
+; X64-XOP-NEXT:    vpmullw %xmm2, %xmm1, %xmm1
+; X64-XOP-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X64-XOP-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; X64-XOP-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-XOP-NEXT:    vpmovsxbw %xmm0, %xmm0
+; X64-XOP-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
+; X64-XOP-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; X64-XOP-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; X64-XOP-NEXT:    retq
 ;
 ; X64-AVX2-LABEL: mul_v16i8_17:
 ; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllw $4, %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
-; X64-AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; X64-AVX2-NEXT:    vpmovsxbw %xmm0, %ymm0
+; X64-AVX2-NEXT:    vpmullw {{.*}}(%rip), %ymm0, %ymm0
+; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; X64-AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; X64-AVX2-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; X64-AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; X64-AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
   %1 = mul <16 x i8> %a0, <i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17, i8 17>
   ret <16 x i8> %1
@@ -362,37 +389,35 @@
 define <2 x i64> @mul_v2i64_17_65(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_17_65:
 ; X86:       # BB#0:
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [17,0,65,0]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $6, %xmm2
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $4, %xmm1
-; X86-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X86-NEXT:    paddq %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_17_65:
 ; X64:       # BB#0:
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [17,65]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $6, %xmm2
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $4, %xmm1
-; X64-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X64-NEXT:    paddq %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i64_17_65:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i64_17_65:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i64_17_65:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,65]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 17, i64 65>
   ret <2 x i64> %1
 }
@@ -400,29 +425,18 @@
 define <4 x i32> @mul_v4i32_5_17_33_65(<4 x i32> %a0) nounwind {
 ; X86-LABEL: mul_v4i32_5_17_33_65:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa {{.*#+}} xmm1 = [4,16,32,64]
-; X86-NEXT:    pmulld %xmm0, %xmm1
-; X86-NEXT:    paddd %xmm1, %xmm0
+; X86-NEXT:    pmulld {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v4i32_5_17_33_65:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa {{.*#+}} xmm1 = [4,16,32,64]
-; X64-NEXT:    pmulld %xmm0, %xmm1
-; X64-NEXT:    paddd %xmm1, %xmm0
+; X64-NEXT:    pmulld {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v4i32_5_17_33_65:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshld {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v4i32_5_17_33_65:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvd {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v4i32_5_17_33_65:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; X64-AVX-NEXT:    retq
   %1 = mul <4 x i32> %a0, <i32 5, i32 17, i32 33, i32 65>
   ret <4 x i32> %1
 }
@@ -510,24 +524,34 @@
 define <2 x i64> @mul_v2i64_7(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_7:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $3, %xmm1
-; X86-NEXT:    psubq %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [7,0,7,0]
+; X86-NEXT:    movdqa %xmm0, %xmm2
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_7:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $3, %xmm1
-; X64-NEXT:    psubq %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [7,7]
+; X64-NEXT:    movdqa %xmm0, %xmm2
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_v2i64_7:
 ; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpsllq $3, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
 ; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 7, i64 7>
   ret <2 x i64> %1
@@ -536,25 +560,24 @@
 define <4 x i32> @mul_v4i32_7(<4 x i32> %a0) nounwind {
 ; X86-LABEL: mul_v4i32_7:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    pslld $3, %xmm1
-; X86-NEXT:    psubd %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmulld {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v4i32_7:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    pslld $3, %xmm1
-; X64-NEXT:    psubd %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmulld {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-AVX-LABEL: mul_v4i32_7:
-; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpslld $3, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
-; X64-AVX-NEXT:    retq
+; X64-XOP-LABEL: mul_v4i32_7:
+; X64-XOP:       # BB#0:
+; X64-XOP-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; X64-XOP-NEXT:    retq
+;
+; X64-AVX2-LABEL: mul_v4i32_7:
+; X64-AVX2:       # BB#0:
+; X64-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; X64-AVX2-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
+; X64-AVX2-NEXT:    retq
   %1 = mul <4 x i32> %a0, <i32 7, i32 7, i32 7, i32 7>
   ret <4 x i32> %1
 }
@@ -562,24 +585,17 @@
 define <8 x i16> @mul_v8i16_7(<8 x i16> %a0) nounwind {
 ; X86-LABEL: mul_v8i16_7:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllw $3, %xmm1
-; X86-NEXT:    psubw %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmullw {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v8i16_7:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllw $3, %xmm1
-; X64-NEXT:    psubw %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_v8i16_7:
 ; X64-AVX:       # BB#0:
-; X64-AVX-NEXT:    vpsllw $3, %xmm0, %xmm1
-; X64-AVX-NEXT:    vpsubw %xmm0, %xmm1, %xmm0
+; X64-AVX-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; X64-AVX-NEXT:    retq
   %1 = mul <8 x i16> %a0, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
   ret <8 x i16> %1
@@ -588,33 +604,58 @@
 define <16 x i8> @mul_v16i8_31(<16 x i8> %a0) nounwind {
 ; X86-LABEL: mul_v16i8_31:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllw $5, %xmm1
-; X86-NEXT:    pand {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    psubb %xmm0, %xmm1
+; X86-NEXT:    pmovsxbw %xmm0, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm2 = [31,31,31,31,31,31,31,31]
+; X86-NEXT:    pmullw %xmm2, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X86-NEXT:    pand %xmm3, %xmm1
+; X86-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X86-NEXT:    pmovsxbw %xmm0, %xmm0
+; X86-NEXT:    pmullw %xmm2, %xmm0
+; X86-NEXT:    pand %xmm3, %xmm0
+; X86-NEXT:    packuswb %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v16i8_31:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllw $5, %xmm1
-; X64-NEXT:    pand {{.*}}(%rip), %xmm1
-; X64-NEXT:    psubb %xmm0, %xmm1
+; X64-NEXT:    pmovsxbw %xmm0, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm2 = [31,31,31,31,31,31,31,31]
+; X64-NEXT:    pmullw %xmm2, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X64-NEXT:    pand %xmm3, %xmm1
+; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-NEXT:    pmovsxbw %xmm0, %xmm0
+; X64-NEXT:    pmullw %xmm2, %xmm0
+; X64-NEXT:    pand %xmm3, %xmm0
+; X64-NEXT:    packuswb %xmm0, %xmm1
 ; X64-NEXT:    movdqa %xmm1, %xmm0
 ; X64-NEXT:    retq
 ;
 ; X64-XOP-LABEL: mul_v16i8_31:
 ; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlb {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpsubb %xmm0, %xmm1, %xmm0
+; X64-XOP-NEXT:    vpmovsxbw %xmm0, %xmm1
+; X64-XOP-NEXT:    vmovdqa {{.*#+}} xmm2 = [31,31,31,31,31,31,31,31]
+; X64-XOP-NEXT:    vpmullw %xmm2, %xmm1, %xmm1
+; X64-XOP-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; X64-XOP-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; X64-XOP-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; X64-XOP-NEXT:    vpmovsxbw %xmm0, %xmm0
+; X64-XOP-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
+; X64-XOP-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; X64-XOP-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; X64-XOP-NEXT:    retq
 ;
 ; X64-AVX2-LABEL: mul_v16i8_31:
 ; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllw $5, %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
-; X64-AVX2-NEXT:    vpsubb %xmm0, %xmm1, %xmm0
+; X64-AVX2-NEXT:    vpmovsxbw %xmm0, %ymm0
+; X64-AVX2-NEXT:    vpmullw {{.*}}(%rip), %ymm0, %ymm0
+; X64-AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; X64-AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; X64-AVX2-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; X64-AVX2-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; X64-AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
   %1 = mul <16 x i8> %a0, <i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31, i8 31>
   ret <16 x i8> %1
@@ -627,37 +668,35 @@
 define <2 x i64> @mul_v2i64_15_63(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_15_63:
 ; X86:       # BB#0:
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [15,0,63,0]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $6, %xmm2
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $4, %xmm1
-; X86-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X86-NEXT:    psubq %xmm0, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_15_63:
 ; X64:       # BB#0:
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [15,63]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $6, %xmm2
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $4, %xmm1
-; X64-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X64-NEXT:    psubq %xmm0, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i64_15_63:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i64_15_63:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i64_15_63:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,63]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 15, i64 63>
   ret <2 x i64> %1
 }
@@ -665,35 +704,43 @@
 define <2 x i64> @mul_v2i64_neg_15_63(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_neg_15_63:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $6, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [4294967281,4294967295,4294967233,4294967295]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $4, %xmm2
-; X86-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; X86-NEXT:    psubq %xmm2, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    movdqa %xmm0, %xmm3
+; X86-NEXT:    psrlq $32, %xmm3
+; X86-NEXT:    pmuludq %xmm1, %xmm3
+; X86-NEXT:    pmuludq {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    paddq %xmm3, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_neg_15_63:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $6, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [18446744073709551601,18446744073709551553]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $4, %xmm2
-; X64-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; X64-NEXT:    psubq %xmm2, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    movdqa %xmm0, %xmm3
+; X64-NEXT:    psrlq $32, %xmm3
+; X64-NEXT:    pmuludq %xmm1, %xmm3
+; X64-NEXT:    pmuludq {{.*}}(%rip), %xmm0
+; X64-NEXT:    paddq %xmm3, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i64_neg_15_63:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpsubq %xmm1, %xmm0, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i64_neg_15_63:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpsubq %xmm1, %xmm0, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i64_neg_15_63:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709551601,18446744073709551553]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm3
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
+; X64-AVX-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 -15, i64 -63>
   ret <2 x i64> %1
 }
@@ -701,43 +748,43 @@
 define <2 x i64> @mul_v2i64_neg_17_65(<2 x i64> %a0) nounwind {
 ; X86-LABEL: mul_v2i64_neg_17_65:
 ; X86:       # BB#0:
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $6, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [4294967279,4294967295,4294967231,4294967295]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $4, %xmm2
-; X86-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; X86-NEXT:    paddq %xmm0, %xmm2
-; X86-NEXT:    pxor %xmm0, %xmm0
-; X86-NEXT:    psubq %xmm2, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    movdqa %xmm0, %xmm3
+; X86-NEXT:    psrlq $32, %xmm3
+; X86-NEXT:    pmuludq %xmm1, %xmm3
+; X86-NEXT:    pmuludq {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    paddq %xmm3, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_neg_17_65:
 ; X64:       # BB#0:
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $6, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [18446744073709551599,18446744073709551551]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $4, %xmm2
-; X64-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; X64-NEXT:    paddq %xmm0, %xmm2
-; X64-NEXT:    pxor %xmm0, %xmm0
-; X64-NEXT:    psubq %xmm2, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    movdqa %xmm0, %xmm3
+; X64-NEXT:    psrlq $32, %xmm3
+; X64-NEXT:    pmuludq %xmm1, %xmm3
+; X64-NEXT:    pmuludq {{.*}}(%rip), %xmm0
+; X64-NEXT:    paddq %xmm3, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i64_neg_17_65:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
-; X64-XOP-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; X64-XOP-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i64_neg_17_65:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
-; X64-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; X64-AVX2-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i64_neg_17_65:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [18446744073709551599,18446744073709551551]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm3
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
+; X64-AVX-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %1 = mul <2 x i64> %a0, <i64 -17, i64 -65>
   ret <2 x i64> %1
 }
@@ -987,41 +1034,35 @@
 define <2 x i64> @mul_v2i64_68_132(<2 x i64> %x) nounwind{
 ; X86-LABEL: mul_v2i64_68_132:
 ; X86:       # BB#0:
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [68,0,132,0]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $5, %xmm2
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $4, %xmm1
-; X86-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X86-NEXT:    paddq %xmm0, %xmm1
-; X86-NEXT:    psllq $2, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i64_68_132:
 ; X64:       # BB#0:
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [68,132]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $5, %xmm2
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $4, %xmm1
-; X64-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X64-NEXT:    paddq %xmm0, %xmm1
-; X64-NEXT:    psllq $2, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i64_68_132:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    vpsllq $2, %xmm0, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i64_68_132:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpsllq $2, %xmm0, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i64_68_132:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [68,132]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %mul = mul <2 x i64> %x, <i64 68, i64 132>
   ret <2 x i64> %mul
 }
@@ -1030,41 +1071,35 @@
 define <2 x i64> @mul_v2i16_60_120(<2 x i64> %x) nounwind {
 ; X86-LABEL: mul_v2i16_60_120:
 ; X86:       # BB#0:
+; X86-NEXT:    movdqa {{.*#+}} xmm1 = [60,0,124,0]
 ; X86-NEXT:    movdqa %xmm0, %xmm2
-; X86-NEXT:    psllq $5, %xmm2
-; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllq $4, %xmm1
-; X86-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X86-NEXT:    psubq %xmm0, %xmm1
-; X86-NEXT:    psllq $2, %xmm1
-; X86-NEXT:    movdqa %xmm1, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm2
+; X86-NEXT:    psrlq $32, %xmm0
+; X86-NEXT:    pmuludq %xmm1, %xmm0
+; X86-NEXT:    psllq $32, %xmm0
+; X86-NEXT:    paddq %xmm2, %xmm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mul_v2i16_60_120:
 ; X64:       # BB#0:
+; X64-NEXT:    movdqa {{.*#+}} xmm1 = [60,124]
 ; X64-NEXT:    movdqa %xmm0, %xmm2
-; X64-NEXT:    psllq $5, %xmm2
-; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllq $4, %xmm1
-; X64-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
-; X64-NEXT:    psubq %xmm0, %xmm1
-; X64-NEXT:    psllq $2, %xmm1
-; X64-NEXT:    movdqa %xmm1, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm2
+; X64-NEXT:    psrlq $32, %xmm0
+; X64-NEXT:    pmuludq %xmm1, %xmm0
+; X64-NEXT:    psllq $32, %xmm0
+; X64-NEXT:    paddq %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;
-; X64-XOP-LABEL: mul_v2i16_60_120:
-; X64-XOP:       # BB#0:
-; X64-XOP-NEXT:    vpshlq {{.*}}(%rip), %xmm0, %xmm1
-; X64-XOP-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-XOP-NEXT:    vpsllq $2, %xmm0, %xmm0
-; X64-XOP-NEXT:    retq
-;
-; X64-AVX2-LABEL: mul_v2i16_60_120:
-; X64-AVX2:       # BB#0:
-; X64-AVX2-NEXT:    vpsllvq {{.*}}(%rip), %xmm0, %xmm1
-; X64-AVX2-NEXT:    vpsubq %xmm0, %xmm1, %xmm0
-; X64-AVX2-NEXT:    vpsllq $2, %xmm0, %xmm0
-; X64-AVX2-NEXT:    retq
+; X64-AVX-LABEL: mul_v2i16_60_120:
+; X64-AVX:       # BB#0:
+; X64-AVX-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,124]
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; X64-AVX-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X64-AVX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; X64-AVX-NEXT:    retq
   %mul = mul <2 x i64> %x, <i64 60, i64 124>
   ret <2 x i64> %mul
 }