X86 Tests: Add Tests for PMADDWD selection. NFC.

Support for ISel to be added.

llvm-svn: 321970

1 files changed, 373 insertions, 99 deletions

diff --git a/llvm/test/CodeGen/X86/madd.ll b/llvm/test/CodeGen/X86/madd.ll
index 44e7b91eef8..2e9237a0ebf 100644
--- a/llvm/test/CodeGen/X86/madd.ll
+++ b/llvm/test/CodeGen/X86/madd.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefix=AVX512
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=AVX,AVX512,AVX512F
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefixes=AVX,AVX512,AVX512BW
 
 define i32 @_Z10test_shortPsS_i(i16* nocapture readonly, i16* nocapture readonly, i32) local_unnamed_addr #0 {
 ; SSE2-LABEL: _Z10test_shortPsS_i:
@@ -30,53 +30,29 @@ define i32 @_Z10test_shortPsS_i(i16* nocapture readonly, i16* nocapture readonly
 ; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    retq
 ;
-; AVX2-LABEL: _Z10test_shortPsS_i:
-; AVX2:       # %bb.0: # %entry
-; AVX2-NEXT:    movl %edx, %eax
-; AVX2-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT:    xorl %ecx, %ecx
-; AVX2-NEXT:    .p2align 4, 0x90
-; AVX2-NEXT:  .LBB0_1: # %vector.body
-; AVX2-NEXT:    # =>This Inner Loop Header: Depth=1
-; AVX2-NEXT:    vmovdqu (%rsi,%rcx,2), %xmm1
-; AVX2-NEXT:    vpmaddwd (%rdi,%rcx,2), %xmm1, %xmm1
-; AVX2-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    addq $8, %rcx
-; AVX2-NEXT:    cmpq %rcx, %rax
-; AVX2-NEXT:    jne .LBB0_1
-; AVX2-NEXT:  # %bb.2: # %middle.block
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512-LABEL: _Z10test_shortPsS_i:
-; AVX512:       # %bb.0: # %entry
-; AVX512-NEXT:    movl %edx, %eax
-; AVX512-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; AVX512-NEXT:    xorl %ecx, %ecx
-; AVX512-NEXT:    .p2align 4, 0x90
-; AVX512-NEXT:  .LBB0_1: # %vector.body
-; AVX512-NEXT:    # =>This Inner Loop Header: Depth=1
-; AVX512-NEXT:    vmovdqu (%rsi,%rcx,2), %xmm1
-; AVX512-NEXT:    vpmaddwd (%rdi,%rcx,2), %xmm1, %xmm1
-; AVX512-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
-; AVX512-NEXT:    addq $8, %rcx
-; AVX512-NEXT:    cmpq %rcx, %rax
-; AVX512-NEXT:    jne .LBB0_1
-; AVX512-NEXT:  # %bb.2: # %middle.block
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX512-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
-; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    vzeroupper
-; AVX512-NEXT:    retq
+; AVX-LABEL: _Z10test_shortPsS_i:
+; AVX:       # %bb.0: # %entry
+; AVX-NEXT:    movl %edx, %eax
+; AVX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    .p2align 4, 0x90
+; AVX-NEXT:  .LBB0_1: # %vector.body
+; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
+; AVX-NEXT:    vmovdqu (%rsi,%rcx,2), %xmm1
+; AVX-NEXT:    vpmaddwd (%rdi,%rcx,2), %xmm1, %xmm1
+; AVX-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    addq $8, %rcx
+; AVX-NEXT:    cmpq %rcx, %rax
+; AVX-NEXT:    jne .LBB0_1
+; AVX-NEXT:  # %bb.2: # %middle.block
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
 entry:
   %3 = zext i32 %2 to i64
   br label %vector.body
@@ -141,55 +117,30 @@ define i32 @test_unsigned_short(i16* nocapture readonly, i16* nocapture readonly
 ; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    retq
 ;
-; AVX2-LABEL: test_unsigned_short:
-; AVX2:       # %bb.0: # %entry
-; AVX2-NEXT:    movl %edx, %eax
-; AVX2-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT:    xorl %ecx, %ecx
-; AVX2-NEXT:    .p2align 4, 0x90
-; AVX2-NEXT:  .LBB1_1: # %vector.body
-; AVX2-NEXT:    # =>This Inner Loop Header: Depth=1
-; 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:    vpmovzxwd {{.*#+}} ymm2 = 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:    vpmulld %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    addq $8, %rcx
-; AVX2-NEXT:    cmpq %rcx, %rax
-; AVX2-NEXT:    jne .LBB1_1
-; AVX2-NEXT:  # %bb.2: # %middle.block
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512-LABEL: test_unsigned_short:
-; AVX512:       # %bb.0: # %entry
-; AVX512-NEXT:    movl %edx, %eax
-; AVX512-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; AVX512-NEXT:    xorl %ecx, %ecx
-; AVX512-NEXT:    .p2align 4, 0x90
-; AVX512-NEXT:  .LBB1_1: # %vector.body
-; AVX512-NEXT:    # =>This Inner Loop Header: Depth=1
-; AVX512-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
-; AVX512-NEXT:    vpmovzxwd {{.*#+}} ymm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
-; AVX512-NEXT:    vpmulld %ymm1, %ymm2, %ymm1
-; AVX512-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
-; AVX512-NEXT:    addq $8, %rcx
-; AVX512-NEXT:    cmpq %rcx, %rax
-; AVX512-NEXT:    jne .LBB1_1
-; AVX512-NEXT:  # %bb.2: # %middle.block
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; AVX512-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
-; AVX512-NEXT:    vmovd %xmm0, %eax
-; AVX512-NEXT:    vzeroupper
-; AVX512-NEXT:    retq
+; AVX-LABEL: test_unsigned_short:
+; AVX:       # %bb.0: # %entry
+; AVX-NEXT:    movl %edx, %eax
+; AVX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    xorl %ecx, %ecx
+; AVX-NEXT:    .p2align 4, 0x90
+; AVX-NEXT:  .LBB1_1: # %vector.body
+; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
+; AVX-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
+; AVX-NEXT:    vpmovzxwd {{.*#+}} ymm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
+; AVX-NEXT:    vpmulld %ymm1, %ymm2, %ymm1
+; AVX-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
+; AVX-NEXT:    addq $8, %rcx
+; AVX-NEXT:    cmpq %rcx, %rax
+; AVX-NEXT:    jne .LBB1_1
+; AVX-NEXT:  # %bb.2: # %middle.block
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vphaddd %ymm0, %ymm0, %ymm0
+; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
 entry:
   %3 = zext i32 %2 to i64
   br label %vector.body
@@ -361,3 +312,326 @@ middle.block:
   %13 = extractelement <16 x i32> %bin.rdx20, i32 0
   ret i32 %13
 }
+
+define <4 x i32> @pmaddwd_8(<8 x i16> %A, <8 x i16> %B) {
+; SSE2-LABEL: pmaddwd_8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pmulhw %xmm1, %xmm2
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE2-NEXT:    paddd %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX-LABEL: pmaddwd_8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX-NEXT:    vpmovsxwd %xmm1, %ymm1
+; AVX-NEXT:    vpmulld %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
+   %a = sext <8 x i16> %A to <8 x i32>
+   %b = sext <8 x i16> %B to <8 x i32>
+   %m = mul nsw <8 x i32> %a, %b
+   %odd = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+   %even = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+   %ret = add <4 x i32> %odd, %even
+   ret <4 x i32> %ret
+}
+
+define <8 x i32> @pmaddwd_16(<16 x i16> %A, <16 x i16> %B) {
+; SSE2-LABEL: pmaddwd_16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm4
+; SSE2-NEXT:    pmulhw %xmm2, %xmm4
+; SSE2-NEXT:    pmullw %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[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 %xmm1, %xmm4
+; SSE2-NEXT:    pmulhw %xmm3, %xmm4
+; SSE2-NEXT:    pmullw %xmm3, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm3
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm3[0,2]
+; SSE2-NEXT:    movdqa %xmm0, %xmm5
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm2[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm3[1,3]
+; SSE2-NEXT:    paddd %xmm4, %xmm1
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
+; SSE2-NEXT:    paddd %xmm5, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX2-LABEL: pmaddwd_16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm2
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm3
+; AVX2-NEXT:    vpmulld %ymm3, %ymm2, %ymm2
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm1
+; AVX2-NEXT:    vpmulld %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2],ymm0[0,2],ymm2[4,6],ymm0[4,6]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,1,3]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[1,3],ymm0[1,3],ymm2[5,7],ymm0[5,7]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; AVX2-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: pmaddwd_16:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpmovsxwd %ymm0, %zmm0
+; AVX512-NEXT:    vpmovsxwd %ymm1, %zmm1
+; AVX512-NEXT:    vpmulld %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[0,2],ymm1[0,2],ymm0[4,6],ymm1[4,6]
+; AVX512-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
+; AVX512-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm1[1,3],ymm0[5,7],ymm1[5,7]
+; AVX512-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; AVX512-NEXT:    vpaddd %ymm0, %ymm2, %ymm0
+; AVX512-NEXT:    retq
+   %a = sext <16 x i16> %A to <16 x i32>
+   %b = sext <16 x i16> %B to <16 x i32>
+   %m = mul nsw <16 x i32> %a, %b
+   %odd = shufflevector <16 x i32> %m, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+   %even = shufflevector <16 x i32> %m, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+   %ret = add <8 x i32> %odd, %even
+   ret <8 x i32> %ret
+}
+
+define <16 x i32> @pmaddwd_32(<32 x i16> %A, <32 x i16> %B) {
+; SSE2-LABEL: pmaddwd_32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm8
+; SSE2-NEXT:    pmulhw %xmm4, %xmm8
+; SSE2-NEXT:    pmullw %xmm4, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm9
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    pmulhw %xmm5, %xmm4
+; SSE2-NEXT:    pmullw %xmm5, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm8
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    pmulhw %xmm6, %xmm4
+; SSE2-NEXT:    pmullw %xmm6, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm6
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    pmulhw %xmm7, %xmm4
+; SSE2-NEXT:    pmullw %xmm7, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm7
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm7[0,2]
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm6[0,2]
+; SSE2-NEXT:    movdqa %xmm1, %xmm10
+; SSE2-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm8[0,2]
+; SSE2-NEXT:    movdqa %xmm0, %xmm11
+; SSE2-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm9[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,3],xmm7[1,3]
+; SSE2-NEXT:    paddd %xmm4, %xmm3
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,3],xmm6[1,3]
+; SSE2-NEXT:    paddd %xmm5, %xmm2
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm8[1,3]
+; SSE2-NEXT:    paddd %xmm10, %xmm1
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm9[1,3]
+; SSE2-NEXT:    paddd %xmm11, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX2-LABEL: pmaddwd_32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm4
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm1
+; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm5
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX2-NEXT:    vpmovsxwd %xmm3, %ymm6
+; AVX2-NEXT:    vpmulld %ymm6, %ymm4, %ymm4
+; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm3
+; AVX2-NEXT:    vpmovsxwd %xmm3, %ymm3
+; AVX2-NEXT:    vpmulld %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm3
+; AVX2-NEXT:    vpmulld %ymm3, %ymm5, %ymm3
+; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm2
+; AVX2-NEXT:    vpmulld %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[0,2],ymm1[0,2],ymm4[4,6],ymm1[4,6]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,1,3]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[0,2],ymm0[0,2],ymm3[4,6],ymm0[4,6]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,1,3]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[1,3],ymm1[1,3],ymm4[5,7],ymm1[5,7]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,1,3]
+; AVX2-NEXT:    vpaddd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm3[1,3],ymm0[1,3],ymm3[5,7],ymm0[5,7]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; AVX2-NEXT:    vpaddd %ymm0, %ymm5, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: pmaddwd_32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm0
+; AVX512F-NEXT:    vpmovsxwd %ymm1, %zmm1
+; AVX512F-NEXT:    vpmovsxwd %ymm2, %zmm2
+; AVX512F-NEXT:    vpmulld %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT:    vpmovsxwd %ymm3, %zmm2
+; AVX512F-NEXT:    vpmulld %zmm2, %zmm1, %zmm1
+; AVX512F-NEXT:    vmovdqa32 {{.*#+}} zmm2 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30]
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
+; AVX512F-NEXT:    vmovdqa32 {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31]
+; AVX512F-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512F-NEXT:    vpaddd %zmm3, %zmm2, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: pmaddwd_32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpmovsxwd %ymm0, %zmm2
+; AVX512BW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512BW-NEXT:    vpmovsxwd %ymm0, %zmm0
+; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm3
+; AVX512BW-NEXT:    vpmulld %zmm3, %zmm2, %zmm2
+; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
+; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm1
+; AVX512BW-NEXT:    vpmulld %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vmovdqa32 {{.*#+}} zmm1 = [0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30]
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm1
+; AVX512BW-NEXT:    vmovdqa32 {{.*#+}} zmm3 = [1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31]
+; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm2, %zmm3
+; AVX512BW-NEXT:    vpaddd %zmm3, %zmm1, %zmm0
+; AVX512BW-NEXT:    retq
+   %a = sext <32 x i16> %A to <32 x i32>
+   %b = sext <32 x i16> %B to <32 x i32>
+   %m = mul nsw <32 x i32> %a, %b
+   %odd = shufflevector <32 x i32> %m, <32 x i32> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+   %even = shufflevector <32 x i32> %m, <32 x i32> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+   %ret = add <16 x i32> %odd, %even
+   ret <16 x i32> %ret
+}
+
+define <4 x i32> @pmaddwd_const(<8 x i16> %A) {
+; SSE2-LABEL: pmaddwd_const:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [32767,32768,0,0,1,7,42,32]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pmulhw %xmm1, %xmm2
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE2-NEXT:    paddd %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX-LABEL: pmaddwd_const:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX-NEXT:    vpmulld {{.*}}(%rip), %ymm0, %ymm0
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
+   %a = sext <8 x i16> %A to <8 x i32>
+   %m = mul nsw <8 x i32> %a, <i32 32767, i32 -32768, i32 0, i32 0, i32 1, i32 7, i32 42, i32 32>
+   %odd = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+   %even = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+   %ret = add <4 x i32> %odd, %even
+   ret <4 x i32> %ret
+}
+
+; Check that there is not selection for unsigned multiplication
+define <4 x i32> @pmaddwd_negative1(<8 x i16> %A, <8 x i16> %B) {
+; SSE2-LABEL: pmaddwd_negative1:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pmulhuw %xmm1, %xmm2
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE2-NEXT:    paddd %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX-LABEL: pmaddwd_negative1:
+; AVX:       # %bb.0:
+; AVX-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
+; AVX-NEXT:    vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX-NEXT:    vpmulld %ymm1, %ymm0, %ymm0
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
+   %a = zext <8 x i16> %A to <8 x i32>
+   %b = zext <8 x i16> %B to <8 x i32>
+   %m = mul nuw <8 x i32> %a, %b
+   %odd = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+   %even = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+   %ret = add <4 x i32> %odd, %even
+   ret <4 x i32> %ret
+}
+
+; Check that there is not selection for out-of-bounds constant
+define <4 x i32> @pmaddwd_negative2(<8 x i16> %A) {
+; SSE2-LABEL: pmaddwd_negative2:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE2-NEXT:    psrad $16, %xmm1
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,7,42,32]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [32768,4294934528,0,0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm4[1,3]
+; SSE2-NEXT:    paddd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX-LABEL: pmaddwd_negative2:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX-NEXT:    vpmulld {{.*}}(%rip), %ymm0, %ymm0
+; AVX-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
+   %a = sext <8 x i16> %A to <8 x i32>
+   %m = mul nsw <8 x i32> %a, <i32 32768, i32 -32768, i32 0, i32 0, i32 1, i32 7, i32 42, i32 32>
+   %odd = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+   %even = shufflevector <8 x i32> %m, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+   %ret = add <4 x i32> %odd, %even
+   ret <4 x i32> %ret
+}