; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=SSE,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,SSE41
; 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 <4 x i16> @zext_mulhuw_v4i16(<4 x i16> %a, <4 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v4i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v4i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = zext <4 x i16> %a to <4 x i32>
  %b1 = zext <4 x i16> %b to <4 x i32>
  %c = mul <4 x i32> %a1, %b1
  %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  %e = trunc <4 x i32> %d to <4 x i16>
  ret <4 x i16> %e
}

define <4 x i16> @and_mulhuw_v4i16(<4 x i64> %a, <4 x i64> %b) {
; SSE2-LABEL: and_mulhuw_v4i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
; SSE2-NEXT:    pslld $16, %xmm2
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    xorps %xmm3, %xmm3
; SSE2-NEXT:    packssdw %xmm3, %xmm2
; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; SSE2-NEXT:    pslld $16, %xmm0
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    packssdw %xmm3, %xmm0
; SSE2-NEXT:    pmulhuw %xmm2, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: and_mulhuw_v4i16:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pxor %xmm4, %xmm4
; SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
; SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; SSE41-NEXT:    packusdw %xmm3, %xmm2
; SSE41-NEXT:    packusdw %xmm4, %xmm2
; SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
; SSE41-NEXT:    packusdw %xmm1, %xmm0
; SSE41-NEXT:    packusdw %xmm4, %xmm0
; SSE41-NEXT:    pmulhuw %xmm2, %xmm0
; SSE41-NEXT:    retq
;
; AVX2-LABEL: and_mulhuw_v4i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7],ymm0[8],ymm1[9,10,11],ymm0[12],ymm1[13,14,15]
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512-LABEL: and_mulhuw_v4i16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    vpmovqw %zmm0, %xmm0
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %a1 = and <4 x i64> %a, <i64 65535, i64 65535, i64 65535, i64 65535>
  %b1 = and <4 x i64> %b, <i64 65535, i64 65535, i64 65535, i64 65535>
  %c = mul <4 x i64> %a1, %b1
  %d = lshr <4 x i64> %c, <i64 16, i64 16, i64 16, i64 16>
  %e = trunc <4 x i64> %d to <4 x i16>
  ret <4 x i16> %e
}

define <4 x i16> @sext_mulhw_v4i16(<4 x i16> %a, <4 x i16> %b) {
; SSE-LABEL: sext_mulhw_v4i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: sext_mulhw_v4i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = sext <4 x i16> %a to <4 x i32>
  %b1 = sext <4 x i16> %b to <4 x i32>
  %c = mul <4 x i32> %a1, %b1
  %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  %e = trunc <4 x i32> %d to <4 x i16>
  ret <4 x i16> %e
}

define <4 x i16> @ashr_mulhw_v4i16(<4 x i32> %a, <4 x i32> %b) {
; SSE2-LABEL: ashr_mulhw_v4i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    packssdw %xmm1, %xmm1
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    packssdw %xmm0, %xmm0
; SSE2-NEXT:    pmulhw %xmm1, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: ashr_mulhw_v4i16:
; SSE41:       # %bb.0:
; SSE41-NEXT:    psrld $16, %xmm0
; SSE41-NEXT:    psrld $16, %xmm1
; SSE41-NEXT:    packusdw %xmm1, %xmm1
; SSE41-NEXT:    packusdw %xmm0, %xmm0
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    retq
;
; AVX-LABEL: ashr_mulhw_v4i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpsrld $16, %xmm0, %xmm0
; AVX-NEXT:    vpsrld $16, %xmm1, %xmm1
; AVX-NEXT:    vpackusdw %xmm1, %xmm1, %xmm1
; AVX-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = ashr <4 x i32> %a, <i32 16, i32 16, i32 16, i32 16>
  %b1 = ashr <4 x i32> %b, <i32 16, i32 16, i32 16, i32 16>
  %c = mul <4 x i32> %a1, %b1
  %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  %e = trunc <4 x i32> %d to <4 x i16>
  ret <4 x i16> %e
}

define <8 x i16> @zext_mulhuw_v8i16(<8 x i16> %a, <8 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v8i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v8i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = zext <8 x i16> %a to <8 x i32>
  %b1 = zext <8 x i16> %b to <8 x i32>
  %c = mul <8 x i32> %a1, %b1
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <8 x i32> %d to <8 x i16>
  ret <8 x i16> %e
}

define <8 x i16> @lshr_mulhuw_v8i16(<8 x i32> %a, <8 x i32> %b) {
; SSE2-LABEL: lshr_mulhuw_v8i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    packssdw %xmm3, %xmm2
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    packssdw %xmm1, %xmm0
; SSE2-NEXT:    pmulhuw %xmm2, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: lshr_mulhuw_v8i16:
; SSE41:       # %bb.0:
; SSE41-NEXT:    psrld $16, %xmm1
; SSE41-NEXT:    psrld $16, %xmm0
; SSE41-NEXT:    packusdw %xmm1, %xmm0
; SSE41-NEXT:    psrld $16, %xmm3
; SSE41-NEXT:    psrld $16, %xmm2
; SSE41-NEXT:    packusdw %xmm3, %xmm2
; SSE41-NEXT:    pmulhuw %xmm2, %xmm0
; SSE41-NEXT:    retq
;
; AVX2-LABEL: lshr_mulhuw_v8i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrld $16, %ymm0, %ymm0
; AVX2-NEXT:    vpsrld $16, %ymm1, %ymm1
; AVX2-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512-LABEL: lshr_mulhuw_v8i16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpsrld $16, %ymm0, %ymm0
; AVX512-NEXT:    vpsrld $16, %ymm1, %ymm1
; AVX512-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    vpmovdw %zmm0, %ymm0
; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %a1 = lshr <8 x i32> %a, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %b1 = lshr <8 x i32> %b, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %c = mul <8 x i32> %a1, %b1
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <8 x i32> %d to <8 x i16>
  ret <8 x i16> %e
}

define <8 x i16> @sext_mulhw_v8i16(<8 x i16> %a, <8 x i16> %b) {
; SSE-LABEL: sext_mulhw_v8i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: sext_mulhw_v8i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = sext <8 x i16> %a to <8 x i32>
  %b1 = sext <8 x i16> %b to <8 x i32>
  %c = mul <8 x i32> %a1, %b1
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <8 x i32> %d to <8 x i16>
  ret <8 x i16> %e
}

define <8 x i16> @sextinreg_mulhw_v8i16(<8 x i32> %a, <8 x i32> %b) {
; SSE-LABEL: sextinreg_mulhw_v8i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pslld $24, %xmm1
; SSE-NEXT:    psrad $24, %xmm1
; SSE-NEXT:    pslld $24, %xmm0
; SSE-NEXT:    psrad $24, %xmm0
; SSE-NEXT:    packssdw %xmm1, %xmm0
; SSE-NEXT:    pslld $25, %xmm3
; SSE-NEXT:    psrad $25, %xmm3
; SSE-NEXT:    pslld $25, %xmm2
; SSE-NEXT:    psrad $25, %xmm2
; SSE-NEXT:    packssdw %xmm3, %xmm2
; SSE-NEXT:    pmulhw %xmm2, %xmm0
; SSE-NEXT:    retq
;
; AVX2-LABEL: sextinreg_mulhw_v8i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpslld $24, %ymm0, %ymm0
; AVX2-NEXT:    vpsrad $24, %ymm0, %ymm0
; AVX2-NEXT:    vpslld $25, %ymm1, %ymm1
; AVX2-NEXT:    vpsrad $25, %ymm1, %ymm1
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
; AVX2-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
; AVX2-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512-LABEL: sextinreg_mulhw_v8i16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512-NEXT:    vpmovdw %zmm1, %ymm1
; AVX512-NEXT:    vpsllw $9, %xmm1, %xmm1
; AVX512-NEXT:    vpsraw $9, %xmm1, %xmm1
; AVX512-NEXT:    vpmovdw %zmm0, %ymm0
; AVX512-NEXT:    vpsllw $8, %xmm0, %xmm0
; AVX512-NEXT:    vpsraw $8, %xmm0, %xmm0
; AVX512-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %a1 = shl <8 x i32> %a, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
  %b1 = shl <8 x i32> %b, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
  %a2 = ashr <8 x i32> %a1, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
  %b2 = ashr <8 x i32> %b1, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
  %c = mul <8 x i32> %a2, %b2
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <8 x i32> %d to <8 x i16>
  ret <8 x i16> %e
}

define <8 x i16> @zext_mulhuw_v8i16_v8i33(<8 x i16> %a, <8 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v8i16_v8i33:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v8i16_v8i33:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = zext <8 x i16> %a to <8 x i33>
  %b1 = zext <8 x i16> %b to <8 x i33>
  %c = mul <8 x i33> %a1, %b1
  %d = lshr <8 x i33> %c, <i33 16, i33 16, i33 16, i33 16, i33 16, i33 16, i33 16, i33 16>
  %e = trunc <8 x i33> %d to <8 x i16>
  ret <8 x i16> %e
}

define <16 x i16> @zext_mulhuw_v16i16(<16 x i16> %a, <16 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v16i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm2, %xmm0
; SSE-NEXT:    pmulhuw %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v16i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX-NEXT:    retq
  %a1 = zext <16 x i16> %a to <16 x i32>
  %b1 = zext <16 x i16> %b to <16 x i32>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <16 x i32> %d to <16 x i16>
  ret <16 x i16> %e
}

define <16 x i16> @and_mulhuw_v16i16(<16 x i32> %a, <16 x i32> %b) {
; SSE2-LABEL: and_mulhuw_v16i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [32767,32767,32767,32767]
; SSE2-NEXT:    pand %xmm8, %xmm3
; SSE2-NEXT:    pand %xmm8, %xmm2
; SSE2-NEXT:    packssdw %xmm3, %xmm2
; SSE2-NEXT:    pand %xmm8, %xmm1
; SSE2-NEXT:    pand %xmm8, %xmm0
; SSE2-NEXT:    packssdw %xmm1, %xmm0
; SSE2-NEXT:    pand %xmm8, %xmm7
; SSE2-NEXT:    pand %xmm8, %xmm6
; SSE2-NEXT:    packssdw %xmm7, %xmm6
; SSE2-NEXT:    pmulhw %xmm2, %xmm6
; SSE2-NEXT:    pand %xmm8, %xmm5
; SSE2-NEXT:    pand %xmm4, %xmm8
; SSE2-NEXT:    packssdw %xmm5, %xmm8
; SSE2-NEXT:    pmulhw %xmm8, %xmm0
; SSE2-NEXT:    movdqa %xmm6, %xmm1
; SSE2-NEXT:    retq
;
; SSE41-LABEL: and_mulhuw_v16i16:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmovsxwd {{.*#+}} xmm8 = [32767,32767,32767,32767]
; SSE41-NEXT:    pand %xmm8, %xmm3
; SSE41-NEXT:    pand %xmm8, %xmm2
; SSE41-NEXT:    packusdw %xmm3, %xmm2
; SSE41-NEXT:    pand %xmm8, %xmm1
; SSE41-NEXT:    pand %xmm8, %xmm0
; SSE41-NEXT:    packusdw %xmm1, %xmm0
; SSE41-NEXT:    pand %xmm8, %xmm7
; SSE41-NEXT:    pand %xmm8, %xmm6
; SSE41-NEXT:    packusdw %xmm7, %xmm6
; SSE41-NEXT:    pmulhw %xmm2, %xmm6
; SSE41-NEXT:    pand %xmm8, %xmm5
; SSE41-NEXT:    pand %xmm4, %xmm8
; SSE41-NEXT:    packusdw %xmm5, %xmm8
; SSE41-NEXT:    pmulhw %xmm8, %xmm0
; SSE41-NEXT:    movdqa %xmm6, %xmm1
; SSE41-NEXT:    retq
;
; AVX2-LABEL: and_mulhuw_v16i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm4 = [32767,32767,32767,32767,32767,32767,32767,32767]
; AVX2-NEXT:    vpand %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpand %ymm4, %ymm1, %ymm1
; AVX2-NEXT:    vpand %ymm4, %ymm2, %ymm2
; AVX2-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm0
; AVX2-NEXT:    vpand %ymm4, %ymm3, %ymm2
; AVX2-NEXT:    vpmulhuw %ymm2, %ymm1, %ymm1
; AVX2-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: and_mulhuw_v16i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmovdw %zmm1, %ymm1
; AVX512F-NEXT:    vpbroadcastw {{.*#+}} ymm2 = [32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767]
; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
; AVX512F-NEXT:    vpmovdw %zmm0, %ymm0
; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: and_mulhuw_v16i16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpbroadcastd {{.*#+}} zmm2 = [32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767,32767]
; AVX512BW-NEXT:    vpandd %zmm2, %zmm0, %zmm0
; AVX512BW-NEXT:    vpandd %zmm2, %zmm1, %zmm1
; AVX512BW-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT:    vpmovdw %zmm0, %ymm0
; AVX512BW-NEXT:    retq
  %a1 = and <16 x i32> %a, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
  %b1 = and <16 x i32> %b, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <16 x i32> %d to <16 x i16>
  ret <16 x i16> %e
}

define <16 x i16> @sext_mulhuw_v16i16(<16 x i16> %a, <16 x i16> %b) {
; SSE-LABEL: sext_mulhuw_v16i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhw %xmm2, %xmm0
; SSE-NEXT:    pmulhw %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-LABEL: sext_mulhuw_v16i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX-NEXT:    retq
  %a1 = sext <16 x i16> %a to <16 x i32>
  %b1 = sext <16 x i16> %b to <16 x i32>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <16 x i32> %d to <16 x i16>
  ret <16 x i16> %e
}

define <16 x i16> @ashr_mulhuw_v16i16(<16 x i32> %a, <16 x i32> %b) {
; SSE2-LABEL: ashr_mulhuw_v16i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrad $16, %xmm5
; SSE2-NEXT:    psrad $16, %xmm4
; SSE2-NEXT:    packssdw %xmm5, %xmm4
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    packssdw %xmm1, %xmm0
; SSE2-NEXT:    pmulhw %xmm4, %xmm0
; SSE2-NEXT:    psrad $16, %xmm7
; SSE2-NEXT:    psrad $16, %xmm6
; SSE2-NEXT:    packssdw %xmm7, %xmm6
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    packssdw %xmm3, %xmm2
; SSE2-NEXT:    pmulhw %xmm6, %xmm2
; SSE2-NEXT:    movdqa %xmm2, %xmm1
; SSE2-NEXT:    retq
;
; SSE41-LABEL: ashr_mulhuw_v16i16:
; SSE41:       # %bb.0:
; SSE41-NEXT:    psrld $16, %xmm3
; SSE41-NEXT:    psrld $16, %xmm2
; SSE41-NEXT:    packusdw %xmm3, %xmm2
; SSE41-NEXT:    psrld $16, %xmm1
; SSE41-NEXT:    psrld $16, %xmm0
; SSE41-NEXT:    packusdw %xmm1, %xmm0
; SSE41-NEXT:    psrld $16, %xmm7
; SSE41-NEXT:    psrld $16, %xmm6
; SSE41-NEXT:    packusdw %xmm7, %xmm6
; SSE41-NEXT:    pmulhw %xmm2, %xmm6
; SSE41-NEXT:    psrld $16, %xmm5
; SSE41-NEXT:    psrld $16, %xmm4
; SSE41-NEXT:    packusdw %xmm5, %xmm4
; SSE41-NEXT:    pmulhw %xmm4, %xmm0
; SSE41-NEXT:    movdqa %xmm6, %xmm1
; SSE41-NEXT:    retq
;
; AVX2-LABEL: ashr_mulhuw_v16i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrld $16, %ymm1, %ymm1
; AVX2-NEXT:    vpsrld $16, %ymm0, %ymm0
; AVX2-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vpsrld $16, %ymm3, %ymm1
; AVX2-NEXT:    vpsrld $16, %ymm2, %ymm2
; AVX2-NEXT:    vpackusdw %ymm1, %ymm2, %ymm1
; AVX2-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT:    retq
;
; AVX512-LABEL: ashr_mulhuw_v16i16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpsrld $16, %zmm0, %zmm0
; AVX512-NEXT:    vpsrld $16, %zmm1, %zmm1
; AVX512-NEXT:    vpmovdw %zmm1, %ymm1
; AVX512-NEXT:    vpmovdw %zmm0, %ymm0
; AVX512-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    retq
  %a1 = ashr <16 x i32> %a, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %b1 = ashr <16 x i32> %b, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <16 x i32> %d to <16 x i16>
  ret <16 x i16> %e
}

define <16 x i16> @zext_mulhuw_v16i16_v16i48(<16 x i16> %a, <16 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v16i16_v16i48:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm2, %xmm0
; SSE-NEXT:    pmulhuw %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v16i16_v16i48:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX-NEXT:    retq
  %a1 = zext <16 x i16> %a to <16 x i48>
  %b1 = zext <16 x i16> %b to <16 x i48>
  %c = mul <16 x i48> %a1, %b1
  %d = lshr <16 x i48> %c, <i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16, i48 16>
  %e = trunc <16 x i48> %d to <16 x i16>
  ret <16 x i16> %e
}

define <32 x i16> @zext_mulhuw_v32i16(<32 x i16> %a, <32 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v32i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm4, %xmm0
; SSE-NEXT:    pmulhuw %xmm5, %xmm1
; SSE-NEXT:    pmulhuw %xmm6, %xmm2
; SSE-NEXT:    pmulhuw %xmm7, %xmm3
; SSE-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v32i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm0
; AVX2-NEXT:    vpmulhuw %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: zext_mulhuw_v32i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm3
; AVX512F-NEXT:    vpmulhuw %ymm2, %ymm3, %ymm2
; AVX512F-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: zext_mulhuw_v32i16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT:    retq
  %a1 = zext <32 x i16> %a to <32 x i32>
  %b1 = zext <32 x i16> %b to <32 x i32>
  %c = mul <32 x i32> %a1, %b1
  %d = lshr <32 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <32 x i32> %d to <32 x i16>
  ret <32 x i16> %e
}

define <32 x i16> @sext_mulhuw_v32i16(<32 x i16> %a, <32 x i16> %b) {
; SSE-LABEL: sext_mulhuw_v32i16:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhw %xmm4, %xmm0
; SSE-NEXT:    pmulhw %xmm5, %xmm1
; SSE-NEXT:    pmulhw %xmm6, %xmm2
; SSE-NEXT:    pmulhw %xmm7, %xmm3
; SSE-NEXT:    retq
;
; AVX2-LABEL: sext_mulhuw_v32i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm2, %ymm0, %ymm0
; AVX2-NEXT:    vpmulhw %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: sext_mulhuw_v32i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm3
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm3, %ymm2
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: sext_mulhuw_v32i16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm1, %zmm0, %zmm0
; AVX512BW-NEXT:    retq
  %a1 = sext <32 x i16> %a to <32 x i32>
  %b1 = sext <32 x i16> %b to <32 x i32>
  %c = mul <32 x i32> %a1, %b1
  %d = lshr <32 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <32 x i32> %d to <32 x i16>
  ret <32 x i16> %e
}

define <64 x i16> @zext_mulhuw_v64i16(<64 x i16> %a, <64 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v64i16:
; SSE:       # %bb.0:
; SSE-NEXT:    movq %rdi, %rax
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm0
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm1
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm2
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm3
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm4
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm5
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm6
; SSE-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm7
; SSE-NEXT:    movdqa %xmm7, 112(%rdi)
; SSE-NEXT:    movdqa %xmm6, 96(%rdi)
; SSE-NEXT:    movdqa %xmm5, 80(%rdi)
; SSE-NEXT:    movdqa %xmm4, 64(%rdi)
; SSE-NEXT:    movdqa %xmm3, 48(%rdi)
; SSE-NEXT:    movdqa %xmm2, 32(%rdi)
; SSE-NEXT:    movdqa %xmm1, 16(%rdi)
; SSE-NEXT:    movdqa %xmm0, (%rdi)
; SSE-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v64i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpmulhuw %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vpmulhuw %ymm6, %ymm2, %ymm2
; AVX2-NEXT:    vpmulhuw %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: zext_mulhuw_v64i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vextracti64x4 $1, %zmm2, %ymm4
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm5
; AVX512F-NEXT:    vpmulhuw %ymm4, %ymm5, %ymm4
; AVX512F-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm0
; AVX512F-NEXT:    vextracti64x4 $1, %zmm3, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
; AVX512F-NEXT:    vpmulhuw %ymm2, %ymm4, %ymm2
; AVX512F-NEXT:    vpmulhuw %ymm3, %ymm1, %ymm1
; AVX512F-NEXT:    vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: zext_mulhuw_v64i16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhuw %zmm2, %zmm0, %zmm0
; AVX512BW-NEXT:    vpmulhuw %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = zext <64 x i16> %a to <64 x i32>
  %b1 = zext <64 x i16> %b to <64 x i32>
  %c = mul <64 x i32> %a1, %b1
  %d = lshr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <64 x i32> %d to <64 x i16>
  ret <64 x i16> %e
}

define <64 x i16> @sext_mulhuw_v64i16(<64 x i16> %a, <64 x i16> %b) {
; SSE-LABEL: sext_mulhuw_v64i16:
; SSE:       # %bb.0:
; SSE-NEXT:    movq %rdi, %rax
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm0
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm3
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm4
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm5
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm6
; SSE-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm7
; SSE-NEXT:    movdqa %xmm7, 112(%rdi)
; SSE-NEXT:    movdqa %xmm6, 96(%rdi)
; SSE-NEXT:    movdqa %xmm5, 80(%rdi)
; SSE-NEXT:    movdqa %xmm4, 64(%rdi)
; SSE-NEXT:    movdqa %xmm3, 48(%rdi)
; SSE-NEXT:    movdqa %xmm2, 32(%rdi)
; SSE-NEXT:    movdqa %xmm1, 16(%rdi)
; SSE-NEXT:    movdqa %xmm0, (%rdi)
; SSE-NEXT:    retq
;
; AVX2-LABEL: sext_mulhuw_v64i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpmulhw %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vpmulhw %ymm6, %ymm2, %ymm2
; AVX2-NEXT:    vpmulhw %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: sext_mulhuw_v64i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vextracti64x4 $1, %zmm2, %ymm4
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm5
; AVX512F-NEXT:    vpmulhw %ymm4, %ymm5, %ymm4
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm0
; AVX512F-NEXT:    vextracti64x4 $1, %zmm3, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm4, %ymm2
; AVX512F-NEXT:    vpmulhw %ymm3, %ymm1, %ymm1
; AVX512F-NEXT:    vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: sext_mulhuw_v64i16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm2, %zmm0, %zmm0
; AVX512BW-NEXT:    vpmulhw %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = sext <64 x i16> %a to <64 x i32>
  %b1 = sext <64 x i16> %b to <64 x i32>
  %c = mul <64 x i32> %a1, %b1
  %d = lshr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %e = trunc <64 x i32> %d to <64 x i16>
  ret <64 x i16> %e
}

define <8 x i16> @zext_mulhuw_v8i16_i64(<8 x i16> %a, <8 x i16> %b) {
; SSE-LABEL: zext_mulhuw_v8i16_i64:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhuw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v8i16_i64:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = zext <8 x i16> %a to <8 x i64>
  %b1 = zext <8 x i16> %b to <8 x i64>
  %c = mul <8 x i64> %a1, %b1
  %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
  %e = trunc <8 x i64> %d to <8 x i16>
  ret <8 x i16> %e
}

define <8 x i16> @sext_mulhuw_v8i16_i64(<8 x i16> %a, <8 x i16> %b) {
; SSE-LABEL: sext_mulhuw_v8i16_i64:
; SSE:       # %bb.0:
; SSE-NEXT:    pmulhw %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: sext_mulhuw_v8i16_i64:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = sext <8 x i16> %a to <8 x i64>
  %b1 = sext <8 x i16> %b to <8 x i64>
  %c = mul <8 x i64> %a1, %b1
  %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
  %e = trunc <8 x i64> %d to <8 x i16>
  ret <8 x i16> %e
}

define <4 x i32> @zext_mulhuw_v4i16_lshr(<4 x i16> %a, <4 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v4i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhuw %xmm1, %xmm0
; SSE2-NEXT:    pxor %xmm1, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v4i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhuw %xmm1, %xmm0
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; SSE41-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v4i16_lshr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; AVX-NEXT:    retq
  %a1 = zext <4 x i16> %a to <4 x i32>
  %b1 = zext <4 x i16> %b to <4 x i32>
  %c = mul <4 x i32> %a1, %b1
  %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  ret <4 x i32> %d
}

define <4 x i32> @mulhsw_v4i16_lshr(<4 x i16> %a, <4 x i16> %b) {
; SSE2-LABEL: mulhsw_v4i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm1, %xmm0
; SSE2-NEXT:    pxor %xmm1, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v4i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; SSE41-NEXT:    retq
;
; AVX-LABEL: mulhsw_v4i16_lshr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; AVX-NEXT:    retq
  %a1 = sext <4 x i16> %a to <4 x i32>
  %b1 = sext <4 x i16> %b to <4 x i32>
  %c = mul <4 x i32> %a1, %b1
  %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  ret <4 x i32> %d
}

define <4 x i32> @mulhsw_v4i16_ashr(<4 x i16> %a, <4 x i16> %b) {
; SSE2-LABEL: mulhsw_v4i16_ashr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm1, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v4i16_ashr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
; SSE41-NEXT:    retq
;
; AVX-LABEL: mulhsw_v4i16_ashr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    vpmovsxwd %xmm0, %xmm0
; AVX-NEXT:    retq
  %a1 = sext <4 x i16> %a to <4 x i32>
  %b1 = sext <4 x i16> %b to <4 x i32>
  %c = mul <4 x i32> %a1, %b1
  %d = ashr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
  ret <4 x i32> %d
}

define <8 x i32> @zext_mulhuw_v8i16_lshr(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v8i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhuw %xmm0, %xmm1
; SSE2-NEXT:    pxor %xmm2, %xmm2
; SSE2-NEXT:    movdqa %xmm1, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v8i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm0, %xmm2
; SSE41-NEXT:    pmulhuw %xmm1, %xmm2
; SSE41-NEXT:    pxor %xmm1, %xmm1
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; SSE41-NEXT:    movdqa %xmm2, %xmm1
; SSE41-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v8i16_lshr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; 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:    retq
  %a1 = zext <8 x i16> %a to <8 x i32>
  %b1 = zext <8 x i16> %b to <8 x i32>
  %c = mul <8 x i32> %a1, %b1
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <8 x i32> %d
}

define <8 x i32> @mulhsw_v8i16_lshr(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: mulhsw_v8i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm0, %xmm1
; SSE2-NEXT:    pxor %xmm2, %xmm2
; SSE2-NEXT:    movdqa %xmm1, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v8i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm0, %xmm2
; SSE41-NEXT:    pmulhw %xmm1, %xmm2
; SSE41-NEXT:    pxor %xmm1, %xmm1
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; SSE41-NEXT:    movdqa %xmm2, %xmm1
; SSE41-NEXT:    retq
;
; AVX-LABEL: mulhsw_v8i16_lshr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; 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:    retq
  %a1 = sext <8 x i16> %a to <8 x i32>
  %b1 = sext <8 x i16> %b to <8 x i32>
  %c = mul <8 x i32> %a1, %b1
  %d = lshr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <8 x i32> %d
}

define <8 x i32> @mulhsw_v8i16_ashr(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: mulhsw_v8i16_ashr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm1, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    movdqa %xmm2, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v8i16_ashr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm2
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm1
; SSE41-NEXT:    movdqa %xmm2, %xmm0
; SSE41-NEXT:    retq
;
; AVX-LABEL: mulhsw_v8i16_ashr:
; AVX:       # %bb.0:
; AVX-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    vpmovsxwd %xmm0, %ymm0
; AVX-NEXT:    retq
  %a1 = sext <8 x i16> %a to <8 x i32>
  %b1 = sext <8 x i16> %b to <8 x i32>
  %c = mul <8 x i32> %a1, %b1
  %d = ashr <8 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <8 x i32> %d
}

define <16 x i32> @zext_mulhuw_v16i16_lshr(<16 x i16> %a, <16 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v16i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm0, %xmm4
; SSE2-NEXT:    pmulhuw %xmm2, %xmm4
; SSE2-NEXT:    pxor %xmm5, %xmm5
; SSE2-NEXT:    movdqa %xmm4, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
; SSE2-NEXT:    pmulhuw %xmm1, %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm2
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
; SSE2-NEXT:    movdqa %xmm4, %xmm1
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v16i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm1, %xmm4
; SSE41-NEXT:    movdqa %xmm0, %xmm1
; SSE41-NEXT:    pmulhuw %xmm2, %xmm1
; SSE41-NEXT:    pxor %xmm5, %xmm5
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
; SSE41-NEXT:    pmulhuw %xmm3, %xmm4
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm2 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
; SSE41-NEXT:    movdqa %xmm4, %xmm3
; SSE41-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v16i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-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
; AVX2-NEXT:    retq
;
; AVX512-LABEL: zext_mulhuw_v16i16_lshr:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512-NEXT:    retq
  %a1 = zext <16 x i16> %a to <16 x i32>
  %b1 = zext <16 x i16> %b to <16 x i32>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <16 x i32> %d
}

; PR109790
define void @PR109790(ptr sret([32 x i8]) %ret, ptr %a) {
; SSE-LABEL: PR109790:
; SSE:       # %bb.0:
; SSE-NEXT:    movq %rdi, %rax
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [32767,32767,32767,32767,32767,32767,32767,32767]
; SSE-NEXT:    movdqa (%rsi), %xmm1
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    pand 16(%rsi), %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [64536,64536,64536,64536,64536,64536,64536,64536]
; SSE-NEXT:    pmulhw %xmm2, %xmm0
; SSE-NEXT:    pmulhw %xmm2, %xmm1
; SSE-NEXT:    movdqa %xmm1, (%rdi)
; SSE-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE-NEXT:    retq
;
; AVX2-LABEL: PR109790:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq %rdi, %rax
; AVX2-NEXT:    vmovdqa (%rsi), %ymm0
; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT:    vpmulhw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 # [64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536]
; AVX2-NEXT:    vmovdqa %ymm0, (%rdi)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: PR109790:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    movq %rdi, %rax
; AVX512F-NEXT:    vmovdqa (%rsi), %ymm0
; AVX512F-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vpmulld {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; AVX512F-NEXT:    vpsrld $16, %zmm0, %zmm0
; AVX512F-NEXT:    vpmovdw %zmm0, (%rdi)
; AVX512F-NEXT:    vzeroupper
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: PR109790:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq %rdi, %rax
; AVX512BW-NEXT:    vmovdqa (%rsi), %ymm0
; AVX512BW-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512BW-NEXT:    vpmaddwd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0 # [64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0,64536,0]
; AVX512BW-NEXT:    vpsrld $16, %zmm0, %zmm0
; AVX512BW-NEXT:    vpmovdw %zmm0, (%rdi)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
  %load = load <16 x i16>, ptr %a, align 32
  %and = and <16 x i16> %load, <i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767>
  %ext = zext nneg <16 x i16> %and to <16 x i32>
  %mul = mul nsw <16 x i32> %ext, <i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000>
  %srl = lshr <16 x i32> %mul, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %res = trunc nuw <16 x i32> %srl to <16 x i16>
  store <16 x i16> %res, ptr %ret, align 32
  ret void
}

; PR109790
define <16 x i16> @zext_mulhuw_v16i16_negative_constant(<16 x i16> %a) {
; SSE-LABEL: zext_mulhuw_v16i16_negative_constant:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [32767,32767,32767,32767,32767,32767,32767,32767]
; SSE-NEXT:    pand %xmm2, %xmm1
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [64536,64536,64536,64536,64536,64536,64536,64536]
; SSE-NEXT:    pmulhw %xmm2, %xmm0
; SSE-NEXT:    pmulhw %xmm2, %xmm1
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v16i16_negative_constant:
; AVX:       # %bb.0:
; AVX-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX-NEXT:    vpmulhw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 # [64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536,64536]
; AVX-NEXT:    retq
  %k = and <16 x i16> %a, <i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767>
  %x = zext nneg <16 x i16> %k to <16 x i32>
  %m = mul nsw <16 x i32> %x, <i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000, i32 -1000>
  %s = lshr <16 x i32> %m, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %t = trunc nuw <16 x i32> %s to <16 x i16>
  ret <16 x i16> %t
}

; PR109790
define <16 x i16> @zext_mulhuw_v16i16_positive_constant(<16 x i16> %a) {
; SSE-LABEL: zext_mulhuw_v16i16_positive_constant:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [32767,32767,32767,32767,32767,32767,32767,32767]
; SSE-NEXT:    pand %xmm2, %xmm1
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [1000,1000,1000,1000,1000,1000,1000,1000]
; SSE-NEXT:    pmulhw %xmm2, %xmm0
; SSE-NEXT:    pmulhw %xmm2, %xmm1
; SSE-NEXT:    retq
;
; AVX-LABEL: zext_mulhuw_v16i16_positive_constant:
; AVX:       # %bb.0:
; AVX-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX-NEXT:    vpmulhw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 # [1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000,1000]
; AVX-NEXT:    retq
  %k = and <16 x i16> %a, <i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767, i16 32767>
  %x = zext nneg <16 x i16> %k to <16 x i32>
  %m = mul nuw nsw <16 x i32> %x, <i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000, i32 1000>
  %s = lshr <16 x i32> %m, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  %t = trunc nuw nsw <16 x i32> %s to <16 x i16>
  ret <16 x i16> %t
}

define <16 x i32> @mulhsw_v16i16_lshr(<16 x i16> %a, <16 x i16> %b) {
; SSE2-LABEL: mulhsw_v16i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm0, %xmm4
; SSE2-NEXT:    pmulhw %xmm2, %xmm4
; SSE2-NEXT:    pxor %xmm5, %xmm5
; SSE2-NEXT:    movdqa %xmm4, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
; SSE2-NEXT:    pmulhw %xmm1, %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm2
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
; SSE2-NEXT:    movdqa %xmm4, %xmm1
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v16i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm1, %xmm4
; SSE41-NEXT:    movdqa %xmm0, %xmm1
; SSE41-NEXT:    pmulhw %xmm2, %xmm1
; SSE41-NEXT:    pxor %xmm5, %xmm5
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
; SSE41-NEXT:    pmulhw %xmm3, %xmm4
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm2 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
; SSE41-NEXT:    movdqa %xmm4, %xmm3
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v16i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm1, %ymm0, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-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
; AVX2-NEXT:    retq
;
; AVX512-LABEL: mulhsw_v16i16_lshr:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512-NEXT:    retq
  %a1 = sext <16 x i16> %a to <16 x i32>
  %b1 = sext <16 x i16> %b to <16 x i32>
  %c = mul <16 x i32> %a1, %b1
  %d = lshr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <16 x i32> %d
}

define <16 x i32> @mulhsw_v16i16_ashr(<16 x i16> %a, <16 x i16> %b) {
; SSE2-LABEL: mulhsw_v16i16_ashr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm2, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; SSE2-NEXT:    psrad $16, %xmm5
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
; SSE2-NEXT:    psrad $16, %xmm4
; SSE2-NEXT:    pmulhw %xmm3, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    movdqa %xmm5, %xmm0
; SSE2-NEXT:    movdqa %xmm4, %xmm1
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v16i16_ashr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm2, %xmm0
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm4
; SSE41-NEXT:    pmulhw %xmm3, %xmm1
; SSE41-NEXT:    pmovsxwd %xmm1, %xmm2
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm5
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm3
; SSE41-NEXT:    movdqa %xmm4, %xmm0
; SSE41-NEXT:    movdqa %xmm5, %xmm1
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v16i16_ashr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm1, %ymm0, %ymm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm0
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: mulhsw_v16i16_ashr:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512-NEXT:    vpmovsxwd %ymm0, %zmm0
; AVX512-NEXT:    retq
  %a1 = sext <16 x i16> %a to <16 x i32>
  %b1 = sext <16 x i16> %b to <16 x i32>
  %c = mul <16 x i32> %a1, %b1
  %d = ashr <16 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <16 x i32> %d
}

define <32 x i32> @zext_mulhuw_v32i16_lshr(<32 x i16> %a, <32 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v32i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhuw %xmm4, %xmm0
; SSE2-NEXT:    pxor %xmm4, %xmm4
; SSE2-NEXT:    movdqa %xmm0, %xmm8
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
; SSE2-NEXT:    pmulhuw %xmm5, %xmm1
; SSE2-NEXT:    movdqa %xmm1, %xmm5
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
; SSE2-NEXT:    pmulhuw %xmm6, %xmm2
; SSE2-NEXT:    movdqa %xmm2, %xmm6
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
; SSE2-NEXT:    pmulhuw %xmm7, %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movdqa %xmm8, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v32i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhuw %xmm4, %xmm0
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm4 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; SSE41-NEXT:    pxor %xmm8, %xmm8
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE41-NEXT:    pmulhuw %xmm5, %xmm1
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm5 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
; SSE41-NEXT:    pmulhuw %xmm6, %xmm2
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm6 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
; SSE41-NEXT:    pmulhuw %xmm7, %xmm3
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm7 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
; SSE41-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE41-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE41-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE41-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE41-NEXT:    movdqa %xmm4, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v32i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vpmulhuw %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vmovdqa %ymm4, %ymm1
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: zext_mulhuw_v32i16_lshr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm2
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: zext_mulhuw_v32i16_lshr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhuw %zmm1, %zmm0, %zmm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    retq
  %a1 = zext <32 x i16> %a to <32 x i32>
  %b1 = zext <32 x i16> %b to <32 x i32>
  %c = mul <32 x i32> %a1, %b1
  %d = lshr <32 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <32 x i32> %d
}

define <32 x i32> @mulhsw_v32i16_lshr(<32 x i16> %a, <32 x i16> %b) {
; SSE2-LABEL: mulhsw_v32i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhw %xmm4, %xmm0
; SSE2-NEXT:    pxor %xmm4, %xmm4
; SSE2-NEXT:    movdqa %xmm0, %xmm8
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
; SSE2-NEXT:    pmulhw %xmm5, %xmm1
; SSE2-NEXT:    movdqa %xmm1, %xmm5
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
; SSE2-NEXT:    pmulhw %xmm6, %xmm2
; SSE2-NEXT:    movdqa %xmm2, %xmm6
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
; SSE2-NEXT:    pmulhw %xmm7, %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movdqa %xmm8, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v32i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhw %xmm4, %xmm0
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm4 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; SSE41-NEXT:    pxor %xmm8, %xmm8
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE41-NEXT:    pmulhw %xmm5, %xmm1
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm5 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
; SSE41-NEXT:    pmulhw %xmm6, %xmm2
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm6 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
; SSE41-NEXT:    pmulhw %xmm7, %xmm3
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm7 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
; SSE41-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE41-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE41-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE41-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE41-NEXT:    movdqa %xmm4, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v32i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm2, %ymm0, %ymm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vpmulhw %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vmovdqa %ymm4, %ymm1
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: mulhsw_v32i16_lshr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm2
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: mulhsw_v32i16_lshr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm1, %zmm0, %zmm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    retq
  %a1 = sext <32 x i16> %a to <32 x i32>
  %b1 = sext <32 x i16> %b to <32 x i32>
  %c = mul <32 x i32> %a1, %b1
  %d = lshr <32 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <32 x i32> %d
}

define <32 x i32> @mulhsw_v32i16_ashr(<32 x i16> %a, <32 x i16> %b) {
; SSE2-LABEL: mulhsw_v32i16_ashr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhw %xmm4, %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
; SSE2-NEXT:    psrad $16, %xmm4
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    pmulhw %xmm5, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; SSE2-NEXT:    psrad $16, %xmm5
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    pmulhw %xmm6, %xmm2
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; SSE2-NEXT:    psrad $16, %xmm6
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    pmulhw %xmm7, %xmm3
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1],xmm7[2],xmm3[2],xmm7[3],xmm3[3]
; SSE2-NEXT:    psrad $16, %xmm7
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movdqa %xmm4, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v32i16_ashr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhw %xmm4, %xmm0
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm4
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
; SSE41-NEXT:    pmulhw %xmm5, %xmm1
; SSE41-NEXT:    pmovsxwd %xmm1, %xmm5
; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm1, %xmm1
; SSE41-NEXT:    pmulhw %xmm6, %xmm2
; SSE41-NEXT:    pmovsxwd %xmm2, %xmm6
; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm2, %xmm2
; SSE41-NEXT:    pmulhw %xmm7, %xmm3
; SSE41-NEXT:    pmovsxwd %xmm3, %xmm7
; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm3, %xmm3
; SSE41-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE41-NEXT:    movdqa %xmm7, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE41-NEXT:    movdqa %xmm6, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE41-NEXT:    movdqa %xmm5, 32(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE41-NEXT:    movdqa %xmm4, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v32i16_ashr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %ymm2, %ymm0, %ymm2
; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm0
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm4
; AVX2-NEXT:    vpmulhw %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm2
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm3
; AVX2-NEXT:    vmovdqa %ymm4, %ymm1
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: mulhsw_v32i16_ashr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm2
; AVX512F-NEXT:    vpmovsxwd %ymm2, %zmm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm1
; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: mulhsw_v32i16_ashr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm1, %zmm0, %zmm1
; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm0
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = sext <32 x i16> %a to <32 x i32>
  %b1 = sext <32 x i16> %b to <32 x i32>
  %c = mul <32 x i32> %a1, %b1
  %d = ashr <32 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <32 x i32> %d
}

define <64 x i32> @zext_mulhuw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v64i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm7, %xmm8
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm0
; SSE2-NEXT:    pxor %xmm10, %xmm10
; SSE2-NEXT:    movdqa %xmm0, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
; SSE2-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm1
; SSE2-NEXT:    movdqa %xmm1, %xmm9
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm2
; SSE2-NEXT:    movdqa %xmm2, %xmm11
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm12
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm4
; SSE2-NEXT:    movdqa %xmm4, %xmm13
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm5
; SSE2-NEXT:    movdqa %xmm5, %xmm14
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm6
; SSE2-NEXT:    movdqa %xmm6, %xmm15
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm10[0],xmm15[1],xmm10[1],xmm15[2],xmm10[2],xmm15[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm10[4],xmm6[5],xmm10[5],xmm6[6],xmm10[6],xmm6[7],xmm10[7]
; SSE2-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm8
; SSE2-NEXT:    movdqa %xmm8, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm10[4],xmm8[5],xmm10[5],xmm8[6],xmm10[6],xmm8[7],xmm10[7]
; SSE2-NEXT:    movdqa %xmm8, 240(%rdi)
; SSE2-NEXT:    movdqa %xmm7, 224(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 208(%rdi)
; SSE2-NEXT:    movdqa %xmm15, 192(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 176(%rdi)
; SSE2-NEXT:    movdqa %xmm14, 160(%rdi)
; SSE2-NEXT:    movdqa %xmm4, 144(%rdi)
; SSE2-NEXT:    movdqa %xmm13, 128(%rdi)
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm12, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm11, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm9, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE2-NEXT:    movaps %xmm0, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v64i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm0, %xmm8
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm8
; SSE41-NEXT:    pxor %xmm11, %xmm11
; SSE41-NEXT:    movdqa %xmm8, %xmm0
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
; SSE41-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm1
; SSE41-NEXT:    movdqa %xmm1, %xmm9
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm2
; SSE41-NEXT:    movdqa %xmm2, %xmm10
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm3
; SSE41-NEXT:    movdqa %xmm3, %xmm12
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm4
; SSE41-NEXT:    movdqa %xmm4, %xmm13
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm5
; SSE41-NEXT:    movdqa %xmm5, %xmm14
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm11[4],xmm14[5],xmm11[5],xmm14[6],xmm11[6],xmm14[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm6
; SSE41-NEXT:    movdqa %xmm6, %xmm15
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm11[4],xmm15[5],xmm11[5],xmm15[6],xmm11[6],xmm15[7],xmm11[7]
; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm7
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm11[4],xmm7[5],xmm11[5],xmm7[6],xmm11[6],xmm7[7],xmm11[7]
; SSE41-NEXT:    movdqa %xmm7, 240(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
; SSE41-NEXT:    movdqa %xmm15, 208(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
; SSE41-NEXT:    movdqa %xmm14, 176(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
; SSE41-NEXT:    movdqa %xmm13, 144(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
; SSE41-NEXT:    movdqa %xmm12, 112(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm10, 80(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm9, 48(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE41-NEXT:    movaps %xmm0, 16(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero
; SSE41-NEXT:    movdqa %xmm0, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v64i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq %rdi, %rax
; AVX2-NEXT:    vpmulhuw %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm4 = 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:    vextracti128 $1, %ymm0, %xmm0
; 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:    vpmulhuw %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm5 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-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
; AVX2-NEXT:    vpmulhuw %ymm6, %ymm2, %ymm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm6 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vpmulhuw %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm7 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm3
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX2-NEXT:    vmovdqa %ymm3, 224(%rdi)
; AVX2-NEXT:    vmovdqa %ymm7, 192(%rdi)
; AVX2-NEXT:    vmovdqa %ymm2, 160(%rdi)
; AVX2-NEXT:    vmovdqa %ymm6, 128(%rdi)
; AVX2-NEXT:    vmovdqa %ymm1, 96(%rdi)
; AVX2-NEXT:    vmovdqa %ymm5, 64(%rdi)
; AVX2-NEXT:    vmovdqa %ymm0, 32(%rdi)
; AVX2-NEXT:    vmovdqa %ymm4, (%rdi)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: zext_mulhuw_v64i16_lshr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm4
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhuw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm5 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vpmulhuw %ymm3, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm3, %ymm0
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vpmulhuw %ymm0, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: zext_mulhuw_v64i16_lshr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhuw %zmm2, %zmm0, %zmm2
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512BW-NEXT:    vpmulhuw %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = zext <64 x i16> %a to <64 x i32>
  %b1 = zext <64 x i16> %b to <64 x i32>
  %c = mul <64 x i32> %a1, %b1
  %d = lshr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <64 x i32> %d
}

define <64 x i32> @mulhsw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
; SSE2-LABEL: mulhsw_v64i16_lshr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm7, %xmm8
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm0
; SSE2-NEXT:    pxor %xmm10, %xmm10
; SSE2-NEXT:    movdqa %xmm0, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
; SSE2-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
; SSE2-NEXT:    movdqa %xmm1, %xmm9
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
; SSE2-NEXT:    movdqa %xmm2, %xmm11
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm3
; SSE2-NEXT:    movdqa %xmm3, %xmm12
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm4
; SSE2-NEXT:    movdqa %xmm4, %xmm13
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm5
; SSE2-NEXT:    movdqa %xmm5, %xmm14
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm6
; SSE2-NEXT:    movdqa %xmm6, %xmm15
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm10[0],xmm15[1],xmm10[1],xmm15[2],xmm10[2],xmm15[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm10[4],xmm6[5],xmm10[5],xmm6[6],xmm10[6],xmm6[7],xmm10[7]
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm8
; SSE2-NEXT:    movdqa %xmm8, %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm10[4],xmm8[5],xmm10[5],xmm8[6],xmm10[6],xmm8[7],xmm10[7]
; SSE2-NEXT:    movdqa %xmm8, 240(%rdi)
; SSE2-NEXT:    movdqa %xmm7, 224(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 208(%rdi)
; SSE2-NEXT:    movdqa %xmm15, 192(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 176(%rdi)
; SSE2-NEXT:    movdqa %xmm14, 160(%rdi)
; SSE2-NEXT:    movdqa %xmm4, 144(%rdi)
; SSE2-NEXT:    movdqa %xmm13, 128(%rdi)
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm12, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm11, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm9, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE2-NEXT:    movaps %xmm0, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v64i16_lshr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movdqa %xmm0, %xmm8
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm8
; SSE41-NEXT:    pxor %xmm11, %xmm11
; SSE41-NEXT:    movdqa %xmm8, %xmm0
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
; SSE41-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
; SSE41-NEXT:    movdqa %xmm1, %xmm9
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
; SSE41-NEXT:    movdqa %xmm2, %xmm10
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm3
; SSE41-NEXT:    movdqa %xmm3, %xmm12
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm4
; SSE41-NEXT:    movdqa %xmm4, %xmm13
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm5
; SSE41-NEXT:    movdqa %xmm5, %xmm14
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm11[4],xmm14[5],xmm11[5],xmm14[6],xmm11[6],xmm14[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm6
; SSE41-NEXT:    movdqa %xmm6, %xmm15
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm11[4],xmm15[5],xmm11[5],xmm15[6],xmm11[6],xmm15[7],xmm11[7]
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm7
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
; SSE41-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm11[4],xmm7[5],xmm11[5],xmm7[6],xmm11[6],xmm7[7],xmm11[7]
; SSE41-NEXT:    movdqa %xmm7, 240(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
; SSE41-NEXT:    movdqa %xmm15, 208(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
; SSE41-NEXT:    movdqa %xmm14, 176(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
; SSE41-NEXT:    movdqa %xmm13, 144(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
; SSE41-NEXT:    movdqa %xmm12, 112(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm10, 80(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm9, 48(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE41-NEXT:    movaps %xmm0, 16(%rdi)
; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero
; SSE41-NEXT:    movdqa %xmm0, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v64i16_lshr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq %rdi, %rax
; AVX2-NEXT:    vpmulhw %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm4 = 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:    vextracti128 $1, %ymm0, %xmm0
; 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:    vpmulhw %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm5 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-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
; AVX2-NEXT:    vpmulhw %ymm6, %ymm2, %ymm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm6 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT:    vpmulhw %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm7 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm3
; AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
; AVX2-NEXT:    vmovdqa %ymm3, 224(%rdi)
; AVX2-NEXT:    vmovdqa %ymm7, 192(%rdi)
; AVX2-NEXT:    vmovdqa %ymm2, 160(%rdi)
; AVX2-NEXT:    vmovdqa %ymm6, 128(%rdi)
; AVX2-NEXT:    vmovdqa %ymm1, 96(%rdi)
; AVX2-NEXT:    vmovdqa %ymm5, 64(%rdi)
; AVX2-NEXT:    vmovdqa %ymm0, 32(%rdi)
; AVX2-NEXT:    vmovdqa %ymm4, (%rdi)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: mulhsw_v64i16_lshr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm0, %ymm4
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm4[0],zero,ymm4[1],zero,ymm4[2],zero,ymm4[3],zero,ymm4[4],zero,ymm4[5],zero,ymm4[6],zero,ymm4[7],zero,ymm4[8],zero,ymm4[9],zero,ymm4[10],zero,ymm4[11],zero,ymm4[12],zero,ymm4[13],zero,ymm4[14],zero,ymm4[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm5 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vpmulhw %ymm3, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vextracti64x4 $1, %zmm3, %ymm0
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vpmulhw %ymm0, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: mulhsw_v64i16_lshr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm2, %zmm0, %zmm2
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
; AVX512BW-NEXT:    vpmulhw %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = sext <64 x i16> %a to <64 x i32>
  %b1 = sext <64 x i16> %b to <64 x i32>
  %c = mul <64 x i32> %a1, %b1
  %d = lshr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <64 x i32> %d
}

define <64 x i32> @mulhsw_v64i16_ashr(<64 x i16> %a, <64 x i16> %b) {
; SSE2-LABEL: mulhsw_v64i16_ashr:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movq %rdi, %rax
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm0
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; SSE2-NEXT:    psrad $16, %xmm8
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm0
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
; SSE2-NEXT:    psrad $16, %xmm9
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm1
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3]
; SSE2-NEXT:    psrad $16, %xmm10
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm2
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm3
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
; SSE2-NEXT:    psrad $16, %xmm11
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm4
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3]
; SSE2-NEXT:    psrad $16, %xmm12
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm4
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm5
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3]
; SSE2-NEXT:    psrad $16, %xmm13
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm5
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm6
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3]
; SSE2-NEXT:    psrad $16, %xmm14
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm6
; SSE2-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm7
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
; SSE2-NEXT:    psrad $16, %xmm15
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4,4,5,5,6,6,7,7]
; SSE2-NEXT:    psrad $16, %xmm7
; SSE2-NEXT:    movdqa %xmm7, 240(%rdi)
; SSE2-NEXT:    movdqa %xmm15, 224(%rdi)
; SSE2-NEXT:    movdqa %xmm6, 208(%rdi)
; SSE2-NEXT:    movdqa %xmm14, 192(%rdi)
; SSE2-NEXT:    movdqa %xmm5, 176(%rdi)
; SSE2-NEXT:    movdqa %xmm13, 160(%rdi)
; SSE2-NEXT:    movdqa %xmm4, 144(%rdi)
; SSE2-NEXT:    movdqa %xmm12, 128(%rdi)
; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE2-NEXT:    movdqa %xmm11, 96(%rdi)
; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE2-NEXT:    movdqa %xmm10, 64(%rdi)
; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE2-NEXT:    movdqa %xmm9, 32(%rdi)
; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE2-NEXT:    movdqa %xmm8, (%rdi)
; SSE2-NEXT:    retq
;
; SSE41-LABEL: mulhsw_v64i16_ashr:
; SSE41:       # %bb.0:
; SSE41-NEXT:    movq %rdi, %rax
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm0
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm8
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
; SSE41-NEXT:    pmovsxwd %xmm1, %xmm9
; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm1, %xmm1
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
; SSE41-NEXT:    pmovsxwd %xmm2, %xmm10
; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm2, %xmm2
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm3
; SSE41-NEXT:    pmovsxwd %xmm3, %xmm11
; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm3, %xmm3
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm4
; SSE41-NEXT:    pmovsxwd %xmm4, %xmm12
; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm4, %xmm4
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm5
; SSE41-NEXT:    pmovsxwd %xmm5, %xmm13
; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm5, %xmm5
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm6
; SSE41-NEXT:    pmovsxwd %xmm6, %xmm14
; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm6, %xmm6
; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm7
; SSE41-NEXT:    pmovsxwd %xmm7, %xmm15
; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,3,2,3]
; SSE41-NEXT:    pmovsxwd %xmm7, %xmm7
; SSE41-NEXT:    movdqa %xmm7, 240(%rdi)
; SSE41-NEXT:    movdqa %xmm15, 224(%rdi)
; SSE41-NEXT:    movdqa %xmm6, 208(%rdi)
; SSE41-NEXT:    movdqa %xmm14, 192(%rdi)
; SSE41-NEXT:    movdqa %xmm5, 176(%rdi)
; SSE41-NEXT:    movdqa %xmm13, 160(%rdi)
; SSE41-NEXT:    movdqa %xmm4, 144(%rdi)
; SSE41-NEXT:    movdqa %xmm12, 128(%rdi)
; SSE41-NEXT:    movdqa %xmm3, 112(%rdi)
; SSE41-NEXT:    movdqa %xmm11, 96(%rdi)
; SSE41-NEXT:    movdqa %xmm2, 80(%rdi)
; SSE41-NEXT:    movdqa %xmm10, 64(%rdi)
; SSE41-NEXT:    movdqa %xmm1, 48(%rdi)
; SSE41-NEXT:    movdqa %xmm9, 32(%rdi)
; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
; SSE41-NEXT:    movdqa %xmm8, (%rdi)
; SSE41-NEXT:    retq
;
; AVX2-LABEL: mulhsw_v64i16_ashr:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq %rdi, %rax
; AVX2-NEXT:    vpmulhw %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm4
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm0
; AVX2-NEXT:    vpmulhw %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm5
; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm1
; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm1
; AVX2-NEXT:    vpmulhw %ymm6, %ymm2, %ymm2
; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm6
; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm2
; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm2
; AVX2-NEXT:    vpmulhw %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    vpmovsxwd %xmm3, %ymm7
; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm3
; AVX2-NEXT:    vpmovsxwd %xmm3, %ymm3
; AVX2-NEXT:    vmovdqa %ymm3, 224(%rdi)
; AVX2-NEXT:    vmovdqa %ymm7, 192(%rdi)
; AVX2-NEXT:    vmovdqa %ymm2, 160(%rdi)
; AVX2-NEXT:    vmovdqa %ymm6, 128(%rdi)
; AVX2-NEXT:    vmovdqa %ymm1, 96(%rdi)
; AVX2-NEXT:    vmovdqa %ymm5, 64(%rdi)
; AVX2-NEXT:    vmovdqa %ymm0, 32(%rdi)
; AVX2-NEXT:    vmovdqa %ymm4, (%rdi)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: mulhsw_v64i16_ashr:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm0, %ymm4
; AVX512F-NEXT:    vpmovsxwd %ymm4, %zmm4
; AVX512F-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
; AVX512F-NEXT:    vpmulhw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm5
; AVX512F-NEXT:    vpmulhw %ymm3, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm2
; AVX512F-NEXT:    vextracti64x4 $1, %zmm3, %ymm0
; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512F-NEXT:    vpmulhw %ymm0, %ymm1, %ymm0
; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm3
; AVX512F-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512F-NEXT:    vmovdqa64 %zmm5, %zmm1
; AVX512F-NEXT:    retq
;
; AVX512BW-LABEL: mulhsw_v64i16_ashr:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpmulhw %zmm2, %zmm0, %zmm2
; AVX512BW-NEXT:    vpmovsxwd %ymm2, %zmm0
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm2, %ymm2
; AVX512BW-NEXT:    vpmovsxwd %ymm2, %zmm4
; AVX512BW-NEXT:    vpmulhw %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm2
; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
; AVX512BW-NEXT:    vpmovsxwd %ymm1, %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512BW-NEXT:    retq
  %a1 = sext <64 x i16> %a to <64 x i32>
  %b1 = sext <64 x i16> %b to <64 x i32>
  %c = mul <64 x i32> %a1, %b1
  %d = ashr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
  ret <64 x i32> %d
}

define <8 x i64> @zext_mulhuw_v8i16_lshr_i64(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: zext_mulhuw_v8i16_lshr_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm0, %xmm3
; SSE2-NEXT:    pmulhuw %xmm1, %xmm3
; SSE2-NEXT:    pxor %xmm4, %xmm4
; SSE2-NEXT:    movdqa %xmm3, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE2-NEXT:    movdqa %xmm1, %xmm0
; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
; SSE2-NEXT:    movdqa %xmm3, %xmm2
; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: zext_mulhuw_v8i16_lshr_i64:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhuw %xmm1, %xmm0
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm4 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm3 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; SSE41-NEXT:    movdqa %xmm4, %xmm0
; SSE41-NEXT:    retq
;
; AVX2-LABEL: zext_mulhuw_v8i16_lshr_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm1
; AVX2-NEXT:    vpmovzxwq {{.*#+}} ymm0 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX2-NEXT:    vpmovzxwq {{.*#+}} ymm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
; AVX2-NEXT:    retq
;
; AVX512-LABEL: zext_mulhuw_v8i16_lshr_i64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0
; AVX512-NEXT:    vpmovzxwq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; AVX512-NEXT:    retq
  %a1 = zext <8 x i16> %a to <8 x i64>
  %b1 = zext <8 x i16> %b to <8 x i64>
  %c = mul <8 x i64> %a1, %b1
  %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
  ret <8 x i64> %d
}

define <8 x i64> @sext_mulhsw_v8i16_lshr_i64(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: sext_mulhsw_v8i16_lshr_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    movdqa %xmm0, %xmm3
; SSE2-NEXT:    pmulhw %xmm1, %xmm3
; SSE2-NEXT:    pxor %xmm4, %xmm4
; SSE2-NEXT:    movdqa %xmm3, %xmm1
; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE2-NEXT:    movdqa %xmm1, %xmm0
; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
; SSE2-NEXT:    movdqa %xmm3, %xmm2
; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; SSE2-NEXT:    retq
;
; SSE41-LABEL: sext_mulhsw_v8i16_lshr_i64:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm4 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm3 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; SSE41-NEXT:    movdqa %xmm4, %xmm0
; SSE41-NEXT:    retq
;
; AVX2-LABEL: sext_mulhsw_v8i16_lshr_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %xmm1, %xmm0, %xmm1
; AVX2-NEXT:    vpmovzxwq {{.*#+}} ymm0 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX2-NEXT:    vpmovzxwq {{.*#+}} ymm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
; AVX2-NEXT:    retq
;
; AVX512-LABEL: sext_mulhsw_v8i16_lshr_i64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX512-NEXT:    vpmovzxwq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; AVX512-NEXT:    retq
  %a1 = sext <8 x i16> %a to <8 x i64>
  %b1 = sext <8 x i16> %b to <8 x i64>
  %c = mul <8 x i64> %a1, %b1
  %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
  ret <8 x i64> %d
}

define <8 x i64> @sext_mulhsw_v8i16_ashr_i64(<8 x i16> %a, <8 x i16> %b) {
; SSE2-LABEL: sext_mulhsw_v8i16_ashr_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pmulhw %xmm1, %xmm0
; 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:    pxor %xmm5, %xmm5
; SSE2-NEXT:    pxor %xmm2, %xmm2
; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
; SSE2-NEXT:    movdqa %xmm1, %xmm4
; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
; SSE2-NEXT:    psrad $16, %xmm3
; SSE2-NEXT:    pcmpgtd %xmm3, %xmm5
; SSE2-NEXT:    movdqa %xmm3, %xmm2
; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; SSE2-NEXT:    movdqa %xmm4, %xmm0
; SSE2-NEXT:    retq
;
; SSE41-LABEL: sext_mulhsw_v8i16_ashr_i64:
; SSE41:       # %bb.0:
; SSE41-NEXT:    pmulhw %xmm1, %xmm0
; SSE41-NEXT:    pmovsxwq %xmm0, %xmm4
; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE41-NEXT:    pmovsxwq %xmm1, %xmm1
; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
; SSE41-NEXT:    pmovsxwq %xmm2, %xmm2
; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; SSE41-NEXT:    pmovsxwq %xmm0, %xmm3
; SSE41-NEXT:    movdqa %xmm4, %xmm0
; SSE41-NEXT:    retq
;
; AVX2-LABEL: sext_mulhsw_v8i16_ashr_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpmulhw %xmm1, %xmm0, %xmm1
; AVX2-NEXT:    vpmovsxwq %xmm1, %ymm0
; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX2-NEXT:    vpmovsxwq %xmm1, %ymm1
; AVX2-NEXT:    retq
;
; AVX512-LABEL: sext_mulhsw_v8i16_ashr_i64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0
; AVX512-NEXT:    vpmovsxwq %xmm0, %zmm0
; AVX512-NEXT:    retq
  %a1 = sext <8 x i16> %a to <8 x i64>
  %b1 = sext <8 x i16> %b to <8 x i64>
  %c = mul <8 x i64> %a1, %b1
  %d = ashr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
  ret <8 x i64> %d
}

define <8 x i16> @sse2_pmulh_w_const(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: sse2_pmulh_w_const:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [0,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    retq
;
; AVX-LABEL: sse2_pmulh_w_const:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovaps {{.*#+}} xmm0 = [0,65535,65535,65535,65535,65535,65535,0]
; AVX-NEXT:    retq
  %res = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 -1, i16 -2, i16 -3, i16 -4, i16 -5, i16 -6, i16 -7, i16 0>, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
  ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>)

define <8 x i16> @sse2_pmulhu_w_const(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: sse2_pmulhu_w_const:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [0,0,1,2,3,4,5,0]
; SSE-NEXT:    retq
;
; AVX-LABEL: sse2_pmulhu_w_const:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovaps {{.*#+}} xmm0 = [0,0,1,2,3,4,5,0]
; AVX-NEXT:    retq
  %res = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> <i16 -1, i16 -2, i16 -3, i16 -4, i16 -5, i16 -6, i16 -7, i16 0>, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
  ret <8 x i16> %res
}
declare <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16>, <8 x i16>)