; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -mtriple=x86_64-unknown-unknown -S | FileCheck %s

;
; UNDEF Elts
;

define <4 x i32> @undef_pmaddwd_128(<8 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_128(
; CHECK-NEXT:    ret <4 x i32> zeroinitializer
;
  %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %a0, <8 x i16> undef)
  ret <4 x i32> %1
}

define <4 x i32> @undef_pmaddwd_128_commute(<8 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_128_commute(
; CHECK-NEXT:    ret <4 x i32> zeroinitializer
;
  %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> undef, <8 x i16> %a0)
  ret <4 x i32> %1
}

define <8 x i32> @undef_pmaddwd_256(<16 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_256(
; CHECK-NEXT:    ret <8 x i32> zeroinitializer
;
  %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %a0, <16 x i16> undef)
  ret <8 x i32> %1
}

define <8 x i32> @undef_pmaddwd_256_commute(<16 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_256_commute(
; CHECK-NEXT:    ret <8 x i32> zeroinitializer
;
  %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> undef, <16 x i16> %a0)
  ret <8 x i32> %1
}

define <16 x i32> @undef_pmaddwd_512(<32 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_512(
; CHECK-NEXT:    ret <16 x i32> zeroinitializer
;
  %1 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> undef)
  ret <16 x i32> %1
}

define <16 x i32> @undef_pmaddwd_512_commute(<32 x i16> %a0) {
; CHECK-LABEL: @undef_pmaddwd_512_commute(
; CHECK-NEXT:    ret <16 x i32> zeroinitializer
;
  %1 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> undef, <32 x i16> %a0)
  ret <16 x i32> %1
}

;
; Zero Elts
;

define <4 x i32> @zero_pmaddwd_128(<8 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_128(
; CHECK-NEXT:    ret <4 x i32> zeroinitializer
;
  %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %a0, <8 x i16> zeroinitializer)
  ret <4 x i32> %1
}

define <4 x i32> @zero_pmaddwd_128_commute(<8 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_128_commute(
; CHECK-NEXT:    ret <4 x i32> zeroinitializer
;
  %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> zeroinitializer, <8 x i16> %a0)
  ret <4 x i32> %1
}

define <8 x i32> @zero_pmaddwd_256(<16 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_256(
; CHECK-NEXT:    ret <8 x i32> zeroinitializer
;
  %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %a0, <16 x i16> zeroinitializer)
  ret <8 x i32> %1
}

define <8 x i32> @zero_pmaddwd_256_commute(<16 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_256_commute(
; CHECK-NEXT:    ret <8 x i32> zeroinitializer
;
  %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> zeroinitializer, <16 x i16> %a0)
  ret <8 x i32> %1
}

define <16 x i32> @zero_pmaddwd_512(<32 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_512(
; CHECK-NEXT:    ret <16 x i32> zeroinitializer
;
  %1 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> zeroinitializer)
  ret <16 x i32> %1
}

define <16 x i32> @zero_pmaddwd_512_commute(<32 x i16> %a0) {
; CHECK-LABEL: @zero_pmaddwd_512_commute(
; CHECK-NEXT:    ret <16 x i32> zeroinitializer
;
  %1 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> zeroinitializer, <32 x i16> %a0)
  ret <16 x i32> %1
}

;
; Constant Folding
;

define <4 x i32> @fold_pmaddwd_128() {
; CHECK-LABEL: @fold_pmaddwd_128(
; CHECK-NEXT:    ret <4 x i32> <i32 19, i32 -229364, i32 -21, i32 -491429>
;
  %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> <i16 -1, i16 2, i16 3, i16 -4, i16 -5, i16 6, i16 7, i16 -8>, <8 x i16> <i16 -5, i16 7, i16 -32768, i16 32765, i16 -9, i16 -11, i16 -32763, i16 32761>)
  ret <4 x i32> %1
}

define <8 x i32> @fold_pmaddwd_256() {
; CHECK-LABEL: @fold_pmaddwd_256(
; CHECK-NEXT:    ret <8 x i32> <i32 -7, i32 32762, i32 91, i32 32750, i32 -239, i32 687938, i32 -451, i32 -32756>
;
  %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> <i16 0, i16 -1, i16 2, i16 3, i16 -4, i16 -5, i16 6, i16 7, i16 -8, i16 9, i16 -10, i16 11, i16 -12, i16 13, i16 -14, i16 -15>, <16 x i16> <i16 -5, i16 7, i16 -32768, i16 32766, i16 -9, i16 -11, i16 -32764, i16 32762, i16 13, i16 -15, i16 -32760, i16 32758, i16 17, i16 -19, i16 -32756, i16 32756>)
  ret <8 x i32> %1
}

define <16 x i32> @fold_pmaddwd_512() {
; CHECK-LABEL: @fold_pmaddwd_512(
; CHECK-NEXT:    ret <16 x i32> <i32 -7, i32 32762, i32 91, i32 32750, i32 -239, i32 687938, i32 -451, i32 -32756, i32 -7, i32 32762, i32 91, i32 32750, i32 -239, i32 687938, i32 -451, i32 -32756>
;
  %1 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> <i16 0, i16 -1, i16 2, i16 3, i16 -4, i16 -5, i16 6, i16 7, i16 -8, i16 9, i16 -10, i16 11, i16 -12, i16 13, i16 -14, i16 -15, i16 -5, i16 7, i16 -32768, i16 32766, i16 -9, i16 -11, i16 -32764, i16 32762, i16 13, i16 -15, i16 -32760, i16 32758, i16 17, i16 -19, i16 -32756, i16 32756>, <32 x i16> <i16 -5, i16 7, i16 -32768, i16 32766, i16 -9, i16 -11, i16 -32764, i16 32762, i16 13, i16 -15, i16 -32760, i16 32758, i16 17, i16 -19, i16 -32756, i16 32756, i16 0, i16 -1, i16 2, i16 3, i16 -4, i16 -5, i16 6, i16 7, i16 -8, i16 9, i16 -10, i16 11, i16 -12, i16 13, i16 -14, i16 -15>)
  ret <16 x i32> %1
}

;
; Demanded Elts
;

define <4 x i32> @elts_pmaddwd_128(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: @elts_pmaddwd_128(
; CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]])
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
;
  %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2>
  %2 = shufflevector <8 x i16> %a1, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
  %3 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %1, <8 x i16> %2)
  %4 = shufflevector <4 x i32> %3, <4 x i32> poison, <4 x i32> zeroinitializer
  ret <4 x i32> %4
}

define <8 x i32> @elts_pmaddwd_256(<16 x i16> %a0, <16 x i16> %a1) {
; CHECK-LABEL: @elts_pmaddwd_256(
; CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> [[A0:%.*]], <16 x i16> [[A1:%.*]])
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> poison, <8 x i32> zeroinitializer
; CHECK-NEXT:    ret <8 x i32> [[TMP2]]
;
  %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %2 = shufflevector <16 x i16> %a1, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %3 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %1, <16 x i16> %2)
  %4 = shufflevector <8 x i32> %3, <8 x i32> poison, <8 x i32> zeroinitializer
  ret <8 x i32> %4
}

define <16 x i32> @elts_pmaddwd_512(<32 x i16> %a0, <32 x i16> %a1) {
; CHECK-LABEL: @elts_pmaddwd_512(
; CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> [[A0:%.*]], <32 x i16> [[A1:%.*]])
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <16 x i32> [[TMP1]], <16 x i32> poison, <16 x i32> zeroinitializer
; CHECK-NEXT:    ret <16 x i32> [[TMP2]]
;
  %1 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 0, i32 1, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
  %2 = shufflevector <32 x i16> %a1, <32 x i16> undef, <32 x i32> <i32 0, i32 1, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
  %3 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %1, <32 x i16> %2)
  %4 = shufflevector <16 x i32> %3, <16 x i32> poison, <16 x i32> zeroinitializer
  ret <16 x i32> %4
}