[llvm] dca7b11 - [X86] Add tests showing failure to reduce the vector width of vpmaddwd/vpmaddubsw/pmulhrsw nodes (#180728)

[via llvm-commits]

Author: Simon Pilgrim
Date: 2026-02-10T12:45:21Z
New Revision: dca7b11a326f523a93268dd17e969d1c26bcf40e

URL: https://github.com/llvm/llvm-project/commit/dca7b11a326f523a93268dd17e969d1c26bcf40e
DIFF: https://github.com/llvm/llvm-project/commit/dca7b11a326f523a93268dd17e969d1c26bcf40e.diff

LOG: [X86] Add tests showing failure to reduce the vector width of vpmaddwd/vpmaddubsw/pmulhrsw nodes (#180728)

Missing demanded elts handling

Added: 
    llvm/test/CodeGen/X86/vector-target-demanded-elts.ll

Modified: 
    

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/vector-target-demanded-elts.ll b/llvm/test/CodeGen/X86/vector-target-demanded-elts.ll
new file mode 100644
index 0000000000000..84dbb03f5e292
--- /dev/null
+++ b/llvm/test/CodeGen/X86/vector-target-demanded-elts.ll
@@ -0,0 +1,71 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v4 | FileCheck %s
+
+define <4 x i32> @demanded_low_pmaddwd256_128(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: demanded_low_pmaddwd256_128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
+  %res = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %a0, <16 x i16> %a1)
+  %ext = shufflevector <8 x i32> %res, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  ret <4 x i32> %ext
+}
+
+define <8 x i32> @demanded_low_pmaddwd512_256(<32 x i16> %x0, <32 x i16> %x1) {
+; CHECK-LABEL: demanded_low_pmaddwd512_256:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmaddwd %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; CHECK-NEXT:    retq
+  %res = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %x0, <32 x i16> %x1)
+  %ext = shufflevector <16 x i32> %res, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i32> %ext
+}
+
+define <8 x i16> @demanded_low_pmaddubs256_128(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: demanded_low_pmaddubs256_128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
+  %res = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %a0, <32 x i8> %a1)
+  %ext = shufflevector <16 x i16> %res, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i16> %ext
+}
+
+define <16 x i16> @demanded_low_pmaddubs512_256(<64 x i8> %x0, <64 x i8> %x1) {
+; CHECK-LABEL: demanded_low_pmaddubs512_256:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; CHECK-NEXT:    retq
+  %res = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %x0, <64 x i8> %x1)
+  %ext = shufflevector <32 x i16> %res, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <16 x i16> %ext
+}
+
+define <8 x i16> @demanded_low_pmulhrsw256_128(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: demanded_low_pmulhrsw256_128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
+  %res = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %a0, <16 x i16> %a1)
+  %ext = shufflevector <16 x i16> %res, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i16> %ext
+}
+
+define <16 x i16> @demanded_low_pmulhrsw512_256(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
+; CHECK-LABEL: demanded_low_pmulhrsw512_256:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpmulhrsw %zmm1, %zmm0, %zmm0
+; CHECK-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; CHECK-NEXT:    retq
+  %res = call <32 x i16> @llvm.x86.avx512.pmul.hr.sw.512(<32 x i16> %x0, <32 x i16> %x1)
+  %ext = shufflevector <32 x i16> %res, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <16 x i16> %ext
+}