[llvm] 7e3aa70 - [X86] Add test coverage for PR26515 / Issue #26889

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2022-05-08T18:19:04+01:00
New Revision: 7e3aa70668aeb28e0d76c0d2807300eab337e0a7

URL: https://github.com/llvm/llvm-project/commit/7e3aa70668aeb28e0d76c0d2807300eab337e0a7
DIFF: https://github.com/llvm/llvm-project/commit/7e3aa70668aeb28e0d76c0d2807300eab337e0a7.diff

LOG: [X86] Add test coverage for PR26515 / Issue #26889

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll b/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
index fd0fd5872229..9fca269ead97 100644
--- a/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
+++ b/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
@@ -1473,3 +1473,43 @@ define <2 x double> @add_sd_mask(<2 x double> %a, <2 x double> %b, <2 x double>
   %8 = insertelement <2 x double> %a, double %7, i64 0
   ret <2 x double> %8
 }
+
+define float @PR26515(<4 x float> %0) nounwind {
+; X86-SSE-LABEL: PR26515:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    pushl %eax
+; X86-SSE-NEXT:    movaps %xmm0, %xmm1
+; X86-SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; X86-SSE-NEXT:    addss %xmm0, %xmm1
+; X86-SSE-NEXT:    movss %xmm1, (%esp)
+; X86-SSE-NEXT:    flds (%esp)
+; X86-SSE-NEXT:    popl %eax
+; X86-SSE-NEXT:    retl
+;
+; X86-AVX-LABEL: PR26515:
+; X86-AVX:       # %bb.0:
+; X86-AVX-NEXT:    pushl %eax
+; X86-AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-AVX-NEXT:    vaddss %xmm0, %xmm1, %xmm0
+; X86-AVX-NEXT:    vmovss %xmm0, (%esp)
+; X86-AVX-NEXT:    flds (%esp)
+; X86-AVX-NEXT:    popl %eax
+; X86-AVX-NEXT:    retl
+;
+; X64-SSE-LABEL: PR26515:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movaps %xmm0, %xmm1
+; X64-SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; X64-SSE-NEXT:    addss %xmm1, %xmm0
+; X64-SSE-NEXT:    retq
+;
+; X64-AVX-LABEL: PR26515:
+; X64-AVX:       # %bb.0:
+; X64-AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-AVX-NEXT:    vaddss %xmm0, %xmm1, %xmm0
+; X64-AVX-NEXT:    retq
+  %2 = shufflevector <4 x float> %0, <4 x float> poison, <4 x i32> <i32 2, i32 undef, i32 undef, i32 undef>
+  %3 = fadd <4 x float> %2, %0
+  %4 = extractelement <4 x float> %3, i64 0
+  ret float %4
+}