[llvm] d6ee91b - [X86] Add test case for Issue #77805

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2024-01-16T17:23:45Z
New Revision: d6ee91b1100a3356506643da37cbd6891a663db9

URL: https://github.com/llvm/llvm-project/commit/d6ee91b1100a3356506643da37cbd6891a663db9
DIFF: https://github.com/llvm/llvm-project/commit/d6ee91b1100a3356506643da37cbd6891a663db9.diff

LOG: [X86] Add test case for Issue #77805

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/fminimum-fmaximum.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/fminimum-fmaximum.ll b/llvm/test/CodeGen/X86/fminimum-fmaximum.ll
index 8905d2bce5e92a..a279f1e937e695 100644
--- a/llvm/test/CodeGen/X86/fminimum-fmaximum.ll
+++ b/llvm/test/CodeGen/X86/fminimum-fmaximum.ll
@@ -1322,3 +1322,47 @@ define <4 x float> @test_fmaximum_vector_zero(<4 x float> %x) {
   %r = call <4 x float> @llvm.maximum.v4f32(<4 x float> %x, <4 x float> <float 0., float 0., float 0., float 0.>)
   ret <4 x float> %r
 }
+
+; PR77805: Check that signed zeroes are handled correctly in this case (FIXME)
+define <4 x float> @test_fmaximum_v4f32_splat(<4 x float> %x, float %y) {
+; SSE2-LABEL: test_fmaximum_v4f32_splat:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE2-NEXT:    movaps %xmm0, %xmm2
+; SSE2-NEXT:    cmpunordps %xmm0, %xmm2
+; SSE2-NEXT:    movaps %xmm0, %xmm3
+; SSE2-NEXT:    andps %xmm2, %xmm3
+; SSE2-NEXT:    maxps %xmm1, %xmm0
+; SSE2-NEXT:    andnps %xmm0, %xmm2
+; SSE2-NEXT:    orps %xmm3, %xmm2
+; SSE2-NEXT:    movaps %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; AVX1-LABEL: test_fmaximum_v4f32_splat:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; AVX1-NEXT:    vmaxps %xmm1, %xmm0, %xmm1
+; AVX1-NEXT:    vcmpunordps %xmm0, %xmm0, %xmm2
+; AVX1-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX512-LABEL: test_fmaximum_v4f32_splat:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX512-NEXT:    vmaxps %xmm1, %xmm0, %xmm1
+; AVX512-NEXT:    vcmpunordps %xmm0, %xmm0, %xmm2
+; AVX512-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; AVX512-NEXT:    retq
+;
+; X86-LABEL: test_fmaximum_v4f32_splat:
+; X86:       # %bb.0:
+; X86-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %xmm1
+; X86-NEXT:    vmaxps %xmm1, %xmm0, %xmm1
+; X86-NEXT:    vcmpunordps %xmm0, %xmm0, %xmm2
+; X86-NEXT:    vblendvps %xmm2, %xmm0, %xmm1, %xmm0
+; X86-NEXT:    retl
+  %splatinsert = insertelement <4 x float> poison, float %y, i64 0
+  %vec = shufflevector <4 x float> %splatinsert, <4 x float> poison, <4 x i32> zeroinitializer
+  %r = call <4 x float> @llvm.maximum.v4f32(<4 x float> %x, <4 x float> %vec) readnone
+  ret <4 x float> %r
+}