[llvm] c7fb1cd - [X86] Add test case for Issue #58546

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2022-10-23T14:27:37+01:00
New Revision: c7fb1cdcbaf3c0e2ae4480ecf7b8516c9ecc5dd3

URL: https://github.com/llvm/llvm-project/commit/c7fb1cdcbaf3c0e2ae4480ecf7b8516c9ecc5dd3
DIFF: https://github.com/llvm/llvm-project/commit/c7fb1cdcbaf3c0e2ae4480ecf7b8516c9ecc5dd3.diff

LOG: [X86] Add test case for Issue #58546

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/bitcast-vector-bool.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
index 8ce36dd3c8ef..94901e665d56 100644
--- a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
+++ b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
@@ -640,3 +640,77 @@ define i64 @bitcast_v128i8_to_v2i64(<128 x i8> %a0) nounwind {
   %5 = add i64 %3, %4
   ret i64 %5
 }
+
+define [2 x i8] @PR58546(<16 x float> %a0) {
+; SSE2-SSSE3-LABEL: PR58546:
+; SSE2-SSSE3:       # %bb.0:
+; SSE2-SSSE3-NEXT:    xorps %xmm4, %xmm4
+; SSE2-SSSE3-NEXT:    cmpunordps %xmm4, %xmm3
+; SSE2-SSSE3-NEXT:    cmpunordps %xmm4, %xmm2
+; SSE2-SSSE3-NEXT:    packssdw %xmm3, %xmm2
+; SSE2-SSSE3-NEXT:    cmpunordps %xmm4, %xmm1
+; SSE2-SSSE3-NEXT:    cmpunordps %xmm4, %xmm0
+; SSE2-SSSE3-NEXT:    packssdw %xmm1, %xmm0
+; SSE2-SSSE3-NEXT:    packsswb %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-SSSE3-NEXT:    movl %eax, %edx
+; SSE2-SSSE3-NEXT:    shrl $8, %edx
+; SSE2-SSSE3-NEXT:    # kill: def $al killed $al killed $eax
+; SSE2-SSSE3-NEXT:    # kill: def $dl killed $dl killed $edx
+; SSE2-SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: PR58546:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vcmpunordps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpackssdw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vcmpunordps %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    shrl $8, %edx
+; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    # kill: def $dl killed $dl killed $edx
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: PR58546:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vcmpunordps %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vcmpunordps %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %edx
+; AVX2-NEXT:    shrl $8, %edx
+; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    # kill: def $dl killed $dl killed $edx
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: PR58546:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX512-NEXT:    vcmpunordps %zmm1, %zmm0, %k0
+; AVX512-NEXT:    kmovw %k0, -{{[0-9]+}}(%rsp)
+; AVX512-NEXT:    vmovdqa -{{[0-9]+}}(%rsp), %xmm0
+; AVX512-NEXT:    vmovd %xmm0, %eax
+; AVX512-NEXT:    vpextrb $1, %xmm0, %edx
+; AVX512-NEXT:    # kill: def $al killed $al killed $eax
+; AVX512-NEXT:    # kill: def $dl killed $dl killed $edx
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
+  %1 = fcmp uno <16 x float> %a0, zeroinitializer
+  %2 = bitcast <16 x i1> %1 to <2 x i8>
+  %3 = extractelement <2 x i8> %2, i64 0
+  %4 = extractelement <2 x i8> %2, i64 1
+  %5 = insertvalue [2 x i8] poison, i8 %3, 0
+  %6 = insertvalue [2 x i8] %5, i8 %4, 1
+  ret [2 x i8] %6
+}