[llvm] 8692e27 - [X86][AVX512] Add PR52500 vXi1 broadcast test case

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2022-03-21T17:25:29Z
New Revision: 8692e27ad671c31514e2ab828974924c149e741b

URL: https://github.com/llvm/llvm-project/commit/8692e27ad671c31514e2ab828974924c149e741b
DIFF: https://github.com/llvm/llvm-project/commit/8692e27ad671c31514e2ab828974924c149e741b.diff

LOG: [X86][AVX512] Add PR52500 vXi1 broadcast test case

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/vector-shuffle-v1.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/vector-shuffle-v1.ll b/llvm/test/CodeGen/X86/vector-shuffle-v1.ll
index 3eadf8812156c..9af454b3cc984 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-v1.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-v1.ll
@@ -908,3 +908,64 @@ define i64 @shuf64i1_zero(i64 %a) {
   %d = bitcast <64 x i1> %c to i64
   ret i64 %d
 }
+
+define <16 x i1> @PR52500(<16 x i1> %msk, i32 %in) {
+; AVX512F-LABEL: PR52500:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; AVX512F-NEXT:    vmovd %edi, %xmm0
+; AVX512F-NEXT:    movl $789, %eax # imm = 0x315
+; AVX512F-NEXT:    vmovd %eax, %xmm1
+; AVX512F-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
+; AVX512F-NEXT:    vptestnmd %zmm0, %zmm0, %k2
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; AVX512F-NEXT:    vpbroadcastd %xmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1 {%k1}
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512VL-LABEL: PR52500:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpmovsxbd %xmm0, %zmm0
+; AVX512VL-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512VL-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; AVX512VL-NEXT:    vmovd %edi, %xmm0
+; AVX512VL-NEXT:    movl $789, %eax # imm = 0x315
+; AVX512VL-NEXT:    vmovd %eax, %xmm1
+; AVX512VL-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
+; AVX512VL-NEXT:    vptestnmd %zmm0, %zmm0, %k2
+; AVX512VL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; AVX512VL-NEXT:    vpbroadcastd %xmm0, %zmm0
+; AVX512VL-NEXT:    vptestmd %zmm0, %zmm0, %k1 {%k1}
+; AVX512VL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VL-NEXT:    vzeroupper
+; AVX512VL-NEXT:    retq
+;
+; VL_BW_DQ-LABEL: PR52500:
+; VL_BW_DQ:       # %bb.0:
+; VL_BW_DQ-NEXT:    vpsllw $7, %xmm0, %xmm0
+; VL_BW_DQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; VL_BW_DQ-NEXT:    vmovd %edi, %xmm2
+; VL_BW_DQ-NEXT:    movl $789, %eax # imm = 0x315
+; VL_BW_DQ-NEXT:    vmovd %eax, %xmm3
+; VL_BW_DQ-NEXT:    vpmulld %xmm3, %xmm2, %xmm2
+; VL_BW_DQ-NEXT:    vptestnmd %zmm2, %zmm2, %k0
+; VL_BW_DQ-NEXT:    vpmovm2d %k0, %zmm2
+; VL_BW_DQ-NEXT:    vpbroadcastd %xmm2, %zmm2
+; VL_BW_DQ-NEXT:    vpmovd2m %zmm2, %k1
+; VL_BW_DQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %k0 {%k1}
+; VL_BW_DQ-NEXT:    vpmovm2b %k0, %xmm0
+; VL_BW_DQ-NEXT:    vzeroupper
+; VL_BW_DQ-NEXT:    retq
+  %insrt = insertelement <16 x i32> undef, i32 %in, i32 0
+  %mul = mul <16 x i32> %insrt, <i32 789, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+  %eq = icmp eq <16 x i32> %mul, zeroinitializer
+  %cmp1 = shufflevector <16 x i1> %eq, <16 x i1> poison, <16 x i32> zeroinitializer
+  %and = and <16 x i1> %cmp1, %msk
+  ret <16 x i1> %and
+}