[X86][AVX] Add PR51226 test case

Author: RKSimon

llvm/test/CodeGen/X86/subvector-broadcast.ll

@@ -1752,3 +1752,80 @@ define <8 x double> @broadcast_v8f64_v2f64_0uuu0101(<2 x double>* %vp) {
   %res = shufflevector <2 x double> %vec, <2 x double> undef, <8 x i32> <i32 0, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 0, i32 1>
   ret <8 x double> %res
 }
+
+define void @PR51226() {
+; X86-AVX1-LABEL: PR51226:
+; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X86-AVX1-NEXT:    vpslld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X86-AVX1-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X86-AVX1-NEXT:    vmovups %ymm0, (%eax)
+; X86-AVX1-NEXT:    vzeroupper
+; X86-AVX1-NEXT:    retl
+;
+; X86-AVX2-LABEL: PR51226:
+; X86-AVX2:       # %bb.0:
+; X86-AVX2-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X86-AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX2-NEXT:    vpslld $16, %ymm0, %ymm0
+; X86-AVX2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X86-AVX2-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X86-AVX2-NEXT:    vmovups %ymm0, (%eax)
+; X86-AVX2-NEXT:    vzeroupper
+; X86-AVX2-NEXT:    retl
+;
+; X86-AVX512-LABEL: PR51226:
+; X86-AVX512:       # %bb.0:
+; X86-AVX512-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X86-AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512-NEXT:    vpslld $16, %ymm0, %ymm0
+; X86-AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X86-AVX512-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X86-AVX512-NEXT:    vmovups %ymm0, (%eax)
+; X86-AVX512-NEXT:    vzeroupper
+; X86-AVX512-NEXT:    retl
+;
+; X64-AVX1-LABEL: PR51226:
+; X64-AVX1:       # %bb.0:
+; X64-AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X64-AVX1-NEXT:    vpslld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X64-AVX1-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X64-AVX1-NEXT:    vmovups %ymm0, (%rax)
+; X64-AVX1-NEXT:    vzeroupper
+; X64-AVX1-NEXT:    retq
+;
+; X64-AVX2-LABEL: PR51226:
+; X64-AVX2:       # %bb.0:
+; X64-AVX2-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X64-AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX2-NEXT:    vpslld $16, %ymm0, %ymm0
+; X64-AVX2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X64-AVX2-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X64-AVX2-NEXT:    vmovups %ymm0, (%rax)
+; X64-AVX2-NEXT:    vzeroupper
+; X64-AVX2-NEXT:    retq
+;
+; X64-AVX512-LABEL: PR51226:
+; X64-AVX512:       # %bb.0:
+; X64-AVX512-NEXT:    vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X64-AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X64-AVX512-NEXT:    vpslld $16, %ymm0, %ymm0
+; X64-AVX512-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; X64-AVX512-NEXT:    vminps %ymm1, %ymm0, %ymm0
+; X64-AVX512-NEXT:    vmovups %ymm0, (%rax)
+; X64-AVX512-NEXT:    vzeroupper
+; X64-AVX512-NEXT:    retq
+  %i = load <4 x i16>, <4 x i16>* undef, align 8
+  %i1 = zext <4 x i16> %i to <4 x i32>
+  %i2 = shl nuw <4 x i32> %i1, <i32 16, i32 16, i32 16, i32 16>
+  %i3 = bitcast <4 x i32> %i2 to <4 x float>
+  %shuffle99 = shufflevector <4 x float> %i3, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+  %i4 = fcmp reassoc nsz contract ogt <8 x float> zeroinitializer, %shuffle99
+  %i5 = select <8 x i1> %i4, <8 x float> %shuffle99, <8 x float> zeroinitializer
+  store <8 x float> %i5, <8 x float>* undef, align 16
+  ret void
+}