1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-linux -opaque-pointers -mcpu=ivybridge -mattr=+avx2 | FileCheck %s --check-prefixes=IVB
3; RUN: llc < %s -mtriple=x86_64-linux -opaque-pointers -mcpu=haswell | FileCheck %s --check-prefixes=HSW
4
5define <2 x i64> @PR55158(ptr %0) {
6; IVB-LABEL: PR55158:
7; IVB:       # %bb.0:
8; IVB-NEXT:    vmovdqa 64(%rdi), %xmm0
9; IVB-NEXT:    vmovdqa 128(%rdi), %xmm1
10; IVB-NEXT:    vpmovsxbd (%rdi), %xmm2
11; IVB-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
12; IVB-NEXT:    vphsubw %xmm0, %xmm0, %xmm0
13; IVB-NEXT:    vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
14; IVB-NEXT:    vpxor %xmm2, %xmm2, %xmm2
15; IVB-NEXT:    vpsrlvq %xmm1, %xmm2, %xmm1
16; IVB-NEXT:    vpunpckhqdq {{.*#+}} xmm0 = xmm1[1],xmm0[1]
17; IVB-NEXT:    retq
18;
19; HSW-LABEL: PR55158:
20; HSW:       # %bb.0:
21; HSW-NEXT:    vmovdqa 64(%rdi), %xmm0
22; HSW-NEXT:    vmovdqa 128(%rdi), %xmm1
23; HSW-NEXT:    vpmovsxbd (%rdi), %xmm2
24; HSW-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
25; HSW-NEXT:    vphsubw %xmm0, %xmm0, %xmm0
26; HSW-NEXT:    vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
27; HSW-NEXT:    vpandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm2
28; HSW-NEXT:    vpsrlvq %xmm1, %xmm2, %xmm1
29; HSW-NEXT:    vpunpckhqdq {{.*#+}} xmm0 = xmm1[1],xmm0[1]
30; HSW-NEXT:    retq
31  %2 = load <16 x i8>, ptr %0, align 16
32  %3 = getelementptr inbounds i32, ptr %0, i64 16
33  %4 = load <8 x i16>, ptr %3, align 16
34  %5 = getelementptr inbounds i32, ptr %0, i64 32
35  %6 = load <4 x i32>, ptr %5, align 16
36  %7 = shufflevector <16 x i8> %2, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
37  %8 = sext <4 x i8> %7 to <4 x i32>
38  %9 = icmp sgt <4 x i32> %6, %8
39  %10 = sext <4 x i1> %9 to <4 x i32>
40  %11 = tail call <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16> %4, <8 x i16> %4)
41  %12 = bitcast <8 x i16> %11 to <2 x i64>
42  %13 = shufflevector <4 x i32> %10, <4 x i32> poison, <2 x i32> <i32 0, i32 1>
43  %14 = zext <2 x i32> %13 to <2 x i64>
44  %15 = bitcast <8 x i16> %11 to <16 x i8>
45  %16 = icmp ne <16 x i8> %15, zeroinitializer
46  %17 = sext <16 x i1> %16 to <16 x i8>
47  %18 = shufflevector <16 x i8> %17, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
48  %19 = zext <2 x i8> %18 to <2 x i64>
49  %20 = insertelement <2 x i64> %19, i64 0, i64 1
50  %21 = tail call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %20, <2 x i64> %14)
51  %22 = shufflevector <2 x i64> %21, <2 x i64> %12, <2 x i32> <i32 1, i32 3>
52  ret <2 x i64> %22
53}
54declare <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16>, <8 x i16>)
55declare <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64>, <2 x i64>)
56