Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / f6fa1d0369c653fffe454f934c24c809d00e7330 / . / llvm / test / CodeGen / X86 / vselect-packss.ll
blob: 15a9b3cc1bfedb632215b3635b2d1890edae7b86 [file] [log] [blame]
Simon Pilgrimf6fa1d02017-09-11 10:50:03 +0000[diff] [blame^]1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE42
4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
5; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
6; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
7; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512VL
8
9define <16 x i8> @vselect_packss(<16 x i16> %a0, <16 x i16> %a1, <16 x i8> %a2, <16 x i8> %a3) {
10; SSE-LABEL: vselect_packss:
11; SSE: # BB#0:
12; SSE-NEXT: pcmpeqw %xmm3, %xmm1
13; SSE-NEXT: pcmpeqw %xmm2, %xmm0
14; SSE-NEXT: packsswb %xmm1, %xmm0
15; SSE-NEXT: pand %xmm0, %xmm4
16; SSE-NEXT: pandn %xmm5, %xmm0
17; SSE-NEXT: por %xmm4, %xmm0
18; SSE-NEXT: retq
19;
20; AVX1-LABEL: vselect_packss:
21; AVX1: # BB#0:
22; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
23; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
24; AVX1-NEXT: vpcmpeqw %xmm4, %xmm5, %xmm4
25; AVX1-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
26; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm0
27; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm1
28; AVX1-NEXT: vpandn %xmm3, %xmm0, %xmm0
29; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
30; AVX1-NEXT: vzeroupper
31; AVX1-NEXT: retq
32;
33; AVX2-LABEL: vselect_packss:
34; AVX2: # BB#0:
35; AVX2-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
36; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
37; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
38; AVX2-NEXT: vpand %xmm2, %xmm0, %xmm1
39; AVX2-NEXT: vpandn %xmm3, %xmm0, %xmm0
40; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
41; AVX2-NEXT: vzeroupper
42; AVX2-NEXT: retq
43;
44; AVX512-LABEL: vselect_packss:
45; AVX512: # BB#0:
46; AVX512-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
47; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
48; AVX512-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
49; AVX512-NEXT: vpand %xmm2, %xmm0, %xmm1
50; AVX512-NEXT: vpandn %xmm3, %xmm0, %xmm0
51; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
52; AVX512-NEXT: vzeroupper
53; AVX512-NEXT: retq
54 %1 = icmp eq <16 x i16> %a0, %a1
55 %2 = sext <16 x i1> %1 to <16 x i16>
56 %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
57 %4 = shufflevector <16 x i16> %2, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
58 %5 = tail call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %3, <8 x i16> %4)
59 %6 = and <16 x i8> %5, %a2
60 %7 = xor <16 x i8> %5, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
61 %8 = and <16 x i8> %7, %a3
62 %9 = or <16 x i8> %6, %8
63 ret <16 x i8> %9
64}
65declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>)