1; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -slp-vectorizer %s | FileCheck -check-prefixes=GCN,GFX7 %s
2; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -slp-vectorizer %s | FileCheck -check-prefixes=GCN,GFX8 %s
3; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -slp-vectorizer %s | FileCheck -check-prefixes=GCN,GFX8 %s
4
5; GCN-LABEL: @bswap_v2i16(
6; GFX7: call i16 @llvm.bswap.i16(
7; GFX7: call i16 @llvm.bswap.i16(
8
9; GFX8: call <2 x i16> @llvm.bswap.v2i16(
10define <2 x i16> @bswap_v2i16(<2 x i16> %arg) {
11bb:
12  %tmp = extractelement <2 x i16> %arg, i64 0
13  %tmp1 = tail call i16 @llvm.bswap.i16(i16 %tmp)
14  %tmp2 = insertelement <2 x i16> undef, i16 %tmp1, i64 0
15  %tmp3 = extractelement <2 x i16> %arg, i64 1
16  %tmp4 = tail call i16 @llvm.bswap.i16(i16 %tmp3)
17  %tmp5 = insertelement <2 x i16> %tmp2, i16 %tmp4, i64 1
18  ret <2 x i16> %tmp5
19}
20
21; GCN-LABEL: @bswap_v2i32(
22; GCN: call i32 @llvm.bswap.i32
23; GCN: call i32 @llvm.bswap.i32
24define <2 x i32> @bswap_v2i32(<2 x i32> %arg) {
25bb:
26  %tmp = extractelement <2 x i32> %arg, i64 0
27  %tmp1 = tail call i32 @llvm.bswap.i32(i32 %tmp)
28  %tmp2 = insertelement <2 x i32> undef, i32 %tmp1, i64 0
29  %tmp3 = extractelement <2 x i32> %arg, i64 1
30  %tmp4 = tail call i32 @llvm.bswap.i32(i32 %tmp3)
31  %tmp5 = insertelement <2 x i32> %tmp2, i32 %tmp4, i64 1
32  ret <2 x i32> %tmp5
33}
34
35declare i16 @llvm.bswap.i16(i16) #0
36declare i32 @llvm.bswap.i32(i32) #0
37
38attributes #0 = { nounwind readnone speculatable willreturn }
39