Mercurial > hg > Members > tobaru > cbc > CbC_llvm

comparison test/CodeGen/AMDGPU/fmax3.ll @ 95:afa8332a0e37

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
LLVM 3.8
author Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp>
date Tue, 13 Oct 2015 17:48:58 +0900
parents
children 1172e4bd9c6f
comparison
equal deleted inserted replaced
84:f3e34b893a5f 95:afa8332a0e37
1 ; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
2 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
3
4 declare float @llvm.maxnum.f32(float, float) nounwind readnone
5
6 ; SI-LABEL: {{^}}test_fmax3_olt_0:
7 ; SI: buffer_load_dword [[REGC:v[0-9]+]]
8 ; SI: buffer_load_dword [[REGB:v[0-9]+]]
9 ; SI: buffer_load_dword [[REGA:v[0-9]+]]
10 ; SI: v_max3_f32 [[RESULT:v[0-9]+]], [[REGC]], [[REGB]], [[REGA]]
11 ; SI: buffer_store_dword [[RESULT]],
12 ; SI: s_endpgm
13 define void @test_fmax3_olt_0(float addrspace(1)* %out, float addrspace(1)* %aptr, float addrspace(1)* %bptr, float addrspace(1)* %cptr) nounwind {
14 %a = load float, float addrspace(1)* %aptr, align 4
15 %b = load float, float addrspace(1)* %bptr, align 4
16 %c = load float, float addrspace(1)* %cptr, align 4
17 %f0 = call float @llvm.maxnum.f32(float %a, float %b) nounwind readnone
18 %f1 = call float @llvm.maxnum.f32(float %f0, float %c) nounwind readnone
19 store float %f1, float addrspace(1)* %out, align 4
20 ret void
21 }
22
23 ; Commute operand of second fmax
24 ; SI-LABEL: {{^}}test_fmax3_olt_1:
25 ; SI: buffer_load_dword [[REGB:v[0-9]+]]
26 ; SI: buffer_load_dword [[REGA:v[0-9]+]]
27 ; SI: buffer_load_dword [[REGC:v[0-9]+]]
28 ; SI: v_max3_f32 [[RESULT:v[0-9]+]], [[REGC]], [[REGB]], [[REGA]]
29 ; SI: buffer_store_dword [[RESULT]],
30 ; SI: s_endpgm
31 define void @test_fmax3_olt_1(float addrspace(1)* %out, float addrspace(1)* %aptr, float addrspace(1)* %bptr, float addrspace(1)* %cptr) nounwind {
32 %a = load float, float addrspace(1)* %aptr, align 4
33 %b = load float, float addrspace(1)* %bptr, align 4
34 %c = load float, float addrspace(1)* %cptr, align 4
35 %f0 = call float @llvm.maxnum.f32(float %a, float %b) nounwind readnone
36 %f1 = call float @llvm.maxnum.f32(float %c, float %f0) nounwind readnone
37 store float %f1, float addrspace(1)* %out, align 4
38 ret void
39 }