150
|
1 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX1032 %s
|
|
2
|
|
3 ; GCN-LABEL: {{^}}test_init_exec:
|
|
4 ; GFX1032: s_mov_b32 exec_lo, 0x12345
|
|
5 ; GFX1064: s_mov_b64 exec, 0x12345
|
|
6 ; GCN: v_add_f32_e32 v0,
|
|
7 define amdgpu_ps float @test_init_exec(float %a, float %b) {
|
|
8 main_body:
|
|
9 %s = fadd float %a, %b
|
|
10 call void @llvm.amdgcn.init.exec(i64 74565)
|
|
11 ret float %s
|
|
12 }
|
|
13
|
|
14 ; GCN-LABEL: {{^}}test_init_exec_from_input:
|
|
15 ; GCN: s_bfe_u32 s0, s3, 0x70008
|
|
16 ; GFX1032: s_bfm_b32 exec_lo, s0, 0
|
|
17 ; GFX1032: s_cmp_eq_u32 s0, 32
|
|
18 ; GFX1032: s_cmov_b32 exec_lo, -1
|
|
19 ; GFX1064: s_bfm_b64 exec, s0, 0
|
|
20 ; GFX1064: s_cmp_eq_u32 s0, 64
|
|
21 ; GFX1064: s_cmov_b64 exec, -1
|
|
22 ; GCN: v_add_f32_e32 v0,
|
|
23 define amdgpu_ps float @test_init_exec_from_input(i32 inreg, i32 inreg, i32 inreg, i32 inreg %count, float %a, float %b) {
|
|
24 main_body:
|
|
25 %s = fadd float %a, %b
|
|
26 call void @llvm.amdgcn.init.exec.from.input(i32 %count, i32 8)
|
|
27 ret float %s
|
|
28 }
|
|
29
|
|
30 declare void @llvm.amdgcn.init.exec(i64)
|
|
31 declare void @llvm.amdgcn.init.exec.from.input(i32, i32)
|