xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/i1-copy.ll (revision e06290e53f2880962fef582f118482d70f1c27f0)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefixes=WAVE64 %s
3; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck -check-prefixes=WAVE32 %s
4
5define amdgpu_ps void @i1_vcc_to_vcc_copy(i32 %val, float %a0, float %a1, float %b0, float %b1) {
6; WAVE64-LABEL: i1_vcc_to_vcc_copy:
7; WAVE64:       ; %bb.0: ; %main_body
8; WAVE64-NEXT:    v_cmp_eq_u32_e32 vcc, 2, v0
9; WAVE64-NEXT:    v_cndmask_b32_e32 v0, v2, v1, vcc
10; WAVE64-NEXT:    v_cndmask_b32_e32 v1, v4, v3, vcc
11; WAVE64-NEXT:    exp mrt0 v0, v1, v0, v0 done vm
12; WAVE64-NEXT:    s_endpgm
13;
14; WAVE32-LABEL: i1_vcc_to_vcc_copy:
15; WAVE32:       ; %bb.0: ; %main_body
16; WAVE32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 2, v0
17; WAVE32-NEXT:    v_cndmask_b32_e32 v0, v2, v1, vcc_lo
18; WAVE32-NEXT:    v_cndmask_b32_e32 v1, v4, v3, vcc_lo
19; WAVE32-NEXT:    exp mrt0 v0, v1, v0, v0 done vm
20; WAVE32-NEXT:    s_endpgm
21main_body:
22  %vcc = icmp eq i32 %val, 2
23  %a = select i1 %vcc, float %a0, float %a1
24  %b = select i1 %vcc, float %b0, float %b1
25  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %a, float %b, float undef, float undef, i1 true, i1 true)
26  ret void
27}
28
29define amdgpu_ps void @i1_sgpr_to_vcc_copy(i32 inreg %val, float %a0, float %a1, float %b0, float %b1) {
30; WAVE64-LABEL: i1_sgpr_to_vcc_copy:
31; WAVE64:       ; %bb.0: ; %main_body
32; WAVE64-NEXT:    s_cmp_eq_u32 s0, 2
33; WAVE64-NEXT:    s_cselect_b32 s0, 1, 0
34; WAVE64-NEXT:    s_and_b32 s0, 1, s0
35; WAVE64-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s0
36; WAVE64-NEXT:    v_cndmask_b32_e32 v0, v1, v0, vcc
37; WAVE64-NEXT:    v_cndmask_b32_e32 v1, v3, v2, vcc
38; WAVE64-NEXT:    exp mrt0 v0, v1, v0, v0 done vm
39; WAVE64-NEXT:    s_endpgm
40;
41; WAVE32-LABEL: i1_sgpr_to_vcc_copy:
42; WAVE32:       ; %bb.0: ; %main_body
43; WAVE32-NEXT:    s_cmp_eq_u32 s0, 2
44; WAVE32-NEXT:    s_cselect_b32 s0, 1, 0
45; WAVE32-NEXT:    s_and_b32 s0, 1, s0
46; WAVE32-NEXT:    v_cmp_ne_u32_e64 vcc_lo, 0, s0
47; WAVE32-NEXT:    v_cndmask_b32_e32 v0, v1, v0, vcc_lo
48; WAVE32-NEXT:    v_cndmask_b32_e32 v1, v3, v2, vcc_lo
49; WAVE32-NEXT:    exp mrt0 v0, v1, v0, v0 done vm
50; WAVE32-NEXT:    s_endpgm
51main_body:
52  %uniform_i1 = icmp eq i32 %val, 2
53  %a = select i1 %uniform_i1, float %a0, float %a1
54  %b = select i1 %uniform_i1, float %b0, float %b1
55  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %a, float %b, float undef, float undef, i1 true, i1 true)
56  ret void
57}
58
59declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1)
60