xref: /llvm-project/llvm/test/CodeGen/AMDGPU/i1-copy-from-loop.ll (revision 2d5f3b0a61fb171617012a2c3ba05fd31fb3bb1d)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=amdgcn-- -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
3
4define amdgpu_ps void @i1_copy_from_loop(ptr addrspace(8) inreg %rsrc, i32 %tid) {
5; SI-LABEL: i1_copy_from_loop:
6; SI:       ; %bb.0: ; %entry
7; SI-NEXT:    s_mov_b32 s14, 0
8; SI-NEXT:    s_mov_b64 s[4:5], 0
9; SI-NEXT:    ; implicit-def: $sgpr6_sgpr7
10; SI-NEXT:    ; implicit-def: $sgpr8_sgpr9
11; SI-NEXT:    s_branch .LBB0_3
12; SI-NEXT:  .LBB0_1: ; in Loop: Header=BB0_3 Depth=1
13; SI-NEXT:    ; implicit-def: $sgpr14
14; SI-NEXT:  .LBB0_2: ; %Flow
15; SI-NEXT:    ; in Loop: Header=BB0_3 Depth=1
16; SI-NEXT:    s_and_b64 s[12:13], exec, s[8:9]
17; SI-NEXT:    s_or_b64 s[4:5], s[12:13], s[4:5]
18; SI-NEXT:    s_andn2_b64 s[6:7], s[6:7], exec
19; SI-NEXT:    s_and_b64 s[10:11], s[10:11], exec
20; SI-NEXT:    s_or_b64 s[6:7], s[6:7], s[10:11]
21; SI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
22; SI-NEXT:    s_cbranch_execz .LBB0_7
23; SI-NEXT:  .LBB0_3: ; %for.body
24; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
25; SI-NEXT:    s_cmp_lt_u32 s14, 4
26; SI-NEXT:    s_cselect_b64 s[10:11], -1, 0
27; SI-NEXT:    s_or_b64 s[8:9], s[8:9], exec
28; SI-NEXT:    s_cmp_gt_u32 s14, 3
29; SI-NEXT:    s_cbranch_scc1 .LBB0_1
30; SI-NEXT:  ; %bb.4: ; %mid.loop
31; SI-NEXT:    ; in Loop: Header=BB0_3 Depth=1
32; SI-NEXT:    v_mov_b32_e32 v1, s14
33; SI-NEXT:    buffer_load_dword v1, v[0:1], s[0:3], 0 idxen offen
34; SI-NEXT:    s_mov_b64 s[10:11], -1
35; SI-NEXT:    s_waitcnt vmcnt(0)
36; SI-NEXT:    v_cmp_le_f32_e32 vcc, 0, v1
37; SI-NEXT:    s_mov_b64 s[8:9], -1
38; SI-NEXT:    s_and_saveexec_b64 s[12:13], vcc
39; SI-NEXT:    s_cbranch_execz .LBB0_6
40; SI-NEXT:  ; %bb.5: ; %end.loop
41; SI-NEXT:    ; in Loop: Header=BB0_3 Depth=1
42; SI-NEXT:    s_add_i32 s14, s14, 1
43; SI-NEXT:    s_xor_b64 s[8:9], exec, -1
44; SI-NEXT:  .LBB0_6: ; %Flow1
45; SI-NEXT:    ; in Loop: Header=BB0_3 Depth=1
46; SI-NEXT:    s_or_b64 exec, exec, s[12:13]
47; SI-NEXT:    s_branch .LBB0_2
48; SI-NEXT:  .LBB0_7: ; %for.end
49; SI-NEXT:    s_or_b64 exec, exec, s[4:5]
50; SI-NEXT:    s_and_saveexec_b64 s[0:1], s[6:7]
51; SI-NEXT:    s_cbranch_execz .LBB0_9
52; SI-NEXT:  ; %bb.8: ; %if
53; SI-NEXT:    exp mrt0 v0, v0, v0, v0 done vm
54; SI-NEXT:  .LBB0_9: ; %end
55; SI-NEXT:    s_endpgm
56entry:
57  br label %for.body
58
59for.body:
60  %i = phi i32 [0, %entry], [%i.inc, %end.loop]
61  %cc = icmp ult i32 %i, 4
62  br i1 %cc, label %mid.loop, label %for.end
63
64mid.loop:
65  %v = call float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8) %rsrc, i32 %tid, i32 %i, i32 0, i32 0)
66  %cc2 = fcmp oge float %v, 0.0
67  br i1 %cc2, label %end.loop, label %for.end
68
69end.loop:
70  %i.inc = add i32 %i, 1
71  br label %for.body
72
73for.end:
74  br i1 %cc, label %if, label %end
75
76if:
77  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float undef, float undef, float undef, float undef, i1 true, i1 true)
78  br label %end
79
80end:
81  ret void
82}
83
84declare float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8), i32, i32, i32, i32 immarg) #0
85declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #1
86
87attributes #0 = { nounwind memory(argmem: read) }
88attributes #1 = { nounwind inaccessiblememonly }
89