xref: /llvm-project/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.buffer.atomic.fadd.ll (revision 6548b6354d1d990e1c98736f5e7c3de876bedc8e)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx908 < %s | FileCheck %s -check-prefix=CHECK
3
4define void @struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset(float %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
5; CHECK-LABEL: struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
6; CHECK:       ; %bb.0:
7; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8; CHECK-NEXT:    buffer_atomic_add_f32 v0, v[1:2], s[16:19], s20 idxen offen offset:24
9; CHECK-NEXT:    s_waitcnt vmcnt(0)
10; CHECK-NEXT:    s_setpc_b64 s[30:31]
11  %voffset.add = add i32 %voffset, 24
12  %ret = call float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset.add, i32 %soffset, i32 0)
13  ret void
14}
15
16; Natural mapping, no voffset
17define void @struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset(float %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 inreg %soffset) {
18; CHECK-LABEL: struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__0_voffset__sgpr_soffset:
19; CHECK:       ; %bb.0:
20; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
21; CHECK-NEXT:    buffer_atomic_add_f32 v0, v1, s[16:19], s20 idxen
22; CHECK-NEXT:    s_waitcnt vmcnt(0)
23; CHECK-NEXT:    s_setpc_b64 s[30:31]
24  %ret = call float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 0, i32 %soffset, i32 0)
25  ret void
26}
27
28define void @struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc(float %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
29; CHECK-LABEL: struct_buffer_atomic_add_f32_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc:
30; CHECK:       ; %bb.0:
31; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
32; CHECK-NEXT:    buffer_atomic_add_f32 v0, v[1:2], s[16:19], s20 idxen offen slc
33; CHECK-NEXT:    s_waitcnt vmcnt(0)
34; CHECK-NEXT:    s_setpc_b64 s[30:31]
35  %ret = call float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 2)
36  ret void
37}
38
39define void @struct_buffer_atomic_add_v2f16_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<2 x half> %val, <4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
40; CHECK-LABEL: struct_buffer_atomic_add_v2f16_noret__vgpr_val__sgpr_rsrc__vgpr_voffset__sgpr_soffset:
41; CHECK:       ; %bb.0:
42; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
43; CHECK-NEXT:    buffer_atomic_pk_add_f16 v0, v[1:2], s[16:19], s20 idxen offen offset:24
44; CHECK-NEXT:    s_waitcnt vmcnt(0)
45; CHECK-NEXT:    s_setpc_b64 s[30:31]
46  %voffset.add = add i32 %voffset, 24
47  %ret = call <2 x half> @llvm.amdgcn.struct.buffer.atomic.fadd.v2f16(<2 x half> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset.add, i32 %soffset, i32 0)
48  ret void
49}
50
51declare float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float, <4 x i32>, i32, i32, i32, i32 immarg) #0
52declare <2 x half> @llvm.amdgcn.struct.buffer.atomic.fadd.v2f16(<2 x half>, <4 x i32>, i32, i32, i32, i32 immarg) #0
53
54attributes #0 = { nounwind }
55