|
221
|
1 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck --check-prefix=GCN %s
|
|
|
2 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefix=GCN %s
|
|
|
3 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-vgpr-index-mode -verify-machineinstrs < %s | FileCheck --check-prefix=GCN %s
|
|
150
|
4
|
|
|
5 ; Tests for indirect addressing on SI, which is implemented using dynamic
|
|
|
6 ; indexing of vectors.
|
|
|
7
|
|
|
8 ; Subtest below moved from file test/CodeGen/AMDGPU/indirect-addressing-si.ll
|
|
|
9 ; to avoid gfx9 scheduling induced issues.
|
|
|
10
|
|
|
11
|
|
|
12 ; GCN-LABEL: {{^}}insert_vgpr_offset_multiple_in_block:
|
|
236
|
13 ; GCN-DAG: s_load_dwordx16 s[[[S_ELT0:[0-9]+]]:[[S_ELT15:[0-9]+]]]
|
|
150
|
14 ; GCN-DAG: {{buffer|flat|global}}_load_dword [[IDX0:v[0-9]+]]
|
|
|
15 ; GCN-DAG: v_mov_b32 [[INS0:v[0-9]+]], 62
|
|
|
16
|
|
|
17 ; GCN-DAG: v_mov_b32_e32 v[[VEC_ELT15:[0-9]+]], s[[S_ELT15]]
|
|
|
18 ; GCN-DAG: v_mov_b32_e32 v[[VEC_ELT0:[0-9]+]], s[[S_ELT0]]
|
|
|
19
|
|
|
20 ; GCN-DAG: v_add_{{i32|u32}}_e32 [[IDX1:v[0-9]+]], vcc, 1, [[IDX0]]
|
|
|
21
|
|
173
|
22 ; GCN: v_cmp_eq_u32_e32
|
|
|
23 ; GCN-COUNT-32: v_cndmask_b32
|
|
150
|
24
|
|
173
|
25 ; GCN-COUNT-4: buffer_store_dwordx4
|
|
150
|
26 define amdgpu_kernel void @insert_vgpr_offset_multiple_in_block(<16 x i32> addrspace(1)* %out0, <16 x i32> addrspace(1)* %out1, i32 addrspace(1)* %in, <16 x i32> %vec0) #0 {
|
|
|
27 entry:
|
|
|
28 %id = call i32 @llvm.amdgcn.workitem.id.x() #1
|
|
|
29 %id.ext = zext i32 %id to i64
|
|
|
30 %gep = getelementptr inbounds i32, i32 addrspace(1)* %in, i64 %id.ext
|
|
|
31 %idx0 = load volatile i32, i32 addrspace(1)* %gep
|
|
|
32 %idx1 = add i32 %idx0, 1
|
|
|
33 %live.out.val = call i32 asm sideeffect "v_mov_b32 $0, 62", "=v"()
|
|
|
34 %vec1 = insertelement <16 x i32> %vec0, i32 %live.out.val, i32 %idx0
|
|
|
35 %vec2 = insertelement <16 x i32> %vec1, i32 63, i32 %idx1
|
|
|
36 store volatile <16 x i32> %vec2, <16 x i32> addrspace(1)* %out0
|
|
|
37 %cmp = icmp eq i32 %id, 0
|
|
|
38 br i1 %cmp, label %bb1, label %bb2
|
|
|
39
|
|
|
40 bb1:
|
|
|
41 store volatile i32 %live.out.val, i32 addrspace(1)* undef
|
|
|
42 br label %bb2
|
|
|
43
|
|
|
44 bb2:
|
|
|
45 ret void
|
|
|
46 }
|
|
|
47
|
|
|
48 declare i32 @llvm.amdgcn.workitem.id.x() #1
|
|
|
49 declare void @llvm.amdgcn.s.barrier() #2
|
|
|
50
|
|
|
51 attributes #0 = { nounwind }
|
|
|
52 attributes #1 = { nounwind readnone }
|
|
|
53 attributes #2 = { nounwind convergent }
|
