1; RUN: llc -march=amdgcn -mcpu=bonaire -show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=ALL %s
2; RUN: llc -march=amdgcn -mcpu=carrizo --show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=ALL %s
3; RUN: llc -march=amdgcn -mcpu=gfx900 --show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=ALL %s
4; RUN: llc -march=amdgcn -mcpu=bonaire -mtriple=amdgcn-unknown-amdhsa < %s -mattr=-flat-for-global | FileCheck -check-prefix=GCNHSA -check-prefix=CIHSA -check-prefix=ALL %s
5; RUN: llc -march=amdgcn -mcpu=carrizo -mtriple=amdgcn-unknown-amdhsa -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCNHSA -check-prefix=VIHSA -check-prefix=ALL %s
6
7; FIXME: align on alloca seems to be ignored for private_segment_alignment
8
9; ALL-LABEL: {{^}}large_alloca_compute_shader:
10
11; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
12; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD0
13; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
14; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD1
15; GCN-DAG: s_mov_b32 s{{[0-9]+}}, -1
16; CI-DAG: s_mov_b32 s{{[0-9]+}}, 0xe8f000
17; VI-DAG: s_mov_b32 s{{[0-9]+}}, 0xe80000
18; GFX9-DAG: s_mov_b32 s{{[0-9]+}}, 0xe00000
19
20
21; GCNHSA: .amd_kernel_code_t
22
23; GCNHSA: enable_sgpr_private_segment_wave_byte_offset = 1
24; GCNHSA: user_sgpr_count = 8
25; GCNHSA: enable_sgpr_workgroup_id_x = 1
26; GCNHSA: enable_sgpr_workgroup_id_y = 0
27; GCNHSA: enable_sgpr_workgroup_id_z = 0
28; GCNHSA: enable_sgpr_workgroup_info = 0
29; GCNHSA: enable_vgpr_workitem_id = 0
30
31; GCNHSA: enable_sgpr_private_segment_buffer = 1
32; GCNHSA: enable_sgpr_dispatch_ptr = 0
33; GCNHSA: enable_sgpr_queue_ptr = 0
34; GCNHSA: enable_sgpr_kernarg_segment_ptr = 1
35; GCNHSA: enable_sgpr_dispatch_id = 0
36; GCNHSA: enable_sgpr_flat_scratch_init = 1
37; GCNHSA: enable_sgpr_private_segment_size = 0
38; GCNHSA: enable_sgpr_grid_workgroup_count_x = 0
39; GCNHSA: enable_sgpr_grid_workgroup_count_y = 0
40; GCNHSA: enable_sgpr_grid_workgroup_count_z = 0
41; GCNHSA: workitem_private_segment_byte_size = 32772
42; GCNHSA: private_segment_alignment = 4
43; GCNHSA: .end_amd_kernel_code_t
44
45
46; GCNHSA: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, s[0:3], s9 offen
47; GCNHSA: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, s[0:3], s9 offen
48
49; Scratch size = alloca size + emergency stack slot, align {{.*}}, addrspace(5)
50; ALL: ; ScratchSize: 32772
51define amdgpu_kernel void @large_alloca_compute_shader(i32 %x, i32 %y) #0 {
52  %large = alloca [8192 x i32], align 4, addrspace(5)
53  %gep = getelementptr [8192 x i32], [8192 x i32] addrspace(5)* %large, i32 0, i32 8191
54  store volatile i32 %x, i32 addrspace(5)* %gep
55  %gep1 = getelementptr [8192 x i32], [8192 x i32] addrspace(5)* %large, i32 0, i32 %y
56  %val = load volatile i32, i32 addrspace(5)* %gep1
57  store volatile i32 %val, i32 addrspace(1)* undef
58  ret void
59}
60
61attributes #0 = { nounwind  }
62