1; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | FileCheck --check-prefix=HSA %s
2; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global | FileCheck --check-prefix=HSA-CI %s
3; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=carrizo  | FileCheck --check-prefix=HSA %s
4; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=carrizo -mattr=-flat-for-global | FileCheck --check-prefix=HSA-VI %s
5; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri -filetype=obj | llvm-readobj -symbols -s -sd | FileCheck --check-prefix=ELF %s
6; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | llvm-mc -filetype=obj -triple amdgcn--amdhsa -mcpu=kaveri | llvm-readobj -symbols -s -sd | FileCheck %s --check-prefix=ELF
7
8; The SHT_NOTE section contains the output from the .hsa_code_object_*
9; directives.
10
11; ELF: Section {
12; ELF: Name: .hsatext
13; ELF: Type: SHT_PROGBITS (0x1)
14; ELF: Flags [ (0xC00007)
15; ELF: SHF_ALLOC (0x2)
16; ELF: SHF_AMDGPU_HSA_AGENT (0x800000)
17; ELF: SHF_AMDGPU_HSA_CODE (0x400000)
18; ELF: SHF_EXECINSTR (0x4)
19; ELF: SHF_WRITE (0x1)
20; ELF: }
21
22; ELF: SHT_NOTE
23; ELF: 0000: 04000000 08000000 01000000 414D4400
24; ELF: 0010: 01000000 00000000 04000000 1B000000
25; ELF: 0020: 03000000 414D4400 04000700 07000000
26; ELF: 0030: 00000000 00000000 414D4400 414D4447
27; ELF: 0040: 50550000
28
29; ELF: Symbol {
30; ELF: Name: simple
31; ELF: Size: 288
32; ELF: Type: AMDGPU_HSA_KERNEL (0xA)
33; ELF: }
34
35; HSA: .hsa_code_object_version 1,0
36; HSA-CI: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
37; HSA-VI: .hsa_code_object_isa 8,0,1,"AMD","AMDGPU"
38
39; HSA: .hsatext
40
41; HSA: .amdgpu_hsa_kernel simple
42; HSA: {{^}}simple:
43; HSA: .amd_kernel_code_t
44; HSA: enable_sgpr_private_segment_buffer = 1
45; HSA: enable_sgpr_kernarg_segment_ptr = 1
46; HSA: .end_amd_kernel_code_t
47; HSA: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x0
48
49; Make sure we are setting the ATC bit:
50; HSA-CI: s_mov_b32 s[[HI:[0-9]]], 0x100f000
51; On VI+ we also need to set MTYPE = 2
52; HSA-VI: s_mov_b32 s[[HI:[0-9]]], 0x1100f000
53; Make sure we generate flat store for HSA
54; HSA: flat_store_dword v{{[0-9]+}}
55
56; HSA: .Lfunc_end0:
57; HSA: .size   simple, .Lfunc_end0-simple
58
59define void @simple(i32 addrspace(1)* %out) {
60entry:
61  store i32 0, i32 addrspace(1)* %out
62  ret void
63}
64