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: .text 13 ; ELF: Type: SHT_PROGBITS (0x1) 14 ; ELF: Flags [ (0x6) 15 ; ELF: SHF_ALLOC (0x2) 16 ; ELF: SHF_EXECINSTR (0x4) 17 ; ELF: } 18 19 ; ELF: SHT_NOTE 20 ; ELF: 0000: 04000000 08000000 01000000 414D4400 21 ; ELF: 0010: 02000000 01000000 04000000 1B000000 22 ; ELF: 0020: 03000000 414D4400 04000700 07000000 23 ; ELF: 0030: 00000000 00000000 414D4400 414D4447 24 ; ELF: 0040: 50550000 25 26 ; ELF: Symbol { 27 ; ELF: Name: simple 28 ; ELF: Size: 288 29 ; ELF: Type: AMDGPU_HSA_KERNEL (0xA) 30 ; ELF: } 31 32 ; HSA: .hsa_code_object_version 2,1 33 ; HSA-CI: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU" 34 ; HSA-VI: .hsa_code_object_isa 8,0,1,"AMD","AMDGPU" 35 36 ; HSA: .text 37 38 ; HSA: .amdgpu_hsa_kernel simple 39 ; HSA: {{^}}simple: 40 ; HSA: .amd_kernel_code_t 41 ; HSA: enable_sgpr_private_segment_buffer = 1 42 ; HSA: enable_sgpr_kernarg_segment_ptr = 1 43 ; HSA: .end_amd_kernel_code_t 44 ; HSA: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x0 45 46 ; Make sure we are setting the ATC bit: 47 ; HSA-CI: s_mov_b32 s[[HI:[0-9]]], 0x100f000 48 ; On VI+ we also need to set MTYPE = 2 49 ; HSA-VI: s_mov_b32 s[[HI:[0-9]]], 0x1100f000 50 ; Make sure we generate flat store for HSA 51 ; HSA: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}} 52 53 ; HSA: .Lfunc_end0: 54 ; HSA: .size simple, .Lfunc_end0-simple 55 56 define amdgpu_kernel void @simple(i32 addrspace(1)* %out) { 57 entry: 58 store i32 0, i32 addrspace(1)* %out 59 ret void 60 } 61
