1; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx803 < %s | FileCheck --check-prefixes=ALL-ASM,OSABI-AMDHSA-ASM %s 2; RUN: llc -filetype=obj -mtriple=amdgcn-amd-amdhsa -mcpu=gfx803 < %s | llvm-readelf -S -r -s --notes - | FileCheck --check-prefix=OSABI-AMDHSA-ELF %s 3 4; ALL-ASM-LABEL: {{^}}fadd: 5 6; OSABI-AMDHSA-ASM-NOT: .hsa_code_object_version 7; OSABI-AMDHSA-ASM-NOT: .hsa_code_object_isa 8; OSABI-AMDHSA-ASM-NOT: .amdgpu_hsa_kernel 9; OSABI-AMDHSA-ASM-NOT: .amd_kernel_code_t 10 11; OSABI-AMDHSA-ASM: s_endpgm 12; OSABI-AMDHSA-ASM: .section .rodata,"a" 13; OSABI-AMDHSA-ASM: .p2align 6 14; OSABI-AMDHSA-ASM: .amdhsa_kernel fadd 15; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_count 12 16; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_private_segment_buffer 1 17; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_kernarg_segment_ptr 1 18; OSABI-AMDHSA-ASM: .amdhsa_next_free_vgpr 3 19; OSABI-AMDHSA-ASM: .amdhsa_next_free_sgpr 10 20; OSABI-AMDHSA-ASM: .amdhsa_reserve_vcc 0 21; OSABI-AMDHSA-ASM: .amdhsa_reserve_flat_scratch 0 22; OSABI-AMDHSA-ASM: .end_amdhsa_kernel 23; OSABI-AMDHSA-ASM: .text 24 25; ALL-ASM-LABEL: {{^}}fsub: 26 27; OSABI-AMDHSA-ASM-NOT: .amdgpu_hsa_kernel 28; OSABI-AMDHSA-ASM-NOT: .amd_kernel_code_t 29 30; OSABI-AMDHSA-ASM: s_endpgm 31; OSABI-AMDHSA-ASM: .section .rodata,"a" 32; OSABI-AMDHSA-ASM: .p2align 6 33; OSABI-AMDHSA-ASM: .amdhsa_kernel fsub 34; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_count 12 35; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_private_segment_buffer 1 36; OSABI-AMDHSA-ASM: .amdhsa_user_sgpr_kernarg_segment_ptr 1 37; OSABI-AMDHSA-ASM: .amdhsa_next_free_vgpr 3 38; OSABI-AMDHSA-ASM: .amdhsa_next_free_sgpr 10 39; OSABI-AMDHSA-ASM: .amdhsa_reserve_vcc 0 40; OSABI-AMDHSA-ASM: .amdhsa_reserve_flat_scratch 0 41; OSABI-AMDHSA-ASM: .end_amdhsa_kernel 42; OSABI-AMDHSA-ASM: .text 43 44; OSABI-AMDHSA-ASM-NOT: .hsa_code_object_version 45; OSABI-AMDHSA-ASM-NOT: .hsa_code_object_isa 46; OSABI-AMDHSA-ASM-NOT: .amd_amdgpu_isa 47; OSABI-AMDHSA-ASM-NOT: .amd_amdgpu_hsa_metadata 48; OSABI-AMDHSA-ASM-NOT: .amd_amdgpu_pal_metadata 49 50; OSABI-AMDHSA-ELF: Section Headers 51; OSABI-AMDHSA-ELF: .text PROGBITS {{[0-9]+}} {{[0-9]+}} {{[0-9a-f]+}} {{[0-9]+}} AX {{[0-9]+}} {{[0-9]+}} 256 52; OSABI-AMDHSA-ELF: .rodata PROGBITS {{[0-9]+}} {{[0-9]+}} {{[0-9a-f]+}} {{[0-9]+}} A {{[0-9]+}} {{[0-9]+}} 64 53 54; OSABI-AMDHSA-ELF: Relocation section '.rela.rodata' at offset 55; OSABI-AMDHSA-ELF: R_AMDGPU_REL64 0000000000000000 fadd + 10 56; OSABI-AMDHSA-ELF: R_AMDGPU_REL64 0000000000000100 fsub + 10 57; OSABI-AMDHSA-ELF: R_AMDGPU_REL64 0000000000000200 empty + 10 58 59; OSABI-AMDHSA-ELF: Symbol table '.symtab' contains {{[0-9]+}} entries 60; OSABI-AMDHSA-ELF: {{[0-9]+}}: 0000000000000000 {{[0-9]+}} FUNC GLOBAL PROTECTED {{[0-9]+}} fadd 61; OSABI-AMDHSA-ELF: {{[0-9]+}}: 0000000000000000 64 OBJECT GLOBAL DEFAULT {{[0-9]+}} fadd.kd 62; OSABI-AMDHSA-ELF: {{[0-9]+}}: 0000000000000100 {{[0-9]+}} FUNC GLOBAL PROTECTED {{[0-9]+}} fsub 63; OSABI-AMDHSA-ELF: {{[0-9]+}}: 0000000000000040 64 OBJECT GLOBAL DEFAULT {{[0-9]+}} fsub.kd 64 65; OSABI-AMDHSA-ELF: Displaying notes found in: .note 66; OSABI-AMDHSA-ELF: AMDGPU 0x{{[0-9a-f]+}} NT_AMDGPU_METADATA (AMDGPU Metadata) 67 68define amdgpu_kernel void @fadd( 69 ptr addrspace(1) %r, 70 ptr addrspace(1) %a, 71 ptr addrspace(1) %b) { 72entry: 73 %a.val = load float, ptr addrspace(1) %a 74 %b.val = load float, ptr addrspace(1) %b 75 %r.val = fadd float %a.val, %b.val 76 store float %r.val, ptr addrspace(1) %r 77 ret void 78} 79 80define amdgpu_kernel void @fsub( 81 ptr addrspace(1) %r, 82 ptr addrspace(1) %a, 83 ptr addrspace(1) %b) { 84entry: 85 %a.val = load float, ptr addrspace(1) %a 86 %b.val = load float, ptr addrspace(1) %b 87 %r.val = fsub float %a.val, %b.val 88 store float %r.val, ptr addrspace(1) %r 89 ret void 90} 91 92; Make sure kernel arguments do not count towards the number of 93; registers used. 94; 95; ALL-ASM-LABEL: {{^}}empty: 96; ALL-ASM: .amdhsa_next_free_vgpr 1 97; ALL-ASM: .amdhsa_next_free_sgpr 1 98define amdgpu_kernel void @empty( 99 i32 %i, 100 ptr addrspace(1) %r, 101 ptr addrspace(1) %a, 102 ptr addrspace(1) %b) { 103entry: 104 ret void 105} 106