1; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -filetype=obj -o - < %s | llvm-readelf --notes - | FileCheck --check-prefix=CHECK %s 2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -amdgpu-dump-hsa-metadata -amdgpu-verify-hsa-metadata -filetype=obj -o - < %s 2>&1 | FileCheck --check-prefix=PARSER %s 3 4; CHECK: --- 5; CHECK: amdhsa.kernels: 6; CHECK: - .args: 7; CHECK-NEXT: - .name: a 8; CHECK-NEXT: .offset: 0 9; CHECK-NEXT: .size: 1 10; CHECK-NEXT: .type_name: char 11; CHECK-NEXT: .value_kind: by_value 12; CHECK-NEXT: - .offset: 8 13; CHECK-NEXT: .size: 8 14; CHECK-NEXT: .value_kind: hidden_global_offset_x 15; CHECK-NEXT: - .offset: 16 16; CHECK-NEXT: .size: 8 17; CHECK-NEXT: .value_kind: hidden_global_offset_y 18; CHECK-NEXT: - .offset: 24 19; CHECK-NEXT: .size: 8 20; CHECK-NEXT: .value_kind: hidden_global_offset_z 21; CHECK-NOT: .value_kind: hidden_default_queue 22; CHECK-NOT: .value_kind: hidden_completion_action 23; CHECK: .language: OpenCL C 24; CHECK-NEXT: .language_version: 25; CHECK-NEXT: - 2 26; CHECK-NEXT: - 0 27; CHECK: .name: test_non_enqueue_kernel_caller 28; CHECK: .symbol: test_non_enqueue_kernel_caller.kd 29define amdgpu_kernel void @test_non_enqueue_kernel_caller(i8 %a) #0 30 !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3 31 !kernel_arg_base_type !3 !kernel_arg_type_qual !4 { 32 ret void 33} 34 35; CHECK: - .args: 36; CHECK-NEXT: - .name: a 37; CHECK-NEXT: .offset: 0 38; CHECK-NEXT: .size: 1 39; CHECK-NEXT: .type_name: char 40; CHECK-NEXT: .value_kind: by_value 41; CHECK-NEXT: - .offset: 8 42; CHECK-NEXT: .size: 8 43; CHECK-NEXT: .value_kind: hidden_global_offset_x 44; CHECK-NEXT: - .offset: 16 45; CHECK-NEXT: .size: 8 46; CHECK-NEXT: .value_kind: hidden_global_offset_y 47; CHECK-NEXT: - .offset: 24 48; CHECK-NEXT: .size: 8 49; CHECK-NEXT: .value_kind: hidden_global_offset_z 50; CHECK-NEXT: - .offset: 32 51; CHECK-NEXT: .size: 8 52; CHECK-NEXT: .value_kind: hidden_hostcall_buffer 53; CHECK-NEXT: - .offset: 40 54; CHECK-NEXT: .size: 8 55; CHECK-NEXT: .value_kind: hidden_default_queue 56; CHECK-NEXT: - .offset: 48 57; CHECK-NEXT: .size: 8 58; CHECK-NEXT: .value_kind: hidden_completion_action 59; CHECK: .language: OpenCL C 60; CHECK-NEXT: .language_version: 61; CHECK-NEXT: - 2 62; CHECK-NEXT: - 0 63; CHECK: .name: test_enqueue_kernel_caller 64; CHECK: .symbol: test_enqueue_kernel_caller.kd 65define amdgpu_kernel void @test_enqueue_kernel_caller(i8 %a) #1 66 !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3 67 !kernel_arg_base_type !3 !kernel_arg_type_qual !4 { 68 ret void 69} 70 71; CHECK: - .args: 72; CHECK-NEXT: - .name: a 73; CHECK-NEXT: .offset: 0 74; CHECK-NEXT: .size: 1 75; CHECK-NEXT: .type_name: char 76; CHECK-NEXT: .value_kind: by_value 77; CHECK-NEXT: - .offset: 8 78; CHECK-NEXT: .size: 8 79; CHECK-NEXT: .value_kind: hidden_global_offset_x 80; CHECK-NEXT: - .offset: 16 81; CHECK-NEXT: .size: 8 82; CHECK-NEXT: .value_kind: hidden_global_offset_y 83; CHECK-NEXT: - .offset: 24 84; CHECK-NEXT: .size: 8 85; CHECK-NEXT: .value_kind: hidden_global_offset_z 86; CHECK-NEXT: - .offset: 32 87; CHECK-NEXT: .size: 8 88; CHECK-NEXT: .value_kind: hidden_hostcall_buffer 89; CHECK-NEXT: - .offset: 40 90; CHECK-NEXT: .size: 8 91; CHECK-NEXT: .value_kind: hidden_default_queue 92; CHECK-NEXT: - .offset: 48 93; CHECK-NEXT: .size: 8 94; CHECK-NEXT: .value_kind: hidden_none 95; CHECK: .language: OpenCL C 96; CHECK-NEXT: .language_version: 97; CHECK-NEXT: - 2 98; CHECK-NEXT: - 0 99; CHECK: .name: test_no_completion_action 100; CHECK: .symbol: test_no_completion_action.kd 101define amdgpu_kernel void @test_no_completion_action(i8 %a) #2 102 !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3 103 !kernel_arg_base_type !3 !kernel_arg_type_qual !4 { 104 ret void 105} 106 107; CHECK: - .args: 108; CHECK-NEXT: - .name: a 109; CHECK-NEXT: .offset: 0 110; CHECK-NEXT: .size: 1 111; CHECK-NEXT: .type_name: char 112; CHECK-NEXT: .value_kind: by_value 113; CHECK-NEXT: - .offset: 8 114; CHECK-NEXT: .size: 8 115; CHECK-NEXT: .value_kind: hidden_global_offset_x 116; CHECK-NEXT: - .offset: 16 117; CHECK-NEXT: .size: 8 118; CHECK-NEXT: .value_kind: hidden_global_offset_y 119; CHECK-NEXT: - .offset: 24 120; CHECK-NEXT: .size: 8 121; CHECK-NEXT: .value_kind: hidden_global_offset_z 122; CHECK-NEXT: - .offset: 32 123; CHECK-NEXT: .size: 8 124; CHECK-NEXT: .value_kind: hidden_hostcall_buffer 125; CHECK-NEXT: - .offset: 40 126; CHECK-NEXT: .size: 8 127; CHECK-NEXT: .value_kind: hidden_none 128; CHECK-NEXT: - .offset: 48 129; CHECK-NEXT: .size: 8 130; CHECK-NEXT: .value_kind: hidden_completion_action 131; CHECK: .language: OpenCL C 132; CHECK-NEXT: .language_version: 133; CHECK-NEXT: - 2 134; CHECK-NEXT: - 0 135; CHECK: .name: test_no_default_queue 136; CHECK: .symbol: test_no_default_queue.kd 137define amdgpu_kernel void @test_no_default_queue(i8 %a) #3 138 !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !3 139 !kernel_arg_base_type !3 !kernel_arg_type_qual !4 { 140 ret void 141} 142 143; CHECK: amdhsa.version: 144; CHECK-NEXT: - 1 145; CHECK-NEXT: - 1 146; CHECK-NOT: amdhsa.printf: 147 148attributes #0 = { optnone noinline "amdgpu-no-default-queue" "amdgpu-no-completion-action" "amdgpu-implicitarg-num-bytes"="48" } 149attributes #1 = { optnone noinline "amdgpu-implicitarg-num-bytes"="48" } 150attributes #2 = { optnone noinline "amdgpu-no-completion-action" "amdgpu-implicitarg-num-bytes"="48" } 151attributes #3 = { optnone noinline "amdgpu-no-default-queue" "amdgpu-implicitarg-num-bytes"="48" } 152 153!llvm.module.flags = !{!0} 154!0 = !{i32 1, !"amdhsa_code_object_version", i32 400} 155 156!1 = !{i32 0} 157!2 = !{!"none"} 158!3 = !{!"char"} 159!4 = !{!""} 160 161!opencl.ocl.version = !{!90} 162!90 = !{i32 2, i32 0} 163 164; PARSER: AMDGPU HSA Metadata Parser Test: PASS 165