1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals 2; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-annotate-kernel-features %s | FileCheck -check-prefix=AKF_GCN %s 3; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-attributor %s | FileCheck -check-prefix=ATTRIBUTOR_GCN %s 4 5; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s 6 7target datalayout = "A5" 8 9define internal void @indirect() { 10; AKF_GCN-LABEL: define {{[^@]+}}@indirect() { 11; AKF_GCN-NEXT: ret void 12; 13; ATTRIBUTOR_GCN-LABEL: define {{[^@]+}}@indirect 14; ATTRIBUTOR_GCN-SAME: () #[[ATTR0:[0-9]+]] { 15; ATTRIBUTOR_GCN-NEXT: ret void 16; 17; GFX9-LABEL: indirect: 18; GFX9: ; %bb.0: 19; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) 20; GFX9-NEXT: s_setpc_b64 s[30:31] 21 ret void 22} 23 24define amdgpu_kernel void @test_simple_indirect_call() { 25; AKF_GCN-LABEL: define {{[^@]+}}@test_simple_indirect_call 26; AKF_GCN-SAME: () #[[ATTR0:[0-9]+]] { 27; AKF_GCN-NEXT: [[FPTR:%.*]] = alloca ptr, align 8, addrspace(5) 28; AKF_GCN-NEXT: [[FPTR_CAST:%.*]] = addrspacecast ptr addrspace(5) [[FPTR]] to ptr 29; AKF_GCN-NEXT: store ptr @indirect, ptr [[FPTR_CAST]], align 8 30; AKF_GCN-NEXT: [[FP:%.*]] = load ptr, ptr [[FPTR_CAST]], align 8 31; AKF_GCN-NEXT: call void [[FP]]() 32; AKF_GCN-NEXT: ret void 33; 34; ATTRIBUTOR_GCN-LABEL: define {{[^@]+}}@test_simple_indirect_call 35; ATTRIBUTOR_GCN-SAME: () #[[ATTR1:[0-9]+]] { 36; ATTRIBUTOR_GCN-NEXT: [[FPTR:%.*]] = alloca ptr, align 8, addrspace(5) 37; ATTRIBUTOR_GCN-NEXT: store ptr @indirect, ptr addrspace(5) [[FPTR]], align 8 38; ATTRIBUTOR_GCN-NEXT: [[FP:%.*]] = load ptr, ptr addrspace(5) [[FPTR]], align 8 39; ATTRIBUTOR_GCN-NEXT: call void @indirect() 40; ATTRIBUTOR_GCN-NEXT: ret void 41; 42; GFX9-LABEL: test_simple_indirect_call: 43; GFX9: ; %bb.0: 44; GFX9-NEXT: s_add_u32 flat_scratch_lo, s12, s17 45; GFX9-NEXT: s_addc_u32 flat_scratch_hi, s13, 0 46; GFX9-NEXT: s_mov_b32 s13, s15 47; GFX9-NEXT: s_mov_b32 s12, s14 48; GFX9-NEXT: s_load_dwordx2 s[14:15], s[4:5], 0x4 49; GFX9-NEXT: s_add_u32 s0, s0, s17 50; GFX9-NEXT: s_addc_u32 s1, s1, 0 51; GFX9-NEXT: s_mov_b32 s32, 0 52; GFX9-NEXT: s_waitcnt lgkmcnt(0) 53; GFX9-NEXT: s_lshr_b32 s14, s14, 16 54; GFX9-NEXT: s_mul_i32 s14, s14, s15 55; GFX9-NEXT: v_mul_lo_u32 v3, s14, v0 56; GFX9-NEXT: s_getpc_b64 s[18:19] 57; GFX9-NEXT: s_add_u32 s18, s18, indirect@rel32@lo+4 58; GFX9-NEXT: s_addc_u32 s19, s19, indirect@rel32@hi+12 59; GFX9-NEXT: s_mov_b32 s14, s16 60; GFX9-NEXT: v_mad_u32_u24 v3, v1, s15, v3 61; GFX9-NEXT: v_add_lshl_u32 v5, v3, v2, 3 62; GFX9-NEXT: v_mov_b32_e32 v3, s18 63; GFX9-NEXT: v_lshlrev_b32_e32 v2, 20, v2 64; GFX9-NEXT: v_lshlrev_b32_e32 v1, 10, v1 65; GFX9-NEXT: v_mov_b32_e32 v4, s19 66; GFX9-NEXT: v_or3_b32 v31, v0, v1, v2 67; GFX9-NEXT: ds_write_b64 v5, v[3:4] 68; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] 69; GFX9-NEXT: s_endpgm 70 %fptr = alloca ptr, addrspace(5) 71 %fptr.cast = addrspacecast ptr addrspace(5) %fptr to ptr 72 store ptr @indirect, ptr %fptr.cast 73 %fp = load ptr, ptr %fptr.cast 74 call void %fp() 75 ret void 76} 77 78 79!llvm.module.flags = !{!0} 80!0 = !{i32 1, !"amdhsa_code_object_version", i32 500} 81;. 82; AKF_GCN: attributes #[[ATTR0]] = { "amdgpu-calls" "amdgpu-stack-objects" } 83;. 84; ATTRIBUTOR_GCN: attributes #[[ATTR0]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "uniform-work-group-size"="false" } 85; ATTRIBUTOR_GCN: attributes #[[ATTR1]] = { "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "uniform-work-group-size"="false" } 86;. 87; AKF_GCN: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500} 88;. 89; ATTRIBUTOR_GCN: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500} 90;. 91