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-prefixes=GCN,AKF_GCN %s 3; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-attributor %s | FileCheck -check-prefixes=GCN,ATTRIBUTOR_GCN %s 4 5; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s 6 7target datalayout = "A5" 8 9; GFX9-LABEL: {{^}}indirect: 10define internal void @indirect() { 11; GCN-LABEL: define {{[^@]+}}@indirect 12; GCN-SAME: () #[[ATTR0:[0-9]+]] { 13; GCN-NEXT: ret void 14; 15 ret void 16} 17 18; GFX9-LABEL: {{^}}test_simple_indirect_call: 19; GFX9: s_add_u32 flat_scratch_lo, s12, s17 20; GFX9-NEXT: s_addc_u32 flat_scratch_hi, s13, 0 21; GFX9-NEXT: s_mov_b32 s13, s15 22; GFX9-NEXT: s_mov_b32 s12, s14 23; GFX9-NEXT: s_load_dwordx2 s[14:15], s[4:5], 0x4 24; GFX9-NEXT: s_add_u32 s0, s0, s17 25; GFX9-NEXT: s_addc_u32 s1, s1, 0 26; GFX9-NEXT: s_mov_b32 s32, 0 27; GFX9-NEXT: s_waitcnt lgkmcnt(0) 28; GFX9-NEXT: s_lshr_b32 s14, s14, 16 29; GFX9-NEXT: s_mul_i32 s14, s14, s15 30; GFX9-NEXT: v_mul_lo_u32 v3, s14, v0 31; GFX9-NEXT: s_getpc_b64 s[18:19] 32; GFX9-NEXT: s_add_u32 s18, s18, indirect@rel32@lo+4 33; GFX9-NEXT: s_addc_u32 s19, s19, indirect@rel32@hi+12 34; GFX9-NEXT: s_mov_b32 s14, s16 35; GFX9-NEXT: v_mad_u32_u24 v3, v1, s15, v3 36; GFX9-NEXT: v_add_lshl_u32 v5, v3, v2, 3 37; GFX9-NEXT: v_mov_b32_e32 v3, s18 38; GFX9-NEXT: v_lshlrev_b32_e32 v2, 20, v2 39; GFX9-NEXT: v_lshlrev_b32_e32 v1, 10, v1 40; GFX9-NEXT: v_mov_b32_e32 v4, s19 41; GFX9-NEXT: v_or3_b32 v0, v0, v1, v2 42; GFX9-NEXT: ds_write_b64 v5, v[3:4] 43; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] 44; GFX9-NEXT: s_endpgm 45define amdgpu_kernel void @test_simple_indirect_call() { 46; AKF_GCN-LABEL: define {{[^@]+}}@test_simple_indirect_call 47; AKF_GCN-SAME: () #[[ATTR1:[0-9]+]] { 48; AKF_GCN-NEXT: [[FPTR:%.*]] = alloca void ()*, align 8, addrspace(5) 49; AKF_GCN-NEXT: [[FPTR_CAST:%.*]] = addrspacecast void ()* addrspace(5)* [[FPTR]] to void ()** 50; AKF_GCN-NEXT: store void ()* @indirect, void ()** [[FPTR_CAST]], align 8 51; AKF_GCN-NEXT: [[FP:%.*]] = load void ()*, void ()** [[FPTR_CAST]], align 8 52; AKF_GCN-NEXT: call void [[FP]]() 53; AKF_GCN-NEXT: ret void 54; 55; ATTRIBUTOR_GCN-LABEL: define {{[^@]+}}@test_simple_indirect_call 56; ATTRIBUTOR_GCN-SAME: () #[[ATTR1:[0-9]+]] { 57; ATTRIBUTOR_GCN-NEXT: call void @indirect() 58; ATTRIBUTOR_GCN-NEXT: ret void 59; 60 %fptr = alloca void()*, addrspace(5) 61 %fptr.cast = addrspacecast void()* addrspace(5)* %fptr to void()** 62 store void()* @indirect, void()** %fptr.cast 63 %fp = load void()*, void()** %fptr.cast 64 call void %fp() 65 ret void 66} 67 68;. 69; AKF_GCN: attributes #[[ATTR0]] = { "amdgpu-dispatch-id" "amdgpu-dispatch-ptr" "amdgpu-implicitarg-ptr" "amdgpu-queue-ptr" "amdgpu-work-group-id-x" "amdgpu-work-group-id-y" "amdgpu-work-group-id-z" "amdgpu-work-item-id-x" "amdgpu-work-item-id-y" "amdgpu-work-item-id-z" } 70; AKF_GCN: attributes #[[ATTR1]] = { "amdgpu-calls" "amdgpu-dispatch-id" "amdgpu-dispatch-ptr" "amdgpu-implicitarg-ptr" "amdgpu-queue-ptr" "amdgpu-stack-objects" "amdgpu-work-group-id-x" "amdgpu-work-group-id-y" "amdgpu-work-group-id-z" "amdgpu-work-item-id-x" "amdgpu-work-item-id-y" "amdgpu-work-item-id-z" } 71;. 72; ATTRIBUTOR_GCN: attributes #[[ATTR0]] = { "amdgpu-dispatch-id" "amdgpu-dispatch-ptr" "amdgpu-implicitarg-ptr" "amdgpu-queue-ptr" "amdgpu-work-group-id-x" "amdgpu-work-group-id-y" "amdgpu-work-group-id-z" "amdgpu-work-item-id-x" "amdgpu-work-item-id-y" "amdgpu-work-item-id-z" "uniform-work-group-size"="false" } 73; ATTRIBUTOR_GCN: attributes #[[ATTR1]] = { "amdgpu-calls" "amdgpu-dispatch-id" "amdgpu-dispatch-ptr" "amdgpu-implicitarg-ptr" "amdgpu-queue-ptr" "amdgpu-stack-objects" "amdgpu-work-group-id-x" "amdgpu-work-group-id-y" "amdgpu-work-group-id-z" "amdgpu-work-item-id-x" "amdgpu-work-item-id-y" "amdgpu-work-item-id-z" "uniform-work-group-size"="false" } 74;. 75