77 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 52; RUN: opt < %s -passes=amdgpu-sw-lower-lds -S -mtriple=amdgcn-amd-amdhsa | FileCheck %s3@lds = internal addrspace(3) global [5 x i32] poison, align 164 5;.6; CHECK: @llvm.amdgcn.sw.lds.k0 = internal addrspace(3) global ptr poison, no_sanitize_address, align 16, !absolute_symbol [[META0:![0-9]+]]7; CHECK: @llvm.amdgcn.sw.lds.k0.md = internal addrspace(1) global %llvm.amdgcn.sw.lds.k0.md.type { %llvm.amdgcn.sw.lds.k0.md.item { i32 0, i32 8, i32 32 }, %llvm.amdgcn.sw.lds.k0.md.item { i32 32, i32 20, i32 64 } }, no_sanitize_address8;.9define amdgpu_kernel void @k0() sanitize_address {10; CHECK-LABEL: define amdgpu_kernel void @k0(11; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {12; CHECK-NEXT: [[WID:.*]]:13; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()14; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()15; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()16; CHECK-NEXT: [[TMP3:%.*]] = or i32 [[TMP0]], [[TMP1]]17; CHECK-NEXT: [[TMP4:%.*]] = or i32 [[TMP3]], [[TMP2]]18; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 019; CHECK-NEXT: br i1 [[TMP5]], label %[[MALLOC:.*]], label %[[BB18:.*]]20; CHECK: [[MALLOC]]:21; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE:%.*]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 1, i32 0), align 422; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 1, i32 2), align 423; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[TMP6]], [[TMP7]]24; CHECK-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i6425; CHECK-NEXT: [[TMP10:%.*]] = call ptr @llvm.returnaddress(i32 0)26; CHECK-NEXT: [[TMP11:%.*]] = ptrtoint ptr [[TMP10]] to i6427; CHECK-NEXT: [[TMP12:%.*]] = call i64 @__asan_malloc_impl(i64 [[TMP9]], i64 [[TMP11]])28; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr addrspace(1)29; CHECK-NEXT: store ptr addrspace(1) [[TMP13]], ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 830; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP13]], i64 831; CHECK-NEXT: [[TMP15:%.*]] = ptrtoint ptr addrspace(1) [[TMP14]] to i6432; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP15]], i64 24)33; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP13]], i64 5234; CHECK-NEXT: [[TMP17:%.*]] = ptrtoint ptr addrspace(1) [[TMP16]] to i6435; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP17]], i64 44)36; CHECK-NEXT: br label %[[BB18]]37; CHECK: [[BB18]]:38; CHECK-NEXT: [[XYZCOND:%.*]] = phi i1 [ false, %[[WID]] ], [ true, %[[MALLOC]] ]39; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()40; CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(1), ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 841; CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 1, i32 0), align 442; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i8, ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, i32 [[TMP20]]43; CHECK-NEXT: [[TMP22:%.*]] = ptrtoint ptr addrspace(3) [[TMP21]] to i3244; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP19]], i32 [[TMP22]]45; CHECK-NEXT: [[TMP24:%.*]] = addrspacecast ptr addrspace(1) [[TMP23]] to ptr46; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds [5 x i32], ptr [[TMP24]], i64 0, i64 047; CHECK-NEXT: store i32 1, ptr [[GEP]], align 448; CHECK-NEXT: br label %[[CONDFREE:.*]]49; CHECK: [[CONDFREE]]:50; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()51; CHECK-NEXT: br i1 [[XYZCOND]], label %[[FREE:.*]], label %[[END:.*]]52; CHECK: [[FREE]]:53; CHECK-NEXT: [[TMP25:%.*]] = call ptr @llvm.returnaddress(i32 0)54; CHECK-NEXT: [[TMP26:%.*]] = ptrtoint ptr [[TMP25]] to i6455; CHECK-NEXT: [[TMP27:%.*]] = ptrtoint ptr addrspace(1) [[TMP19]] to i6456; CHECK-NEXT: call void @__asan_free_impl(i64 [[TMP27]], i64 [[TMP26]])57; CHECK-NEXT: br label %[[END]]58; CHECK: [[END]]:59; CHECK-NEXT: ret void60;61 %gep = getelementptr inbounds [5 x i32], ptr addrspacecast (ptr addrspace(3) @lds to ptr), i64 0, i64 062 store i32 1, ptr %gep, align 463 ret void64}65 66!llvm.module.flags = !{!0}67!0 = !{i32 4, !"nosanitize_address", i32 1}68;.69; CHECK: attributes #[[ATTR0]] = { sanitize_address "amdgpu-lds-size"="16" }70; CHECK: attributes #[[ATTR1:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }71; CHECK: attributes #[[ATTR2:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(none) }72; CHECK: attributes #[[ATTR3:[0-9]+]] = { convergent nocallback nofree nounwind willreturn }73;.74; CHECK: [[META0]] = !{i32 0, i32 1}75; CHECK: [[META1:![0-9]+]] = !{i32 4, !"nosanitize_address", i32 1}76;.77