102 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 52; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-sw-lower-lds < %s | FileCheck %s3@lds = external addrspace(3) global [5 x i8], align 84declare void @non_kernel_declaration() sanitize_address5 6;.7; CHECK: @llvm.amdgcn.sw.lds.k1 = internal addrspace(3) global ptr poison, no_sanitize_address, align 8, !absolute_symbol [[META0:![0-9]+]]8; CHECK: @llvm.amdgcn.sw.lds.k1.md = internal addrspace(1) global %llvm.amdgcn.sw.lds.k1.md.type { %llvm.amdgcn.sw.lds.k1.md.item { i32 0, i32 8, i32 32 }, %llvm.amdgcn.sw.lds.k1.md.item { i32 32, i32 5, i32 32 } }, no_sanitize_address9; @llvm.amdgcn.sw.lds.base.table = internal addrspace(1) constant [1 x ptr addrspace(3)] [ptr addrspace(3) @llvm.amdgcn.sw.lds.k1], no_sanitize_address10; @llvm.amdgcn.sw.lds.offset.table = internal addrspace(1) constant [1 x [1 x ptr addrspace(1)]] [[1 x ptr addrspace(1)] [ptr addrspace(1) getelementptr inbounds (%llvm.amdgcn.sw.lds.k1.md.type, ptr addrspace(1) @llvm.amdgcn.sw.lds.k1.md, i32 0, i32 1, i32 0)]], no_sanitize_address11;.12define void @non_kernel_function() sanitize_address {13; CHECK-LABEL: define void @non_kernel_function(14; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {15; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()16; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x ptr addrspace(3)], ptr addrspace(1) @llvm.amdgcn.sw.lds.base.table, i32 0, i32 [[TMP1]]17; CHECK-NEXT: [[TMP3:%.*]] = load ptr addrspace(3), ptr addrspace(1) [[TMP2]], align 418; CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(1), ptr addrspace(3) [[TMP3]], align 819; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x [1 x ptr addrspace(1)]], ptr addrspace(1) @llvm.amdgcn.sw.lds.offset.table, i32 0, i32 [[TMP1]], i32 020; CHECK-NEXT: [[TMP6:%.*]] = load ptr addrspace(1), ptr addrspace(1) [[TMP5]], align 821; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr addrspace(1) [[TMP6]], align 422; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr addrspace(3) [[TMP3]], i32 [[TMP7]]23; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr addrspace(3) [[TMP8]] to i3224; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP4]], i32 [[TMP14]]25; CHECK-NEXT: [[TMP11:%.*]] = addrspacecast ptr addrspace(1) [[TMP10]] to ptr26; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr addrspace(3) [[TMP8]] to i3227; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP4]], i32 [[TMP12]]28; CHECK-NEXT: [[TMP9:%.*]] = addrspacecast ptr addrspace(1) [[TMP13]] to ptr29; CHECK-NEXT: store i8 5, ptr [[TMP9]], align 830; CHECK-NEXT: ret void31;32 %Y = addrspacecast ptr addrspace(3) @lds to ptr33 store i8 5, ptr addrspacecast( ptr addrspace(3) @lds to ptr), align 834 ret void35}36 37define amdgpu_kernel void @k1() sanitize_address {38; CHECK-LABEL: define amdgpu_kernel void @k1(39; CHECK-SAME: ) #[[ATTR1:[0-9]+]] !llvm.amdgcn.lds.kernel.id [[META2:![0-9]+]] {40; CHECK-NEXT: [[WID:.*]]:41; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()42; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()43; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()44; CHECK-NEXT: [[TMP3:%.*]] = or i32 [[TMP0]], [[TMP1]]45; CHECK-NEXT: [[TMP4:%.*]] = or i32 [[TMP3]], [[TMP2]]46; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 047; CHECK-NEXT: br i1 [[TMP5]], label %[[MALLOC:.*]], label %[[BB18:.*]]48; CHECK: [[MALLOC]]:49; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K1_MD_TYPE:%.*]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k1.md, i32 0, i32 1, i32 0), align 450; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K1_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k1.md, i32 0, i32 1, i32 2), align 451; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[TMP6]], [[TMP7]]52; CHECK-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i6453; CHECK-NEXT: [[TMP10:%.*]] = call ptr @llvm.returnaddress(i32 0)54; CHECK-NEXT: [[TMP11:%.*]] = ptrtoint ptr [[TMP10]] to i6455; CHECK-NEXT: [[TMP12:%.*]] = call i64 @__asan_malloc_impl(i64 [[TMP9]], i64 [[TMP11]])56; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr addrspace(1)57; CHECK-NEXT: store ptr addrspace(1) [[TMP13]], ptr addrspace(3) @llvm.amdgcn.sw.lds.k1, align 858; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP13]], i64 859; CHECK-NEXT: [[TMP15:%.*]] = ptrtoint ptr addrspace(1) [[TMP14]] to i6460; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP15]], i64 24)61; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP13]], i64 3762; CHECK-NEXT: [[TMP17:%.*]] = ptrtoint ptr addrspace(1) [[TMP16]] to i6463; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP17]], i64 27)64; CHECK-NEXT: br label %[[BB18]]65; CHECK: [[BB18]]:66; CHECK-NEXT: [[XYZCOND:%.*]] = phi i1 [ false, %[[WID]] ], [ true, %[[MALLOC]] ]67; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()68; CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(1), ptr addrspace(3) @llvm.amdgcn.sw.lds.k1, align 869; CHECK-NEXT: call void @non_kernel_function()70; CHECK-NEXT: call void @non_kernel_declaration()71; CHECK-NEXT: br label %[[CONDFREE:.*]]72; CHECK: [[CONDFREE]]:73; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()74; CHECK-NEXT: br i1 [[XYZCOND]], label %[[FREE:.*]], label %[[END:.*]]75; CHECK: [[FREE]]:76; CHECK-NEXT: [[TMP20:%.*]] = call ptr @llvm.returnaddress(i32 0)77; CHECK-NEXT: [[TMP21:%.*]] = ptrtoint ptr [[TMP20]] to i6478; CHECK-NEXT: [[TMP22:%.*]] = ptrtoint ptr addrspace(1) [[TMP19]] to i6479; CHECK-NEXT: call void @__asan_free_impl(i64 [[TMP22]], i64 [[TMP21]])80; CHECK-NEXT: br label %[[END]]81; CHECK: [[END]]:82; CHECK-NEXT: ret void83;84 call void @non_kernel_function()85 call void @non_kernel_declaration()86 ret void87}88 89!llvm.module.flags = !{!0}90!0 = !{i32 4, !"nosanitize_address", i32 1}91;.92; CHECK: attributes #[[ATTR0]] = { sanitize_address }93; CHECK: attributes #[[ATTR1]] = { sanitize_address "amdgpu-lds-size"="8" }94; CHECK: attributes #[[ATTR2:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }95; CHECK: attributes #[[ATTR3:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(none) }96; CHECK: attributes #[[ATTR4:[0-9]+]] = { convergent nocallback nofree nounwind willreturn }97;.98; CHECK: [[META0]] = !{i32 0, i32 1}99; CHECK: [[META1:![0-9]+]] = !{i32 4, !"nosanitize_address", i32 1}100; CHECK: [[META2]] = !{i32 0}101;.102