Skip to content

Commit abfcc87

Browse files
skc7skc7
committed
[AMDGPU] Introduce amdgpu-sw-lower-lds pass to lower LDS accesses. (#87265)
1 parent 0916ae4 commit abfcc87

27 files changed

+5278
-0
lines changed

‎llvm/lib/Target/AMDGPU/AMDGPU.h

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -268,6 +268,17 @@ struct AMDGPUAlwaysInlinePass : PassInfoMixin<AMDGPUAlwaysInlinePass> {
268268
bool GlobalOpt;
269269
};
270270

271+
void initializeAMDGPUSwLowerLDSLegacyPass(PassRegistry &);
272+
extern char &AMDGPUSwLowerLDSLegacyPassID;
273+
ModulePass *
274+
createAMDGPUSwLowerLDSLegacyPass(const AMDGPUTargetMachine *TM = nullptr);
275+
276+
struct AMDGPUSwLowerLDSPass : PassInfoMixin<AMDGPUSwLowerLDSPass> {
277+
const AMDGPUTargetMachine &TM;
278+
AMDGPUSwLowerLDSPass(const AMDGPUTargetMachine &TM_) : TM(TM_) {}
279+
PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
280+
};
281+
271282
class AMDGPUCodeGenPreparePass
272283
: public PassInfoMixin<AMDGPUCodeGenPreparePass> {
273284
private:

‎llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,7 @@ MODULE_PASS("amdgpu-always-inline", AMDGPUAlwaysInlinePass())
2020
MODULE_PASS("amdgpu-lower-buffer-fat-pointers",
2121
AMDGPULowerBufferFatPointersPass(*this))
2222
MODULE_PASS("amdgpu-lower-ctor-dtor", AMDGPUCtorDtorLoweringPass())
23+
MODULE_PASS("amdgpu-sw-lower-lds", AMDGPUSwLowerLDSPass(*this))
2324
MODULE_PASS("amdgpu-lower-module-lds", AMDGPULowerModuleLDSPass(*this))
2425
MODULE_PASS("amdgpu-perf-hint",
2526
AMDGPUPerfHintAnalysisPass(

‎llvm/lib/Target/AMDGPU/AMDGPUSwLowerLDS.cpp

Lines changed: 1335 additions & 0 deletions
Large diffs are not rendered by default.

‎llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -412,6 +412,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
412412
initializeSILoadStoreOptimizerPass(*PR);
413413
initializeAMDGPUCtorDtorLoweringLegacyPass(*PR);
414414
initializeAMDGPUAlwaysInlinePass(*PR);
415+
initializeAMDGPUSwLowerLDSLegacyPass(*PR);
415416
initializeAMDGPUAttributorLegacyPass(*PR);
416417
initializeAMDGPUAnnotateKernelFeaturesPass(*PR);
417418
initializeAMDGPUAnnotateUniformValuesLegacyPass(*PR);

‎llvm/lib/Target/AMDGPU/CMakeLists.txt

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -74,6 +74,7 @@ add_llvm_target(AMDGPUCodeGen
7474
AMDGPULowerKernelArguments.cpp
7575
AMDGPULowerKernelAttributes.cpp
7676
AMDGPULowerModuleLDSPass.cpp
77+
AMDGPUSwLowerLDS.cpp
7778
AMDGPUMachineFunction.cpp
7879
AMDGPUMachineModuleInfo.cpp
7980
AMDGPUMacroFusion.cpp

‎llvm/test/CodeGen/AMDGPU/amdgpu-sw-lower-lds-dynamic-indirect-access-asan.ll

Lines changed: 260 additions & 0 deletions
Large diffs are not rendered by default.

‎llvm/test/CodeGen/AMDGPU/amdgpu-sw-lower-lds-dynamic-indirect-access.ll

Lines changed: 142 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,142 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 4
2+
; RUN: opt < %s -passes=amdgpu-sw-lower-lds -amdgpu-asan-instrument-lds=false -S -mtriple=amdgcn-amd-amdhsa | FileCheck %s
3+
4+
; Test to check indirect dynamic LDS access through a non-kernel from kernel is lowered correctly.
5+
@lds_1 = internal addrspace(3) global [1 x i8] poison, align 1
6+
@lds_2 = internal addrspace(3) global [1 x i32] poison, align 2
7+
@lds_3 = external addrspace(3) global [0 x i8], align 4
8+
@lds_4 = external addrspace(3) global [0 x i8], align 8
9+
10+
;.
11+
; CHECK: @llvm.amdgcn.sw.lds.k0 = internal addrspace(3) global ptr poison, no_sanitize_address, align 8, !absolute_symbol [[META0:![0-9]+]]
12+
; CHECK: @llvm.amdgcn.k0.dynlds = external addrspace(3) global [0 x i8], no_sanitize_address, align 8, !absolute_symbol [[META1:![0-9]+]]
13+
; 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 1, i32 32 }, %llvm.amdgcn.sw.lds.k0.md.item { i32 64, i32 4, i32 32 }, %llvm.amdgcn.sw.lds.k0.md.item { i32 96, i32 0, i32 32 }, %llvm.amdgcn.sw.lds.k0.md.item { i32 128, i32 0, i32 32 } }, no_sanitize_address
14+
; @llvm.amdgcn.sw.lds.base.table = internal addrspace(1) constant [1 x ptr addrspace(3)] [ptr addrspace(3) @llvm.amdgcn.sw.lds.k0], no_sanitize_address
15+
; @llvm.amdgcn.sw.lds.offset.table = internal addrspace(1) constant [1 x [2 x ptr addrspace(1)]] [[2 x ptr addrspace(1)] [ptr addrspace(1) getelementptr inbounds (%llvm.amdgcn.sw.lds.k0.md.type, ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 3, i32 0), ptr addrspace(1) getelementptr inbounds (%llvm.amdgcn.sw.lds.k0.md.type, ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 4, i32 0)]], no_sanitize_address
16+
;.
17+
define void @use_variables() sanitize_address {
18+
; CHECK-LABEL: define void @use_variables(
19+
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
20+
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()
21+
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x ptr addrspace(3)], ptr addrspace(1) @llvm.amdgcn.sw.lds.base.table, i32 0, i32 [[TMP1]]
22+
; CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(3), ptr addrspace(1) [[TMP2]], align 4
23+
; CHECK-NEXT: [[TMP7:%.*]] = load ptr addrspace(1), ptr addrspace(3) [[TMP4]], align 8
24+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x [2 x ptr addrspace(1)]], ptr addrspace(1) @llvm.amdgcn.sw.lds.offset.table, i32 0, i32 [[TMP1]], i32 0
25+
; CHECK-NEXT: [[TMP5:%.*]] = load ptr addrspace(1), ptr addrspace(1) [[TMP6]], align 8
26+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(1) [[TMP5]], align 4
27+
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr addrspace(3) [[TMP4]], i32 [[TMP8]]
28+
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x [2 x ptr addrspace(1)]], ptr addrspace(1) @llvm.amdgcn.sw.lds.offset.table, i32 0, i32 [[TMP1]], i32 1
29+
; CHECK-NEXT: [[TMP12:%.*]] = load ptr addrspace(1), ptr addrspace(1) [[TMP11]], align 8
30+
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr addrspace(1) [[TMP12]], align 4
31+
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i8, ptr addrspace(3) [[TMP4]], i32 [[TMP10]]
32+
; CHECK-NEXT: [[TMP13:%.*]] = ptrtoint ptr addrspace(3) [[TMP9]] to i32
33+
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP7]], i32 [[TMP13]]
34+
; CHECK-NEXT: store i8 3, ptr addrspace(1) [[TMP14]], align 4
35+
; CHECK-NEXT: [[TMP30:%.*]] = ptrtoint ptr addrspace(3) [[TMP15]] to i32
36+
; CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP7]], i32 [[TMP30]]
37+
; CHECK-NEXT: store i8 3, ptr addrspace(1) [[TMP31]], align 8
38+
; CHECK-NEXT: ret void
39+
;
40+
store i8 3, ptr addrspace(3) @lds_3, align 4
41+
store i8 3, ptr addrspace(3) @lds_4, align 8
42+
ret void
43+
}
44+
45+
define amdgpu_kernel void @k0() sanitize_address {
46+
; CHECK-LABEL: define amdgpu_kernel void @k0(
47+
; CHECK-SAME: ) #[[ATTR1:[0-9]+]] !llvm.amdgcn.lds.kernel.id [[META2:![0-9]+]] {
48+
; CHECK-NEXT: WId:
49+
; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
50+
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
51+
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
52+
; CHECK-NEXT: [[TMP3:%.*]] = or i32 [[TMP0]], [[TMP1]]
53+
; CHECK-NEXT: [[TMP4:%.*]] = or i32 [[TMP3]], [[TMP2]]
54+
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0
55+
; CHECK-NEXT: br i1 [[TMP5]], label [[MALLOC:%.*]], label [[TMP21:%.*]]
56+
; CHECK: Malloc:
57+
; CHECK-NEXT: [[TMP9:%.*]] = 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 2, i32 0), align 4
58+
; 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 2, i32 2), align 4
59+
; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[TMP9]], [[TMP7]]
60+
; CHECK-NEXT: [[TMP6:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
61+
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds ptr addrspace(4), ptr addrspace(4) [[TMP6]], i64 15
62+
; CHECK-NEXT: store i32 [[TMP8]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 3, i32 0), align 4
63+
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr addrspace(4) [[TMP10]], align 4
64+
; CHECK-NEXT: store i32 [[TMP11]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 3, i32 1), align 4
65+
; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[TMP11]], 7
66+
; CHECK-NEXT: [[TMP13:%.*]] = udiv i32 [[TMP12]], 8
67+
; CHECK-NEXT: [[TMP14:%.*]] = mul i32 [[TMP13]], 8
68+
; CHECK-NEXT: store i32 [[TMP14]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 3, i32 2), align 4
69+
; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[TMP8]], [[TMP14]]
70+
; CHECK-NEXT: store i32 [[TMP15]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 4, i32 0), align 4
71+
; CHECK-NEXT: [[TMP27:%.*]] = load i32, ptr addrspace(4) [[TMP10]], align 4
72+
; CHECK-NEXT: store i32 [[TMP27]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 4, i32 1), align 4
73+
; CHECK-NEXT: [[TMP17:%.*]] = add i32 [[TMP27]], 7
74+
; CHECK-NEXT: [[TMP18:%.*]] = udiv i32 [[TMP17]], 8
75+
; CHECK-NEXT: [[TMP19:%.*]] = mul i32 [[TMP18]], 8
76+
; CHECK-NEXT: store i32 [[TMP19]], ptr addrspace(1) getelementptr inbounds ([[LLVM_AMDGCN_SW_LDS_K0_MD_TYPE]], ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, i32 0, i32 4, i32 2), align 4
77+
; CHECK-NEXT: [[TMP28:%.*]] = add i32 [[TMP15]], [[TMP19]]
78+
; CHECK-NEXT: [[TMP26:%.*]] = zext i32 [[TMP28]] to i64
79+
; CHECK-NEXT: [[TMP22:%.*]] = call ptr @llvm.returnaddress(i32 0)
80+
; CHECK-NEXT: [[TMP23:%.*]] = ptrtoint ptr [[TMP22]] to i64
81+
; CHECK-NEXT: [[TMP35:%.*]] = call i64 @__asan_malloc_impl(i64 [[TMP26]], i64 [[TMP23]])
82+
; CHECK-NEXT: [[TMP20:%.*]] = inttoptr i64 [[TMP35]] to ptr addrspace(1)
83+
; CHECK-NEXT: store ptr addrspace(1) [[TMP20]], ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 8
84+
; CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP20]], i64 8
85+
; CHECK-NEXT: [[TMP37:%.*]] = ptrtoint ptr addrspace(1) [[TMP36]] to i64
86+
; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP37]], i64 24)
87+
; CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP20]], i64 33
88+
; CHECK-NEXT: [[TMP73:%.*]] = ptrtoint ptr addrspace(1) [[TMP53]] to i64
89+
; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP73]], i64 31)
90+
; CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP20]], i64 68
91+
; CHECK-NEXT: [[TMP75:%.*]] = ptrtoint ptr addrspace(1) [[TMP74]] to i64
92+
; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP75]], i64 28)
93+
; CHECK-NEXT: br label [[TMP21]]
94+
; CHECK: 32:
95+
; CHECK-NEXT: [[XYZCOND:%.*]] = phi i1 [ false, [[WID:%.*]] ], [ true, [[MALLOC]] ]
96+
; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()
97+
; CHECK-NEXT: [[TMP31:%.*]] = load ptr addrspace(1), ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 8
98+
; CHECK-NEXT: [[TMP24:%.*]] = 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 4
99+
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i8, ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, i32 [[TMP24]]
100+
; CHECK-NEXT: [[TMP29:%.*]] = 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 2, i32 0), align 4
101+
; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds i8, ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, i32 [[TMP29]]
102+
; CHECK-NEXT: call void @llvm.donothing() [ "ExplicitUse"(ptr addrspace(3) @llvm.amdgcn.k0.dynlds) ]
103+
; CHECK-NEXT: call void @use_variables()
104+
; CHECK-NEXT: [[TMP38:%.*]] = ptrtoint ptr addrspace(3) [[TMP25]] to i32
105+
; CHECK-NEXT: [[TMP39:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP31]], i32 [[TMP38]]
106+
; CHECK-NEXT: store i8 7, ptr addrspace(1) [[TMP39]], align 1
107+
; CHECK-NEXT: [[TMP55:%.*]] = ptrtoint ptr addrspace(3) [[TMP30]] to i32
108+
; CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP31]], i32 [[TMP55]]
109+
; CHECK-NEXT: store i32 8, ptr addrspace(1) [[TMP56]], align 2
110+
; CHECK-NEXT: br label [[CONDFREE1:%.*]]
111+
; CHECK: CondFree:
112+
; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()
113+
; CHECK-NEXT: br i1 [[XYZCOND]], label [[FREE:%.*]], label [[END:%.*]]
114+
; CHECK: Free:
115+
; CHECK-NEXT: [[TMP32:%.*]] = call ptr @llvm.returnaddress(i32 0)
116+
; CHECK-NEXT: [[TMP33:%.*]] = ptrtoint ptr [[TMP32]] to i64
117+
; CHECK-NEXT: [[TMP34:%.*]] = ptrtoint ptr addrspace(1) [[TMP31]] to i64
118+
; CHECK-NEXT: call void @__asan_free_impl(i64 [[TMP34]], i64 [[TMP33]])
119+
; CHECK-NEXT: br label [[END]]
120+
; CHECK: End:
121+
; CHECK-NEXT: ret void
122+
;
123+
call void @use_variables()
124+
store i8 7, ptr addrspace(3) @lds_1, align 1
125+
store i32 8, ptr addrspace(3) @lds_2, align 2
126+
ret void
127+
}
128+
129+
!llvm.module.flags = !{!0}
130+
!0 = !{i32 4, !"nosanitize_address", i32 1}
131+
132+
;.
133+
; CHECK: attributes #[[ATTR0]] = { sanitize_address }
134+
; CHECK: attributes #[[ATTR1]] = { sanitize_address "amdgpu-lds-size"="8,8" }
135+
; CHECK: attributes #[[ATTR2:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(none) }
136+
; CHECK: attributes #[[ATTR3:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
137+
; CHECK: attributes #[[ATTR4:[0-9]+]] = { convergent nocallback nofree nounwind willreturn }
138+
;.
139+
; CHECK: [[META0]] = !{i32 0, i32 1}
140+
; CHECK: [[META1]] = !{i32 8, i32 9}
141+
; CHECK: [[META2]] = !{i32 0}
142+
;.

‎llvm/test/CodeGen/AMDGPU/amdgpu-sw-lower-lds-dynamic-lds-test-asan.ll

Lines changed: 113 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,113 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 4
2+
; RUN: opt < %s -passes=amdgpu-sw-lower-lds -S -mtriple=amdgcn-amd-amdhsa | FileCheck %s
3+
4+
; Test to check if direct access of dynamic LDS in kernel is lowered correctly.
5+
@lds_1 = external addrspace(3) global [0 x i8]
6+
@lds_2 = external addrspace(3) global [0 x i8]
7+
8+
;.
9+
; CHECK: @llvm.amdgcn.sw.lds.k0 = internal addrspace(3) global ptr poison, no_sanitize_address, align 1, !absolute_symbol [[META0:![0-9]+]]
10+
; CHECK: @llvm.amdgcn.k0.dynlds = external addrspace(3) global [0 x i8], no_sanitize_address, align 1, !absolute_symbol [[META1:![0-9]+]]
11+
; 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 0, i32 32 } }, no_sanitize_address
12+
;.
13+
define amdgpu_kernel void @k0() sanitize_address {
14+
; CHECK-LABEL: define amdgpu_kernel void @k0(
15+
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
16+
; CHECK-NEXT: WId:
17+
; CHECK-NEXT: [[TMP0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
18+
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
19+
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
20+
; CHECK-NEXT: [[TMP3:%.*]] = or i32 [[TMP0]], [[TMP1]]
21+
; CHECK-NEXT: [[TMP4:%.*]] = or i32 [[TMP3]], [[TMP2]]
22+
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0
23+
; CHECK-NEXT: br i1 [[TMP5]], label [[MALLOC:%.*]], label [[TMP7:%.*]]
24+
; CHECK: Malloc:
25+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(1) @llvm.amdgcn.sw.lds.k0.md, align 4
26+
; CHECK-NEXT: [[TMP9:%.*]] = 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 0, i32 2), align 4
27+
; CHECK-NEXT: [[TMP24:%.*]] = add i32 [[TMP8]], [[TMP9]]
28+
; CHECK-NEXT: [[TMP20:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
29+
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds ptr addrspace(4), ptr addrspace(4) [[TMP20]], i64 15
30+
; CHECK-NEXT: store i32 [[TMP24]], 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 4
31+
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr addrspace(4) [[TMP18]], align 4
32+
; CHECK-NEXT: store i32 [[TMP13]], 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 1), align 4
33+
; CHECK-NEXT: [[TMP14:%.*]] = add i32 [[TMP13]], 0
34+
; CHECK-NEXT: [[TMP15:%.*]] = udiv i32 [[TMP14]], 1
35+
; CHECK-NEXT: [[TMP16:%.*]] = mul i32 [[TMP15]], 1
36+
; CHECK-NEXT: store i32 [[TMP16]], 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 4
37+
; CHECK-NEXT: [[TMP17:%.*]] = add i32 [[TMP24]], [[TMP16]]
38+
; CHECK-NEXT: [[TMP21:%.*]] = zext i32 [[TMP17]] to i64
39+
; CHECK-NEXT: [[TMP22:%.*]] = call ptr @llvm.returnaddress(i32 0)
40+
; CHECK-NEXT: [[TMP23:%.*]] = ptrtoint ptr [[TMP22]] to i64
41+
; CHECK-NEXT: [[TMP19:%.*]] = call i64 @__asan_malloc_impl(i64 [[TMP21]], i64 [[TMP23]])
42+
; CHECK-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP19]] to ptr addrspace(1)
43+
; CHECK-NEXT: store ptr addrspace(1) [[TMP6]], ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 8
44+
; CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP6]], i64 8
45+
; CHECK-NEXT: [[TMP44:%.*]] = ptrtoint ptr addrspace(1) [[TMP42]] to i64
46+
; CHECK-NEXT: call void @__asan_poison_region(i64 [[TMP44]], i64 24)
47+
; CHECK-NEXT: br label [[TMP7]]
48+
; CHECK: 23:
49+
; CHECK-NEXT: [[XYZCOND:%.*]] = phi i1 [ false, [[WID:%.*]] ], [ true, [[MALLOC]] ]
50+
; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()
51+
; CHECK-NEXT: [[TMP28:%.*]] = load ptr addrspace(1), ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, align 8
52+
; CHECK-NEXT: [[TMP10:%.*]] = 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 4
53+
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr addrspace(3) @llvm.amdgcn.sw.lds.k0, i32 [[TMP10]]
54+
; CHECK-NEXT: call void @llvm.donothing() [ "ExplicitUse"(ptr addrspace(3) @llvm.amdgcn.k0.dynlds) ]
55+
; CHECK-NEXT: [[TMP45:%.*]] = ptrtoint ptr addrspace(3) [[TMP11]] to i32
56+
; CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP28]], i32 [[TMP45]]
57+
; CHECK-NEXT: [[TMP29:%.*]] = ptrtoint ptr addrspace(1) [[TMP46]] to i64
58+
; CHECK-NEXT: [[TMP30:%.*]] = lshr i64 [[TMP29]], 3
59+
; CHECK-NEXT: [[TMP31:%.*]] = add i64 [[TMP30]], 2147450880
60+
; CHECK-NEXT: [[TMP32:%.*]] = inttoptr i64 [[TMP31]] to ptr
61+
; CHECK-NEXT: [[TMP33:%.*]] = load i8, ptr [[TMP32]], align 1
62+
; CHECK-NEXT: [[TMP34:%.*]] = icmp ne i8 [[TMP33]], 0
63+
; CHECK-NEXT: [[TMP35:%.*]] = and i64 [[TMP29]], 7
64+
; CHECK-NEXT: [[TMP36:%.*]] = trunc i64 [[TMP35]] to i8
65+
; CHECK-NEXT: [[TMP37:%.*]] = icmp sge i8 [[TMP36]], [[TMP33]]
66+
; CHECK-NEXT: [[TMP38:%.*]] = and i1 [[TMP34]], [[TMP37]]
67+
; CHECK-NEXT: [[TMP39:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 [[TMP38]])
68+
; CHECK-NEXT: [[TMP40:%.*]] = icmp ne i64 [[TMP39]], 0
69+
; CHECK-NEXT: br i1 [[TMP40]], label [[ASAN_REPORT:%.*]], label [[TMP43:%.*]], !prof [[PROF2:![0-9]+]]
70+
; CHECK: asan.report:
71+
; CHECK-NEXT: br i1 [[TMP38]], label [[TMP41:%.*]], label [[CONDFREE:%.*]]
72+
; CHECK: 41:
73+
; CHECK-NEXT: call void @__asan_report_store1(i64 [[TMP29]]) #[[ATTR6:[0-9]+]]
74+
; CHECK-NEXT: call void @llvm.amdgcn.unreachable()
75+
; CHECK-NEXT: br label [[CONDFREE]]
76+
; CHECK: 42:
77+
; CHECK-NEXT: br label [[TMP43]]
78+
; CHECK: 43:
79+
; CHECK-NEXT: store i8 7, ptr addrspace(1) [[TMP46]], align 4
80+
; CHECK-NEXT: br label [[CONDFREE1:%.*]]
81+
; CHECK: CondFree:
82+
; CHECK-NEXT: call void @llvm.amdgcn.s.barrier()
83+
; CHECK-NEXT: br i1 [[XYZCOND]], label [[FREE:%.*]], label [[END:%.*]]
84+
; CHECK: Free:
85+
; CHECK-NEXT: [[TMP25:%.*]] = call ptr @llvm.returnaddress(i32 0)
86+
; CHECK-NEXT: [[TMP26:%.*]] = ptrtoint ptr [[TMP25]] to i64
87+
; CHECK-NEXT: [[TMP27:%.*]] = ptrtoint ptr addrspace(1) [[TMP28]] to i64
88+
; CHECK-NEXT: call void @__asan_free_impl(i64 [[TMP27]], i64 [[TMP26]])
89+
; CHECK-NEXT: br label [[END]]
90+
; CHECK: End:
91+
; CHECK-NEXT: ret void
92+
;
93+
store i8 7, ptr addrspace(3) @lds_1, align 4
94+
;store i8 8, ptr addrspace(3) @lds_2, align 8
95+
ret void
96+
}
97+
98+
!llvm.module.flags = !{!0}
99+
!0 = !{i32 4, !"nosanitize_address", i32 1}
100+
101+
;.
102+
; CHECK: attributes #[[ATTR0]] = { sanitize_address "amdgpu-lds-size"="8,8" }
103+
; CHECK: attributes #[[ATTR1:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(none) }
104+
; CHECK: attributes #[[ATTR2:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
105+
; CHECK: attributes #[[ATTR3:[0-9]+]] = { convergent nocallback nofree nounwind willreturn }
106+
; CHECK: attributes #[[ATTR4:[0-9]+]] = { convergent nocallback nofree nounwind willreturn memory(none) }
107+
; CHECK: attributes #[[ATTR5:[0-9]+]] = { convergent nocallback nofree nounwind }
108+
; CHECK: attributes #[[ATTR6]] = { nomerge }
109+
;.
110+
; CHECK: [[META0]] = !{i32 0, i32 1}
111+
; CHECK: [[META1]] = !{i32 8, i32 9}
112+
; CHECK: [[PROF2]] = !{!"branch_weights", i32 1, i32 1048575}
113+
;.

0 commit comments

Comments
 (0)