Skip to content

Commit 2d67cb1

Browse files
skc7skc7
authored
[AMDGPU][Attributor] Stop inferring amdgpu-no-flat-scratch-init in sanitized functions. (#161319)
This PR stops the attributor pass to infer `amdgpu-no-flat-scratch-init` for functions marked with `sanitize_*` attribute.
1 parent 73651ba commit 2d67cb1

File tree

3 files changed

+38
-10
lines changed

3 files changed

+38
-10
lines changed

‎llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp‎

Lines changed: 12 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -131,10 +131,8 @@ static bool isDSAddress(const Constant *C) {
131131
return AS == AMDGPUAS::LOCAL_ADDRESS || AS == AMDGPUAS::REGION_ADDRESS;
132132
}
133133

134-
/// Returns true if the function requires the implicit argument be passed
135-
/// regardless of the function contents.
136-
static bool funcRequiresHostcallPtr(const Function &F) {
137-
// Sanitizers require the hostcall buffer passed in the implicit arguments.
134+
/// Returns true if sanitizer attributes are present on a function.
135+
static bool hasSanitizerAttributes(const Function &F) {
138136
return F.hasFnAttribute(Attribute::SanitizeAddress) ||
139137
F.hasFnAttribute(Attribute::SanitizeThread) ||
140138
F.hasFnAttribute(Attribute::SanitizeMemory) ||
@@ -469,15 +467,21 @@ struct AAAMDAttributesFunction : public AAAMDAttributes {
469467

470468
// If the function requires the implicit arg pointer due to sanitizers,
471469
// assume it's needed even if explicitly marked as not requiring it.
472-
const bool NeedsHostcall = funcRequiresHostcallPtr(*F);
473-
if (NeedsHostcall) {
470+
// Flat scratch initialization is needed because `asan_malloc_impl`
471+
// calls introduced later in pipeline will have flat scratch accesses.
472+
// FIXME: FLAT_SCRATCH_INIT will not be required here if device-libs
473+
// implementation for `asan_malloc_impl` is updated.
474+
const bool HasSanitizerAttrs = hasSanitizerAttributes(*F);
475+
if (HasSanitizerAttrs) {
474476
removeAssumedBits(IMPLICIT_ARG_PTR);
475477
removeAssumedBits(HOSTCALL_PTR);
478+
removeAssumedBits(FLAT_SCRATCH_INIT);
476479
}
477480

478481
for (auto Attr : ImplicitAttrs) {
479-
if (NeedsHostcall &&
480-
(Attr.first == IMPLICIT_ARG_PTR || Attr.first == HOSTCALL_PTR))
482+
if (HasSanitizerAttrs &&
483+
(Attr.first == IMPLICIT_ARG_PTR || Attr.first == HOSTCALL_PTR ||
484+
Attr.first == FLAT_SCRATCH_INIT))
481485
continue;
482486

483487
if (F->hasFnAttribute(Attr.second))

‎llvm/test/CodeGen/AMDGPU/amdgpu-attributor-flat-scratch-init-asan.ll‎

Lines changed: 24 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,24 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 6
2+
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -passes='amdgpu-attributor' %s -o - | FileCheck %s
3+
4+
@lds_1 = internal addrspace(3) global [1 x i8] poison, align 4
5+
6+
;.
7+
; CHECK: @lds_1 = internal addrspace(3) global [1 x i8] poison, align 4
8+
;.
9+
define amdgpu_kernel void @k0() #0 {
10+
; CHECK: Function Attrs: sanitize_address
11+
; CHECK-LABEL: define amdgpu_kernel void @k0(
12+
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
13+
; CHECK-NEXT: store i8 7, ptr addrspace(3) @lds_1, align 4
14+
; CHECK-NEXT: ret void
15+
;
16+
store i8 7, ptr addrspace(3) @lds_1, align 4
17+
ret void
18+
}
19+
20+
attributes #0 = { sanitize_address }
21+
; "amdgpu-no-flat-scratch-init" attribute should not be present in attribute list
22+
;.
23+
; CHECK: attributes #[[ATTR0]] = { sanitize_address "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-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" "target-cpu"="gfx90a" "uniform-work-group-size"="false" }
24+
;.

‎llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa-call.ll‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -707,8 +707,8 @@ attributes #6 = { "enqueued-block" }
707707
; ATTRIBUTOR_HSA: attributes #[[ATTR14]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "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-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "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" "target-cpu"="gfx900" "uniform-work-group-size"="false" }
708708
; ATTRIBUTOR_HSA: attributes #[[ATTR15:[0-9]+]] = { nounwind "uniform-work-group-size"="false" }
709709
; ATTRIBUTOR_HSA: attributes #[[ATTR16]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "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" }
710-
; ATTRIBUTOR_HSA: attributes #[[ATTR17]] = { nounwind sanitize_address "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "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-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" }
711-
; ATTRIBUTOR_HSA: attributes #[[ATTR18]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "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-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" }
710+
; ATTRIBUTOR_HSA: attributes #[[ATTR17]] = { nounwind sanitize_address "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-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" }
711+
; ATTRIBUTOR_HSA: attributes #[[ATTR18]] = { nounwind "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-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" }
712712
; ATTRIBUTOR_HSA: attributes #[[ATTR19:[0-9]+]] = { nounwind sanitize_address "amdgpu-no-implicitarg-ptr" "uniform-work-group-size"="false" }
713713
; ATTRIBUTOR_HSA: attributes #[[ATTR20:[0-9]+]] = { "enqueued-block" "uniform-work-group-size"="false" }
714714
; ATTRIBUTOR_HSA: attributes #[[ATTR21]] = { "amdgpu-agpr-alloc"="0" "amdgpu-no-cluster-id-x" "amdgpu-no-cluster-id-y" "amdgpu-no-cluster-id-z" "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" "enqueued-block" "uniform-work-group-size"="false" }

0 commit comments

Comments
 (0)