Skip to content

AMDGPU: Fix attributor not handling all trap intrinsics #131758

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
arsenm merged 1 commit into from
Mar 19, 2025
+83 −0
Merged

AMDGPU: Fix attributor not handling all trap intrinsics #131758

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -110,6 +110,8 @@ intrinsicToAttrMask(Intrinsic::ID ID, bool &NonKernelOnly, bool &NeedsImplicit,
return CodeObjectVersion >= AMDGPU::AMDHSA_COV5 ? IMPLICIT_ARG_PTR
: QUEUE_PTR;
case Intrinsic::trap:
case Intrinsic::debugtrap:
case Intrinsic::ubsantrap:
if (SupportsGetDoorBellID) // GetDoorbellID support implemented since V4.
return CodeObjectVersion >= AMDGPU::AMDHSA_COV4 ? NOT_IMPLICIT_INPUT
: QUEUE_PTR;
Expand Down
6 changes: 6 additions & 0 deletions llvm/test/CodeGen/AMDGPU/abi-attribute-hints-undefined-behavior.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -383,6 +383,12 @@ define void @debugtrap_requires_queue() #0 {
unreachable
}

; FIXME: Broken in globalisel
; define void @ubsantrap_requires_queue() #0 {
; call void @llvm.ubsantrap(i8 0)
; unreachable
; }

declare i32 @llvm.amdgcn.workitem.id.x()
declare i32 @llvm.amdgcn.workitem.id.y()
declare i32 @llvm.amdgcn.workitem.id.z()
Expand Down
31 changes: 31 additions & 0 deletions llvm/test/CodeGen/AMDGPU/hsa-metadata-queue-ptr-v4.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,14 @@
; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1
; GFX9: .amdhsa_user_sgpr_queue_ptr 0

; CHECK: debugtrap_requires_queue_ptr
; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1
; GFX9: .amdhsa_user_sgpr_queue_ptr 0

; CHECK: ubsantrap_requires_queue_ptr
; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1
; GFX9: .amdhsa_user_sgpr_queue_ptr 0

; CHECK: amdgcn_queue_ptr_requires_queue_ptr
; CHECK: .amdhsa_user_sgpr_queue_ptr 1

Expand Down Expand Up @@ -71,6 +79,28 @@ define amdgpu_kernel void @trap_requires_queue_ptr() {
unreachable
}

; CHECK: - .args:

; CHECK-NOT: hidden_shared_base
; CHECK-NOT: hidden_private_base
; CHECK-NOT: hidden_queue_ptr
; CHECK-LABEL: .name: debugtrap_requires_queue_ptr
define amdgpu_kernel void @debugtrap_requires_queue_ptr() {
call void @llvm.debugtrap()
unreachable
}

; CHECK: - .args:

; CHECK-NOT: hidden_shared_base
; CHECK-NOT: hidden_private_base
; CHECK-NOT: hidden_queue_ptr
; CHECK-LABEL: .name: ubsantrap_requires_queue_ptr
define amdgpu_kernel void @ubsantrap_requires_queue_ptr() {
call void @llvm.ubsantrap(i8 0)
unreachable
}

; CHECK: - .args:
; CHECK-NOT: hidden_queue_ptr
; CHECK-NOT: hidden_shared_base
Expand All @@ -96,6 +126,7 @@ declare i1 @llvm.amdgcn.is.shared(ptr)
declare i1 @llvm.amdgcn.is.private(ptr)
declare void @llvm.trap()
declare void @llvm.debugtrap()
declare void @llvm.ubsantrap(i8 immarg)

!llvm.module.flags = !{!0}
!0 = !{i32 1, !"amdhsa_code_object_version", i32 400}
21 changes: 21 additions & 0 deletions llvm/test/CodeGen/AMDGPU/hsa-metadata-queue-ptr-v5.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,26 @@ define amdgpu_kernel void @trap_requires_queue_ptr() {
unreachable
}

; CHECK: - .args:

; PRE-GFX9: hidden_queue_ptr
; GFX9-NOT: hidden_queue_ptr
; CHECK-LABEL: .name: debugtrap_requires_queue_ptr
define amdgpu_kernel void @debugtrap_requires_queue_ptr() {
call void @llvm.debugtrap()
unreachable
}

; CHECK: - .args:

; PRE-GFX9: hidden_queue_ptr
; GFX9-NOT: hidden_queue_ptr
; CHECK-LABEL: .name: ubsantrap_requires_queue_ptr
define amdgpu_kernel void @ubsantrap_requires_queue_ptr() {
call void @llvm.ubsantrap(i8 0)
unreachable
}

; CHECK: - .args:
; CHECK: hidden_queue_ptr
; CHECK-LABEL: .name: amdgcn_queue_ptr_requires_queue_ptr
Expand All @@ -84,6 +104,7 @@ declare i1 @llvm.amdgcn.is.shared(ptr)
declare i1 @llvm.amdgcn.is.private(ptr)
declare void @llvm.trap()
declare void @llvm.debugtrap()
declare void @llvm.ubsantrap(i8 immarg)

!llvm.module.flags = !{!0}
!0 = !{i32 1, !"amdhsa_code_object_version", i32 500}
23 changes: 23 additions & 0 deletions llvm/test/CodeGen/AMDGPU/implicit-kernarg-backend-usage.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -260,6 +260,29 @@ define amdgpu_kernel void @llvm_debugtrap() {
unreachable
}

define amdgpu_kernel void @llvm_ubsantrap() {
; GFX8V4-LABEL: llvm_ubsantrap:
; GFX8V4: ; %bb.0:
; GFX8V4-NEXT: s_mov_b64 s[0:1], s[6:7]
; GFX8V4-NEXT: s_trap 2
;
; GFX8V5-LABEL: llvm_ubsantrap:
; GFX8V5: ; %bb.0:
; GFX8V5-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0xc8
; GFX8V5-NEXT: s_waitcnt lgkmcnt(0)
; GFX8V5-NEXT: s_trap 2
;
; GFX9V4-LABEL: llvm_ubsantrap:
; GFX9V4: ; %bb.0:
; GFX9V4-NEXT: s_trap 2
;
; GFX9V5-LABEL: llvm_ubsantrap:
; GFX9V5: ; %bb.0:
; GFX9V5-NEXT: s_trap 2
call void @llvm.ubsantrap(i8 0)
unreachable
}

define amdgpu_kernel void @llvm_amdgcn_queue_ptr(ptr addrspace(1) %ptr) {
; GFX8V4-LABEL: llvm_amdgcn_queue_ptr:
; GFX8V4: ; %bb.0:
Expand Down
Loading