Skip to content

[AMDGPU] Fix debug line table for MSG_DEALLOC_VGPRS optimization #88924

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
merged 3 commits into from
Apr 18, 2024
Merged

[AMDGPU] Fix debug line table for MSG_DEALLOC_VGPRS optimization #88924

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f2da128
[AMDGPU] Fix debug line table for MSG_DEALLOC_VGPRS optimization
epilk Apr 16, 2024
6692558
Update llvm/test/CodeGen/AMDGPU/release-vgprs-dbg-loc.mir
epilk Apr 16, 2024
c057427
Update llvm/test/CodeGen/AMDGPU/release-vgprs-dbg-loc.mir
epilk Apr 16, 2024
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
5 changes: 3 additions & 2 deletions llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2665,10 +2665,11 @@ bool SIInsertWaitcnts::runOnMachineFunction(MachineFunction &MF) {
// instructions.
for (MachineInstr *MI : ReleaseVGPRInsts) {
if (ST->requiresNopBeforeDeallocVGPRs()) {
BuildMI(*MI->getParent(), MI, DebugLoc(), TII->get(AMDGPU::S_NOP))
BuildMI(*MI->getParent(), MI, MI->getDebugLoc(), TII->get(AMDGPU::S_NOP))
.addImm(0);
}
BuildMI(*MI->getParent(), MI, DebugLoc(), TII->get(AMDGPU::S_SENDMSG))
BuildMI(*MI->getParent(), MI, MI->getDebugLoc(),
TII->get(AMDGPU::S_SENDMSG))
.addImm(AMDGPU::SendMsg::ID_DEALLOC_VGPRS_GFX11Plus);
Modified = true;
}
Expand Down
40 changes: 40 additions & 0 deletions llvm/test/CodeGen/AMDGPU/release-vgprs-dbg-loc.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -run-pass=si-insert-waitcnts -o - %s | FileCheck %s

# Verify that si-insert-waitcnts copies debug locations from the s_endpgm to the
# "dealloc vgprs" s_sendmsg. If these are not present, the debugger will be
# unable to break at the end of the shader.
Comment on lines +5 to +6
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you add something like this description to the commit message when you merge? I guess the problem is that when you set a source breakpoint at the end of the program, you want the debugger to put it before the sendmsg not after it, so the vgprs are still available??

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, I'll add more context to the commit. I just asked @lancesix about this, its not just that the VGPRs are unavailable, the problem is also that deallocating VGPRs makes it impossible to do a context save, which GDB needs to report a breakpoint. See: https://github.com/ROCm/ROCdbgapi/blob/7e22c4d4a8852e1950aea7f9d3e49899e4929fb6/include/amd-dbgapi.h.in#L358-L361


--- |
define amdgpu_ps void @test() !dbg !5 { ret void, !dbg !8 }

!llvm.dbg.cu = !{!0}
!llvm.debugify = !{!2, !3}
!llvm.module.flags = !{!4}

!0 = distinct !DICompileUnit(language: DW_LANG_C, file: !1, producer: "debugify", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
!1 = !DIFile(filename: "t.ll", directory: "/")
!2 = !{i32 1}
!3 = !{i32 0}
!4 = !{i32 2, !"Debug Info Version", i32 3}
!5 = distinct !DISubprogram(name: "test", linkageName: "test", scope: null, file: !1, line: 1, type: !6, scopeLine: 1, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0)
!6 = !DISubroutineType(types: !7)
!7 = !{}
!8 = !DILocation(line: 1, column: 1, scope: !5)
...

---
name: test
machineFunctionInfo:
isEntryFunction: true
body: |
bb.0:
; CHECK-LABEL: name: test
; CHECK: GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
; CHECK-NEXT: S_NOP 0, debug-location !8
; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0, debug-location !8
; CHECK-NEXT: S_ENDPGM 0, debug-location !8
GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
S_ENDPGM 0, debug-location !8
...