Skip to content

[LLDB][AMDGPU] Refine GPU core comparison checks#113

Draft
thechenli wants to merge 2 commits into
llvm-server-pluginsfrom
chenlii/gpu-core-comparison-framework-pr-clean
Draft

[LLDB][AMDGPU] Refine GPU core comparison checks#113
thechenli wants to merge 2 commits into
llvm-server-pluginsfrom
chenlii/gpu-core-comparison-framework-pr-clean

Conversation

@thechenli

Copy link
Copy Markdown
Collaborator

Changes:

  • Add module comparison and GPU backtrace comparison tests.
  • Auto-load debug info for module/backtrace comparisons.
  • Select the ROCLLDB GPU wave matching ROCGDB's selected AMDGPU wave id before reading registers, locals, and backtraces.
  • Compare modules by normalized containment instead of exact raw list equality.
  • Treat LLDB-only modules and duplicate module counts as diagnostics.
  • Normalize debugger-specific module spelling for memory code objects, file-backed code objects, URL-escaped paths, and vdso.
  • Avoid failing backtrace comparison for demangler spelling differences or ROCGDB-unknown/ROCLLDB-symbolized frames.

Tested:

  • python3 -m py_compile on the modified comparison framework files.
  • Eval runs against production merged GPU cores.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant