[Bugfix] Fix pdl related intrin handling to avoid strict annotation codegen #1650
+8
−24
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request updates the handling of CUDA synchronization and trigger calls in the PDL lowering and warp specialization code, and updates the TVM submodule. The changes simplify and clarify how these synchronization points are detected and marked in the codebase.
Refactoring of CUDA Sync/Trigger Call Detection:
MarkCudaSyncCallsinlower_pdl.ccto directly check fortl::pdl_trigger()andtl::pdl_sync()instead of matching external call names, improving robustness and maintainability.WarpSpecializedRoleMarkerinwarp_specialized_rewriter.ccto directly check forpdl_sync()andpdl_trigger()instead of string-based matching, ensuring more reliable role assignment.Testing Improvements:
test_pdl_synctest intest_tilelang_language_pdl.pyto assert that"__restrict__"is not present in the generated code, in addition to checking for"cudaGridDependencySynchronize".Dependency Update:
Summary by CodeRabbit
Release Notes
Improvements
Tests
✏️ Tip: You can customize this high-level summary in your review settings.