[Bugfix] Do not consider local.var as local buffer during LowerTileOP

[LeiWang1999]

as title, which may lead to significant performance regression when we assign value into var.

Summary by CodeRabbit

Release Notes

• Bug Fixes

  • Enhanced local buffer detection in parallel loop optimization, enabling thread binding and vectorization in additional scenarios.

• Tests

  • Added runtime verification of generated kernel source code in test suite.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[github-actions]

👋 Hi! Thank you for contributing to the TileLang project.

Please remember to run pre-commit run --all-files in the root directory of the project to ensure your changes are properly linted and formatted. This will help ensure your contribution passes the format check.

We appreciate you taking this step! Our team will review your contribution, and we look forward to your awesome work! 🚀

[coderabbitai]

Caution

Review failed

The pull request is closed.

📝 Walkthrough

Walkthrough

Broadens local buffer recognition in lower_tile_op.cc by replacing IsLocalBuffer(..., true) calls with IsLocalBuffer(...), affecting store detection, register-only checks, and non-local buffer validation. This propagates into parallel-loop partitioning and vectorization logic. Additionally, adds runtime kernel source verification in a test.

Changes

Cohort / File(s)	Summary
Core compiler logic src/transform/lower_tile_op.cc	Removes second argument from all IsLocalBuffer() calls across multiple functions (store_into_local, local_register_only, has_non_local, parallel_loop, should_vectorize). Broadens buffer classification to enable parallel-loop partitioning and vectorization in additional cases by relaxing local-buffer detection criteria.
Test verification testing/python/issue/test_tilelang_issue_1549.py	Adds post-execution runtime introspection: retrieves generated kernel source via get_kernel_source() and verifies that a specific multi-line loop snippet is present in the compiled code.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• [BugFix] Fix some bugs in lowering ParallelOp and VectorizeLoop #1607: Directly modifies IsLocalBuffer usage in src/transform/lower_tile_op.cc with overlapping scope.
• Improve memory access safety and T.assume handling #1292: Changes IsLocalBuffer semantics or logic definition, affecting the underlying buffer classification function.
• [Enhancement] Enhance Cast operations Vectorization #1156: Modifies parallel-loop and vectorization lowering logic based on local buffer detection criteria.

Suggested reviewers

• kurisu6912

Poem

🐰 Buffers hop with newfound cheer,
Local checks grow crystal clear,
Thread partitions dance with glee,
Vectorized loops wild and free,
No strict guards to hold us back—
Compilation speeds on track! 🚀

✨ Finishing touches

• 📝 Generate docstrings

---

📜 Recent review details

Configuration used: defaults

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between a8bf4f6 and b3297d1.

📒 Files selected for processing (2)

• src/transform/lower_tile_op.cc
• testing/python/issue/test_tilelang_issue_1549.py

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}