Use TargetIsCuda for all cuda target

[oraluben]

Summary by CodeRabbit

• Refactor
  • Simplified GPU dispatch logic for GEMM operations. Replaced architecture-specific checks with a unified CUDA-target approach, improving code maintainability while preserving existing functionality.

_{✏️ 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]

📝 Walkthrough

Walkthrough

This change simplifies GEMM target dispatch logic in GemmPyNode::getGemmInst by replacing architecture-specific checks (Volta, Ampere, Turing, Hopper, Sm100/SM120) with a single generic CUDA-target check. The return behavior remains unchanged—all CUDA targets now return GemmInst::kMMA.

Changes

Cohort / File(s)	Summary
GEMM dispatch logic simplification src/op/gemm_py.cc	Replaced enumerated architecture-specific conditions with a single CUDA-target check in GemmPyNode::getGemmInst; functionally equivalent but now applies to all CUDA targets rather than pre-defined architectures

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• [Typo] fix typo for SM120 #1408 — Directly modifies the same GEMM target-dispatch logic in GemmPyNode::getGemmInst

Suggested reviewers

• LeiWang1999

Poem

🐰 Hop, hop, the CUDA dispatch takes flight,
One check replaces many—so sleek, so right!
Volta, Ampere, Hopper fade away,
A single path for kMMA's day. ✨

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title directly describes the main change: replacing specific architecture checks with a unified TargetIsCuda check for all CUDA targets, which matches the core modification in the changeset.

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

📜 Recent review details

Configuration used: defaults

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between cfccd63 and f61d39a.

📒 Files selected for processing (1)

• src/op/gemm_py.cc

🧰 Additional context used

🧬 Code graph analysis (1)

src/op/gemm_py.cc (2)

src/target/utils.cc (2)

• TargetIsCuda (14-16)
• TargetIsCuda (14-14)

tilelang/language/ast/ir.py (1)

• target (1677-1707)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (2)

• GitHub Check: Test for Python 3.12 with Nightly-ROCm-7.1 (on self-hosted-amd)
• GitHub Check: Test for Python 3.12 with Metal (on macos-latest)

🔇 Additional comments (1)

src/op/gemm_py.cc (1)

136-137: The change is correct and safe. The codebase only supports CUDA SM70 (Volta) and newer architectures, all of which have MMA instruction capability. The TargetIs* helper functions in src/target/utils.cc only exist for SM70+ (Volta, Turing, Ampere, Hopper, SM100, SM120), confirming no pre-Volta architectures are supported. The generic TargetIsCuda check appropriately returns kMMA for all supported CUDA targets, with earlier checks handling specialized instructions (TCGEN5MMA for SM100, WGMMA for Hopper).

---

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.}

[oraluben]

This PR is trivial, but I believe there's somewhere missing that causing #1498