Repository navigation
[BugFix][Profiler] Use the current CUDA device when no device is specified - #3357
Conversation
|
👋 Hi! Thank you for contributing to the TileLang project. Please remember to run We appreciate you taking this step! Our team will review your contribution, and we look forward to your awesome work! 🚀 |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Merge Risk: ⚪ Minimal · up to
|
| Check name | Status | Explanation |
|---|---|---|
| Docstring Coverage |  |
Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 7 functions across 2 files. |
| Linked Issues check | |
Check skipped because no linked issues were found for this pull request. |
| Out of Scope Changes check | |
Check skipped because no linked issues were found for this pull request. |
| Description Check | |
Check skipped - CodeRabbit’s high-level summary is enabled. |
| Title check | |
The title clearly and concisely describes the main change: using the current CUDA device when no device is specified in the profiler. |
✨ Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Create a new PR
-
- Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Autopilot is currently an internal CodeRabbit preview.
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.
Comment @coderabbitai help to get the list of available commands.
|
Added docstrings for the touched helper and test methods in 7c99e7e. All five host regression tests and the targeted formatting checks pass. |
Afloat16
changed the title
SiriusNEO
added
bug
When the current CUDA device is nonzero and
do_benchreceives no explicit device,_cache_device(None)uses the module'scuda:0default. The cache-flush allocation can therefore target a different GPU from the benchmarked work.Use the unindexed
"cuda"default so PyTorch resolves the active device when the buffer is allocated. Explicit device selection and the NPU/MPS defaults are preserved. The current diff is one line; the earlier standalone host tests were removed during maintainer revision.Validation:
971c4c4: Quick Lint, CUDA, ROCm, Metal, Ascend, and CuTeDSL examples.No local CUDA hardware or timing benchmarks were run.
do_benchnow uses"cuda"as its default device string when CUDA is available. This allows the current CUDA device to be resolved at each call. The NPU and MPS defaults remain unchanged.No test results or formatting checks are established by the supplied repository evidence. The change does not touch C++, CI, lint tooling, or C++ style documentation.