perf: Add attribute caching to reduce runtime patching overhead #23
+706
−26
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
Collaborator
Author
|
Also tested along with SGLang, everything works as expected. |
yafengio
approved these changes
Jan 29, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Description
This PR adds caching optimizations to reduce the overhead of torchada's runtime patching for frequently accessed attributes.
Changes
_CudaModuleWrapperfortorch.cuda.*access_CudartWrapperfortorch.cuda.cudart()function lookups_CDLLWrapperfor ctypes function name translationtorch.backends.cuda.is_built()result (constant at runtime)_translate_device()for common device stringsPerformance Improvements
torch.cuda.Stream(attr)torch.cuda.device_count()cudart.cudaHostRegistertorch.backends.cuda.is_built()All operations now complete in <700ns, which is negligible compared to GPU kernel launch times (5,000-20,000ns).