Simplify _compute_symbolic_stride() #138844
Closed
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138844
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 3bee778 with merge base 82ce888 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
cc ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv [ghstack-poisoned]
cc ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv [ghstack-poisoned]
cc ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv [ghstack-poisoned]
cc ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv [ghstack-poisoned]
Rewrite _compute_symbolic_stride() to make it simpler and faster. The existing code involves several inner loops in an attempt to process the common case faster - but in reality this effort is actually slower than the simpler code. Testing: The initial version of this PR (which passed all tests) ran both the old algorithm and new algorithm and compared the results to make sure that results were substantially the same (they weren't the same simply because the algorithm allocates new dynamic symbols as part of it). I also measured the timing of both methods and from the cases I checked the simpler algorithm was generally about 30% faster (which was usually the "fast path" of the old algorithm). cc ezyang SherlockNoMad EikanWang jgong5 wenzhe-nrv [ghstack-poisoned]
|
I am deferring to Bob as he is also monkeying around with this code recently |
bobrenjc93
approved these changes
Dec 17, 2024
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
aditew01
pushed a commit
to aditew01/pytorch
that referenced
this pull request
Dec 18, 2024
Rewrite _compute_symbolic_stride() to make it simpler and faster. The existing code involves several inner loops in an attempt to process the common case faster - but in reality this effort is actually slower than the simpler code. Testing: The initial version of this PR (which passed all tests) ran both the old algorithm and new algorithm and compared the results to make sure that results were substantially the same (they weren't the same simply because the algorithm allocates new dynamic symbols as part of it). I also measured the timing of both methods and from the cases I checked the simpler algorithm was generally about 30% faster (which was usually the "fast path" of the old algorithm). Pull Request resolved: pytorch#138844 Approved by: https://github.com/bobrenjc93 ghstack dependencies: pytorch#138843
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.
Rewrite _compute_symbolic_stride() to make it simpler and faster.
The existing code involves several inner loops in an attempt to process the common case faster - but in reality this effort is actually slower than the simpler code.
Testing:
The initial version of this PR (which passed all tests) ran both the old algorithm and new algorithm and compared the results to make sure that results were substantially the same (they weren't the same simply because the algorithm allocates new dynamic symbols as part of it).
I also measured the timing of both methods and from the cases I checked the simpler algorithm was generally about 30% faster (which was usually the "fast path" of the old algorithm).
Stack from ghstack (oldest at bottom):
cc @ezyang @SherlockNoMad @EikanWang @jgong5 @wenzhe-nrv