Omit CUDA sync events on unused CUDA streams #795
Closed
Conversation
This file contains 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
Summary:
CUDA WaitforStream sync events can lead to a lot of verbosity in the traces.
It looks like PyTorch cycles through all the streams to add these wait calls but they are no-ops on stream that are not in active use.
{F1063309394}
This change will simply filter out WaitforStream events whenever a CUDA stream has not been used.
* Track device, stream used in a set, this updated by an GPU kernel or memcpy or memset.
* Filter waitforstream events.
Reviewed By: chaekit, davidberard98
Differential Revision: D47975637
fbshipit-source-id: 367b1e0c7f41d1ee606d36cec0bb9441dd258c2b
|
This pull request was exported from Phabricator. Differential Revision: D47975637 |
|
This pull request has been merged in f6c845e. |
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Outdated suggestions cannot be applied.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.
Summary:
CUDA WaitforStream sync events can lead to a lot of verbosity in the traces.
It looks like PyTorch cycles through all the streams to add these wait calls but they are no-ops on stream that are not in active use.
{F1063309394}
This change will simply filter out WaitforStream events whenever a CUDA stream has not been used.
Reviewed By: chaekit, davidberard98
Differential Revision: D47975637