[runtime][hip][cuda] Fix waiting on a semaphore on the host #17073
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
When waiting on a semaphore we first attempt to wait on a timepoint event on the dispatch stream. This is not correct as the semaphore is signaled in a callback on the host stream when signaling that an action is complete. This callback happens after the dispatch stream event, so the semaphore value may not be update yet. The fix is to directly wait on the host semaphore and not on the dispatch stream event. Also add a test that catches this failure.
sogartar
requested review from
antiagainst,
nithinsubbiah and
ScottTodd
as code owners
antiagainst
approved these changes
Apr 17, 2024
There was a problem hiding this comment.
Awesome, thanks for fixing this! This was meant to be an optimization (#14876) becuase it was faster to wait the device event, at least on cuda. But yeah; it's problematic..
| // The signaling relationship is | ||
| // command_buffer1 -> semaphore1 -> command_buffer2 -> semaphore2 | ||
|
|
||
| // Create second command_buffer1. |
LLITCHEV
pushed a commit
to LLITCHEV/iree
that referenced
this pull request
Jul 30, 2024
…#17073) When waiting on a semaphore we first attempt to wait on a timepoint event on the dispatch stream. This is not correct as the semaphore is signaled in a callback on the host stream when signaling that an action is complete. This callback happens after the dispatch stream event, so the semaphore value may not be update yet. The fix is to directly wait on the host semaphore and not on the dispatch stream event. Also add a test that catches this failure. Signed-off-by: Lubo Litchev <lubol@google.com>
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.
When waiting on a semaphore we first attempt to wait on a timepoint event on the dispatch stream.
This is not correct as the semaphore is signaled in a callback on the host stream when signaling that an action is complete.
This callback happens after the dispatch stream event, so the semaphore value may not be update yet.
The fix is to directly wait on the host semaphore and not on the dispatch stream event.
Also add a test that catches this failure.