Skip to content

Conversation

@RossBrunton
Copy link
Contributor

urQueueRelease now waits for all work in the queue to complete, which
is consistent with other plugins.

urQueueRelease now waits for all work in the queue to complete, which
is consistent with other plugins.
break;
}
OL_RETURN_ON_ERR(olSyncQueue(Q));
OL_RETURN_ON_ERR(olDestroyQueue(Q));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be less error-prone for the plugins to do synchronization as part of destroy implicitly. So I think we should change this in liboffload.
But lgtm for now to unblock testing.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not syncing was an intentional design decision. It allows queue objects to be created and destroyed for a single operation without requiring the liboffload user to track lifetimes of each queue object.

@RossBrunton
Copy link
Contributor Author

@intel/llvm-gatekeepers Should be mergable.

The offload backend isn't checked as part of any of the failing CI, so these failures are unrelated.

@steffenlarsen steffenlarsen merged commit dc2edea into intel:sycl Sep 17, 2025
35 of 37 checks passed
YixingZhang007 pushed a commit to YixingZhang007/llvm that referenced this pull request Sep 22, 2025
urQueueRelease now waits for all work in the queue to complete, which
is consistent with other plugins.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants