-
Notifications
You must be signed in to change notification settings - Fork 797
[UR][SYCL] Implement USM prefetch from device to host in SYCL runtime and UR #19437
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: sycl
Are you sure you want to change the base?
Conversation
Seems to just be my browser not updating, for some reason. Disregard this comment! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
Hi @intel/unified-runtime-reviewers, mind if I get a review? Thanks in advance! |
sycl/test-e2e/Graph/RecordReplay/ext_oneapi_enqueue_functions_prefetch.cpp
Outdated
Show resolved
Hide resolved
Co-authored-by: Pablo Reble <pablo@reble.org>
There is a UR precommit failure for CUDA: it is unrelated to the changes here. The same test fails on latest commit of |
Add the ability to control USM prefetch direction (host-to-device, device-to-host) in the enqueue_function extension:
Note:
There is a test failure regarding a new ABI symbol: In order to not break the ABI, I added a new handler function to represent prefetch from device-to-host. Despite the precommit failures, this should not be an ABI-breaking change and thus be okay to merge.I modified the ABI symbols tests as a part of this PR.