-
Notifications
You must be signed in to change notification settings - Fork 787
[SYCL][NFC] Add E2E test for gpu AOT to JIT fallback #18492
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
Conversation
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
Signed-off-by: Tikhomirova, Kseniya <[email protected]>
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, just a nit.
@jzc can you take a look at this? |
Co-authored-by: Marcos Maronas <[email protected]>
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
failing check is unrelated, same failure is present in another PRs. @intel/llvm-gatekeepers this PR is ready to be merged. |
I believe the feature of AOT to JIT fallback was implemented and enabled in this commit 38e588d
commit includes UT for jit fallback for kernel compiled for older platform 38e588d
this fallback works when
-fsycl-targets
is set for compilation and AOT images contain info about architecture it is built for. In order to build program SYCL RT choose images by compile target (if property is set) and adds spirv (if available) to the image set passed to backend later.Choice what image to use is done in UR adapter for specific backend. Where it chooses AOT image if available and uses SPIRV as fallback.
While RT part is completely tested by UT, adding small E2E test to verify stack behavior.