-
Notifications
You must be signed in to change notification settings - Fork 42
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
Add XeTLA
FA backward implementation to benchmark
#2367
Conversation
d38497e
to
1b1064c
Compare
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.
Can we run it on CI and check if the result can be compared with torch?
Co-authored-by: Anatoly Myachev <[email protected]>
CI tutorial seems not ready for backward (include torch's implication) yet. |
set(XETLA_KERNEL_FLAGS ${XETLA_KERNEL_FLAGS} -fsycl) | ||
set(XETLA_KERNEL_FLAGS ${XETLA_KERNEL_FLAGS} | ||
-fsycl | ||
-fsycl-device-code-split=per_kernel |
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.
What does it change?
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.
This change adds additional flag to perform sycl kernel splitting and helps to resolve RuntimeError below.
No perf regression for xetla in my local env.
RuntimeError: The program was built for 1 devices
Build program log for 'Intel(R) Data Center GPU Max 1100':
-11 (PI_ERROR_BUILD_PROGRAM_FAILURE)
Implementation comes from IPEX
https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/11089057652