-
Notifications
You must be signed in to change notification settings - Fork 787
[SYCL][Matrix] Move elementwise operation under intel namespace and add joint_matrix_apply. #8417
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
This patch moves the wi_data and wi_element class (and corresponding operations) in the matrix-intel.hpp file, under the intel::experimental::matrix namespace. It also removes the miarray based implementation of the wi_data class, as the miarray based implementation was needed to support CUDA. The CUDA support will be established through joint_matrix_apply() API and a cuda based get_wi_data() support in tensorcore.cpp file (coming up next).
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 other than reverting the nvptx-tf32 test changes.
matrix-unified.hpp, also adding NVIDIA implementation of joint_matrix_apply.
I think that this is the normal way of outputing the deprecation: llvm/sycl/include/sycl/accessor.hpp Line 776 in cf245f6
So I think if you just refer users to joint_matrix_apply in the deprecated message that would be good. |
I added the message, can you please double check if I added in the right place? Thank you |
LGTM! |
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
@intel/llvm-reviewers-runtime Please help merge this patch ASAP, as there are two feature PRs dependent on this. Thank you |
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!
@arnamoy10, please, fix these warnings emitted by
|
Thanks for the report. PR here: #8684 |
…ing namespace for get_wi_data() (#1636) for both ATS-M and PVC. Regarding the namespace change, the tests will pass once [this PR](intel/llvm#8417) gets approved.
intel/llvm#8417 is merged so this is ready for review. --------- Signed-off-by: JackAKirk <[email protected]>
…ing namespace for get_wi_data() (intel/llvm-test-suite#1636) for both ATS-M and PVC. Regarding the namespace change, the tests will pass once [this PR](intel#8417) gets approved.
…suite#1655) intel#8417 is merged so this is ready for review. --------- Signed-off-by: JackAKirk <[email protected]>
This patch moves the wi_data and wi_element class (and corresponding
operations) in the matrix-intel.hpp file, under the
intel::experimental::matrix namespace. The original implementation is kept (but soon will be deprecated) to make the existing CUDA test cases work.