Skip to content

[SYCL][CUDA] Allow joint_matrix to be loaded from const T #6532

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

Merged
merged 3 commits into from
Oct 5, 2022

Conversation

JackAKirk
Copy link
Contributor

@JackAKirk JackAKirk commented Aug 5, 2022

Fixes a bug where if joint_matrix_load attempts to load joint_matrix from an array of const Tincorrect behaviour will occur or an error will be thrown. To fix this we make use of std::remove_const_t<T> in appropriate places. This is important functionality for integrating joint_matrix with existing SYCL-DNN routines.
I think that similar problems might occur in the intel backends for their existing impl: I have not made corresponding changes because I do not have the hardware to test it.

Signed-off-by: JackAKirk [email protected]

@JackAKirk JackAKirk changed the title [SYCL[CUDA]]Allow joint_matrix to be loaded from const. [SYCL][CUDA] Allow joint_matrix to be loaded from const. Aug 5, 2022
Signed-off-by: JackAKirk <[email protected]>
@JackAKirk JackAKirk marked this pull request as ready for review August 5, 2022 13:50
@JackAKirk JackAKirk requested a review from a team as a code owner August 5, 2022 13:50
@JackAKirk JackAKirk requested a review from v-klochkov August 5, 2022 13:50
@JackAKirk JackAKirk changed the title [SYCL][CUDA] Allow joint_matrix to be loaded from const. [SYCL][CUDA] Allow joint_matrix to be loaded from const T Aug 5, 2022
@JackAKirk JackAKirk requested a review from dkhaldi September 1, 2022 09:13
Copy link
Contributor

@dkhaldi dkhaldi left a comment

Choose a reason for hiding this comment

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

LGTM

@JackAKirk
Copy link
Contributor Author

btw the CUDA test failure is the now XFAIL'ed known flaky test Assert/assert_in_simultaneously_multiple_tus.cpp

@JackAKirk
Copy link
Contributor Author

ping @v-klochkov

Copy link
Contributor

@v-klochkov v-klochkov left a comment

Choose a reason for hiding this comment

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

Looks good to me. Thank you.
Before proceeding to merge. Please add a test verifying the test case being fixed by this PR.

If there is an existing test, but it is XFAIL due to being flaky, please add a simpler test or a compile-only test.

@JackAKirk
Copy link
Contributor Author

/verify with intel/llvm-test-suite#1280

@JackAKirk
Copy link
Contributor Author

/verify with intel/llvm-test-suite#1280

@JackAKirk
Copy link
Contributor Author

Looks good to me. Thank you. Before proceeding to merge. Please add a test verifying the test case being fixed by this PR.

If there is an existing test, but it is XFAIL due to being flaky, please add a simpler test or a compile-only test.

Thanks. I've added all new possible use cases in the test here: intel/llvm-test-suite#1280

@pvchupin
Copy link
Contributor

/verify with intel/llvm-test-suite#1280

@JackAKirk
Copy link
Contributor Author

Can this be merge now?

@pvchupin pvchupin merged commit 134618f into intel:sycl Oct 5, 2022
pvchupin pushed a commit to intel/llvm-test-suite that referenced this pull request Oct 5, 2022
This tests new cases made possible in intel/llvm#6532, such that A,B, Accumulator accessors that are loaded to registers in joint_matrix can be of type const T.
aelovikov-intel pushed a commit to aelovikov-intel/llvm that referenced this pull request Mar 27, 2023
…st-suite#1280)

This tests new cases made possible in intel#6532, such that A,B, Accumulator accessors that are loaded to registers in joint_matrix can be of type const T.
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.

4 participants