[SYCL][CUDA] Refactor joint_matrix tests and remove legacy cuda extension #8951
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
cuda joint_matrix e2e tests refactored so that they will pass on any tensor cores supporting gpu. Signed-off-by: JackAKirk <[email protected]>
Signed-off-by: JackAKirk <[email protected]>
JackAKirk
changed the title
bader
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The new joint_matrix portable extension has been live for a while now and has made it into the latest release. In this PR I've removed the legacy Cuda only version of the extension to avoid future confusion.
The Cuda joint_matrix e2e tests are also refactored so that they will pass on any tensor cores supporting gpu. Previously sm_70, sm_72, and sm_80 cases were included in the same test file. The target builtins required compilation for sm_80 or above, and despite checks for compute capability for each sm_xx test case within the main test file, since the ptx does not jit if compiled for sm_80 but ran a lower compute capability, e.g. sm_75, this led to an invalid binary error in such a case. Now the tests are split into files referencing the minimum compute capability for test cases included in the file, meaning that the test cases will run and pass on any tensor cores supporting gpu.
get_wi_data()is deprecated in the cuda backend and we can remove this support before the next release and remove corresponding test cases from joint_matrix_apply_cuda.hpp.