Skip to content

Conversation

hdelan
Copy link
Contributor

@hdelan hdelan commented Feb 8, 2024

The guessLocalWorkSize func for cuda adapter was erroneously giving a factor of 2 wg size for the x dimension, even if it creates a global dim which is not the same as the user range. This may have caused out of bounds accesses in the CUDA adapter. This reverts that.

Also refactor the HIP version a little bit to remove an unnecessary check.

The guessLocalWorkSize func for cuda adapter was erroneously giving a factor
of 2 even if it creates a global dim which is not the same as the user range
without using any range rounding kernels. This reverts that. Also refactor the
HIP version a little bit to remove an unnecessary check.
@hdelan hdelan requested review from a team as code owners February 8, 2024 13:46
@hdelan hdelan requested a review from keyradical February 8, 2024 13:46
@hdelan hdelan changed the title Refactor guess local worksize funcs [CUDA][HIP] Refactor guess local worksize funcs Feb 8, 2024
@hdelan hdelan closed this Feb 8, 2024
@hdelan
Copy link
Contributor Author

hdelan commented Feb 8, 2024

I couldn't reopen this PR since I force pushed to the branch so this is the new PR #1326 .

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.

3 participants