-
Notifications
You must be signed in to change notification settings - Fork 787
[SYCL] Disable inlining kernel lambda operator at -O0 #7578
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
PR intel#6977 enabled always inliling kernel lambda operators. This PR disables this at -O0 as it was leading to a poor debugging experience.
The three failing tests in the CUDA LLVM test suite: |
#7567 should fix fails. |
@pvchupin, the CUDA LLVM test suite is failing for reasons not clear to me. |
Temporarily moving this to draft state since it looks like passing the flag to additional components may be necessary. |
Sorry, that was a false alarm due to incorrect environment setup. This is ready for review. |
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, thanks!
@pvchupin, the tests seem to be failing again, though there were no compiler changes between yesterday (when it all succeeded) and now. (I did change a test based on code-review, and that test does pass.) Could you please review the doc change and if you are okay with that, merge the change? |
sycl/doc/UsersManual.md
Outdated
Enabled by default; disabled when optimizations are disabled (-O0 or | ||
equivalent). |
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.
As far as I remember clang has -O0 by default?
Enabled by default; disabled when optimizations are disabled (-O0 or | |
equivalent). | |
Disabled when optimizations are disabled (-O0 or | |
equivalent). Enabled otherwise. |
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.
Done, thanks.
@premanandrao, error when toolchain archive is not found happens usually when you restart failed task manually. When you update the patch you don't need to do anything actually, testing will restart automatically. In the cases when you really want to restart testing the only option is to "Restart all tasks". The reason is that toolchain artifact is erased on the runner as soon as testing is over. Restarting just llvm-test-suite testing is not possible today. |
Thanks, that is helpful. In my case, testing had started automatically after patch update, and that testing failed cryptically - like with a failed cloud service message. I realize now that I should not have attempted to restart as it seems the original failure info is lost. |
Failing tests are unrelated to this change. @pvchupin / @intel/llvm-gatekeepers, this is ready for a merge. |
PR #6977 enabled always inlining kernel lambda operators.
This PR disables this at -O0 as it was leading to a poor
debugging experience.