Skip to content
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

[CUDA] Fix SYCL NVPTX passes if function operand is null-ref #3272

Merged
merged 1 commit into from
Mar 10, 2021

Conversation

rolandschulz
Copy link
Contributor

Fixes #3271

@rolandschulz rolandschulz requested a review from bader as a code owner February 28, 2021 06:37
@rolandschulz
Copy link
Contributor Author

Note: I tested that this fixes the issue. But I don't really understand the code and why it is null in this case. Therefore I don't know whether this is the correct fix and isn't just working around the actual problem.

@bader bader changed the title Fix SYCL PI passes if function operand is null-ref [CUDA] Fix SYCL PI passes if function operand is null-ref Mar 3, 2021
@bader bader changed the title [CUDA] Fix SYCL PI passes if function operand is null-ref [CUDA] Fix SYCL NVPTX passes if function operand is null-ref Mar 3, 2021
@bader bader added the cuda CUDA back-end label Mar 3, 2021
@bader
Copy link
Contributor

bader commented Mar 3, 2021

+@steffenlarsen, @Naghasan

@steffenlarsen
Copy link
Contributor

I do not have an explanation as to why the function operand is null-ref, and frankly I am slightly concerned. That said, I am okay with this fix.

Copy link
Contributor

@bader bader left a comment

Choose a reason for hiding this comment

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

@rolandschulz, could you add a small regression test exposing the issue, please?
You can try bugpoint tool to reduce the test case from #3271.

@Naghasan
Copy link
Contributor

Naghasan commented Mar 8, 2021

The null ref is probably coming from the dead argument elimination pass. As it builds a new function and removes the old one, this would turns into a nullref in the MD later on.

Regardless, the checks this patch adds make sense and is safe.

@bader bader merged commit 20c5aea into intel:sycl Mar 10, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
cuda CUDA back-end
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[SYCL][CUDA] pass 'Add implicit SYCL global offset': Assertion `Val && "isa<> used on a null pointer"' failed.
4 participants