Skip to content

Conversation

AlexeySachkov
Copy link
Contributor

No description provided.

Refactored GlobalOffset pass to use new pass manager interface and added
a wrapper to use it with the legacy pass manager.

At the moment this is a non-functional change, because CodeGen passes
are still scheduled through the legacy pass manager and some additional
changes will be needed to include the pass into a pipeline when it will
be composed using the new pass manager.
Refactored LocalAccessorToSharedMemory pass to use new pass manager
interface and added a wrapper to use it with the legacy pass manager.

At the moment this is a non-functional change, because CodeGen passes
are still scheduled through the legacy pass manager and some additional
changes will be needed to include the pass into a pipeline when it will
be composed using the new pass manager.
@AlexeySachkov AlexeySachkov requested a review from bader as a code owner February 25, 2021 12:18
@AlexeySachkov
Copy link
Contributor Author

Tagging @Alexander-Johnston and @steffenlarsen for awareness as well

@bader bader changed the title [NFC] refactor CUDA-specific passes to work with new pass manager [CUDA][NFC] Refactor CUDA-specific passes to work with new pass manager Mar 3, 2021
@bader bader added the cuda CUDA back-end label Mar 3, 2021
@bader
Copy link
Contributor

bader commented Mar 3, 2021

+@Naghasan

@bader
Copy link
Contributor

bader commented Mar 3, 2021

Please, fix CI checks.

@bader
Copy link
Contributor

bader commented May 11, 2021

Please, fix CI checks.

and merge conflicts.

@bader bader marked this pull request as draft July 14, 2021 14:28
@bader
Copy link
Contributor

bader commented Jan 13, 2022

@mlychkov, @AlexeySachkov, should we close this PR?

@AlexeySachkov
Copy link
Contributor Author

@bader, those passes are still using legacy pass manager interface, but at the same time, my understanding is that llvm CodeGen has not yet migrated to a new pass manager, meaning that updating of this passes is not so important at the moment.

Let's close this PR as I don't work on it actively anyway

@AlexeySachkov AlexeySachkov deleted the private/asachkov/update-cuda-passes-to-new-pm branch May 22, 2024 09:49
vmaksimo pushed a commit that referenced this pull request Jul 29, 2025
There's no SPIRV translation for `llvm.fake.use`, and currently it's
causing the translator to crash, so just ignore it to prevent crashes.

Original commit:
KhronosGroup/SPIRV-LLVM-Translator@daada5684da36fd
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.

2 participants