Explicitly build for all GPU architectures #15959
Merged
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Outdated suggestions cannot be applied.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.
Description
The libcudf conda package is not specifying to build for all supported architectures and is instead letting build.sh fall back to NATIVE. However, because the default behavior of rapids-cmake is to build SASS for all supported architectures if NATIVE is specified but no local architecture is detected, we're still ending up with all of the RAPIDS architectures having SASS built for them. The problem is that we are failing to build PTX for the latest version, which would be produced if we used RAPIDS instead of NATIVE. This PR should resolve that issue.
Checklist