Set NCCL_CUMEM_HOST_ENABLE=0 for ARM with CUDA 12.8 #735
+8
−0
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
raydouglass
requested changes
Feb 14, 2025
Co-authored-by: Ray Douglass <3107146+raydouglass@users.noreply.github.com>
raydouglass
approved these changes
Feb 14, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
25.02 ARM containers on CUDA 12.8 are running into an issue of a get_mempolicy call that requires elevated privileges. This very likely will not be desirable for many users, so this change sets the environment NCCL_CUMEM_HOST_ENABLE=0 inside the container that removes the call and makes the containers work similar to prior versions.
This bug only affects RAPIDS libraries that use RAFT comms with NCCL, not NCCL standalone tests, so we are still investigating the root cause. This fix does not seem to have any ill effects during testing and benchmarking though, so it's the fastest way to restore behavior in ARM 12.8 behaviors without needing user intervention.