Use native NVSHMEM synchronization APIs in NVSHMEM backends#111
Open
romerojosh wants to merge 2 commits intomainfrom
Open
Use native NVSHMEM synchronization APIs in NVSHMEM backends#111romerojosh wants to merge 2 commits intomainfrom
romerojosh wants to merge 2 commits intomainfrom
Conversation
Signed-off-by: Josh Romero <joshr@nvidia.com>
Collaborator
Author
|
/build |
|
🚀 Build workflow triggered! View run |
|
✅ Build workflow passed! View run |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR is a second attempt at #107.
The implementation is unchanged from the original PR, but this PR contains a workaround for an NVSHMEM bug impacting versions <= 3.2.5. During testing, it was determined that there are scenarios where remote (i.e. non-NVLINK) signaling operations can trigger segfaults due to a caching bug in the symmetric heap management. The workaround is to enforce a symmetric heap granularity of at least 1 GiB (up from the default of 512 MiB) for impacted NVSHMEM versions.
This can cause an increase of up to 512 MiB in memory usage for users currently using NVSHMEM backends and NVSHMEM version <= 3.2.5. To avoid this increase, users can upgrade to 3.3.24 or greater (included in NVHPC SDK 25.9+).