Skip to content

Improve Bounce Buffer Allocation Strategy#1222

Merged
markbrown314 merged 3 commits intoSandia-OpenSHMEM:performancefrom
markbrown314:1221/development
Feb 24, 2026
Merged

Improve Bounce Buffer Allocation Strategy#1222
markbrown314 merged 3 commits intoSandia-OpenSHMEM:performancefrom
markbrown314:1221/development

Conversation

@markbrown314
Copy link
Collaborator

@markbrown314 markbrown314 commented Feb 3, 2026

Currently bounce buffers are allocated via malloc, which causes some performance issues with OFI providers that required registered memory. This change introduces a bounce buffer allocation strategy which allocates buffers using a reserved region of the symmetric memory heap which is registered memory. There is also a feature to pin the bounce buffers if desired.

Addresses Issue #1221
For evaluation purposes only.

Do not force hard polling when XPMEM is enabled.
This conflicts with counter based polling when used with OFI transport.

Issue Sandia-OpenSHMEM#1217

Signed-off-by: Mark F. Brown <mark.f.brown@intel.com>
Replaced complex polling with simpler OFI counter wait

Issue Sandia-OpenSHMEM#1217

Signed-off-by: Mark F. Brown <mark.f.brown@intel.com>
Issue Sandia-OpenSHMEM#1221

Signed-off-by: Mark F. Brown <mark.f.brown@intel.com>
@markbrown314 markbrown314 self-assigned this Feb 3, 2026
@markbrown314 markbrown314 changed the base branch from main to performance February 24, 2026 17:32
@markbrown314 markbrown314 marked this pull request as ready for review February 24, 2026 17:33
@markbrown314 markbrown314 merged commit 1e28dc8 into Sandia-OpenSHMEM:performance Feb 24, 2026
31 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant