UCT/CUDA_IPC: Implement uct_ep_put_sgl_zcopy#11359
Draft
michal-shalev wants to merge 1 commit intoopenucx:masterfrom
Draft
UCT/CUDA_IPC: Implement uct_ep_put_sgl_zcopy#11359michal-shalev wants to merge 1 commit intoopenucx:masterfrom
michal-shalev wants to merge 1 commit intoopenucx:masterfrom
Conversation
michal-shalev
force-pushed
the
uct-cuda-ipc-sgl
branch
2 times, most recently
from
81ed122 to
2101677
Compare
michal-shalev
added
Ready for Review
and removed
WIP-DNM
michal-shalev
force-pushed
the
uct-cuda-ipc-sgl
branch
from
2101677 to
e66fb59
Compare
michal-shalev
force-pushed
the
uct-cuda-ipc-sgl
branch
from
e66fb59 to
f56085d
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What?
Implement
uct_ep_put_sgl_zcopyfor thecuda_ipctransport.Why?
Allow UCP SGL put offload to run over
cuda_ipc, replacing N separate per-buffer copies with a single batched submission and reducing per-element overhead on multi-buffer transfers.How?
On CUDA 13+, issue one
cuMemcpyBatchAsyncfor the whole SGL, with all mappings tracked in a per-eventuct_cuda_ipc_sgl_mapping_tand unmapped on completion. On older CUDA, fall back to per-elementuct_cuda_ipc_ep_put_zcopy, arming the user completion only on the last element.