Add NVLS switch-channel cross-rank barrier api - #863
Open
RJ Souza (Empyreus) wants to merge 2 commits into
Open
Conversation
Introduce SwitchDevice2DeviceSemaphore, a switch (NVLS multicast) group barrier backed by multimem atomics: one signal() issues a multimem add on the shared arrival counter and wait() spins on this rank's reflected copy until all ranks arrive. Ordered (release/acquire) and relaxed variants are provided. Wire it into SwitchChannel as the barrier backend so no separate memory-channel semaphores or host-side barrier are needed; the barrier flag is carved from the connection's own multicast region. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 6f77eb0b-a070-4b73-b532-58ce7e39c93e
Cover the NVLS switch-channel barrier semaphore correctness and add relaxed/ordered barrier performance microbenchmarks under test/mp_unit. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 6f77eb0b-a070-4b73-b532-58ce7e39c93e
|
Azure Pipelines: There may be pipelines that require an authorized user to comment /azp run to 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a device-side group barrier built directly on the NVLS switch (NVLink SHARP multicast). A barrier becomes a single signal() / wait() pair backed by one multimem atomic on a shared arrival counter that the switch reflects into every rank — replacing the O(n²) mesh of memory-channel semaphores previously needed to synchronize a multicast group.