[NIXL][PCP][DCP] Expose PCP producer KV shards as transfer ranks - #56645
Merged
LucasWilkinson merged 4 commits intoSep 13, 2026
Merged
Conversation
mergify
Bot
added
deepseek
LucasWilkinson
force-pushed
the
codex/nixl-pcp-dcp-main
branch
from
September 12, 2026 23:28
84a3509 to
0b9f0e4
Compare
LucasWilkinson
marked this pull request as ready for review
September 12, 2026 23:41
LucasWilkinson
requested review from
ApostaC,
NickLucche,
ProExpertProg,
WoosukKwon,
houseroad,
ivanium,
mgoin,
njhill,
orozery,
robertgshaw2-redhat,
tlrmchlsmth,
xuechendi,
yewentao256 and
youkaichao
as code owners
September 12, 2026 23:41
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
LucasWilkinson
added
the
ready
Collaborator
Author
|
/ci run |
|
✅ Triggered Buildkite CI #88573 for commit |
LucasWilkinson
enabled auto-merge (squash)
September 13, 2026 05:56
Rebase vllm-project#54496 onto vllm-project#56157. Publish each TP1 PCP+DCP producer shard through the existing DCP transfer topology and retain upstream completion aggregation. Share transfer-size calculation between the pull and push schedulers and validate the PCP/DCP group in the connector. Cover shard publication, completion, push scheduling, and advertised transfer topology for replicated and sharded producers. Co-authored-by: OpenAI Codex <noreply@openai.com> Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
LucasWilkinson
force-pushed
the
codex/nixl-pcp-dcp-main
branch
from
September 13, 2026 06:35
0b9f0e4 to
c2ef591
Compare
Collaborator
Author
|
/ci run |
|
✅ Triggered Buildkite CI #88660 for commit |
Keep the synthetic worker fixture aligned with the transfer-rank fields used by NIXL cache registration. Co-authored-by: OpenAI Codex <noreply@openai.com> Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
…ex/nixl-pcp-dcp-main Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> # Conflicts: # tests/v1/kv_connector/unit/test_nixl_desc_geometry.py
Collaborator
Author
|
/ci run |
|
✅ Triggered Buildkite CI #88672 for commit |
Co-authored-by: OpenAI Codex <noreply@openai.com> Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Collaborator
Author
|
/ci run |
|
✅ Triggered Buildkite CI #88675 for commit |
ItsRoy69
pushed a commit
to ItsRoy69/vllm
that referenced
this pull request
Sep 15, 2026
…m-project#56645) Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: OpenAI Codex <noreply@openai.com>
keneoneth
pushed a commit
to keneoneth/vllm
that referenced
this pull request
Sep 16, 2026
…m-project#56645) Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: OpenAI Codex <noreply@openai.com>
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
2 participants
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.
Expose TP1 PCP+DCP KV shards as NIXL transfer ranks. Preserve completion reports from every sharded PCP rank in
get_transfer_results(); replicated PCP still reports only from rank zero.Builds on merged #56157; replaces #54496. Distinct from #45340’s block accounting.
Validation: 18 PCP producer/topology tests passed on B300 with standard fixtures, including NCCL initialization. 72 TP-mapping/push tests passed on CPU. Before the guard, a CPU run of the corrected producer test failed all four nonzero PCP+DCP cases; afterward all eight cases passed. Pre-commit/mypy passed.
Combined-head GLM-5.2-NVFP4 PCP4+TP1+EP4+DCP4 → TP4 tests passed in eager and CUDA-graph modes: GPU and CPU-offload each 28/32 GSM8K, direct TP4 26/32 in both runs. All 32 GPU/CPU pairs matched tokens/log probabilities in each run, with confirmed CPU hits; 6,155-token retrieval also passed (6,144 CPU-restored tokens; continuations differed). Zero transfer failures or pending stores. PCP requires
-cc.cudagraph_mode=PIECEWISE; its default FULL_AND_PIECEWISE fails startup. Graph run captured prefiller sizes 4–1024 and logged 100 PIECEWISE prefiller / 8,432 FULL decoder dispatches. Uses default breakable graphs, not torch.compile. Clean merge: mainb6e2aa748b, #56645f995e33f61, #5580210afd0150e, #56715899a24c572; merged #56621, managed precompiled native extensions, no optional top-k changes. Small sample, not a full accuracy benchmark. UCX errors occurred only during teardown; GPUs released.AI assistance: OpenAI Codex. Human review and validation remain required.