[GG] fix(distributed): isolate semantic B12X PCIe graph channels

[voipmonitor]

Purpose

Port the reviewed semantic PCIe channel lifecycle from #216 to the current b12x package contract. A process-local CUDA stream handle is not a stable identity for independently captured target, draft, encoder, profiling, production, and eager communication paths.

What changes

• Threads stable semantic channel_id values through V1/V2 CUDA graph capture, custom all-reduce, and DCP A2A.
• Separates target/draft/encoder and profile/production owners.
• Rolls disposable profiling graph and channel state back on success or failure.
• Closes custom all-reduce before NCCL and process groups while keeping __del__ non-collective.
• Uses only the renamed b12x package and updates the distributed regression tests.

This preserves the behavior and review fixes from #216 without retaining the obsolete sparkinfer import contract. It is intended to supersede #216.

Validation

Paired with b12x master plus PR local-inference-lab/b12x#125 on 2x RTX PRO 6000 Blackwell (remote GPUs 4-5):

• B12X fused all-reduce eager/capture/replay: 1 passed.
• B12X DCP A2A eager/capture/replay against the reference: 1 passed.
• Ruff lint, Python byte-compilation, and git diff --check: pass.
• Clean three-way composition after vLLM [GG] chore(b12x): port integration to renamed package #246: pass.

No throughput or VRAM gain is claimed; this is a communication ownership and teardown correctness change.

[coderabbitai]

Warning

Review limit reached

@voipmonitor, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 13 minutes

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5307890f-a02b-43de-aac8-a55f7375800d

📥 Commits

Reviewing files that changed from the base of the PR and between e2666d9 and 42055eb.

📒 Files selected for processing (19)

• tests/distributed/test_b12x_fused_all_reduce.py
• tests/distributed/test_custom_allreduce_lifecycle.py
• tests/distributed/test_dcp_a2a.py
• tests/v1/cudagraph/test_breakable_cudagraph.py
• tests/v1/spec_decode/test_dflash_cudagraph_lifetime.py
• tests/v1/worker/test_gpu_autoregressive_speculator.py
• tests/v1/worker/test_gpu_model_runner.py
• vllm/distributed/device_communicators/cuda_communicator.py
• vllm/distributed/device_communicators/custom_all_reduce.py
• vllm/distributed/parallel_state.py
• vllm/v1/attention/ops/dcp_alltoall.py
• vllm/v1/worker/gpu/cudagraph_utils.py
• vllm/v1/worker/gpu/model_runner.py
• vllm/v1/worker/gpu/spec_decode/autoregressive/cudagraph_utils.py
• vllm/v1/worker/gpu/spec_decode/autoregressive/speculator.py
• vllm/v1/worker/gpu/spec_decode/dflash/cudagraph.py
• vllm/v1/worker/gpu/spec_decode/dflash/speculator.py
• vllm/v1/worker/gpu/spec_decode/speculator.py
• vllm/v1/worker/gpu_model_runner.py

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[voipmonitor]

Superseded by #251. The DSpark context-KV FULL graph creates an additional distributed graph owner and therefore must land with this semantic channel lifecycle in one source state. #251 was rebuilt directly from the current dev/gilded-gnosis head, contains this PR plus the context graph feature, is non-draft and mergeable, and has no hidden stacked integration commit. Closing this duplicate avoids applying the same channel changes twice in release composition.