Perf: skip redundant CSA scoring and fuse gather unpack

[Little-oil]

• Bypass query scoring and sorting when the ratio-4 cache fits Top-K
  in distributed CSA, retaining causal visibility, cache updates and
  the long-cache path.
• Read CSA gather payloads directly into typed consumers before epoch
  retirement, removing full packed buffers and separate unpack waves.
  Distributed CSA calls both typed gathers directly; TP1 uses its
  separate decode_csa_tp1 entry. Standalone transport fixtures exercise
  the same typed gathers, replacing their legacy packed implementations.

TP4 CSA at 16x256 on a2a3: minimum per-rank median latency falls from
1010.05 µs to 880.500 µs (12.83%). Measured with PYPTO_BENCH=1,
swimlane0, 100 timed rounds and 5 warmups, with all 400 timed samples
retained.

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: b5d50e66-7f11-49e8-92e1-8cab2420681d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 7fa23168-b998-419d-bf38-013d2aeeb961

📥 Commits

Reviewing files that changed from the base of the PR and between c80a7e3 and 3daaee2.

📒 Files selected for processing (3)

• models/deepseek_v4_flash_dspark/decode_cp_allgather.py
• models/deepseek_v4_flash_dspark/decode_csa.py
• models/deepseek_v4_flash_dspark/decode_indexer.py

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

The change adds typed CSA all-gather pipelines and a TP_SIZE==4 decode fast path. Cache-fitting requests bypass redundant indexer scoring and projection stages. Indexer helper annotations now expose task dependencies required by the new orchestration.

Changes

CSA TP4 decode path

Layer / File(s)	Summary
Typed CSA all-gather pipelines models/deepseek_v4_flash_dspark/decode_cp_allgather.py	Adds main and auxiliary typed all-gather steps. The steps decode gathered payload segments, synchronize workers, retire signals, and return completion dependencies.
TP4 dependency and projection bypasses models/deepseek_v4_flash_dspark/decode_csa.py, models/deepseek_v4_flash_dspark/decode_indexer.py	For TP_SIZE==4 requests whose compressed cache fits within IDX_TOPK, the orchestration uses typed gathers, skips selected indexer stages, computes bounded indices, and propagates task dependencies through updated return annotations. Existing paths remain for other cases.

Estimated code review effort: 4 (Complex) | ~45 minutes

Sequence Diagram(s)

sequenceDiagram
  participant decode_csa
  participant typed_allgather
  participant indexer
  participant outputs
  decode_csa->>typed_allgather: gather typed CSA payloads
  typed_allgather->>outputs: write projection and cache tensors
  decode_csa->>indexer: propagate task dependencies
  indexer->>outputs: write bounded Top-K indices

Loading

Merge Risk: ⚪ Minimal · up to 3daae

The TP4 decode optimization preserves fallback behavior, and no merge-blocking correctness or runtime risk is currently identified.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 8 functions across 3 files.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly summarizes the two main changes: skipping redundant CSA scoring and fusing gather unpacking for performance.
Description check	✅ Passed	The description is directly related to the changeset. It explains the TP4 fast path, typed gather integration, retained fallback behavior, and measured performance impact.

---

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

_{A rabbit reads each line,
The patch grows clear beneath the moon,
Small changes hop in place,
Tests guard the garden path,
Reviews bloom before the dawn.}

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