EXAMPLES/DEVICE/EP/CSRC: Removed extra params from ht functions.

[rakhmets]

What?

Removed unused low_latency_mode parameter from the functions from ht namespace.
Removed type parameter from ht::combine, and moved the assert from ht::combine to Buffer::ht_combine.

Why?

Reduced hanges in #1793 to facilitate review.

Summary by CodeRabbit

Summary

• Refactor

  • Simplified high-throughput notification/dispatch/combine interfaces by removing the low-latency mode option.
  • Reduced template branching so fewer execution variants are selected at launch.

• Bug Fixes

  • Added stricter scalar-type validation during combine operations to ensure incompatible input types are rejected earlier.

[github-actions]

👋 Hi rakhmets! Thank you for contributing to ai-dynamo/nixl.

Your PR reviewers will review your contribution then trigger the CI to test your changes.

🚀

[coderabbitai]

📝 Walkthrough

Walkthrough

This change removes low-latency-mode parameters from HT dispatch and combine kernel declarations, launch paths, and call sites, and adds a host-side BF16 validation before ht_combine calls combine.

Changes

HT dispatch and combine cleanup

Layer / File(s)	Summary
Dispatch API and launch path examples/device/ep/csrc/kernels/api.cuh, examples/device/ep/csrc/kernels/nixl_ep_ht.cu, examples/device/ep/csrc/nixl_ep.cpp	notify_dispatch and dispatch declarations and templates drop low_latency_mode; launch selection is reduced to cached/non-cached dispatch variants, and the ht_dispatch call sites stop passing the removed argument.
Combine API, launch path, and validation examples/device/ep/csrc/kernels/api.cuh, examples/device/ep/csrc/kernels/nixl_ep_ht.cu, examples/device/ep/csrc/nixl_ep.cpp	cached_notify and combine declarations and templates drop low_latency_mode, combine drops its leading datatype parameter, the launch wiring is updated to the new signatures, and ht_combine adds a BF16 scalar-type assertion before calling combine.

Sequence Diagram(s)

sequenceDiagram
  participant Buffer
  participant notify_dispatch
  participant dispatch
  Buffer->>notify_dispatch: launch without low_latency_mode
  Buffer->>dispatch: launch without low_latency_mode

Loading

sequenceDiagram
  participant Buffer
  participant cached_notify
  participant combine
  Buffer->>cached_notify: enqueue barrier without low_latency_mode
  Buffer->>combine: call with BF16-validated input and no type argument

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

🐇 I hopped through kernels, swift and bright,
With one less flag to dim the flight.
BF16 got its tiny test,
Then combine sailed on with cheerful zest.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 8.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly matches the main change: removing extra parameters from ht functions in the examples/device/ep/csrc area.
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.
Description check	✅ Passed	The PR description includes the required What and Why sections and clearly matches the changes; the optional How section is omitted, which is acceptable.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[ColinNV]

Also hard-wiring scalar type?

The merge-base changed after approval.

[rakhmets]

/build