[voltjia]

Summary

• Re-target the already-reviewed InfiniOps dependency pin from build: pin NineToothed fused_add_rms_norm #606 to main.
• Update third_party/InfiniCore from 2b4f2f20584a69b074f62a455e3de7368e3446c2 to a6ba44fa568244c5dcc0f12b799f609483c06205.
• Introduce no InfiniLM source changes.

Motivation

#605 was merged into main as commit cc6d52ae04f4c9ff17229bc7fc13a9ad105aca6d. #606 was then accidentally merged into its old base branch, perf/initialize-ops-auto-tuning, rather than into main.

The resulting #606 merge commit is:

711054ce05267ad80245222763cbda69b91e3fe9

That commit contains the intended dependency pin, but it is not reachable from main:

origin/main -> cc6d52a  # contains #605 only
perf/initialize-ops-auto-tuning -> 711054c  # contains #605 plus #606

The complete diff from current main to 711054c is exactly one gitlink change:

third_party/InfiniCore 2b4f2f20584a69b074f62a455e3de7368e3446c2
                      -> a6ba44fa568244c5dcc0f12b799f609483c06205

This PR therefore carries the reviewed #606 change to main without rewriting merged history. It also keeps #607 stacked on the intended dependency state.

Closes N/A - replacement for the incorrectly targeted merge in #606.

Type of Change

• feat — new feature / new model
• fix — bug fix
• perf — performance improvement (no behavioral change)
• refactor — code restructuring without behavior change
• test — adding or fixing tests only
• docs — documentation only
• build / ci — build system or CI configuration
• chore — tooling, formatting, or other non-code changes
• Breaking change

Test Results of Involved Models on Supported Platforms (Please attach screenshots)

Platform	Result
NVIDIA A100-SXM4-80GB / CUDA 12.6.77	Dependency build and runtime validation passed

Validation inherited from the reviewed #606 change:

InfiniOps a6ba44fa:
- CUDA smoke wheel build with WITH_NINETOOTHED=ON: passed
- Full CPU/NVIDIA/Torch/Flash-Inference/W4-Marlin install build: passed
- Focused FusedAddRmsNorm suite: 180 passed, 52 skipped
- InfiniOps NVIDIA CI: passed

InfiniLM:
- _infinilm build against the dependent full InfiniOps prefix: passed
- Native nvidia:0 runtime creation: passed
- Operator tuning initialization smoke: passed
- Host-provided tuning cache regression from #605: passed

This replacement PR changes only the destination branch for the already reviewed gitlink; it does not alter the dependency content.

Benchmark / Performance Impact

N/A - manifest-only dependency update. The motivating provider benchmark remains:

Selection	Average launch time
Native FusedAddRmsNorm index 0	0.157911 ms
NineToothed index 9	0.093901 ms

Notes for Reviewers

• This is the follow-up to build: pin NineToothed fused_add_rms_norm #606 being merged into the wrong base branch.
• The head branch is intentionally perf/initialize-ops-auto-tuning, because that branch already contains the reviewed build: pin NineToothed fused_add_rms_norm #606 result and is also the current base for perf: add NineToothed fused QK norm #607.
• Reviewing this PR against main shows only the intended one-line submodule revision change.
• perf: add NineToothed fused QK norm #607 should remain open until this replacement dependency PR and perf: initialize InfiniOps auto tuning #605's merged changes are both present in main.

CI / ChatOps

CI will be requested manually on head 711054ce05267ad80245222763cbda69b91e3fe9.

---

Checklist

Title, Branch, and Commits

• PR title follows Conventional Commits.
• N/A - Replacement PR intentionally reuses the already-reviewed dependency-stack branch after build: pin NineToothed fused_add_rms_norm #606 was merged into the wrong base.
• Each commit message follows Conventional Commits.
• The branch is a small, reviewable dependency update.
• No stray merge commits from main.
• No fixup! / squash! / wip commits remain.
• N/A - Existing PR/branch/commit does not use the legacy issue format.

Scope and Design

• Changes are minimal and limited to the dependency manifest.
• No dead code, debug output, or TODOs.
• No unrelated formatting churn.
• N/A - No public InfiniLM API changes.

General Code Hygiene

• N/A - No source files were modified.
• N/A - No source-file newline changes were needed.
• N/A - No source whitespace or BOM changes were needed.
• N/A - No comments or identifiers were added.

C++ Specific

• N/A - No C++ files were changed.

Python Specific

• N/A - No Python files were changed.

Testing

• N/A - No requested platform was unavailable.
• N/A - Single request test: manifest-only dependency update; no model code path changed.
• N/A - Offline performance test: manifest-only dependency update; provider performance is documented above.
• Sanity test: the dependent InfiniOps suite and InfiniLM native runtime smoke passed.
• N/A - Service test: manifest-only update with no service-layer changes.

Build, CI, and Tooling

• Dependency manifest and recursive submodule state were checked.
• The dependent project builds cleanly with the pinned component revision on NVIDIA.
• CI was triggered manually.

Documentation

• N/A - No developer workflow, user-facing behavior, or build flag changed beyond the dependency pin.
• N/A - No breaking change.

Security and Safety

• No secrets, tokens, internal URLs, customer data, or personal hardware identifiers were committed.
• No third-party code was copied into this repository.
• N/A - No executable InfiniLM code was changed.