[None][fix] Enforce NCCL >= 2.28 at CMake configure time

[eopXD]

Summary

• Bumps the CMake find_package(NCCL ...) minimum version from 2 to 2.28
• allreduceOp.cpp uses NCCLWindowAllocator and createNCCLWindowTensor unconditionally, but these symbols are guarded by #if NCCL_VERSION_CODE >= NCCL_VERSION(2, 28, 0) in ncclUtils.h
• Without this fix, building with NCCL < 2.28 fails deep into compilation with cryptic undeclared symbol errors. With this fix, the build fails immediately at CMake configure time with a clear version mismatch message

Test plan

• Build with NCCL >= 2.28 succeeds (no change in behavior)

Summary by CodeRabbit

• Chores
  • Updated minimum NCCL dependency version requirement to 2.28 for multi-device support builds.

[coderabbitai]

📝 Walkthrough

Walkthrough

The minimum required version of NCCL dependency was updated from version 2 to version 2.28 in the CMake build configuration for multi-device support builds. No control flow or functional changes were introduced.

Changes

Cohort / File(s)	Summary
NCCL Dependency Version Constraint cpp/CMakeLists.txt	Updated NCCL minimum version requirement from 2 to 2.28 when ENABLE_MULTI_DEVICE is enabled.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main change: enforcing NCCL >= 2.28 at CMake configure time, which matches the primary modification in CMakeLists.txt.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description check	✅ Passed	The PR description clearly explains the issue, rationale, and test plan following the repository template structure with Summary and Test plan sections.

_{✏️ 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 and usage tips.}

[eopXD]

/bot run --disable-fail-fast

[bobboli]

Hi @eopXD , could the version be further pinned to >= 2.29 ? There are some device APIs added.
TRTLLM image has also bumped to 2.29.2 #11227

[eopXD]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #43714 [ run ] triggered by Bot. Commit: 926ddb6 Link to invocation

[tensorrt-cicd]

PR_Github #43714 [ run ] completed with state FAILURE. Commit: 926ddb6
/LLM/main/L0_MergeRequest_PR pipeline #34197 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[eopXD]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #44308 [ run ] triggered by Bot. Commit: 8c41ccb Link to invocation

[tensorrt-cicd]

PR_Github #44308 [ run ] completed with state FAILURE. Commit: 8c41ccb
/LLM/main/L0_MergeRequest_PR pipeline #34729 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[eopXD]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #44329 [ run ] triggered by Bot. Commit: 8c41ccb Link to invocation

[tensorrt-cicd]

PR_Github #44329 [ run ] completed with state FAILURE. Commit: 8c41ccb
/LLM/main/L0_MergeRequest_PR pipeline #34747 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[eopXD]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #44454 [ run ] triggered by Bot. Commit: 8c41ccb Link to invocation

[tensorrt-cicd]

PR_Github #44454 [ run ] completed with state SUCCESS. Commit: 8c41ccb
/LLM/main/L0_MergeRequest_PR pipeline #34858 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

[eopXD]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #44597 [ run ] triggered by Bot. Commit: 8c41ccb Link to invocation

[tensorrt-cicd]

PR_Github #44597 [ run ] completed with state SUCCESS. Commit: 8c41ccb
/LLM/main/L0_MergeRequest_PR pipeline #34983 completed with status: 'SUCCESS'

CI Report

Link to invocation