[None][perf] Use FlashInfer MXFP8 GEMM for MiniMax-M3 decode

[peihu-nv]

@coderabbitai summary

Description

MiniMax-M3 decode CUDA graphs execute 240 MXFP8 linear GEMMs per replay. This PR uses autotuned FlashInfer mm_mxfp8 for those captured GEMMs while reusing TensorRT-LLM's existing quantized activations, weights, and scales without repacking.

The automatic path is enabled only for MiniMax-M3 decode CUDA-graph capture. Eager execution, CTX/prefill, and piecewise prefill graphs remain on the native TensorRT-LLM backend; an environment override and native fallback are preserved. Matched GB200 serving improved total/output throughput by 10.4% and reduced median TPOT by 9.8%, with median TTFT unchanged.

Test Coverage

• Focused GB200 tests covering native default, explicit override, zero-copy FlashInfer dispatch, auto eager/decode selection, invalid configuration, and autotuner warmup: 5 passed.
• tests/integration/defs/accuracy/test_llm_api_pytorch.py::TestMiniMaxM3::test_nvfp4[use_msa=True]: passed on GB200 with MMLU 84.844% and GSM8K 90.637%.
• Matched GB200 disaggregated serving A/B, TP4/EP4 with attention DP, FP8 KV cache, ISL 8192 / OSL 1024, concurrency 64, and 1,000 requests: 21,599.7 to 23,840.6 total token/s; median TPOT 25.269 to 22.798 ms.

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

• Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.