Add vLLM-metax vllm metax cmake presets compiler arg

[ghangz]

Summary

• Adds a focused vllm metax cmake presets compiler arg improvement for MetaX-MACA/vLLM-metax.
• The change targets MetaX MACA development and validation workflows, with emphasis on earlier diagnostics, reproducible logs, or safer benchmark tooling.
• Existing default behavior is kept compatible; the new logic is scoped to explicit checks, helper tools, or validation metadata.

Validation

• Verified on Gitee.AI MetaX GPU resources: vLLM-metax_vLLM image batch, 10/10 PASS; PyTorch-MACA batch also covered vLLM-metax runtime tools.
• Branch validation command: python -m pytest tests/tools/test_generate_cmake_presets.py
• Pull request text is intentionally ASCII-only to avoid encoding issues on web forms and API clients.

Review notes

• Source branch: ghangz:mengz/vllm-metax-cmake-presets-compiler-arg
• Target branch: MetaX-MACA/vLLM-metax:master
• Maintainers can modify this branch if follow-up adjustments are needed.

[gemini-code-assist]

Code Review

This pull request introduces the ability to specify an explicit CUDA-compatible compiler (such as cucc or mxcc in MACA containers) via a new --cuda-compiler CLI argument and a corresponding parameter in generate_presets. It also adds a new --output argument to specify the output path of the generated presets file, along with unit tests verifying this functionality. The review feedback highlights that if a relative path or command name is provided for the CUDA compiler, it should be resolved to an absolute path to prevent configuration failures when CMake is executed from a different working directory.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

If the cuda_compiler argument is provided as a relative path or as a command name, it is written directly to CMakeUserPresets.json without resolution. Since CMake configure presets are executed with the working directory set to the build directory (e.g., ${sourceDir}/cmake-build-release), any relative path will be resolved relative to the build directory rather than the project root, leading to configuration failures. Resolving the compiler path to an absolute path (similar to how python_executable is handled) ensures robust behavior.

    nvcc_path = cuda_compiler
    if nvcc_path:
        if not os.path.isabs(nvcc_path) and which(nvcc_path):
            nvcc_path = os.path.abspath(which(nvcc_path))
        elif not os.path.isabs(nvcc_path):
            nvcc_path = os.path.abspath(nvcc_path)
        print(f"Using CUDA-compatible compiler from argument: {nvcc_path}")