Add support for getting cpu info on Windows for llama_bench

[kylo5aby]

• I have read the contributing guidelines
• Self-reported review complexity:
  • Low
  • Medium
  • High

When run llama-bench -m model_name -o json on Windows, before this change, output:

    "build_commit": "a15ef8f8",
    "build_number": 3416,
    "cuda": false,
    "vulkan": false,
    "kompute": false,
    "metal": false,
    "sycl": false,
    "rpc": "0",
    "gpu_blas": false,
    "blas": false,
    "cpu_info": "",
    ....

after this change, output:

    "build_commit": "68504f09",
    "build_number": 3455,
    "cuda": false,
    "vulkan": false,
    "kompute": false,
    "metal": false,
    "sycl": false,
    "rpc": "0",
    "gpu_blas": false,
    "blas": false,
    "cpu_info": "Intel(R) Core(TM) Ultra 5 125H",

[slaren]

Looks like this is failing on Windows ARM.

[kylo5aby]

Looks like this is failing on Windows ARM.

Updated to support ARM, PTAL

[kylo5aby]

I see in the CI/CD pipeline, Server / server (Release) (pull_request) is fail, but I have run CI/Server on my forked repo on target branch, in which CI and Server are passed, and the change won't affect llama-server.

here is my CI link

[slaren]

The CI failure is not related to this change, it can be ignored. The code looks good but there are two issues:

• The hKey is leaked on success
• RegQueryValueEx is not guaranteed to return a NUL-terminated string, and that needs to be handled correctly to avoid buffer overflows

[kylo5aby]

Fixed. Thanks the feedback