how to run test-backend-ops in single threaded mode

[amritahs]

I am trying to debug llamafile_sgemm API and I am invoking just MUL_MAT operations on CPU backend. I would like to know if it is possible to run test-backend-ops in single threaded mode so that debugging becomes easier

[slaren]

You would have to modify the call to ggml_backend_set_n_threads_fn in test-backend-ops.cpp to set the number of threads to 1 instead of std::thread::hardware_concurrency().

[nikhil-arm]

You can set this to 1 and recompile as well
https://github.com/ggerganov/llama.cpp/blob/afd9909a6481402844aecefa8a8908afdd7f52f1/ggml/include/ggml.h#L229