b3750
llama : llama_perf + option to disable timings during decode (#9355) * llama : llama_perf + option to disable timings during decode ggml-ci * common : add llama_arg * Update src/llama.cpp Co-authored-by: Xuan Son Nguyen <[email protected]> * perf : separate functions in the API ggml-ci * perf : safer pointer handling + naming update ggml-ci * minor : better local var name * perf : abort on invalid sampler pointer ggml-ci --------- Co-authored-by: Xuan Son Nguyen <[email protected]>