llama.cpp

tools: add quant-bench for profiling raw kernel performance

#19434

Open