llama.cpp

ggml-cpu : add AVX2 and AVX optimization for nvfp4 dot product

#23961

Open