llama.cpp

ggml-cpu: extend RVV quantization vec dot to higher VLENs

#22754

Merged