llama.cpp

b5fcf8ef - ggml : fix and optimize ppc64le (ggml/849)

ggml : fix and optimize ppc64le (ggml/849) * fix compile issues introduced by loongarch_asx * restore quant changes to merge * fix compile issues introduced by loongarch_asx * further optimize by using vec_msum & vec_sum4s on ppc64le