Optimize reduction stage of dot product of q4_L/q5_K to q8_K on AVX2#22181

nariox wants to merge 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

nariox:no-vphaddw-on-avx2nariox/llama.cpp:no-vphaddw-on-avx2Copy head branch name to clipboard