cuda: speed up Q1_0 extraction with byte permutes#73

khosravipasha merged 2 commits into

prismPrismML-Eng/llama.cpp:prismfrom

perf/cuda-q1-byte-permPrismML-Eng/llama.cpp:perf/cuda-q1-byte-permCopy head branch name to clipboard