A follow-up pull request in the llama.cpp repository has optimized low-level CPU dot product operations for the q1_0 ...