llama.cpp

Vulkan: Tune Vulkan mmq int dot shader for performance

#12767

Merged