llama.cpp

vulkan: optimize coopmat2 q2_k dequant function

#11130

Merged