llama.cpp

opencl: add optimized q8_0 mm kernel for adreno

#18871

Merged