llama.cpp

opencl: add optimized q4_1 mm kernel for adreno

#19840

Merged