llama.cpp

Porting the improved K-Quant CUDA kernels to OpenCL

#1966

Merged