llama.cpp

23ee8797 - opencl: q8_0 gemv precision improvement (#24923)