llama.cpp

518a0148 - sycl: remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor (#12734)