llama.cpp

Support multiple GPUs (split mode) on SYCL backend

#5806

Merged