llama.cpp

llama : do not skip iGPU when only RPC devices are present

#23868

Merged