Will llama.cpp support NPU in new AI CPUs?

[victorx98]

We will have multiple CPUs that are equipped with NPU and more power GPU over 40 TOPS, like Snapdragon X Elite, Intel Lunar lake and AMD Ryzen 9 AI HX 370. Is it possible for llama.cpp to use the combined power from their NPU and GPU?

Thanks

[hpvd]

see links to issues in #9181