Detect NPUs and variant core specs?

Running on a Rockchip RK3588 board, llmfit shows "GPU: none". But that chip has an onboard Arm Mali-G610 MP4 GPU as well as a 6 TOPS NPU which can accelerate model performance (though it does require preprocessing of models to NPU format). Also the CPU cores are divided into 4 each of two performance classes, just like Apple's M* chips.

Is it feasible for llmfit to incorporate this data in its model capacity and speed estimates?

For reference here is a ChatGPT answer on detection of these system features, FWIW:

https://chatgpt.com/s/t_69a316ce1b0481918746b2fd988fd584