I saw this complaint everywhere. This is basically code for "AMD NPU Support." AMD prioritized support on Windows first like that's where all the ML devs were...
To crickets.
No one so much as looked at it.
There might be a chance in heck the NPUs actually work with something now, on both Windows and Linux. Still an off chance, because apparently a $384,000,000,000 company can't spare a single engineer to work on GGML/llama.cpp...