detail | 详细描述 | 詳細な説明

Hi, thank your excellent work！

i do the inference speed test with ncnn on jetson nano and find ncnn has faster speed in cpu mode than other mode (vulcan and cuda), i confuse to this phenomenon.

Anyone could tell me the reason, thank u!

In cuda mode, i use ncnn-with-cuda ( https://github.com/atanmarko/ncnn-with-cuda )