How could I accelerate FFT on Nano?

I need to use FFT to process data in python on Nano, and I currently use the scipy.fftpack.fft()。
But the speed is so slow and I want to utilize the GPU to accelerate this process.
I have tried cupy, but it takes more time than before.
Does there exist any other way to do FFT on GPU in Nano? I know that pycuda could, but implement a FFT in C seems hard to me.

Hi,

You can try our VPI library that supports both C++ and python interface.
https://docs.nvidia.com/vpi/algo_fft.html

/opt/nvidia/vpi1/samples/07-fft

Thanks.

AastaLLL,
thank you very much, I’ll try it.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.