Performance optimization strategies

koenderinklab / OpenDDM

https://koenderinklab.github.io/OpenDDM/

Apache License 2.0

3 stars 2 forks source link

Closed mwakok closed 2 years ago

mwakok commented 2 years ago

We can look into various methods to optimize the performance of the code:

[x] Vectorization with numpy
[x] Optimize dataset (resolution, number of frames, bitdepth)
[x] CPU parallelization with Dask
[x] Just in time compiler with numba
[x] Evalute different fft implementations (numpy, scipy, pyfftw, tensorflow, pytorch)
[x] GPU processing

mwakok commented 2 years ago

I cam across this notebook benchmarking various fft implementations: https://thomasaarholt.github.io/fftspeedtest/fftspeedtest.html