You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
One way for huge optimizations would be to support natively Cupy or Cuspatial. I tried replacing the numpy import on the sys level with cupy however a few small inconsistencies arise. Although cupy requires Nvidia and cuda I feel the headline 10 times performance boost on most numpy operations feels highly worth it for the amount of work you'd need to facilitate native support.
The text was updated successfully, but these errors were encountered:
Thanks @dan771. I am interested in supporting GPU processing, especially for the highly parallelizable functions like flow directions. Numba also supports GPU processing, and so I may leverage numba for this purpose in the future.
One way for huge optimizations would be to support natively Cupy or Cuspatial. I tried replacing the numpy import on the sys level with cupy however a few small inconsistencies arise. Although cupy requires Nvidia and cuda I feel the headline 10 times performance boost on most numpy operations feels highly worth it for the amount of work you'd need to facilitate native support.
The text was updated successfully, but these errors were encountered: