search

matejak / imreg_dft

Image registration using discrete Fourier transform.
Other
245 stars 68 forks source link

GPU Support #22

Open sherjilozair opened 8 years ago

sherjilozair commented 8 years ago

It seems like this is the right kind of application which a GPU can speed up. Have you considered this?

matejak commented 8 years ago

I use 3rd party libs for computationally intensive operations - numpy and for the DFT, pyfftw is used if it is available. If you know 3rd party libs that can do the job faster, let me know. I consider imreg_dft a high-level project that should not contain any low-level code since it doesn't do anything really special under the hood. However, some of the algorithms (esp. the tiling one) can be improved substantially in terms of performance.

sherjilozair commented 8 years ago

Thanks for your comment. I'll try to switch in a GPU implementation of the FFT functions and benchmark it.

cissoidx commented 5 years ago

Thanks for your comment. I'll try to switch in a GPU implementation of the FFT functions and benchmark it.

hi, have you guys implemented a GPU version of this algorithm? It will be very helpful.

damien-ls commented 3 years ago

Any update on this? GPU implementation would be great.