Open tomgoddard opened 1 year ago
Okay - certainly for cryoSPARC 3DVA this will be quite feasible I think (albeit not as responsive as on the GPU). I will need to do some tests for cryoDRGN, this may be very slow.
I will put this on my to do list. Wiggle should check the system for GPU installation and default back to CPU in the absence of GPU.
I've never tested Wiggle on Mac - so if you have issues I would be interested to hear what these are - I'm already aware that the UI needs to be able to scale as currently it is fixed and too large for some screens.
I am looking forward to moving my cryoDRGN ribosome tutorial output to the Mac to try this out. I'll let you know how it goes.
It would be great if Wiggle could compute maps with CUDA if cupy is available or without CUDA if not available. Although cryoDRGN users will no doubt have an Nvidia GPU linux system for running cryoDRGN they may analyze the data on other computers. I do this on a Mac. I think this would be widely used without CUDA, although I don't have a timing of how slow a map computation is without CUDA and it could be quite bad. Typically GPU acceleration speeds are on the order of 5-10 for neural net calculations (pytorch) and numpy type operations like FFT.