search

CodeSchmiedeHGW / BLITZ

Bulk Loading and Interactive Time series Zonal analysis
GNU General Public License v3.0
1 stars 0 forks source link

Performance / GPU #16

Closed PiMaV closed 1 week ago

PiMaV commented 7 months ago

These can be easily done: pg.setConfigOptions(useNumba=True) pg.setConfigOptions(useOpenGL=True)

this is a bit more complex: (Maybe as an option to set in the ini File) User needs to install cupy manually beforehands. pg.setConfigOptions(useCupy=True) # Requires CUDA and CuPy ... Bad for small laptops

irkri commented 7 months ago

I don't think that any of these do something for BLITZ. Right now, we do all the heavy stuff on our own. Pyqtgraph only shows the images we manipulate using plain numpy. Calculating the mean inside the ROI or the normalization stuff doesn't use numba either. (This is something that pyqtgraph handles.)

PiMaV commented 7 months ago

Lets drop this for now

PiMaV commented 2 weeks ago

Can we make (more) use of GPU? building a special cupy version? Might not be worth the trouble for BLITZ, but for heavy pca use it might be.

PiMaV commented 1 week ago

Will be taken care when dealing with the PCA properly