Hey Michael, When using generic filter, the function call overhead is huge. Are you using LowLevelCallable? I presume that is a significant bottleneck.
After you’ve done that, assuming LLC allows release of the GIL, you can use the dask-based skimage.util.apply_parallel to parallelise your code. HTH! At any rate, feel free to post more details about your problem in a new thread! =) Juan. On 28 Jun 2017, 7:11 AM +1000, K.-Michael Aye <kmichael....@gmail.com>, wrote: > Awesome effort! ;) > > Is it maybe worth, during the refactor, to evaluate how NDimage could be > supportive for auto-parallelization? > I’m thinking that it somehow would be able to split off into embarrassingly > parallel n tasks, if thrown to a n-core machine? > > I’m sorry if I talk non-sense, I’m just fighting to get GLCM faster, when > running it via “generic_filter” over large-ish images, so far not successful, > even trying to use numba, but as many things are already running in Cython, > so I guess I can’t expect much auto-improvements. ;) > > So, I’m anticipating to have to go to a cluster for my hundreds of images I > want to run multiple/many GLCM analyses on. > > Thanks again for the developer docs, they are great! > > Regards, > Michael > > > > On Jun 27, 2017, at 09:06, Stefan van der Walt <stef...@berkeley.edu> wrote: > > > > On Tue, Jun 27, 2017, at 07:49, Egor Panfilov wrote: > > > - Not sure that I'll find enough time to contribute to the code by > > > myself, but it won't harm if you could provide to the public some > > > directions on where to start with `scipy.ndimage` C-code. > > > > Making ndimage developer notes along the way would be extremely helpful! > > > > Stéfan > > > > _______________________________________________ > > scikit-image mailing list > > scikit-image@python.org > > https://mail.python.org/mailman/listinfo/scikit-image > > _______________________________________________ > scikit-image mailing list > scikit-image@python.org > https://mail.python.org/mailman/listinfo/scikit-image
_______________________________________________ scikit-image mailing list scikit-image@python.org https://mail.python.org/mailman/listinfo/scikit-image
