#29 ,as per the pull request, I note here some thoughts about cythonization.
By just declaring with cdef the integers i and j in the for loops it would be possible to improve a speed-up with a very limited code modification. Still that would require a cythonization (a .pyx file, importing cython, transpilation build and then importing the eventual .so extension... i suppose).
I leave it here as a side note, knowing that a quick benchmark allowed to have a 30% performance speedup in a benchmark I recall I made. I suppose the code already leverages at best numpy.
#29 ,as per the pull request, I note here some thoughts about cythonization.
By just declaring with cdef the integers
iandjin the for loops it would be possible to improve a speed-up with a very limited code modification. Still that would require a cythonization (a .pyx file, importing cython, transpilation build and then importing the eventual .so extension... i suppose).I leave it here as a side note, knowing that a quick benchmark allowed to have a 30% performance speedup in a benchmark I recall I made. I suppose the code already leverages at best numpy.