You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think we could actually get away without having the interputil_cython extension if you used np.searchsorted as that function accepts a list of points to insert. Unless you wanted to further optimize the matrix creation routines themselves. As they stand now, there isn't any speedup from cython except very little on the _interp_point_1D function.
From @rowanc1 on November 17, 2015 22:0
In the interp utils we should be using
np.searchsorted
rather thannp.argmin
over the vectors:Speed up of O(log(n)) vs O(n) for each point you are interpolating, could be significant, especially on large meshes.
Copied from original issue: simpeg/simpeg#170
The text was updated successfully, but these errors were encountered: