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
Yes, to replace "mask" and fancy-indexing, generically, we would likely need to replace the Python-level functions. There might be a way to speed things up a bit by replacing functions on the "PyArray_ArrFuncs *f" member of the PyArray_Descr ("dtype") object. There are functions there that also need to be swapped out to speed up sorting, and such. See lines near 443 to 553 of ndarraytypes.h (though I may have an old version of the NumPy code checked-out).
boolean and fancy index masking related routines can be sped up and are used very often.
Be nice if we could hook these routines.
The text was updated successfully, but these errors were encountered: