Ilhan Polat
Ilhan Polat
option # 1 is for all intentions and purposes, impossible. That ship has sailed a few decades ago. > Some users could end up faced with runtime failures ( INFO...
> Would it be thinkable that the reference implements C
> I think by "this function" you mean C
Indeed, if we have these per array vendor implementations for such operations, then probably SciPy won't be needing all that home-baked Cythonized array utilities to shave off some microseconds. Then...
> which I believe is all there is to it for exact symmetry. Ah yes, this is one of the examples I have. So imagine you have a 5000x5000 array...
Yes exactly here is what we are doing currently https://github.com/scipy/scipy/blob/main/scipy/linalg/_cythonized_array_utils.pyx#L239-L365
Ah I should clarify, the np.allclose is used only if the user wants an inexact comparison, that is if things can be off a little bit in the lower and...
The SciPy signature is going to have another keyword here denoted by `p_indices` for lack of a better word to return the permutation as a 1D array. See https://github.com/scipy/scipy/pull/18358 `lu(a,...
The `lu_solve`, `lu_factor` are practically historical hacks for the lack of a better API so I would really discourage any standardization around it. Also LU is the most essential starting...
A couple of quick points; - The `dimension too large` is data dependent. If decorations are used in tight places then you will get errors. The typical source for this...