Ilhan Polat
Ilhan Polat
Does `--keep-cr` help for you? I mean I don't think anybody cares about the format so we can regenerate the file on a linux box and be done with it.
I guess this won't hurt anyways until someone else comes and fixes back. Thanks @hoodmane ,all!
Thanks for the report @projekter Could you please attach the `error.txt` to the issue too if you can? This error is generated by the low-level LAPACK driver so might be...
Oh nevermind you already did. My bad sorry for the noise.
Quick update; this is definitely a bug. I can't quite figure out whether this is a new f2py bug or a problem in the wrapper yet but I'll update if...
Sorry for the delay on this one; so far I confused myself even more. For example if I complete your matrix to a full symmetric matrix or just use transpose...
This took me quite some time to figure out but I've prepared some Cython code just to bypass all the wrapper and `eigh` machinery and this is still failing. Hence...
> Yes, I just mean to say that R's approach is one that was already mentioned in the original post. I interpreted the mention of R's implementation as a suggestion...
First the technical part, the standard build of SciPy if I am not mistaken is built with ILP64 disabled. Hence you need to use a BLAS implementation that uses ILP64....
The copying mostly comes from the f2py overhead (f2py handles the conversion internally). > The overwrite_a keyword is kind of misleading, as it should be on ALWAYS Not really. Overwriting...