lapackpp-devel

[Lapackpp-devel] Re: Hello, thank you for lapack++

[Christian S. <sti...@tu...>]

Dear Maho,

thank you very much for your contribution. I have added this to CVS.

NAKATA Maho schrieb:
> 1. void LaEigSolveVecIP(LaSymmMatDouble &S, LaVectorDouble &eigvals);
> solves symmetric matrices in place and also return eigenvectors
> as we specified.

I don't use that function and I don't quite understand what it should 
do. However, if you say it segfaults then that's a bad thing and your 
provided patch is a good thing :-) Therefore I directly applied this to CVS.

> 2. Definition of erf for FreeBSD is incorrect.
> is it possible to remove throw() for FreeBSD?

It is not possible to simply remove the throw(), since on Linux the 
declaration in math.h includes this throw(). However, I just checked 
whether this declaration can be removed anyway, and in fact this seems 
to be the case. I have therefore removed this declaration at all, so it 
shouldn't give you any problems.

If you have any further questions, do not hesitate to ask me. Even 
better, you can subscribe to the mailing list lapackpp-devel and discuss 
these questions there. This has the additional benefit that the postings 
will be archived.

Best regards,

Christian Stimming

[Lapackpp-devel] Re: Hello, thank you for lapack++

[NAKATA M. <ch...@ma...>]

In Message-ID: <413...@tu...> 
Christian Stimming <sti...@tu...> wrote:

Dear Christian,

> I don't use that function and I don't quite understand what it should 
> do. However, if you say it segfaults then that's a bad thing and your 
> provided patch is a good thing :-) Therefore I directly applied this to CVS.

thanks! However, my investigation seems to be somewhat odd, however
I don't have better clue which I sent you the last e-mail.
o LaSymmMatDouble is inherted by triangle matrices, and class of
triangle matrices is inherted by lagenmatdouble. so apparently we have
full matrix for LaSymmMatDouble. However, my stupid patch just copies
from LaSymmMatDouble to LaGenMatDouble without thought.

> If you have any further questions, do not hesitate to ask me. Even 
> better, you can subscribe to the mailing list lapackpp-devel and discuss 
> these questions there. This has the additional benefit that the postings 
> will be archived.

okay, I'll subscribe this ML. 

And I noticed before that Lapack++ (also ver 1.1 and 2.0)
is not usable for FreeBSD/amd64, passing a invalid type of pointer. 
Now I'm investigating... 

thank you very much for your effort, since I have been
a fun of lapack++, which had not been maintained for long time...

See you at mailing list!
--nakata maho