|
|
Ublas : |
From: Thomas Klimpel (Thomas.Klimpel_at_[hidden])
Date: 2008-06-16 15:41:27
> As such, I'm not particularly unhappy with the setup as it is now.
> And Thomas is free to cherrypick patches off my tree [1] as he sees fit.
I have to admit that I'm not unhappy with the current setup either.
My current goal is to bring the regression tests into good shape and run regression tests on a variety of different setups. I'm not really spending much time into this, so it will take quite a while until this task is finished.
If Andreas would not provide well packaged releases, I would have to worry about migrating the organization structure of the boost-numeric-bindings to the structure described at http://www.boost.org/community/sandbox.html. This would invalidate some existing links, without providing any real benefit (only that libs/numeric/bindings would be more visible).
Regards,
Thomas
-----Original Message-----
From: ublas-bounces_at_[hidden] [mailto:ublas-bounces_at_[hidden]] On Behalf Of Andreas Klöckner
Sent: Montag, 16. Juni 2008 19:13
To: ublas mailing list
Subject: Re: [ublas] LAPACK Bindings HSEQR addition
On Mittwoch 11 Juni 2008, Karl Meerbergen wrote:
Hi Karl, all,
> Another option is to stop contributing to the lapack bindings in
> Boost.Sandbox and have Andreas repository as the 'official' lapack
> bindings repository. We could add a link to Andreas' repository. That
> would save others (including me) a lot of time trying to keep the
> 'official' bindings up to date.
Hmm. In principle I don't mind, but I think Thomas might not like that, and
he's recently done some significant work on the svn that needs to be
acknowledged. Whatever solution we choose should make optimal use of the time
both Thomas and I put into this.
As such, I'm not particularly unhappy with the setup as it is now. And Thomas
is free to cherrypick patches off my tree [1] as he sees fit.
> In my opinion, what is most important is that the vector and matrix
> traits classes are stable and are fixed in the Boost.Sandbox. I am less
> concerned with the bindings themselves (LAPACK, MUMPS, UMFPACK, etc.),
> as long as they use the traits classes from the official bindings.
I fully agree. But the past seems to have shown that the traits classes, as
infrastructure to the rest of the bindings, are mature and seem to have
required very little change.
Andreas
[1] http://git.tiker.net/?p=boost-numeric-bindings.git;a=summary (changed)