|Re: [eigen] News on sparse modules and 3.1 beta 1 on Monday|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] News on sparse modules and 3.1 beta 1 on Monday
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Sat, 3 Dec 2011 18:13:41 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type:content-transfer-encoding; bh=MoykgSx1/uAoHbwj9Y2K+4On0nlMWbhoUZIbrXfeAZ8=; b=bWCxjWV6H+QLJ/1lYoLgBZMswT5FX5y0eYZvLzjoS4hCr33404ispRWwBkM8X6iY8Q YTD1vmKKb77aIS8t75OjpW9LdGDv2z+sjEEIf6X95dVGZ+sEwH/amwLQbGe2neyZL5tI nuG72j5m48WALd85m3IpgFftTuvZuesPEYwYo=
On Sat, Dec 3, 2011 at 2:47 PM, Christoph Hertzberg
> Do you plan to keep _valuePtr(), etc? They are marked as internal, however
> they are very handy when interacting with other libraries.
Good point, I forgot about them. Actually I plan to make them official
and remove the "_" prefix.
> I guess that solves http://eigen.tuxfamily.org/bz/show_bug.cgi?id=284 ?
> (Except for re-computation of matrix*matrix, but that was a rather exotic
> There might be other bugs already fixed (some very old bugs just have a
> non-working link to bitbucket and have quite likely been fixed meanwhile)
sure we should have a more careful look at the bug list.
> The example code in:
> uses 3 different identifiers for the same matrix.
oops, no comment!
> And I think you should mention that it is also possible to insert entries at
> the back, keeping the compressed-ness of the matrix.
That's on purpose. For now I prefer to keep such feature internal
because the fast insertion into a compressed form requires to put the
matrix into a state where you can only do "compressed insertion"
before you call finalize(). This is rather complicated and dangerous
so => internal only !
In the near future I plan to add the possibility to append a raw list
of triplets and let Eigen figures out itself what's the best insertion
strategy. This is the cleanest solution I can think of.
>> I think that now they are sane enough to release a first beta of the
>> future 3.1 version of Eigen for which stable sparse modules will be
>> the major feature.
> As I asked in IRC already:
> Can you consider applying the patch to bug
> into 3.1 beta? (Or give me suggestions how to modify/improve it)
>> I plan to do the release on Monday. Though this is less critical for a
>> beta release, reports of the results of the unit tests on various
>> systems, especially with MSVC, would still be appreciated.
> Dipl.-Inf. Christoph Hertzberg
> Cartesium 0.051
> Universität Bremen
> Enrique-Schmidt-Straße 5
> 28359 Bremen
> Tel: (+49) 421-218-64252