RE: [eigen] [PATCH] Toeplitz matrix specialization |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: "eigen@xxxxxxxxxxxxxxxxxxx" <eigen@xxxxxxxxxxxxxxxxxxx>
- Subject: RE: [eigen] [PATCH] Toeplitz matrix specialization
- From: Ben Axelrod <baxelrod@xxxxxxxxxxxx>
- Date: Fri, 30 Jan 2009 07:43:33 -0800
- Accept-language: en-US
- Acceptlanguage: en-US
- Thread-index: AcmCqSsj3uuBOngSQgebfXPNTKiiSAAR/QzQ
- Thread-topic: [eigen] [PATCH] Toeplitz matrix specialization
FYI: If you are interested in storing only half of a triangular matrix, I wrote an article about how to do this a while ago: http://www.codeguru.com/cpp/cpp/algorithms/general/article.php/c11211/
-Ben
-----Original Message-----
From: Listengine [mailto:listengine@xxxxxxxxxxxxxxxxx] On Behalf Of Gael Guennebaud
Sent: Friday, January 30, 2009 2:04 AM
To: eigen@xxxxxxxxxxxxxxxxxxx
Subject: Re: [eigen] [PATCH] Toeplitz matrix specialization
On Fri, Jan 30, 2009 at 12:54 AM, Jitse Niesen <jitse@xxxxxxxxxxxxxxxxx> wrote:
> On Thu, 29 Jan 2009, Gael Guennebaud wrote:
>
>> to start the debate I put some thougths on the wiki:
>> http://eigen.tuxfamily.org/index.php?title=SpecialMatrix
>
> Does "we assume we don't need a special class for triangular storage" mean
> that triangular matrices are stored as general matrices (and thus half the
> memory is wasted)? That strikes me as an assumption that could be regretted
> later. So perhaps we should make sure that a special class for triangular
> storage can be added later.
in short yes for both. actually, in all our current cases (LU, LLt,
QR, Hessenberg, etc.) the triangular/hessenberg matrix are packed
with other data so that there is no overhead. So what we primarily
need is "view" to triangular part of a dense matrix. In the future, I
don't any reason why we could not add a compact triangular storage.