|Re: [eigen] Getting Householder reflections right|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Getting Householder reflections right
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Mon, 11 May 2009 21:47:48 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=xeUqEDwvECx9j1K4sxuswkyCX9nrgecVZpOqcOUXwf4=; b=b+uVoc1WOsPJLfkbpiAapiqdWVSBmA0cBGgGCaFOiJTybAqVKktgVqH99Vk1tatJ2y KyWIPzXFbWPBBMHo46q5xxsTozu2zoU/uNsYWfVdTkhk8l5ets8m5qP+GWBKCDxdD6I+ B829hP4dFd2xeWSF2UQelXL7MzKy8QhRJD4kQ=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:content-transfer-encoding; b=HE6JZtJ7B7OfExHGGfsPthwIDRbbIgAKa+qrppO3l+YwMCMoAGZ7ZDEjnxOmICiXV0 7kRMypjPLNVmBhKe/qkuRlDxtMvaTuVu2f4vspW9/J55v7amvpMzeK1s8jTMDuynU0mX AaXRC4p6eqt/f3MrtaJD/LeQdDySFlmyWtMu4=
On Mon, May 11, 2009 at 4:10 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
> 2009/5/11, Gael Guennebaud <gael.guennebaud@xxxxxxxxx>:
>> 39 secs to extract Q ! there is clearly something wrong here ! anyway,
>> Triadiag is going to be slightly (?), significantly (?) updated by
>> Benoit, and we'll try not to forget to also bench the extraction !!
> I was going to ask you, in the QR module, what do you consider already
> polished and what is a quick port of external code?
that's a good question. Actually nothing is really polished as it
could be. The best parts are probably Tridiag and
SelfAdjointEigenSolver because I needed them. However I wrote them
from scratch quite quickly and there is still of work to be done. In
particular, one could factorize the Householder code, extend
Triadiag/Hessenberg to make them suitable for sub matrices, for in
place computation, make them better suited for their main use case:
Eigen decomposition, add specializations for small sizes (there is
already one for the tridiagonalization of 3x3 matrices), fix "equal to
zero" tests, etc.
> Yes I plan to redo whatever needs to be redone in QR (and
> eigensolvers) as really that's the same basic idea as in SVD.