|Re: [eigen] Eigen2 in Krita|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Eigen2 in Krita
- From: "Gael Guennebaud" <gael.guennebaud@xxxxxxxxx>
- Date: Sat, 23 Aug 2008 17:00:02 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:in-reply-to:mime-version:content-type :content-transfer-encoding:content-disposition:references; bh=tLUqbmsdfEhFYR6+bBcriT/OcgUsW5gJmQ3o87u38Sc=; b=ite7ZiR6uFQw29bP8QAUr5jUWnVrHSn/GFEdcJfXM1VX3Mj+V9wPp7kfnKbwyMs022 d+Qo+7F9BimhPxJgaBsVcHRmXOGuIjsnRhz6JBg91dKCCneoa4mJ12V6vKoVFc8davo9 z7pVoyF3KJETagYYTj4WyaNjA+/NGPDzduZZE=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=L8I4Q4YepmIjFUvHyNjF52zsKZrKon4PVSxB+1QTBmADKPe1Ki4dO97w6KXOiwvLt0 29SZZmwVeIfqB0v/f4y1P6fLzK81CuYnw7c6yZY3Lt7X4PogRPFrhvwzj7Xoq3YQ3Lg2 djJs/efprYbmiJEbaoHYwFL8FJi7FjdMG/p38=
that's great new ! It's very nice to see Eigen2 starting to be used
here and there...
On Sat, Aug 23, 2008 at 6:35 AM, <jacob@xxxxxxxxxxxxxxx> wrote:
> Their code lends itself very well to this porting, for example they
> have a lot of
> center = (p1 + p2) / 2;
> which is benefiting a lot from expression templates (only one vector
> traversal instead of 3). Another nice thing is that Vector2d is of course
> nice for vectorization. One of the places where this should bring large
> performance improvements is in the Bezier curves.
For this example I don't expect any improvement from expression
templates, these vectors are much too small for that. The compiler is
smart enough to optimize that itself. On the other hand the
vectorization might help a bit, sure.
> There are other usages, for example Cyrille wrote a perspective
> transformation tool a while ago, using a 9x9 LU solver which is now provided
> by Eigen.
> The most useful thing in the future would be a complete Sparse module with
> some features such as least squares (Ben and Cyrille could refresh us on
> what exactly is needed). That would allow to remove the GMM dependency, and
> hopefully to improve performance. Perhaps for KOffice 2.1...
yes that would be cool to know what is really needed here such that we
can focus on that part first.