|Re: [eigen] Eigen "views"|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Eigen "views"
- From: Jose Luis Blanco <joseluisblancoc@xxxxxxxxx>
- Date: Tue, 16 Aug 2011 17:29:43 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type; bh=6qZbVEwjYQpl6+9W+TfK7xdiPUDnQG6F6ciWFiez/VI=; b=K6S9CaOFxAO9pD6b22zZKMUoTatbdxyd+mYSrWvoRLr2uBUfKSDOFqq0PUUAXqQnvi Rfi+L9xEaw/fnUrdrV4qZWlG9wvjGIWus8Hsy56hoKP1nsg+PRDgQdAfzOMOcr3LuVY0 wu1Djgbv3qcbP9/xKt8fEd5JNMfEu6EUSJAgk=
On Tue, Aug 16, 2011 at 5:03 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
> ...or rather, just the fact that each packet load would have to do 4
> (or 2) random memory accesses.
Exactly... that's why I think it would be more efficient to first
"lose" time arranging all the sparse coefficients together into a
temporary Matrix<> from the new class MyViewExpression<>, and then
operating with that dense matrix that would vectorize as usual.
But probably all is a matter of benchmarking and seeing the pros/cons
of using the temporary dense matrix for each situation.