| Re: [eigen] slow sparse dense product |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] slow sparse dense product
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Thu, 1 Mar 2012 10:17:23 +0100
- Authentication-results: mr.google.com; spf=pass (google.com: domain of gael.guennebaud@xxxxxxxxx designates 10.42.151.193 as permitted sender) smtp.mail=gael.guennebaud@xxxxxxxxx; dkim=pass header.i=gael.guennebaud@xxxxxxxxx
- 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; bh=l6Qd/imRbkZmHPzPJx+L6gzK8z3ibLNIwLu8Dnz6Lw8=; b=IrxJkDmhL0MjUVO6km/hUHlloM7YrygkpN/8H9OqMo1M9LkdsW00d0Fl+pe3GfcV92 jkI5tE4T2Bd5YI1dsQ+2WvK6i7V0bazq2vRiPhQqrxSPmL4H4IHzc256qrGoeiFhiDKS o9nWoPG7P2BGGamPfv1u4echtJKOvZWw0Jjh8=
Hi,
sorry for not replying earlier. This performance issue is now fixed in
the devel branch (changeset 4544:0b51d2327999). When the right hand
side is column major we now systematically performs sparse *
dense_vector products rather than looping over the row in the most
inner loop. This is indeed always faster even though this means
looping over the sparse matrix multiple times.
thanks for your report.
Gael
On Wed, Feb 22, 2012 at 10:30 PM, Sebastian Birk
<birk@xxxxxxxxxxxxxxxxxxxxx> wrote:
> Hi,
>
> again I encountered an unexpected slow product of a sparse matrix A and
> a dense matrix X. As can be seen in the attached file, computing A*X or
> A^H*X is about 50% slower for X being col-major than X being row-major
> and even computing A*c for all columns c of X. Obviously, the col-major
> storage order is non-optimal, but 50% slower seems to be too much.
>
> Sebastian
>