This is the mail archive of the
libc-alpha@sources.redhat.com
mailing list for the glibc project.
Re: PATCH: ldconfig.c considers ld.so.conf entries before /lib,/usr/lib
- To: Dirk Eddelbuettel <edd at debian dot org>
- Subject: Re: PATCH: ldconfig.c considers ld.so.conf entries before /lib,/usr/lib
- From: Jakub Jelinek <jakub at redhat dot com>
- Date: Thu, 30 Aug 2001 10:15:50 +0200
- Cc: libc-alpha at sources dot redhat dot com, Ben Collins <bcollins at debian dot org>, Camm Maguire <camm at enhanced dot com>
- References: <15245.45665.961678.23004@sonny.eddelbuettel.com>
- Reply-To: Jakub Jelinek <jakub at redhat dot com>
On Wed, Aug 29, 2001 at 10:26:25PM -0500, Dirk Eddelbuettel wrote:
> III. Illustration of Atlas' benfits
>
> This shows some echo'ed code for R and Octave (doing simple matrix cross
> products) along with timings in seconds.
>
> Evaluating Atlas libs at /usr/lib/atlas on size 1500
>
> GNU R: m<-matrix(rnorm(1500*1500),ncol=1500); cat(system.time(crossprod(m))[1])
> Without Atlas: 92.66
> With Atlas : 9.5
>
> GNU Octave: X=randn(1500,1500); t=cputime(); Y=X'*X; disp(cputime-t)
> Without Atlas: 96.100
> With Atlas : 9.8200
>
> Evaluating Atlas libs at /usr/lib/3dnow/atlas on size 1500
>
> GNU R: m<-matrix(rnorm(1500*1500),ncol=1500); cat(system.time(crossprod(m))[1])
> Without Atlas: 92.7
> With Atlas : 7.86
>
> GNU Octave: X=randn(1500,1500); t=cputime(); Y=X'*X; disp(cputime-t)
> Without Atlas: 96.060
> With Atlas : 8.1900
If libblas.so from Atlas is that much better, I don't see the point why you
just don't package it up separately as standard libblas.so.
Also, if you place it under say /usr/lib/`uname -m`/libblas.so, ldconfig
will pick it up in the cache as architecture optimized and thus ld.so will
pick it up preferably to /usr/lib/libblas.so.
Jakub