From: Pjotr Prins <pjotr.public12@thebird.nl>
To: Ricardo Wurmus <rekado@elephly.net>
Cc: guix-devel@gnu.org, Dave Love <fx@gnu.org>
Subject: Re: OpenBLAS and performance
Date: Wed, 20 Dec 2017 20:28:02 +0100 [thread overview]
Message-ID: <20171220192802.GA8426@thebird.nl> (raw)
In-Reply-To: <87d139xo3v.fsf@elephly.net>
On Wed, Dec 20, 2017 at 07:15:16PM +0100, Ricardo Wurmus wrote:
> Is it really single-threaded? I remember having a couple of problems
> with OpenBLAS on our cluster when it is used with Numpy as both would
> spawn lots of threads. The solution was to limit OpenBLAS to at most
> two threads.
Looks like 1 on my system.
> > If I compile for a target it
> > makes a large difference.
>
> The FAQ document[1] says this:
>
> The environment variable which control the kernel selection is
> OPENBLAS_CORETYPE (see driver/others/dynamic.c) e.g. export
> OPENBLAS_CORETYPE=Haswell. And the function char*
> openblas_get_corename() returns the used target.
>
> [1]: https://github.com/xianyi/OpenBLAS/wiki/Faq
>
> Have you tried this and compared the performance?
About 10x difference on 24+ cores for matrix multiplication (my
version vs what comes with Guix).
I do think we need to default to a conservative openblas for general
use. Question is how we make it fly on dedicated hardware.
package python-numpy:openblas-haswellp
for the parallel version?
also for R and others. Problem is that we blow up the types of
packages.
Pj.
next prev parent reply other threads:[~2017-12-20 19:31 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-12-19 10:49 OpenBLAS and performance Pjotr Prins
2017-12-19 17:12 ` Ludovic Courtès
2017-12-20 11:50 ` Dave Love
2017-12-20 14:48 ` Dave Love
2017-12-20 15:06 ` Ricardo Wurmus
2017-12-22 12:24 ` Dave Love
2017-12-20 17:22 ` Pjotr Prins
2017-12-20 18:15 ` Ricardo Wurmus
2017-12-20 19:28 ` Pjotr Prins [this message]
2017-12-20 20:00 ` Ricardo Wurmus
2017-12-20 20:32 ` Pjotr Prins
2017-12-20 19:02 ` Eric Bavier
2017-12-21 16:38 ` Dave Love
2017-12-20 23:02 ` Ricardo Wurmus
2017-12-21 10:36 ` Pjotr Prins
2017-12-21 14:43 ` Ludovic Courtès
2017-12-22 14:35 ` Dave Love
2017-12-21 16:17 ` Dave Love
2017-12-21 16:46 ` Ricardo Wurmus
2017-12-21 14:55 ` Ludovic Courtès
2017-12-22 12:45 ` Dave Love
2017-12-22 15:10 ` Ludovic Courtès
2017-12-22 16:08 ` Pjotr Prins
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://guix.gnu.org/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20171220192802.GA8426@thebird.nl \
--to=pjotr.public12@thebird.nl \
--cc=fx@gnu.org \
--cc=guix-devel@gnu.org \
--cc=rekado@elephly.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/guix.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).