[Numpy-discussion] performance matrix multiplication vs. matlab

David Paul Reichert D.P.Reichert at sms.ed.ac.uk
Fri Jun 5 10:39:23 EDT 2009


Thanks for the suggestion.

Unfortunately I'm using university managed machines here, so
I have no control over the distribution, not even root access.

However, I just downloaded the latest Enthought distribution,
which uses numpy 1.3, and now numpy is only 30% to 60% slower
than matlab, instead of 5 times slower. I can live with that.
(whether it uses atlas now or not, I don't know).



Quoting Jason Rennie <jrennie at gmail.com>:

> Hi David,
> Let me suggest that you try the latest version of Ubuntu (9.04/Jaunty),
> which was released two months ago.  It sounds like you are effectively using
> release 5 of RedHat Linux which was originally released May 2007.  There
> have been updates (5.1, 5.2, 5.3), but, if my memory serves me correctly,
> RedHat updates are more focused on fixing bugs and security issues rather
> than improving functionality.  Ubuntu does a full, new release every 6
> months so you don't have to wait as long to see improvements.  Ubuntu also
> has a tremendously better package management system.  You generally
> shouldn't be installing packages by hand as it sounds like you are doing.
> This post suggests that the latest version of Ubuntu is up-to-date wrt
> http://www.mail-archive.com/numpy-discussion@scipy.org/msg13102.html
> Jason
> On Fri, Jun 5, 2009 at 5:44 AM, David Paul Reichert <
> D.P.Reichert at sms.ed.ac.uk> wrote:
>> Thanks for the replies so far.
>> I had already tested using an already transposed matrix in the loop,
>> it didn't make any difference. Oh and btw, I'm on (Scientific) Linux.
>> I used the Enthought distribution, but I guess I'll have to get
>> my hands dirty and try to get that Atlas thing working (I'm not
>> a Linux expert though). My simulations pretty much consist of
>> matrix multiplications, so if I don't get rid of that factor 5,
>> I pretty much have to get back to Matlab.
>> When you said Atlas is going to be optimized for my system, does
>> that mean I should compile everything on each machine separately?
>> I.e. I have a not-so-great desktop machine and one of those bigger
>> multicore things available...
>> Cheers
>> David
> --
> Jason Rennie
> Research Scientist, ITA Software
> 617-714-2645
> http://www.itasoftware.com/

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

More information about the NumPy-Discussion mailing list