help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: benchmarking - evidence that there's no problem


From: David Bateman
Subject: Re: benchmarking - evidence that there's no problem
Date: Tue, 16 Nov 2004 09:57:11 +0100
User-agent: Mutt/1.4.1i

According to address@hidden <address@hidden> (on 11/15/04):
> On Mon, Nov 15, 2004 at 02:56:13PM +0000, Michael Creel wrote:
> > Here are some benchmark results using my USOBS.m script, which is attached.
> > This is certainly not conclusive, but it seems that the problem I had last 
> > week may have been due to something other than Octave.
> > Michael
> 
> You don't have difference but I do (only in  your matrix create)
> 
> 2.1.50  matrix create       5.702     100.000       0.057
> 2.1.57  matrix create       5.620     100.000       0.056
> 2.1.61  matrix create       8.653     100.000       0.087
> 
> I don't know why rand in my 2.1.61 is 1.5x slower (all version use
> some octave-forge, 2.1.61 has recent (about 7 days) from cvs)

Check the 32/53 bit issues in the octave-forge rand functions. You can
either use the 53-bit versions and get better coverage of all possible
random values, or the 32-bit versions and have some double floating 
point values that can not be obtained... The cost is speed. If you
don't care that all floating point values between 0 and 1 can be obtained
by rand use the 32-bit versions and get back the 1.5 factor you observed.


> But my real (deterministic) application is going to infinity :(
> 
> 2.1.50  68.4  seconds
> 2.1.57 179.37
> 2.1.61 261.90
> 
> My application is (badly) complicated - above 180 m-files.
> Today I will finish my code instrumentation to find drawbacks.

Can you give a simple test case. My experience is exactly the reverse,
and the newer versions are faster.. A slow-up like this is a clear 
regression.

> Another issue with my code is that results from 2.1.50 are different to
> 2.1.57/2.1.61 (some rounding problem???).

Again a simple example would help in deciding if this is an issue...

Regards
David

-- 
David Bateman                                address@hidden
Motorola CRM                                 +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 1 69 35 77 01 (Fax) 
91193 Gif-Sur-Yvette FRANCE

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------



reply via email to

[Prev in Thread] Current Thread [Next in Thread]