[Top][All Lists]

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

Re: moving toward a 3.0 release

From: David Grohmann
Subject: Re: moving toward a 3.0 release
Date: Thu, 28 Sep 2006 09:35:17 -0500
User-agent: Thunderbird (Windows/20060909)

Sebastien Loisel wrote:
I think Octave needs a new, stable release.

On graphics and Windows versions: the reason I've vanished for the last while was because my chosen platform (Windows + MinGW) was too buggy.

Finally, someone recently noted that Matlab has fast loops. Let me just add: holy matrices, Batman! Pointwise operations in Matlab are faster in loop style than in vector style. There's bound to be more Matlab code that won't work in Octave, not because of language incompatibility, but because of huge performance differences. For instance, my loop-style implementation of the Game of Life is faster than my matrix-style implementation.

Sebastien Loisel

Yes I can definitely say that as far as my organization goes, all the functionality is there or could be easily added to by our staff, *but* the performance of octave vs matlab in loops is so abysmal that we had to trash the idea of switching over. So if your goal is to increase the number of users of octave the most efficient way to do that is to simply (ok so I know its not simple) work on some kind of byte code or JIT backend for octave that speeds up loops.

If loops were fast, my entire research lab would switch over in an instant, and if any features were missing we would build them ourselves. We are dying to get away from matlab and all of its licensing issues. Vectorization is just not an option for most of the code we have.

David Grohmann
Senior Student Associate
Applied Research Lab : UT Austin : ESL - S206
Office: 512-835-3237

reply via email to

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