Re: are there plans for an octave profiler?

[Joe Koski]

on 5/13/05 11:05 AM, Ben Barrowes at address@hidden wrote:

> Has there been a discussion about implementing a profiler for octave? I
> find the profiler to be very useful when trying to optimize matlab code
> for speed.
> 
> In the mean time, are there any tools besides cputime and tic/toc that
> can help to identify performance bottlenecks in octave code?
> 
> Ben

Different computer languages have different strengths and disadvantages.
Octave is interpreted, which will never be blazingly fast because inside a
loop because it just keeps reinterpreting the same octave instructions into
machine code again and again.

The strengths that I've discovered are that Octave is easy to debug because
it tells you exactly where it is having problems, and it is easy to try,
test and change until you get things right. That is a good byproduct of
interpreting line-by-line.
 
If you want octave to go fast, write your routine in Fortran or C, put a
wrapper on it, and call it from Octave. That is essentially what many octave
and octave-forge routines do.

To me Octave is a day-to-day working language for things like data analysis
of results from Fortran-based computational fluid dynamics codes and
experimental data. It's also a very good teaching tool. For these purposes,
Octave is fast enough.

My personal opinion is that a profiler for Octave would not be that useful
unless an Octave compiler also exists.

Joe




-------------------------------------------------------------
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
-------------------------------------------------------------