Date: Feb 24, 1997 1:56 PM
Author: John W. Eaton
Subject: Re: Octave vs Matlab ?

In article <blattner-2402971446520001@mac03-14.unine.ch>,
Blattner Peter <blattner@imt.unine.ch> wrote:

: is there any significant difference between Octave and Matlab in respect
: of speed ?

I suppose it depends on what you want to do.  If you want to know what
the differences are for problems you might care about, why not run
some tests for yourself?

Here is a comparision of a couple of simple things.  I ran this on an
HP 700 series machine.

		      Octave 2.0.x   Matlab 4.2
		      ------------   ----------

  3 nested loops          31.81         9.85      (CPU seconds)

  rand, svd, inv, lu       0.14         0.18
  (once)

  rand, svd, inv, lu       9.59        18.33
  (1 loop, 100 times)

  rand, svd, inv, lu       9.57        18.32
  (2 nested loops,
   100 times total)

Here is the code I used for the test:

  t = cputime;
    for n = 1:10; for m = 1:100; for k = 1:100;
      x = n*m^2*k^3;
    end; end; end
  cputime-t

  k = 50; n = 100;
  t = cputime;
    x = rand (k); [u, s, v] = svd (x); y = inv (x); [l, u, p] = lu (x);
  cputime-t

  t = cputime;
    for i = 1:n
      x = rand (k); [u, s, v] = svd (x); y = inv (x); [l, u, p] = lu (x);
    end
  cputime-t

  n = 10; m = 10;
  t = cputime;
    for i = 1:n; for i = 1:m
      x = rand (k); [u, s, v] = svd (x); y = inv (x); [l, u, p] = lu (x);
    end; end
  cputime-t

I'm sure that Octave's interpreter could be made to run a lot faster,
particularly for loops and function calls.  So far, I've concentrated
on making it produce correct results with code that is reasonably easy
to maintain.

I'm somewhat surprised by the difference in performance for the simple
linear algebra functions.  In this particular case, Octave is using
standard Fortran LAPACK and BLAS routines, and I've done nothing to
try to improve their performance.

BTW, while I was trying this, I found that Matlab would not accept

  cputime - t

or

  cputime() - t

so I was forced to write something like

  cputime-t

or

  (cputime) - t

Is this a bug?


Thanks,

jwe

-- 
In the beginning, Ken Thompson | Octave: http://www.che.wisc.edu/octave
wrote the searching tool grep. |     Me: http://www.che.wisc.edu/~jwe
       -- A. Hume, SP&E (1988) |