Re: Slow sparse matrix multiplication

[Andy Adler]

On Wed, 9 Mar 2005, John W. Eaton wrote:
> Those are excellent numbers, but I don't trust tic/toc because they
> also measure time spent working on other processes.  So can you run
>
 for d=[1e-5,3e-5,1e-4,3e-5,1e-3,2e-3,3e-3,5e-3,7e-3,1e-2];
     a=sprand(1e4,1e4,d);
     t=cputime; a*a; t1=cputime-t;
     t=cputime; a+a; t2=cputime-t;
     fprintf('%f %f %f\n',d,t1,t2);
 end

Octave (cvs):
    0.000010  0.720000 0.000000
    0.000030  0.680000 0.000000
    0.000100  0.680000 0.000000
    0.000030  0.670000 0.000000
    0.001000  0.750000 0.010000
    0.002000  0.940000 0.020000
    0.003000  1.250000 0.030000
    0.005000  2.100000 0.090000
    0.007000  3.250000 0.140000
    0.010000  4.920000 0.170000

Matlab (7.0.1):
    0.000010  0.000000 0.000000
    0.000030  0.000000 0.000000
    0.000100  0.010000 0.000000
    0.000030  0.000000 0.000000
    0.001000  0.270000 0.030000
    0.002000  1.110000 0.060000
    0.003000  2.560000 0.080000
    0.005000  6.560000 0.160000
    0.007000 12.060000 0.240000
    0.010000 20.110000 0.350000


As an aside, to get more precision I tried
    t=cputime; for i=1:10; a+a; end; t2=cputime-t;

This always gives an out of memory error in Matlab (but not octave)
  >> for i=1:10;a*a;end
??? Error using ==> mtimes
Out of memory. Type HELP MEMORY for your options.

while
  >> a*a
is ok.  No idea why.

--
Andy Adler




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