Re: A faster sum

[Mike Miller]

On Sat, 21 May 2005, Keith Goodman wrote:

> Most of the time in your examples, I think, is spent in rand. Try moving 
> that outside of your timing.

You are quite right, Keith.  Here are the results:

GNU Octave, version 2.1.50 (i686-pc-linux-gnu).

octave:140> N=1000; X=rand(N,N); t=cputime; ones(1,N)*X; cputime-t
ans = 0.0100000
octave:141> N=2000; X=rand(N,N); t=cputime; ones(1,N)*X; cputime-t
ans = 0.040000
octave:142> N=3000; X=rand(N,N); t=cputime; ones(1,N)*X; cputime-t
ans = 0.090000
octave:143> N=4000; X=rand(N,N); t=cputime; ones(1,N)*X; cputime-t
ans = 0.16000
octave:144> N=5000; X=rand(N,N); t=cputime; ones(1,N)*X; cputime-t
ans = 0.25000

octave:135> N=1000; X=rand(N,N); t=cputime; sum(X); cputime-t
ans = 0.020000
octave:136> N=2000; X=rand(N,N); t=cputime; sum(X); cputime-t
ans = 0.080000
octave:137> N=3000; X=rand(N,N); t=cputime; sum(X); cputime-t
ans = 0.17000
octave:138> N=4000; X=rand(N,N); t=cputime; sum(X); cputime-t
ans = 0.30000
octave:139> N=5000; X=rand(N,N); t=cputime; sum(X); cputime-t
ans = 0.47000


It is clearer now that "sum" takes twice as long as "ones" and that, for 
both, CPU times are porportional to n^2.

Why the inconsistency between my results and some of the results reported 
by other people?  I think my sysadmin installed Octave from an RPM without 
a blas.  Could that be the difference?  Would blas improve "ones" much 
more than "sum"?

Anyway, so far, "ones" is a clear winner.

Mike



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