Re: Octave 3.6.0 on Windows XP plot fails.

[Martin Helm]

Am 29.02.2012 16:45, schrieb Michael Goffioul:
> Does you simple example refers to the "rand" statement? If not, could
> you try with a matrix multiplication like the following: n=2000;
> A=randn(n); B=randn(n);tic; C=A*B; t=toc, MFLOPS=2*n^3/t*1e-6 Michael. 
It refered to the a*a of a random matrix.
Here with your example (my PC, octave self compiled, atlas self
compiled, openSUSE 64 bit)

LD_PRELOAD=/usr/lib64/atlas/libsatlas.so.3 octave -qf --eval "n=2000;
A=randn(n); B=randn(n);tic; C=A*B; t=toc, MFLOPS=2*n^3/t*1e-6"
t =  1.3142
MFLOPS =  1.2174e+04

LD_PRELOAD=/usr/lib64/atlas/libtatlas.so.3 octave -qf --eval "n=2000;
A=randn(n); B=randn(n);tic; C=A*B; t=toc, MFLOPS=2*n^3/t*1e-6"
t =  0.42290
MFLOPS =  3.7834e+04

on quad core (factor 3.11)

and on the very slow atom on my netbook (octave 3.6.1 from repository,
self compiled atlas, openSUSE 64 bit)

LD_PRELOAD=/usr/lib64/atlas/libsatlas.so.3 octave -qf --eval "n=2000;
A=randn(n); B=randn(n);tic; C=A*B; t=toc, MFLOPS=2*n^3/t*1e-6"
t =  19.305
MFLOPS =  828.82

LD_PRELOAD=/usr/lib64/atlas/libtatlas.so.3 octave -qf --eval "n=2000;
A=randn(n); B=randn(n);tic; C=A*B; t=toc, MFLOPS=2*n^3/t*1e-6"
t =  8.2825
MFLOPS =  1931.8

dual core with HT (factor 2.33)

I have a dual boot with windows 7 on the netbook and will install the
newest octave tonight and compare the numbers with the same hardware for
the windows version of octave.