Re: [OctDev] randn() broken

[Justus P. <Jus...@UL...>]

I have some more news:

It seems that compiling octave-forge-2004.09.09/FIXES/rand.cc (and
perhaps later versions) using gcc 3.3.x produces a dysfunctional
randn(). Older and newer compilers seem to work fine.

Is this compatible with our experience out there?

Specifically:
- My faulty setup was compiled using gcc 3.3.2 (i686).
- Simply recompiling rand.cc using gcc 3.4.1 produced a correct randn().
- Another user reported similar problems with gcc 3.3.3 and gcc 3.3.5.
- The same user reported a working version with gcc 3.2.2.


Since the thread is already more than two months old, here's a
reminder of the problem.

Output (faulty) of the following sequence of commands is available at
http://www.montefiore.ulg.ac.be/~piater/Demos/randn-dots.eps[.gz]:

octave> randn("seed", 12345);
octave> plot(1:10000, randn(1, 10000), ".");

The figure demonstrates that the samples are nowhere near Gaussian for
sample sizes of at least up to several thousands, due to a strange
sequential clustering effect.

--=20
Justus H. Piater, Ph.D.         http://www.montefiore.ulg.ac.be/~piater/
Institut Montefiore, B28        Phone: +32-4-366-2279
Universit=E9 de Li=E8ge, Belgium    Fax:   +32-4-366-2620