Andreas,
I can't investigate until next week, but my first
guess is that the problem is related to the way
gcl handles floating point numbers. What I remember
from testing the mt19937 rng is that gcl did not
seem to distinguish single and double precision floats.
Given an argument of N bits, Maxima $random grabs
ceiling(N/32) chunks (32 bits each) to create a
return value. So if arguments are different sizes,
that throws off the sequence of chunks.
However, I am surprised to see this come up
since I was under the impression that all Maxima
floats are double precision, therefore all arguments
are the same size and the sequences should be the
same (or so I thought).
If you are interested, you may wish to review my
notes which I posted at http://maxima.sf.net/wiki --
follow the link on the front page to "MT19937 rng"
or something like that. I will investigate next
week if this problem isn't resolved by then.
All the best,
Robert Dodier
__________________________________
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!
http://smallbusiness.yahoo.com/resources/