Comments for Jorge Tavares weblog

Comment on Exploring pseudo-random numbers in Lisp by mcheema1

mcheema1 — Sat, 23 Mar 2013 14:06:05 +0000

Thanks and I enjoyed this article. I am a R user and not a programmer by training but really want to learn Common lisp and use it for statistics/programming in addition to R so appreciate this starter article.

Comment on Exploring pseudo-random numbers in Lisp by Jorge Tavares

Jorge Tavares — Sat, 23 Mar 2013 13:30:46 +0000

Hi,

Yes, I did several tests and even used the code for a personal library of random number generators. The generator is fine and is the fastest/best one in a CL implementation. I must confess I never understood why the previous comment was made since no code/experiments were given to show the problem.

Comment on Exploring pseudo-random numbers in Lisp by mcheema

mcheema — Sat, 23 Mar 2013 11:53:03 +0000

Did you ever test SBCL random generator? I tried some simple tests and looks fine to me.

Comment on More random number generators in Lisp: using GSLL by Jorge Tavares

Jorge Tavares — Wed, 13 Feb 2013 21:55:53 +0000

Hi!

Yes, the same results can only be guaranteed if it’s the same platform, same SBCL version, etc. Try to make some small experiments first to see if what kind of differences you’re getting on your settings. AT least you will know more or less what to expect. SBCL is a very good implementation for numerical computing so you won’t go wrong on that :-)

Best,
Jorge

Comment on More random number generators in Lisp: using GSLL by Frederic

Frederic — Wed, 13 Feb 2013 21:05:42 +0000

Thanks for the quick reply and your insights!
I had found the sbcl extension for setting the seed state, but the prospect of using the same code on different machines and implementations, and still getting reproducable results sounded nice :)
At least the manual for sbcl 1.1.4 states that:

“The sequence of numbers produced by repeated calls to random starting with the same random state and using the same sequence of limit arguments is guaranteed to be reproducible only in the same version of SBCL on the same platform, using the same code under the same evaluator mode and compiler optimization qualities.”

So on different machines (32/64bit), the same seed might result in different outcomes. But sbcl seems to be the fastest implementation, so for now I will stick with it, and keeping note which results were computed on which machine is useful anyway…