rand_rate

["Robert G. Brown" <rgb at phy dot duke dot edu>]

I mentioned my random number tester program, which inspired me to
actually work on it some to clean it up and post it where it can be
reached by humans other than myself.

It can be found here:

  http://www.phy.duke.edu/~rgb/General/rand_rate.php

If you unpack the tarball it should just build on most linux or unixoid
systems with a make.  It links to GSL, and lets you a) time and b) run
at least some tests of quality on all the GSL rand number routines,
outputting a "report" that includes a p-value (or set of p-values) and a
fairly arbitrary conclusion about whether or not the rng is likely to be
questionable for people who can't interpret p-values, for the test being
run.

It is still very much a project in progress -- I'm missing lots (most)
of the STS and diehard tests -- but it does provide a framework for
testing into which it should be very easy to insert new tests as I write
them, and does have some very useful bitlevel tests that (I think)
improve on what is out there in some ways.

I do plan to continue working on this and will add more tests as time
permits, because I think it is important to have a fully GPL random
number testing harness out where people can criticize it and add to it
and modify it; random numbers are very important to so many people.
Just being able to time all the routines is a moderately useful thing.

The tool also has both /dev/random and /dev/urandom wrapped up GSL
style, and includes an "add_my_types" routine that walks the available
list of types to the end and adds user defined rng's at the end by name.
This too might be useful to somebody, as you should be able to lift just
the wrapping code and add_my_types routine and add it to your own code
to extend the list of GSL rng's to include or test new ones in your own
code without having to rewrite anything if you're using GSL routines
now.

As always with an open source project, I cherish feedback from anybody
who tries it and finds it useful, or useless, or broken, or wrong, or...

    rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb@phy.duke.edu