On Thu, 24 Dec 2009, Talha Yalta wrote:
>> > We could add a "set" option, but first I think we'd better
>> > both "crush" and "big crush" (many hours' worth) on
>> > ziggurat code and Box-Muller. �Right now we don't really have
>> > any basis for thinking that Box-Muller is "safer".
>> I think we would be safe to employ any of the two (or another one for
>> that matter) as long as we can say that the RNG in gretl passes the
>> big crush test in TestU01.
> OK, any volunteers to run "big crush" on CVS gretl?
I have experience testing various software (including gretl) using the
earlier and now irrelevant DIEHARD test suite of randomness. TESTU01
is now the standard, however, AFAIK it requires the knowledge of
programming in the C language. I am not sure how difficult it would be
to program TESTU01 to test gretl on this front but I can promise to
learn C and attempt to do this in the summer. Given how busy you are
trying to address all sorts of points we keep raising, this is the
least I can do.
As I mentioned, I ran the Crush suite on gretl without any
failures. I've now run most of Big Crush. I had to unplug my
laptop after about 14 hours, and got through 80 out of 106 tests,
again with no failures. The completed tests include all of the
"collisions" variants, on which Doornik said that standard
ziggurat failed. Obviously, though, it would be nice to run the
whole thing, which would require about 16 hours on my machine.
I'm attaching the source for the test program I used. I built the
CC = gcc -Wall -O2
CFLAGS = `pkg-config --cflags glib-2.0 gretl`
LIBS = `pkg-config --libs glib-2.0 gretl`
$(CC) $(CFLAGS) -o $@ $< -ltestu01 $(LIBS)