I am seeing a failure of the R random number generator using testu01 SmallCrush. This surprises me as the documentation says runif
uses Mersenne Twister. I wonder what I am doing wrong.
I run
set.seed(42)
rs <- runif(n=51320000)
write(x=rs,file="foo.bar")
and then SmallCrush
via
#include "bbattery.h"
int main(int argc, char *argv[])
{
bbattery_SmallCrushFile("foo.bar");
}
and get a failure:
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Starting SmallCrush
Version: TestU01 1.2.3
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
***********************************************************
HOST = Hogsmill.local, Darwin
ufile_CreateReadText: foo.bar
smarsa_BirthdaySpacings test:
-----------------------------------------------
N = 1, n = 5000000, r = 0, d = 1073741824, t = 2, p = 1
Number of cells = d^t = 1152921504606846976
Lambda = Poisson mean = 27.1051
----------------------------------------------------
Total expected number = N*Lambda : 27.11
Total observed number : 67744
p-value of test : eps *****
-----------------------------------------------
CPU time used : 00:00:03.04
Generator state:
10000000 numbers have been read
Aucun commentaire:
Enregistrer un commentaire