Re: rng question

[Jari Häkkinen <jari at chiralcomp dot com>]

No, the attached program does not return from the call to 
gsl_rng_uniform_int.

The problem shows up on my powermac running MacOSX 10.4.3 with gcc 
version powerpc-apple-darwin8-gcc-4.0.0 (GCC) 4.0.0 20041026 (Apple 
Computer, Inc. build 4061).
I compile GSL from CVS (20051208 = today) with with flags -NDEBUG -O2. 
The test program is compiled with a plain 'gcc -o rnd_test rnd_test.c 
-lgsl'.

Now, I also run the test program on a SuSE based machine, and the call 
fails with an floating point exception as expected.


Jari


Brian Gough wrote:
> Jari Häkkinen writes:
>  > According to the documentation gsl_rng_max(rng_ returns the maximum 
>  > random number the underlying rng can give, and 
>  > gsl_rng_uniform_int(rng,gsl_rng_max(rng)) will return 
>  > [0,gsl_rng_max(rng)-1]. This will not give the maximum number from the 
>  > underlying generator, so I tried 
>  > gsl_rng_uniform_int(rng,gsl_rng_max(rng)+1) but this call does not 
>  > return (since the second argument becomes 0).
>
> Doesn't it give an exception? (due to division by zero in the expression scale = range / n;)
> What platform are you using?
#include <gsl/gsl_rng.h>

int main(const int argc,const char* argv[])
{
	gsl_rng_env_setup();
	gsl_rng* rng=gsl_rng_alloc(gsl_rng_default);
	unsigned long max=gsl_rng_max(rng);
	printf("%u\n",max+1);
	unsigned long number=gsl_rng_uniform_int(rng,max+1);
	printf("%u\n",number);
}