[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gsl] gsl_max and nested random variables

From: Frank Reininghaus
Subject: Re: [Bug-gsl] gsl_max and nested random variables
Date: Mon, 9 Jun 2008 13:05:32 +0200


I'm attaching a reduced testcase that can be compiled. I'm getting a
negative result in the first iteration.

The reason appears to be that the macro GSL_MAX is defined like this in

#define GSL_MAX(a,b) ((a) > (b) ? (a) : (b))

The preprocessor therefore translates the line

double max = GSL_MAX (0.0, gsl_ran_gaussian (r, 2));


double max = (0.0 > gsl_ran_gaussian (r, 2) ? 0.0 : gsl_ran_gaussian (r,

Because gsl_ran_gaussian () does not always return the same value for the
same input arguments, it is possible that the first function call yields
something positive, so the ( ? : ) expression decides not to return 0.0, but
the result of a second call to the function, which might be negative.

I can't see any obvious way to fix this right now (except for replacing the
macro by a function call). Maybe one could say in the documentation that one
should use the workaround

double random = gsl_ran_gaussian (r, 2);
double max = GSL_MAX (0.0, random);

in cases like this.


2008/6/9 K Okamoto <address@hidden>:

> Hello,
> I searched gsl_max in the bug archives to no avail so I am guessing
> this is the first time? In any event I found GSL_MAX to occasionally
> give the wrong answer. Here's the code:

Attachment: testcase.c
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]