Re: wierd random number behavior

[Benedikt Stefansson]

Sirs,

Before further research into compilers and processor types, I was
wondering if Paul could fix some potential syntax errors in the
offending method which started this thread. Since Paul didn't include
the full source code I can't replicate the situation, would need to know
what type and value the instance variables "urgency","load" and
"moisture" take.

Note the exp() function call here:

    threshold = urgency * load * exp(-moisture / 4);

On Linux I've had problems when dividing a floating point value by an
integer (i.e using 4 instead of 4.0). The "-moisture" is not proper C
and  should be written "-1.0*moisture". I remember programs behaving
strangely when I forgot to make this adjustment. These two factors taken
together could very probably account for the large values returned by
the print statements for threshold, although it is harder to explain the
overrun for randNum. However it could be a strange byproduct of the
variable threshold going to infinity. 

I've copied some of Paul's code and output below so you can figure out
what I'm talking about:

- makeDecisionAtUrgency: (float) urgency {

double randNum;
double threshold;

threshold = urgency * load * exp(-moisture / 4);
randNum = [uniformDblRand getDoubleWithMin: 0.0 withMax: 1.0];

if (randNum < threshold)
[self setBurning: 1];
else
[self setBurning: 0];

printf ("\t\t fuel compared %f and %f \n", randNum, threshold);

return self;
}

The printf statement was added as a way to check why things were going 
wrong when this ran on my laptop. At this part of the simulation, the 
printf statement gave the following output:

fuel compared 0.353868 and -27255296557.392986 
fuel compared 0.313851 and 9945314261.998653 
fuel compared 0.007383 and 18195143603.152775 
fuel compared 0.236007 and -12823928380.525589 
fuel compared 0.304299 and 4679383970.345970 
fuel compared 0.628275 and 8561022917.100419 
fuel compared NaN and -28163390113.579769 
fuel compared NaN and 10276672821.115992 
fuel compared NaN and 18801370456.165474 
fuel compared NaN and -8273947266.455919 
fuel compared NaN and 3019119809.569216 
fuel compared NaN and 5523537722.697916 


Regards,
Benedikt
-- 
---
Present coordinates: 
Dep. of Economics, Univ. of Trento, Via Inama 1, 38100 Trento, ITALY
Office: +39 0461 882246 Mobile: +39 0347 0415721 Fax: +39 0461 882222

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.