Re: [Swarm-Modelling] Floating point arithmetic

[Marcus G. Daniels]

Russell Standish wrote:

> I would disagree that using Java is easier than using C++. I wouldn't
> claim the opposite either. My experience with Java has been quite
> negative - but that is probably explained by my relative inexperience
> in Java development environment vs C++ ones.
>  
I've observed the perspective of many users or potential users of 
packages like EcoLab, NetLogo, Repast, Starlogo or Swarm tends to be:  
"Coming from near zero which is easier to learn?"
That then can get equated to "easier to use", which is may be a mistake 
for some, since a package that is easy to learn can end up being a 
package that doesn't do enough to be useful for some  kinds of 
modeling.  As far as the Java vs. C++ comparison goes, I think 
empirically most people would report Java is easier to learn.  I found 
Java was very easy to learn.  Trivial even.  I too prefer C++ between 
the two.  Subjectively, Java is tiresome to me for being too 
milquetoast.  It's nothing in particular.. It does work though!

> I have had to revise my conclusions obtained from play on my laptop after 
> running my model on a faster computer for more timesteps,
> or more entities. More computational power opens up more modelled
> phenomena.
Amen.  Understanding a rugged landscape of model predictive performance 
means lots of variant runs.  Gaining intuition about some dynamics of a 
system may not, but there is a definitely a distinction between 
experimenting with a simulation and really measuring how it behaves and 
validating it against data.  To be practical, these longer runs and 
parameter sweeps typically involve a computer (or a set of computers) 
that fill a room.