[Help-gsl] siman: fixed vs variable configurations

[John Gehman]

Hi all,

I'm currently looking into using the simulated annealing routine to 
search parameter space that doesn't not appear to behave nicely using 
neither the usual newtonian nor more customized approaches. My 
objective function requires much more data than a simple array of 
variables, even though it is only the simple array of variables that 
need to be optimized. I currently have all of this data, including the 
variables array, tied together into a single struct for easy passing to 
different functions. By the time I call the siman routine, the struct 
is fully defined and therefore of a fixed size. Hence I'm inclined to 
use the fixed-size type approach to siman. I'm not sure, though, what 
to do about the size argument to the siman_solve routine.

In the examples from the manual, the array of variables is either a 
single double value (in the trivial example), or an array of integers 
(in the travelling salesman example), passed as a void * argument into 
any component functions. Hence it is easy to define the size of the 
elements in the variables structure -- sizeof double or integer.

As my void * argument references a complicated structure, and not a 
simple array of fixed-size elements, am I forced to use the "variable 
size" approach to siman, where I define copy, constructor, and 
destructor functions instead of just saying NULL, and set the size to 0 
instead of sizeof(double) ?

Cheers,
john