Re: [Swarm-Support] Re: [Swarm-Modelling] xmalloc problem

[Marcus G. Daniels]

Steve Railsback wrote:

> If you are running under Cygwin in Windows, we found a help page that 
> shows how to increase the system's limit on how much memory Cygwin can 
> allocate. It seems likely that this would solve the xmalloc error, if 
> it occurs when you have not yet used up ALL your system's memory. We 
> haven't tested this trick yet.

The xmalloc things means you are either using or leaking memory like 
crazy.   Increasing the amount of memory may delay the crash (perhaps 
enough), but keep in mind that the more you allocate past your physical 
RAM, the more will be hitting disk and slowing everything down.

You can either track the memory allocates by putting a break point in 
the debugger on xmalloc, or use the Zone diagnostics to figure out what 
kind of objects are live and numerous.  These things have been asked and 
answered many times in the mailing list archives.

Usually the only way to use breakpoints for xmalloc effecitively is to 
make the conditions under which there are active delayed or 
conditional.  For example, run the model for a while (until you think it 
is likely the memory consumption is welll underway), and then interrupt 
the debugger or stop by way of break point carefully-chosen point in the 
code.  To make it easier/faster, you can even modify your code to have 
special lines that only execute and thus are easy to trap with the 
debugger (gdb).   Once trapped, begin monitoring memory allocations and 
look for obvious regularities in the origin in the backtraces.