Is this after long running times creating and destroying lots of
geometry? My current theory is memory fragmentation, some interaction
between little allocations from the racket gc and larger primitive
buffer allocations.
i hope you are right. i have the feeling that memory fragmentation should
not decrease the available free memory, although free memory is
decresing during long fluxus runs as far as i remember.
we can check if memory fragmentation is the issue with collectl for
example: http://collectl.sourceforge.net/BuddyInfo.html
best,
gabor
