If I load an experiment state from a previous experiment,
whenever I hit 'Go' the simulation runs for 10 secs, pauses for
15, runs for 10, pauses for 15.... resulting in a 'steps' rate much much
lower than normal. I have world ''speed' set to every 1000, and the
'individuals' window updating every 10000.
I have tested this simply by loading frams, starting a simple experiment and
seeing it jump to 25 / 30,000 steps a second and stay that way. Then, load
an experiment, press go and get the pausing problem.
I have also loaded frams, then imediatley loaded an experiment (without just
running a basic one first), press go and get the pausing problem...
This is really causing me a headache - I now have 5 or 6 experiments that I
ran for the first billion and a half steps with results that looked
promising (definite speciation of various creatures evolved for speed over
rough&mixed terrain, with varying approaches to locomotion)
Is this a known problem?
To see details of the eperiment I am currently running, try
http://yab.yellowalienbabies.co.uk/page_frams.html
This uses a perl script to examine the latest backup file generated by
framsticks ;P Nothing too spectacular at the moment, but I'm working on
it...
Re: Annoying problem? - Low performance for big gene pools
> If I load an experiment state from a previous experiment,
> whenever I hit 'Go' the simulation runs for 10 secs, pauses for
> 15, runs for 10, pauses for 15.... resulting in a 'steps' rate much much
> lower than normal. I have world ''speed' set to every 1000, and the
> 'individuals' window updating every 10000.
The reason for low performance in such cases is
that similarity is computed, for 800 genotypes
there are 320,000 comparisons needed!
This is the 'delay' when the program 'freezes'.
Similarity must be re-computed after every
change in the gene pool.