Hi Carlo,
I'm currently contemplating trying to get some simulation results from a 1,000,000 robot swarm (I will have access to a cluster this fall with some nodes with 128 cores/256GB memory, so I think it might be possible for some simple controllers!). My question is this: are there are limitations in ARGoS revolving around swarm sizes? (i.e. things that are counted per-robot each timestep, and are unsigned ints that won't roll over with reasonable swarm sizes, but with 1,000,000 robots they very well might and cause all sorts of havoc).
Also, have you ever tried building ARGoS with profiling, running some simulations, and then feeding the profile results back into ARGoS for a recompile (e.g. gcc's -fprofile-arcs and -fprofile-use)? If so, were the results noticeably faster?
Thanks!
- John