OK lovely, I have a little demo of my nascent agent modeling framework thinger that can run 100,000 agents doing non-trivial (but faked for now) computations at about 1/3 of my screen refresh rate, meaning near real-time. I haven’t tried optimizing it yet, just tinkering so far. That’s pretty promising.
#mzw5ema
(#mzw5ema) for instance, I’m just firing all the agents’ workloads asynchronously at the CPU and hoping for the best, where it’d probably be more efficient to batch up the work. I’m using scala and haven’t done any jvm heap of GC tuning yet, so that’s another way to improve performance.
#fww2yra
(#mzw5ema) @prologic@twtxt.net a little experimental framework for helping to build agent-based models. I’m unhappy with the existing frameworks I know about so I’m trying a different tack and so far I’m pretty happy with the results.
#x25a74a