 |
 |
Hi Sean, thanks for the reply once again. It sounds like my agents don't have quite as much info as those you referred to which each had a map in their brains, I just need to make a lot of them because i'm modeling 5 species of fish off of Florida, where each species abundance is about 5 million fish (i.e. about 25 million fish agents). Each fish has about 16 int parameters, 33 doubles, and a few booleans. The fish species don't interact with one another, but instead are fished by a fishing fleet (~600 agents so no big deal). As a result, I'm thinking now that I should put each of the 5 species in its own grid and run each species on a seperate processor. In this way, in each step, I'll just wait until the fish are finished being processed, and end the parallel portion of that step before I do fishing so that the vessels and fish are not competing for the same grid cell. Any thoughts on the best way to set up each species of fish (type of agent) on its own thread? This would be analogous to having 5 different types of heat bugs and running the agents from each bug type on their own thread/processor. Also, I have studied the 16 StupidModel tutorials that were implemented in MASON and noticed that they are setup kind of differently, where instead of placing the methods of each agent (i.e. StupidBug) in the step method call of that object, each method from the agent (StupidBug) was instead called in its own steppable. Is there any advantage for setting up the model this way verses the way you have set up the MASON tutorials and examples? Sorry to keep asking about this and thanks so much for your time! Steve > The double loop isn't really a big deal. But the overhead of > parallelism may or may not be worth it to you depending on the amount > of stuff you can do parallelized. > > One gotcha: we recently did a very heavily parallelized model with > many agents which each kept big maps in their heads. And it turned > out that our parallelized version was slower than a single-threaded > version when we ran it on multi-core Intel boxes (as opposed to actual > multi-CPU boxes, where it ran great). The reason: though the code was > easily parallelized, the amount of data that each parallel process had > to manipulate was larger than cache. This forced a lot of memory > contention and though Intel's chips have multiple cores they only have > one, generally crummy, memory controller, which turned out to be the > big bottleneck. Long story short, you may or may not benefit from > parallelization. You'll have to try and see. > > Do watch out for race conditions though: they're really fun to debug. > > Sean > > On May 13, 2010, at 12:21 PM, Steven Saul wrote: > >> Hi Sean, >> >> Thanks for your reply! From the reading I have recently done on >> concurrent programming, that actually makes a lot of sense and >> avoiding >> race conditions are my primary concern. It seems like I don't have >> much >> of a choice but to parallelize my own simulation because it involves >> modeling millions of individual fish that are fished by a fishing >> fleet. >> It sounds like I can process in parallel all of the things that each >> fish >> does itself without dependency on anything else i.e. (things like >> growth, >> maturity, etc.). From what you said, it also sounds like the fish >> can go >> through their movement algorithm in parallel (which tends to take some >> processing juice), store their new x and y location, then within >> that time >> step, after all fish went through the parallel sequence, they can go >> serial again at which point the fish are assigned to their new grid >> location. I wonder if this would be computationally efficient as you >> would kind of need to loop through the objects twice, right, once in >> parallel and then once serially? >> >> Thanks, >> Steve >> >>> On May 12, 2010, at 3:34 PM, Steven Saul wrote: >>> >>>> After looking at HeatBugs and the ThreadedDiffuser class, I was >>>> wondering >>>> if there was a way to use ParallelSequence to split the processing >>>> of >>>> agents (in this case the actual heat bugs) in each time step. For >>>> example, if you have 800 heatbugs in any given time step, a computer >>>> with >>>> 8 cores would process 100 bugs on each core simultaneously in that >>>> time >>>> step. >>> >>> Sort of. You just schedule a ParallelSequence on the schedule. In >>> the ParallelSequence you put, say, eight RandomSequences. Each of >>> the >>> RandomSequences >>> holds 1/8 of all your bugs. >>> >>> However you have to be careful there. Each HeatBug reads >>> information, >>> moves, and writes information. This presents opportunities for race >>> conditions. First, if one HeatBug is reading a location while >>> another >>> is writing the same location, you'll get messed up results. Second, >>> if two HeatBugs access the SparseGrid2D at the same time to move, >>> they'll break the hash table. This means that realistically you can >>> only do the read portion of these operations in parallel. So you'd >>> have the parallel sequence call step() methods which do the reads and >>> internal computation for each bug in parallel; but after that you'd >>> have to schedule the bugs serially (as before, maybe in a later >>> priority) to move themselves and then write to the heat array. >>> >>> This is advanced threading stuff for people who are really familiar >>> with the perils of threaded coding. If the above paragraph causes >>> you >>> to say "huh?", then the answer is NO, you SHOULD NOT parallelize your >>> agents. :-) >>> >>> Sean >>> >> >> >> -- >> Steven Saul, M.A. >> Graduate Assistant, Marine Biology and Fisheries >> Cooperative Institute for Marine and Atmospheric Studies >> Cooperative Unit for Fisheries Education and Research >> University of Miami - RSMAS >> 4600 Rickenbacker Cswy. >> Miami, Florida 33149 >> + 1 305-421-4831 >> http://cufer.rsmas.miami.edu > -- Steven Saul, M.A. Graduate Assistant, Marine Biology and Fisheries Cooperative Institute for Marine and Atmospheric Studies Cooperative Unit for Fisheries Education and Research University of Miami - RSMAS 4600 Rickenbacker Cswy. Miami, Florida 33149 + 1 305-421-4831 http://cufer.rsmas.miami.edu