A very interesting paper about accelerating agent-based models with graphics cards (GPU). I placed the paper at http://www.friam.org/DSouza_GPUacceleratedABM.pdf
-Stephen > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > Sent: Friday, December 14, 2007 7:19 AM > To: [EMAIL PROTECTED] > Subject: Simulating millions of agents on desktops at over 50 > frames a sec > > Steve, > > I though the people in the this friam.org may be interested > in this topic. My group has developed what we believe are the > first data-parallel algorithms for agent-based model > simulation. We have implemented these algorithmson a GeForce > 8800GTX GPU. To the best of my knowledge, this prototype is > the fastest and most capable ABM simulator in the world. > > We have achieved huge increases in performance. For example, > our implementation of the SugarScape model can handle 2 > million + agents on a > 2 million + enviroment grid size with visualization at 56 > frames a second. > Our implementation of Railsback's StupidModel v16 can handle > 2 million+ bugs (predators, reproduction, movement etc) on a > 4 million + environment grid at 30 frames a sec. All these > run on a COTS desktop costing about $1,400. It is faster than > the typical ABM toolkit such as NetLogo, MASON, Repast, Swarm > etc by nearly 4 order of magnitude for large problem sizes. > I am including a paper that was presented at the Agent2007 conference. > Also here is a link to the website with a few more details. > > http://www.me.mtu.edu/~rmdsouza/ABM_GPU.html > > Regards, > Dr. Roshan M. D'Souza PhD > Asst. Professor > Dept. of ME-EM > Michigan Tech University > Houghton MI 49931 > Ph: 906-487-1001 > ============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
