Parallel Genesis Available

We are please to announce that pGenesis, the parallel form of Genesis, is available for use at the NSF-supported Pittsburgh Supercomputing Center. Documentation, including examples, is here.

pGenesis is designed for modelers who have large scale applications and have reached the limit of their local computing resources. Two types of investigations are most suitable:

  • Network models - contain many cells which communicate via spikes.
  • Parameter estimation - many instances of the model are run in parallel.
  • In simple examples we have found that speedup of a factor of close to 16 on 16 processors is achievable in both cases, and we anticipate that much greater speedup is possible especially for parameter estimation. We are actively seeking real world simulations to refine our implementation and documentation of pGenesis. Our large scale machine is a Cray T3D with 512 DEC Alpha processors. Imagine if your 3 day simulation ran in 10 minutes...

    Access and Assistance

    During this experimental phase of pGenesis installation, for development and debugging of a selected set of parallel models, we can waive the usual procedures governing use of PSC supercomputers. For these models we can provide substantial assistance during the development and debugging tasks. If you have a large scale application and would like to explore this streamlined and assisted access method, please send mail to pgenesis@psc.edu. It will help if you include the following information:

  • what is the time and memory requirement for a single run and on what hardware?
  • if parameter estimation, what automated search technique are you using, if any? Do you have an estimate of how many runs are needed to converge?
  • if a network model, what is the network structure (number of cells, connectivity pattern and type - spikes, electrotonic,...). What are the axonal delays, if any, as a multiple of the timestep? Do you need to do parameter estimation on this model?
  • How would you scale this model up if unlimited computing resources were available?
  • All inquiries regarding pGenesis should be sent to pgenesis@psc.edu.