Cog Ex Machina (Cog), a project lead by Greg Snider at HP Labs, is a software framework the lab uses both as a simulation tool and as a medium for enforcing consistency between modeling and hardware work. As a simulation tool, Cog allows modelers to build their initial prototypes at large scale with minimal difficulty. With conventional tools, large-scale simulation requires too much effort to allow rapid exploration.
Beyond the simplification of large-scale simulation, Cog is a a key product of the joint HP-BU research. Cog captures a year's worth of work towards finding the right abstractions to match modeling needs to hardware capabilities. Both HP and BU have contributed a great deal towards the design assumptions built into the system, so the abstractions are a good fit both for the models we need to build and the hardware we can fabricate. This common set of abstractions is the only way to ensure that the hardware and modeling research are on track to converge to a single physical artifact capable of intelligent behavior.
Large scale simulations, such as the one typical of MoNETA models, will leverage high performance computing resources, such as the new GPU cluster hosted at HP Labs under the direction of Greg Snider and Dick Carter. The cluster, called Simcity, features a total of 144 GPUs, 576 GB of conventional memory, 432 GB of GPU memory, and an Infiniband interconnect. A prototype cluster containing three nodes and six GPUs, called Simtown, is also available for testing and debugging. By the end of 2011, the Neuromorphics Lab will finish building out a twin system with roughly half the computing power of Simcity.
IEEE Computer cover page featured article on Cog
Snider G., Amerson R., Carter D., Abdalla H., Qureshi S., Leveille J., Versace M., Ames H., Patrick S., Chandler B., Gorchetchnikov A., and Mingolla E. (2011) Adaptive Computation with Memristive Memory. IEEE Computer 44(2), 21-28. PDF