Abstract
State-of-the-art software tools for neuronal network simulations scale to the largest computing systems available today and enable investigations of large-scale networks of up to 10 % of the human cortex at a resolution of individual neurons and synapses. Due to an upper limit on the number of incoming connections of a single neuron, network connectivity becomes extremely sparse at this scale. To manage computational costs, simulation software ultimately targeting the brain scale needs to fully exploit this sparsity. Here we present a two-tier connection infrastructure and a framework for directed communication among compute nodes accounting for the sparsity of brain-scale networks. We demonstrate the feasibility of this approach by implementing the technology in the NEST simulation code and we investigate its performance in different scaling scenarios of typical network simulations. Our results show that the new data structures and communication scheme prepare the simulation kernel for post-petascale high-performance computing facilities without sacrificing performance in smaller systems.
Highlights
Modern neuroscience has established numerical simulation as a third pillar supporting the investigation of the dynamics and function of neuronal networks, next to experimental and theoretical approaches
It was previously suggested that synapse objects should be stored on the compute node on which their target neuron resides (Morrison et al, 2005). This choice reduces the amount of information to be communicated during simulation of the network, while on the other hand it allows for completely parallel network construction, a practical necessity for large-scale simulations
The previous kernel of NEST uses a source-based addressevent-representation (AER) scheme (Boahen, 2000; Lansner and Diesmann, 2012), employing MPI_Allgather to communicate spike events among compute nodes: Each MPI process receives the global ids (GIDs) of all neurons that emitted a spike since the last MPI communication took place, and each thread needs to determine the thread-local neurons to which it needs to deliver the events
Summary
Modern neuroscience has established numerical simulation as a third pillar supporting the investigation of the dynamics and function of neuronal networks, next to experimental and theoretical approaches. Most neuronal network simulation software, is based on the hypothesis that the main processes of brain function can be captured at the level of individual nerve cells and their interactions through electrical pulses Scalable Neuronal Network Simulation Code show little variation in shape, it is generally believed that they convey information only through their timing or rate of occurrence Simulators in this area that follow a general-purpose approach employ simplified models of neurons and synapses with individually configurable parameters and connectivity (Carnevale and Hines, 2006; Bower and Beeman, 2007; Gewaltig and Diesmann, 2007; Bekolay et al, 2013; Goodman and Brette, 2013).
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
