sPyNNaker

sPyNNaker 4.0.0 is a software package that simulates PyNN-defined spiking neural networks (SNNs) on the SpiNNaker neuromorphic computing platform. It introduces a range of functionalities to ensure efficient and practical simulation of SNNs, which are crucial for advancing the field of neuromorphic computing.

Key Features and Functionalities:

- Real-time SNN Execution: It employs an event-based operating system that facilitates time-driven updates of neuron states and pipelined processing of spike events, ensuring real-time execution of SNNs.

- Comprehensive Simulation Support: The package covers all critical operations necessary for SNN simulation, including preprocessing, real-time execution, and detailed neuron/synapse model implementations.

- Performance Profiling: sPyNNaker 4.0.0 includes performance profiling capabilities, providing users with insights into the interaction between software and hardware, helping understand how the system achieves real-time execution.

- Efficient System Performance: It demonstrates system performance that closely aligns with the original design target, highlighting the efficiency of the SpiNNaker platform in handling synaptic events. Performance is shown significantly influences by SNN topology, emphasizing the importance of network design.

- Cost Model Development: A novel cost model is introduced, enabling users to estimate simulation performance based on network connectivity and SNN partitioning. This model is instrumental for planning and optimizing SNN simulations on SpiNNaker.