SpiNNTools: The Execution Engine for the SpiNNaker Platform

TL;DR

This paper introduces SpiNNTools, a set of tools designed to efficiently map and execute computational graphs on the highly scalable SpiNNaker neuromorphic platform, simplifying the use of its millions of cores.

Contribution

The paper presents SpiNNTools, a novel software framework that automates the mapping and execution of computational graphs on the SpiNNaker architecture, addressing scalability challenges.

Findings

Successfully mapped complex neural models to SpiNNaker

Improved data loading and retrieval efficiency

Enabled broader use of SpiNNaker for large-scale neural simulations

Abstract

Distributed systems are becoming more common place, as computers typically contain multiple computation processors. The SpiNNaker architecture is such a distributed architecture, containing millions of cores connected with a unique communication network, making it one of the largest neuromorphic computing platforms in the world. Utilising these processors efficiently usually requires expert knowledge of the architecture to generate executable code. This work introduces a set of tools (SpiNNTools) that can map computational work described as a graph in to executable code that runs on this novel machine. The SpiNNaker architecture is highly scalable which in turn produces unique challenges in loading data, executing the mapped problem and the retrieval of data. In this paper we describe these challenges in detail and the solutions implemented.