Brian2CUDA: Flexible and Efficient Simulation of Spiking Neural Network Models on GPUs

Abstract

Graphics processing units (GPUs) are widely available and have been used with great success to accelerate scientific computing in the last decade. These advances, however, are often not available to researchers interested in simulating spiking neural networks, but lacking the technical knowledge to write the necessary low-level code. Writing low-level code is not necessary when using the popular Brian simulator, which provides a framework to generate efficient CPU code from high-level model definitions in Python. Here, we present Brian2CUDA, an open-source software that extends the Brian simulator with a GPU backend. Our implementation generates efficient code for the numerical integration of neuronal states and for the propagation of synaptic events on GPUs, making use of their massively parallel arithmetic capabilities. We benchmark the performance improvements of our software for several model types and find that it can accelerate simulations by up to three orders of magnitude compared to Brian's CPU backend. Currently, Brian2CUDA is the only package that supports Brian's full feature set on GPUs, including arbitrary neuron and synapse models, plasticity rules, and heterogeneous delays. When comparing its performance with Brian2GeNN, another GPU-based backend for the Brian simulator with fewer features, we find that Brian2CUDA gives comparable speedups, while being typically slower for small and faster for large networks. By combining the flexibility of the Brian simulator with the simulation speed of GPUs, Brian2CUDA enables researchers to efficiently simulate spiking neural networks with minimal effort and thereby makes the advancements of GPU computing available to a larger audience of neuroscientists.

Introducing the Dendrify framework for incorporating dendrites to spiking neural networks

“…Notably, the proposed approach requires a relatively small number of free parameters, resulting in straightforward model development and calibration. Another advantage of our implementation is its compatibility with all popular operating systems running on CPUs and GPUs 72,73 . Finally, our approach allows testing new algorithms compatible with neuromorphic hardware [88][89][90] , which has seen impressive resource-saving benefits by including dendrites 91 .…”

“…6). It is important to note that since simulation performance depends on multiple factors such as model complexity, hardware specifications, and case-specific optimizations (e.g., C++ code generation 42 or GPU acceleration 72,73 ), designing a single most representative test is unrealistic. For the sake of simplicity and to replicate a real-world usage scenario, all simulations presented in this section were performed on an average laptop using standard and widely used Python tools (Supplementary Table 4).…”

A scalable implementation of the recursive least-squares algorithm for training spiking neural networks

“…Advances in GPU computing and strong interest in neuromorphic computing have led to various efficient implementations of spiking neural networks. Recent work that implements simulations of spiking neural networks in GPUs include the following: a code-generation based system that generates CUDA code for GPU (GeNN [28]) and a popular Python-based simulator for spiking neural networks, Brian2, extended for generating CUDA code directly (Brian2CUDA [29]) or through GeNN (Brian2GeNN [30]). Similarly, highly efficient CPU-based simulations of spiking neural networks can be implemented in NEST (NEural Simulation Tool [31]).…”

Introducing the Dendrify framework for incorporating dendrites to spiking neural networks

“…6). It is important to note that since simulation performance depends on multiple factors such as model complexity, hardware specifications, and case-specific optimizations (e.g., C++ code generation 42 or GPU acceleration 72,73 ), designing a single most representative test is unrealistic.…”