Code Generation in Computational Neuroscience: A Review of Tools and Techniques

Abstract

Advances in experimental techniques and computational power allowing researchers to gather anatomical and electrophysiological data at unprecedented levels of detail have fostered the development of increasingly complex models in computational neuroscience. Large-scale, biophysically detailed cell models pose a particular set of computational challenges, and this has led to the development of a number of domain-specific simulators. At the other level of detail, the ever growing variety of point neuron models increases the implementation barrier even for those based on the relatively simple integrate-and-fire neuron model. Independently of the model complexity, all modeling methods crucially depend on an efficient and accurate transformation of mathematical model descriptions into efficiently executable code. Neuroscientists usually publish model descriptions in terms of the mathematical equations underlying them. However, actually simulating them requires they be translated into code. This can cause problems because errors may be introduced if this process is carried out by hand, and code written by neuroscientists may not be very computationally efficient. Furthermore, the translated code might be generated for different hardware platforms, operating system variants or even written in different languages and thus cannot easily be combined or even compared. Two main approaches to addressing this issues have been followed. The first is to limit users to a fixed set of optimized models, which limits flexibility. The second is to allow model definitions in a high level interpreted language, although this may limit performance. Recently, a third approach has become increasingly popular: using code generation to automatically translate high level descriptions into efficient low level code to combine the best of previous approaches. This approach also greatly enriches efforts to standardize simulator-independent model description languages. In the past few years, a number of code generation pipelines have been developed in the computational neuroscience community, which differ considerably in aim, scope and functionality. This article provides an overview of existing pipelines currently used within the community and contrasts their capabilities and the technologies and concepts behind them.

Brian 2: an intuitive and efficient neural simulator

“…Brian 2, a complete rewrite of the Brian simulator, solves the apparent dichotomy between flexibility and performance using the technique of code generation, which transparently transforms high-level user-defined models into efficient compiled code (Goodman, 2010;Stimberg et al, 2014;Blundell et al, 2018). This generated code is inserted within the flow of the simulation script, which makes it compatible with the procedural approach.…”

“…, but only in parts of the simulation. This technique is now being increasingly widely used in other simulators, seeBlundell et al (2018) for a review.In brief, from the high level abstract description of the model, we generate independent blocks of code (in C++ and other languages) that, when run in sequence, carry out the simulation. To generate this code, we make use of a combination of various techniques from symbolic mathematics and compilers that are available in third party Python libraries, as well as…”

Larger GPU-accelerated brain simulations with procedural connectivity

“…While the procedural connectivity presented in the previous section allows simulating models which would otherwise not fit into the memory of a GPU, there are additional problems when using code generation for models with a large number of neuron and synapse populations. GeNN and other SNN simulators which use code generation to generate all of their simulation code (21) (as opposed to, for example NESTML (22), which uses code generation only to generate neuron simulation code) generate seperate pieces of code for each population of neurons and synapses. This allows optimizations such as hard-coding constant parameters and, although generating code for models with many populations will result in large code size, C++ CPU code can easily be divided between multiple modules and compiled in parallel, minimizing the effects on build time.…”

Brian2GeNN: accelerating spiking neural network simulations with graphics hardware

“…www.nature.com/scientificreports www.nature.com/scientificreports/ The technique of code generation allows us to solve this apparent conflict, and has been used by both the GeNN and Brian simulators 9,10,19 as well as a number of other neural simulators 11 . In the case of GeNN, when writing a new model users need to write only a very small section of generic C++ code that defines how the variables of a neuron model are updated, and this is then inserted into a detailed template that allows that model to be simulated efficiently on a GPU.…”