Abstract
Simulations of electrical activity of networks of morphologically detailed neuron models allow for a better understanding of the brain. State-of-the-art simulations describe the dynamics of ionic currents and biochemical processes within branching topological representations of the neurons. Acceleration of such simulation is possible in the weak scaling limit by modeling neurons as indivisible computation units and increasing the computing power. Strong scaling and simulations close to biological time are difficult, yet required, for the study of synaptic plasticity and other use cases requiring simulation of neurons for long periods of time. Current methods rely on parallel Gaussian Elimination, computing triangulation and substitution of many branches simultaneously. Existing limitations are: (a) high heterogeneity of compute time per neuron leads to high computational load imbalance; and (b) difficulty in providing a computation model that fully utilizes the computing resources on distributed multi-core architectures with Single Instruction Multiple Data (SIMD) capabilities. To address these issues, we present a strategy that extracts flow-dependencies between parameters of the ODEs and the algebraic solver of individual neurons. Based on the resulting map of dependencies, we provide three techniques for memory, communication, and computation reorganization that yield a load-balanced distributed asynchronous execution. The new computation model distributes datasets and balances computational workload across a distributed memory space, exposing a tree-based parallelism of neuron topological structure, an embarrassingly parallel execution model of neuron subtrees, and a SIMD acceleration of subtree state updates. The capabilities of our methods are demonstrated on a prototype implementation developed on the core compute kernel of the NEURON scientific application, built on the HPX runtime system for the ParalleX execution model. Our implementation yields an asynchronous distributed and parallel simulation that accelerates single neuron to medium-sized neural networks. Benchmark results display better strong scaling properties, finer-grained parallelism, and lower time to solution compared to the state of the art, on a wide range of distributed multi-core compute architectures.
Highlights
Interest in the simulation of large neural network activity has been steadily increasing in recent years (Kandel et al, 2013)
We detailed the method for the numerical resolution of our problem, and showed that (1) the activity of the electrical current at the level of neuron topological trees depends on three parameters of the numerical solver that include current contributions between connecting tree sections; (2) the previous dependency allows for the grouping of connecting compartments into subtrees, where each subtree is a subset of the initial problem, and the set of substrees holds the same data representation as the initial neuron; and (3) subtrees can be grouped into a tree of subtrees—holding the previous cover and distinct set properties—and allocated to any locality on the network in order to allow for accurate distributed load balancing
Our analysis showed that a numerical resolution with full usage of compute resources on a distributed network of compute nodes is possible at three layers of parallelism: (1) at the level of compute nodes, a load balancing method delegates sections of neuron arborizations to localities at the onset of execution; (2) at the compute node level, load balancing follows analogously with the clustering of compartments into subtrees, allowing a multicore acceleration by dynamically delegating subtrees to compute cores throughout the execution; and (3) at the core level, where Single Instruction Multiple Data (SIMD)-based acceleration of state variable updates and numerical solver acceleration is possible by realigning the memory layout of each subtree
Summary
Interest in the simulation of large neural network activity has been steadily increasing in recent years (Kandel et al, 2013). Experimental advances such as high resolution recording of neurons in vivo and in vitro have supported quantitative modeling. Recent efforts from Markram et al (2015) presented for the first time a simulation of a morphologically detailed model of a part of the neocortex, simulated in the NEURON simulation environment (Hines and Carnevale, 1997). The multicompartment HH model involves the analytical resolution of stiff, coupled, continuous differential equations on each individual cell, with high variability in time and space scales (Carnevale and Rosenthal, 1992)
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 call
