An applied waveform relaxation method for distributed gap junctions in the Arbor simulation library

With morphologically detailed neurons at the forefront, the Arbor simulation library provides a framework for the simulation of large-scale spiking neural networks. Simulating the temporal behaviour of a neuron’s biophysical properties is heavily dependent on the implementation of connectivity between neurons via synapses and gap junctions. While the former intrinsically induce a signal delay, the latter implement an instantaneous influence on a cell’s membrane potential, calling for state updates in every time step during a simulation. Due to high levels of paral- lelization, the zero-delay communication of such state vectors across processes with MPI would be highly inefficient. The aim of this thesis is to implement and evalu- ate Waveform Relaxation as a method to iteratively solve systems of gap-junction- coupled neurons without the requirement of continuous data exchange across processes