Performance Optimization and Statistical Analysis of Basic Immune Simulator (BIS) Using the FLAME GPU Environment

Abstract

Agent-based models (ABMs) are increasingly being used to study population dynamics in complex systems such as the human immune system. Previously, Folcik et al. developed a Basic Immune Simulator (BIS) and implemented it using the RePast ABM simulation framework. However, frameworks such as RePast are designed to execute serially on CPUs and therefore cannot efficiently handle large simulations. In this thesis, we developed a parallel implementation of immune simulator using FLAME GPU, a parallel ABM simulation framework designed to execute of Graphics Processing Units(GPUs). The parallel implementation was tested against the original RePast implementation for accuracy by running a simulation of immune response to a viral infection of generic tissue cells. Finally, a performance benchmark done against the original RePast implementation demonstrated a significant performance gain 13X for the parallel FLAME GPU implementation.