Hybrid multi-threaded simulation of agent-based pandemic modeling using multiple GPUs

Abstract

Epidemiology computation models are crucial for the assessment and control of public health crises. Agent-based simulations of pandemic influenza forecast the infectious disease spreading in order to help public health policy makers during emergencies. In such emergencies, decisions are required for public health preparedness in cycles of less than a day, and the agent-based model should be adaptable and tractable for quick and simple calibration with low computational overhead. GPU accelerated computing involves the use of graphics processing units (GPUs) in combination with the CPU to perform heterogeneous computing by offloading compute-intensive portions of the program to the GPU while the remaining program runs on the CPU. In this paper, we demonstrate the utilization of the hardware environment and software tools and discuss strategies for porting agent-based simulations to multiple GPUs. We further compare the performance of simulations using two or four GPUs with the sequential execution on the CPU, in terms of time and speedup. The multi-GPU implementations exhibit great performance and support populations with up to 100 million individuals.