Benchmarking multi-GPU communication using the shallow water equations
Daniel Peter Playne, Ken A. Hawick · International Journal of Big Data Intelligence · 2015
The shallow water model equations provide a simple yet realistic benchmark problem in computational fluid dynamics (CFD) that can be implemented on a variety of computational platforms. Graphical processing units (GPUs) can be used to accelerate such applications on a single device using a data parallel decompositional scheme or with multiple devices using a domain decompositional approach. A shallow water equation simulation is implemented on a range of modern GPU architectures and multi-GPU systems. The typical performance of these systems using the two main device-device communication methods is reported on for single- and double-precision calculations.