Parallel construction of large circular cartograms using graphics processing units

Cartograms are a cartographic approach that is based on the geometric transformation of enumeration units to enhance spatial presentation of variables of interest. The automatic construction of cartograms within digital computing environments, as recognized in the literature, is impeded by the issue of computational efficiency. Parallel computing is a potential solution for tackling the computational issue of cartogram construction, and its importance is increasingly acknowledged with the wider spread and application of advanced cyberinfrastructure into geographic studies. The purpose of this article is to investigate the use of a state-of-the-art parallel computing approach, graphics processing units (GPUs), to accelerate the construction of large circular cartograms, a special form of area cartograms. GPUs, built on many-core computing architecture and thread parallelism, provide tremendous high-performance computing and even supercomputing support for general-purpose geocomputation. In this study, the construction of a circular cartogram is divided into a large number of fine-grained subtasks that can be efficiently and simultaneously computed on the underlying many-core GPUs. Simulated and real data were used in experiments to examine the computational performance of GPU-enabled construction of circular cartograms. Experimental results suggest that the GPU-enabled parallel computing approach offers significant acceleration performance for the construction of large circular cartograms, and thus sheds further light on the resolution of computationally intensive spatial and spatiotemporal problems using advanced cartographic and geovisualization approaches.