作者: Marianne M. Francois , Li-Ta Lo , Christopher Sewell
关键词: Coprocessor 、 Xeon Phi 、 Data structure 、 Computer science 、 Parallel computing 、 Software 、 Interface (computing) 、 Initialization 、 Algorithm 、 Reconstruction algorithm 、 Volume (computing)
摘要: With the increasing heterogeneity and on-node parallelism of high-performance computing hardware, a major challenge to computational physicists is work in close collaboration with computer scientists develop portable efficient algorithms software. The objective our implement code perform interface reconstruction using NVIDIA’s Thrust library. Interface technique commonly used volume tracking methods for simulations interfacial flows. For that, we have designed two-dimensional mesh data structure that easily mapped 1D vectors by at same time simple familiar structures terminology (such as cell, vertices edges). this new place, implemented recursive volume-of-fluid initialization algorithm standard piecewise algorithm. Our makes use table look-up identify all intersection cases, design on parallel architectures such GPUs. Finally, report performance results which show single implementation these can be compiled multiple backends (specifically, multi-core CPUs, NVIDIA GPUs, Intel Xeon Phi coprocessors), making available each.© 2014 ASME