Performance analysis of the OP2 framework on many-core architectures

This paper presents a performance analysis and benchmark-ing study of the OP2 “active ” library, which provides an abstraction framework for the solution of parallel unstruc-tured mesh applications. OP2 aims to decouple the scientific specification of the application from its parallel implemen-tation, and thereby achieve code longevity and near-optimal performance through re-targeting the back-end to different hardware. Runtime performance results are presented for a repre-sentative unstructured mesh application written using OP2 on a variety of many-core processor systems, including the traditional X86 architectures from Intel (Xeon based on the older Penryn and current Nehalem micro-architectures) and GPU offerings from NVIDIA (GTX260, Tesla C2050). Our analysis demonstrates the contrasting performance between the use of CPU (OpenMP) and GPU (CUDA) parallel im-plementations for the solution on an industrial sized unstruc-tured mesh consisting of about 1.5 million edges. Results show the significance of choosing the correct par-tition and thread-block configuration, the factors limiting the GPU performance and insights into optimizations for improved performance