Efficient Transfer of C++ Objects on Intel Xeon Phi KNC in Offload Mode

by

Efficient Transfer of C++ Objects on Intel Xeon Phi KNC in Offload Mode

Vít Hanousek, Tomáš Oberhuber

Department of Mathematics, Faculty of Nuclear Sciences and Physical Engineering, Czech Technical University in Prague.

DOI:

https://doi.org/10.7494/cmms.2017.2.0594

Abstract:

Intel Xeon Phi KNC is a modern coprocessor designed for the high performance computing. In this paper we describe efficient method for transferring C++ objects in the Offloading mode. Our aim is to get consistent interface with NVidia CUDA framework in Template Numerical Library (TNL). As working example we use this library and the heat equation problem to demonstrate efficiency of implementation on Intel Xeon Phi and compare CPU with this coprocessor.

Cite as:

Hanousek, V., Oberhuber, T. (2017). Efficient Transfer of C++ Objects on Intel Xeon Phi KNC in Offload Mode. Computer Methods in Materials Science, 17(2), 94 – 100. https://doi.org/10.7494/cmms.2017.2.0594

Article (PDF):

Keywords:

Intel Xeon Phi, HPC, MIC, Offload

References:

Davis, K. D., 2013, Effective Use of the Intel Compiler’s OffloadFeatures. Accessed: 30. 9 2016, Available form: IntelDeveloper Zone: https://software.intel.com/en-us/articles/effective-use-of-the-intel-compilers-offloadfeatures

Chris J. Newburn, R. D., 2013, Offload Compiler Runtime forthe Intel® Xeon Phi™ Coprocessor. Accessed: 11. 102016, Available form: Intel Developer Zone:https://software.intel.com/sites/default/files/article/366893/offload-runtime-for-the-intelr-xeon-phitmcoprocessor.pdf

Chrysos, G., 2012, Intel® Xeon Phi™ X100 FamilyCoprocessor – the Architecture. Accessed: 11. 10 2016,Available form: Intel Developer Zone:https://software.intel.com/en-us/articles/intel-xeon-phicoprocessor-codename-knights-corner11122

Intel Corporation., 2015, Generating an Offload Report.Accessed: 11. 10 2016, Available form: Intel DeveloperZone: https://software.intel.com/en-us/node/522521

Oberhuber T., K. J. (in preparation). TNL: Framework for thefinite difference method on modern parallelarchitectures.

OpenMP., 2013, OpenMP Application Program Interface.Accessed: 12. 12 2016, Available form: OpenMP:http://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf.