Distributed Computing Architecture on Epiphany MIMD Accelerators
Last few years have seen introduction of more and more advanced manycore processors. Both very well known devices like GPGPU and Intel MIC and less popular, but still very interesting, like Epiphany. There is also a growing popularity of cheap, credit-card-sized devices offering advanced features and high computational power. One of this kind of devices is the Parallella board that focuses on the parallel computing and features the Epiphany coprocessor. In this paper we propose an architecture for distributed computational systems based on Parallella and the “multiple instruction, multiple data” (MIMD) coprocessor Epiphany. This manycore processor consists of sixteen cores connected by a mesh network-on-a-chip. The presented architecture enables the usage of multiple Parallella boards in a single system with a possibility to also use other computing units. The target usage of this system are multi-agent systems (MAS) and we present selected scenarios that could be easily implemented and would benefit from the properties provided by multiple MIMD devices.
KeywordsDistributed computing Parallel computing Epiphany Multi-agent systems
The research reported in the paper was supported by the grant “Hybrid model of the early detection of internal diseases based on the paradigm of interacting particles and multi-agent system” (No. DEC-2013/09/N/ST6/01011) from the Polish National Science Centre.
- 2.Chrysos, G.: Intel Xeon Phi TM Coprocessor-the Architecture. Intel Whitepaper (2014)Google Scholar
- 3.Harding, S.L., Banzhaf, W.: Distributed genetic programming on GPUs using CUDA. In: Workshop on Parallel Architectures and Bioinspired Algorithms. Raleigh, USA. Citeseer (2009)Google Scholar
- 4.Heinecke, A., Vaidyanathan, K., Smelyanskiy, M., Kobotov, A., Dubtsov, R., Henry, G., Shet, A.G., Chrysos, G., Dubey, P.: Design and implementation of the linpack benchmark for single and multi-node systems based on Intel Xeon Phi coprocessor. In: IEEE 27th International Symposium on Parallel Distributed Processing (IPDPS), pp. 126–137 (May 2013)Google Scholar
- 5.Olofsson, A., Nordström, T., Ul-Abdin, Z.: Kickstarting High-performance Energy-efficient Manycore Architectures with Epiphany, 2–5 Nov 2014Google Scholar
- 6.Epihany SDK Reference, rev. 5.13.09.10, http://adapteva.com/docs/epiphany_sdk_ref.pdf
- 7.Ross, J.A., Richie, D.A., Park, S.J., Shires, D.R.: Parallel programming model for the epiphany many-core coprocessor using threaded MPI. In: Proceedings of the 3rd International Workshop on Many-core Embedded Systems, pp. 41–47. ACM (2015)Google Scholar
- 8.Papadogiannakis, A., Agathos, S.N., Dimakopoulos, V.V.: OpenMP: Heterogenous execution and data movements. In: Proceedings of 11th International Workshop on OpenMP, IWOMP 2015, Aachen, Germany, 1–2 Oct 2015, chap. OpenMP 4.0 Device Support in the OMPi Compiler, pp. 202–216. Springer International Publishing, Cham (2015), http://dx.doi.org/10.1007/978-3-319-24595-9_15
- 9.Pratt-Szeliga, P.C., Fawcett, J.W., Welch, R.D.: Rootbeer: seamlessly using GPUs from Java. In: 2012 IEEE 14th International Conference on High Performance Computing and Communication 2012 IEEE 9th International Conference on Embedded Software and Systems (HPCC-ICESS), pp. 375–380 (2012)Google Scholar
- 10.Byrski, A., Dreżewski, R., Siwik, L., Kisiel-Dorohinicki, M.: Evolutionary multi-agent systems. Knowl. Eng. Rev. 30(02), 171–186 (Mar 2015), http://www.journals.cambridge.org/abstract_S0269888914000289
- 11.Faber, Ł., Boryczko, K., Kisiel-Dorohinicki, M.: Hybrid architecture for simulation of blood flow with foreign bodies. In: ECMS, pp. 523–529. Brescia, Italy (2014), http://www.scs-europe.net/dlib/2014/ecms14papers/dis_ECMS2014_0138.pdf
- 12.Varghese, A., Edwards, B., Mitra, G., Rendell, A.P.: Programming the Adapteva Epiphany 64-core network-on-chip coprocessor. Int. J. High Perform. Comput. Appl. (2015)Google Scholar