A.P. Antonov – Ph.D.(Eng.), Associate Professor, Peter The Great St. Petersburg Polytechnic University

E-mail: alexander.antonov.ru@yandex.ru

V.S. Zaborovskij – Dr.Sc.(Eng.), Professor, Director of Institute of Computer Science and Technology,  Peter The Great St. Petersburg Polytechnic University

E-mail: vlad2tu@yandex.ru

I.O. Kiselev – Post-graduate Student, Peter The Great St. Petersburg Polytechnic University E-mail: kio.93@mail.ru

The requirements to modern supercomputer systems get tougher each year. The most significant values are: performance to power consumption rate (TFlops/Wt), effective performance, i.e. actual performance to peak performance rate. In addition, data transfer within computational unit without host computer management is required. Paper shows how reconfigurable computers allow to increase the effectiveness of supercomputer systems in order to suit those requirements. Besides, we offer and justify an idea of «artificial intelligence», that implements the procedure of profiling the executing programs and automatically reconfigures the hardware architecture to fit the exact algorithm. This approach can increase supercomputer system parameters by an order of magnitude. Paper also shows a custom-designed reconfigurable platform for high performance computing. The solution of these problems will allow creating a prototype of a self-adjusting hybrid cluster system that will become part of the instrumental cloud environment of the «Politehnicheskiy» SCC.

1. Top 500 list [Elektronnyj resurs] URL = https://www.top500.org/ (data obrashcheniya: 01.05.2018).
2. OpenCL specification 1.2. Khronos Group. 2012. 380 s.
3. Kalyaev I.A., Levin I.I., Semernikov E.A.,Shmojlov V.I. Rekonfiguriruemye mul'tikonvejernye vychislitel'nye struktury: Uchebnik dlya vuzov. Izd. 1-e. Rostov-na-Donu: YUNC RAN. 2008. 393 s.
4. Kalyaev I.A., Levin I.I., Semernikov E.A.,Shmojlov V.I. Razvitie otechestvennyh mnogokristall'nyh rekonfiguriruemyh vychislitel'nyh sistem: ot vozdushnogo k zhidkostnomu ohlazhdeniyu // Trudy SPIIRAN. 2017. № 1. S. 27−39.
5. Zaborovsky V.S., Mulukha V.A. et al. Preemptive Queuing System with Randomized Push-Out Mechanism // Communications in Nonlinear Science and Numerical Simulation. April 2015. V. 21. № 1−3. P. 147−158.
6. Rumyancev Yu.A., Zaharov P.N., Abrashitova N.A., Shmatok A.V., Ryzhih V.O., Gudimchuk N.B., Ataullahanov F.I. Primenenie PLIS dlya rascheta depolimerizacii mikrotrubochki metodom brounovskoj dinamiki // Trudy Instituta sistemnogo programmirovaniya RAN. 2016. T. 28. № 3. S. 241−266.
7. Wang Z., Zhang S., He B., Zhang W. Melia: A MapReduce framework on OpenCL-based FPGA // IEEE Transactions on Parallel and Distributed Systems. 2015. S. 27−39.
8. Kobayashi R., Oobata Y., Fujita N., Yamaguchi Y., Boku T. OpenCL-ready High Speed FPGA Network for Reconfigurable High Performance Computing // Proc. of the International Conference on High Performance Computing in Asia-Pacific Region (HPC Asia 2018). 2018. S. 192−201.
9. Zwagerman M. High Level Synthesis, a Use Case Comparison with Hardware Description Language // Grand Valley State University Masters Theses. 2015. 36 s.