43-50

UDC 004.272
DOI: 10.15350/2306-2819.2018.2.43

RESEARCH ON OpenMP AND OpenCL TECHNOLOGIES PERFORMANCE

R. F. Gibadullin¹, R. Sh. Minyazev¹, A .A. Baev²
¹Kazan National Research Technical University named after A. N. Tupolev – KAI,
K. Marx Street, 10, Kazan, 420111, Russian Federation
E-mail: rfgibadullin@kai.ru; rshminyazev@kai.ru
²Volga State University of Technology,
3, Lenin Square, Yoshkar-Ola, 424000, Russian Federation
E-mail: krtmbs@volgatech.net

ABSTRACT

Introduction. The control of flows, consisting of creating, starting, merging and deleting threads, is considered the applied problem in parallel programming. There are a great number of investigations, comparing the efficiency of different means of parallel programming on multi-core processors. The general recommendation for processors is minimizing the flow time and reducing flow control time. Since modern processors usually have no more than several tens of cores, the control of flows is not a chief task in parallel applications, created for the prolonged work. OpenMP is widely used for multi-core processors on different hardware-software platforms, which require the minimum setting and allow achieving the parallelization of code sections such as loops (cycles) easily, provided parallel code sections are developed taking into account the use of threads. This is a platform-independent software solution for multi-core shared-memory systems, unlike own flow solutions, such as Linux pThreads and Windows threads. For programming on OpenCL, a large number of settings and templates are required, since it supports not only central processors, but also graphics processors. The support of graphics processors implies the need for detecting and initializing accessible GPU, sending and obtaining data from them, compiling a kernel for a specific graphics processor, kernel response, synchronization of GPU and its code sections. All these operations are applied, but the code for most of them can be encapsulated within auxiliary procedures that allow avoiding the clutter of other code sections. The purpose of the article is the research on the performance of parallel technologies OpenMP and OpenCL. The following problems are solved in the work: 1) to describe the key features of parallel programming; 2) to develop software modules for obtaining time estimates; 3) to estimate the performance of OpenMP and OpenCL technologies by testing the developed modules. Practical significance of the work is that obtained estimates of the performance of parallel technologies can be used as practical recommendations for the choice of approaches to the implementation of parallel algorithms for solving the applied problems, having the great computational complexity.

KEYWORDS

OpenMP; OpenCL; parallel programming.

FULL TEXT (pdf)

ACKNOWLEDGMENT

The work is executed at financial support of the Ministry of education and science of Russian Federation, project RFMEFI577170254 "System intraoperative navigation technology to support augmented reality-based virtual 3D models of organs obtained from the results of CT diagnostics, minimally invasive surgeries".

REFERENCES

1.   Sharma M., Soni P. Comparative study of parallel programming models to compute complex algorithm. International Journal of Computer Appli-cations. 2014. Vol. 96 (19) Pp. 9-12.
2.   Zhong W., Altun G., Tian X. et al. Parallel protein secondary structure prediction schemes using Pthread and OpenMP over hyper-threading technolo-gy. The Journal of Supercomputing. 2007. Vol. 41(1). Pp. 1-16.
3. Klimov А. V., Levchenko N. N., Okunev А. S. et al. Superkomp'yutery, ierarkhiya pamyati i potokovaya model' vychisleniy [Supercomputers, Memory Hierarchy and a Flow-Oriented Model]. Programmnye sistemy: teoriya i prilozheniya: elektron. nauchn. zhurn [Software Systems: Theory and Applications: Electronic Scientific Journal]. 2014. Vol. 5, № 1(19), Pp. 15-36, URL: http://psta.psiras.ru/read/psta2014_1_15-36.pdf (reference date: 20.11.2017) (In Russ.).
4.   Levchenko N. N., Okunev А. S., Stempkovsky А. L. Ispol'zovanie modeli vychisleniy s upravleniem potokom dannykh i realizuyushchey ee arkhitektury dlya sistem eksaflopsnogo urovnya proizvoditel'nosti [The Use of a Computing Model with Dataflow Control and Architecture, Implementing it for Systems of the Exascale Performance Level]. Vserossiyskaya nauchno-tekhnicheskaya konferentsiya «Problemy razrabotki perspektivnykh mikro- i nanoelektronnykh sistem – 2012» (MEHS-2012): sbornik trudov [Proceedings of All-Russian Scientific and Technical Conference «Problems of Development of Advanced Micro- and Nanoelectronic Systems - 2012»]. Moscow: Institute of Design Problems in Microelectronics, 2012. Pp. 459-462. (In Russ.).
5.   Gibadullin R.F., Novikov А.А., Smirnov I.N. et al. Razrabotka modulya monitoringa setevoy aktivnosti pri obrashchenii k zashchishhennoy kartograficheskoy baze dannykh [Development of a Network Activity Monitoring Module When Accessing the Protected Cartographic Database]. Vestnik Kazanskogo tekhnologicheskogo universiteta [Bulletin of Kazan Technological University]. 2016. No 8. Pp.95-99. (In Russ.).
6.   Gibadullin R.F., Novikov А.А., Smirnov I.N. and others. Upravlenie dostupom i zashchita besprovodnoy peredachi kartograficheskoy informatsii [Access Control and the Protection of Wireless Cartographic Information Transmission]. Vestnik Kazanskogo tekhnologicheskogo universiteta [Bulletin of Kazan Technological University]. 2016. No 8. Pp. 105-109. (In Russ.).
7.   Boreskov А.V., Kharlamov А.А. Osnovy raboty s tekhnologiey CUDA [Basics of Work with CUDA Technology]. Moscow: DMK-Press, 2010. 232 p. (In Russ.).
8.   Karpov Yu.G. Teoriya i tekhnologiya programmirovaniya. Osnovy postroeniya translyatorov [Programming Theory and Technology. Basics of Designing Translators]. StPb.: BHV-Petersburg, 2012. 272 p. (In Russ.).
10. Аsharina I.V. Osnovy programmirovaniya na yazykakh C i C++ [Fundamentals of Programming in C and C++]. Moscow: GLT, 2012. 208 p. (In Russ.).
11. Podval'ny S.L. Mnogoal'ternativnye sistemy: obzor i klassifikatsiya [Multi-Alternative Systems: Review and Classification]. Sistemy upravleniya i informatsionnye tekhnologii [Control Systems and Information Technologies]. 2012 Vol. 48. No 2.
Pp. 4-13. (In Russ.).
12. Podvalny S.L., Ledeneva T.M. Intellektual'nye sistemy modelirovaniya: printsipy razrabotki [Intelligent Modeling Systems: Development Principles]. Sistemy upravleniya i informatsionnye tekhnologii [Control Systems and Information Technologies]. 2013. Vol. 51. No 1. Pp.4-10. (In Russ.).

For citation: Gibadullin R. F., Minyazev R. Sh., Baev A. A. Research on OpenMP and OpenCL Technologies Performance. Vestnik of Volga State University of Technology. Ser.: Radio Engineering and Infocommunication Systems. 2018. No 2 (38). Pp. 43-50. DOI: 10.15350/2306-2819.2018.2.43