A.A. Petrov1, S.V. Chernomordov2, M.A. Kabanov3

1–3 Bunin Yelets State University (Yelets, Russia)

Problem setting. Currently, heterogeneous architectures are found in many areas of computing technology – from cluster supercomputers and high-performance servers to low-power embedded devices, including mobile phones and tablets. In this regard, the task of analyzing the prospects for the use of instrumental software for the implementation of heterogeneous computing is very relevant. To evaluate the performance of heterogeneous computing, the use of neural network modeling and machine learning methods is of particular theoretical and applied interest.

Target. Perform a comparative analysis of the heterogeneous computing instrumentation and develop an approach to performance evaluation using neural network modeling.

Results. The features of the instrumental support of heterogeneous computing are considered and the possibilities of using appropriate computing platforms for modeling controlled technical systems are analyzed. The characteristic of modern heterogeneous computing tools for solving machine learning problems is given. A neural network model has been developed to evaluate computing performance. Examples of the implementation of operations on data arrays are considered. A number of computational experiments were carried out and the interpretation of the results was given.

Practical significance. The results obtained can be used in computer modeling of controlled systems, in providing support for heterogeneous computing, as well as in various tasks related to the use of neural network algorithms and machine learning.

Petrov A.A., Chernomordov S.V., Kabanov M.A. Analysis of the instrumental support for heterogeneous computing using neural network modeling. Neurocomputers. 2023. V. 25. № 2. Р. 21-29. DOI: https://doi.org/10.18127/j19998554-202302-02 (In Russian)

1. Antonov A.S., Afanasyev I.V., Voevodin V.V. High-performance computing platforms: current status and development trends. Computational methods and programming. 2021. V. 22. № 2. P. 135–177. (In Russian).
2. Starostin N.V., Pankratova M.A. Multilevel data graph decomposition algorithms for parallel computing on a heterogeneous computing system. Issues of atomic science and technology. Series: Mathematical modeling of physical processes. 2016. № 1. P. 60–68. (In Russian).
3. Druzhinina O.V., Masina O.N., Petrov A.A. Up-to-date software and methodological support for studying models of controlled dynamic systems using artificial intelligence. Lecture Notes in Networks and Systems (LNNS). Springer. 2021. V. 1225. P. 470–483.
4. Aggarwal C. Neural Networks and Deep Learning. Springer International Publishing. 2019. DOI: 10.1007/978-3-319-94463-0.
5. Geron A. Hands-on Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. OReilly. 2019.
6. Haikin S. Neural networks: full course. М.: Williams. 2006. (In Russian).
7. Dulhare U.N., Bin Ahmad K.A., Ahmad K. Machine Learning and Big Data: Concepts, Algorithms, Tools and Applications. Wiley. 2020.
8. Makharadze G.T. MPI-model of parallel computing organization in heterogeneous clusters. Problems of development of promising micro- and nanoelectronic systems (MES). 2021. № 3. P. 106–111. (In Russian).
9. Kurapov P.A., Melik-Adamyan A.F. Optimization of analytical queries in heterogeneous systems. Modern information technologies and IT education. 2021. V. 17. № 4. P. 972–987. (In Russian).
10. Kurapov P.A., Kulikov D.V., Melik-Adamyan A.F. Cost model for optimization of analytical queries in heterogeneous systems. International Journal of Open Information Technologies. 2022. V. 10. № 4. P. 61–70. (In Russian).
11. Bystrov O., Kačeniauskas A., Pacevič R., Starikovicius V. Performance evaluation of parallel haemodynamic computations on heterogeneous clouds. Computing and Informatics. 2021. V. 39. № 4. P. 695–723.
12. Antonov A., Zaborovskij V., Kisilev I. Developing a new generation of reconfigurable heterogeneous distributed high performance computing system. Smart Innovation, Systems and Technologies. 2021. V. 220. P. 255–265.
13. Rodriguez D., Alvarez D., Rivera S., Gomez D. A review of parallel heterogeneous computing algorithms in power systems. Algorithms. 2021. V. 14. № 10. DOI: 10.3390/a14100275.
14. Lumpp F., Aldegheri S., Bombieri N., Patel H.D. Task mapping and scheduling for OpenVX applications on heterogeneous multi/many-core architectures. IEEE Transactions on Computers. 2021. V. 70. № 8. P. 1148–1159.
15. Zu Y. Deep learning parallel computing and evaluation for embedded system clustering architecture processor. Design Automation for Embedded Systems. 2020. V. 24. № 3. P. 145–159.
16. Druzhinina O.V., Korepanov E.R., Belousov V.V., Masina O.N., Petrov A.A. Experience in developing methods and tools for neural network modeling of nonlinear systems based on the domestic computing platform "Elbrus 801-PC". A non-linear world. 2020. V. 18.
    № 2. P. 5–17. DOI: 10.18127/j20700970-202002-01. (In Russian).
17. Chernomordov S.V., Druzhinina O.V., Masina O.N., Petrov A.A. Application of machine learning methods in the tasks of neural network modeling of controlled technical systems. Neurocomputers: development, application. 2022. V. 24. № 1. P. 25–35. DOI: 10.18127/j19998554-202201-03. (In Russian).
18. Masina O.N., Petrov A.A., Druzhinina O.V. Fundamentals of the methodology of scientific research in the field of modeling of complex controlled systems. Study guide. Yelets: Yelets State University named after I.A. Bunin. 2022. 86 p. (In Russian).
19. Druzhinina O.V., Masina O.N., Petrov A.A. Highly parallel learning algorithms for neural network models of technical systems. Materials of the V International Scientific and Practical Conference dedicated to the 70th anniversary of the birth of Professor Yu.N. Merenkov. Yelets. 2019. P. 37–41. (In Russian).