Preview

Herald of Dagestan State Technical University. Technical Sciences

Advanced search

A new method for determining the characteristics of block algorithms

https://doi.org/10.21822/2073-6185-2025-52-2-90-97

Abstract

Objective. In the field of high-performance computing, optimization of software systems for specific computing architectures is of great importance. The aim of the study is to improve the methods for determining the parameters of block algorithms. Method. The study is based on the methodology of computer modeling and automated design in engineering and technology, including the formulation, formalization and typification of design procedures and algorithms. Result. An analytical approach to improving the efficiency of designed software systems is proposed, within the framework of which a method for determining the parameters of block algorithms is considered. The advantage of this method is the possibility of its implementation based on static analysis of the source code of the program at the initial stages of design; the obtained parameters can be used in the optimization process. Computational experiments confirming the efficiency of the proposed method were conducted. Conclusion. The proposed method allows determining the block sizes at which no cache misses will be generated at a certain level of cache memory or the number of these misses will remain at an acceptable level. The method can be used at the initial stages of designing software systems without resorting to dynamic analysis of the application; in this case, static analysis of the program source code is used.

About the Author

V. A. Egunov
Volgograd State Technical University
Russian Federation

Vitaly A. Egunov, Cand. Sci. (Eng.), Assoc. Prof., Assoc. Prof., Computers and Systems Department

28 Lenin Ave., Volgograd 400005



References

1. Lawson C.L., Hanson R.J., Kincaid D.R., Krogh F.T. Basic Linear Algebra Subprograms for Fortran Usage. ACM Transactions on Mathematical Software. 1979;5(3):308—323. DOI: 10.1145/355841.355847.

2. Dongarra J.J., Du Croz J., Hammarling S., Hanson R.J. An Extended Set of FORTRAN Basic Linear Algebra Subprograms. ACM Transactions on Mathematical Software. 1988;14(1):1-17. DOI: 10.1145/42288.42291.

3. Xianyi Z., Qian W., Yunquan Z. Model-driven level 3 BLAS performance optimization on Loongson 3A processor // 2012 IEEE 18th International Conference on Parallel and Distributed Systems (Singapore, December 17—19, 2012). Boston, Massachusetts, USA, IEEE Xplore Digital Library, 2012:684-691. DOI: 10.1109/ICPADS.2012.97

4. Intel Math Kernel Library Reference Manual. URL: https://www.bu.edu/tech/files/2010/02/mklman61.pdf (date of request: 06.02.2025).

5. Goto K., Van De Geijn R. High-Performance Implementation of the Level-3 BLAS. ACM Transactions on Mathematical Software. 2008;35(1):4:1-4:14. DOI: 10.1145/1377603.1377607.

6. Goto K., Geijn R.A. Anatomy of High-Performance Matrix Multiplication. ACM Transactions on Mathematical Software. 2008;34(158(3)):12:1—12:25. DOI: 10.1145/1356052.1356053.

7. K. Czarnecki, U. W. Eisenecker, R. Glück, D. Vandevoorde, and T. L. Veldhuizen. Generative programming and active libraries. In Selected Papers from the International Seminar on Generic Programming, pages 25–39, London, UK, 2000. Springer-Verlag.

8. Low T.M., Igual F.D., Smith T.M., Quintana-Orti E.S. Analytical Modeling Is Enough for High-Performance BLIS. ACM Transactions on Mathematical Software. 2016; 43(2):12:1—12:18. DOI: 10.1145/2925987.

9. Kravets A.G., Egunov V. The software cache optimization-based method for decreasing energy consumption of computational clusters. Energies. 2022;15(20):7509.

10. Egunov V.A. A method for improving caching strategies for shared memory computing systems / Egunov V.A., Kravets A.G.. Software Engineering. 2023;14(7): 329-338. (In Russ.)

11. Egunov V.A. A new method to increase the efficiency of vectorization of BLAS operations / Egunov V.A., Kravets A.G. Information technology. 2024;30(6): 318-328. (In Russ.)


Review

For citations:


Egunov V.A. A new method for determining the characteristics of block algorithms. Herald of Dagestan State Technical University. Technical Sciences. 2025;52(2):90-97. (In Russ.) https://doi.org/10.21822/2073-6185-2025-52-2-90-97

Views: 4


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.


ISSN 2073-6185 (Print)
ISSN 2542-095X (Online)