Информационные технологии интеллектуальной поддержки принятия решений, Информационные технологии интеллектуальной поддержки принятия решений 2016

Размер шрифта: 
Software for exact computations in heterogeneous computing environment and its applications
A. V. Panyukov, V. A. Golodov

Изменена: 2019-12-12

Аннотация


The paper introduces usage of exact rationalcomputations in heterogeneous computerenvironments, cloud computing is considered asone of the area that intensively uses differenthardware combined in one cluster complex.Rational computations avoid the overflow androunding mistakes of IEEE754 floating pointnumbers. Exact rational computations in theheterogeneous environments are provided by libraryof classes “Exact Computation 2.0”, the libraryprovides overlong and rational data types on wideamount of modern architectures of centralprocessors and coprocessors. The library providesMPI communications out-of-the-box. Someexamples of software based on developed librarydescribed below.The library “Exact Computation 2.0” is the basis forimplementation of linear equation systems solverwith extended accuracy, developed data types andused calculation methods provide effective solverwith polynomial computational and spacecomplexity. Another example considered in thepaper is linear programming task solver withabsolute accuracy it also has polynomialcomputational and space complexity. Newapproach to linear systems of equation with intervaluncertainty of the coefficients named intervalregularization substantially uses exact computationability.Further research will be designed to improvingsoftware efficiency with the parallel algorithms ofbasic arithmetic operations and cooperation withand/or adoption to existing industrial software.

Литература


1. The GNU MP Bignum Library. March 24 2014. (http://gmplib.org/)

2. Intel® Xeon Phi™ Product Family. October 20 2015. (http://www.intel.com/content/www/us/en/processors/xeon/xeon-phi-detail.html)

3. High Performance Computing (HPC) and Supercomputing | NVIDIA Tesla; October 20 2015. (http://www.nvidia.com/object/tesla-supercomputingsolutions.html )

4. Boreskov A.V., Kharlamov A.A., et all Parallel'nye vychisleniya na GPU. Arkhitektura i programmnaya model' CUDA. [Parallel Computing on GPU. Architecture and CUDA Programming Model]. Moskva, Izdatel'stvo moskovskogo universiteta [Moscow, Lomonosov State University Publishing], 2012. P. 336. (in Russian).

5. Parallel Programming and Computing Platform | CUDA | NVIDIA; October 20 2015. (http://www.nvidia.com/object/cuda_home_new.html)

6. Liniov A.V., Bogolepov D.K., Bastrakov S.I. Tekhnologii parallel'nogo programmirovaniya dlya protsessorov novykh arkhitektur. [Parallel programming for new prosessor architectures]. Izdatel'stvo moskovskogo universiteta [Lomonosov State University Publishing], 2010. P. 160. (in Russian).

7. Golodov V.A., Panyukov A.V. Library of Classes Exact Computation 2.0 Programs, Data Bases and Topologies of VLIS. Official bulletin of Russian Agensy of Patients and Trademarks, Moscow, FIPS, (3) 2013. P. 251. (in Russian).

8. Monakov A., Platonov V. Optimizatsiya metoda resheniya lineinykh sistem uravnenii v OpenFOAM dlya platformy MPI + CUDA [Optimizations for linear solvers in OpenFOAM for MPI + CUDA platform]. Trudy ISP RAN [The Proceedings of ISP RAS], 2014, vol. 26, no. 3, pp. 91–102 (in Russian). doi: 10.15514/ISPRAS-2014-26(3)-4

9. Golodov V.A., Panyukov A.V Programma ExactLinPSolutor 1.0 Svidetel'stvo o gosudarstvennoi registratsii programmy dlya EVM 2014610445 ot 09 yanvarya.[ Program ExactLinPSolutor 1.0] Programmy dlya EVM, bazy dannykh, topologii integral'nykh mikroskhem. Ofitsial'nyi byulleten' Rossiiskogo agentstva po patentam i tovarnym znakam [Programs, Data Bases and Topologies of VLIS. Official bulletin of Russian Agensy of Patients and Trademarks]. M.: FIPS, (2), 2014. (in Russian).

10. Golodov V.A. Interval'nyi podkhod k regulyarizatsii netochno zadannykh sistem lineinykh uravnenii Dissertatsiya na soiskanie uchenoi stepeni kandidata fiziko-matematicheskikh nauk [Interval regularization approach to uncertain linear equations systems. PhD thesis for the degree of candidate of physical and mathematical sciences] VTS RAN im. A. A. Dorodnitsyna,[Institution of Russian Academy of Sciences Dorodnicyn Computing Centre of RAS], 2014. P 107. (in Russian).

11. Panyukov A.V., Golodov V.A. Computing Best Possible Pseudo-Solutions to Interval Linear Systems of Equations. Reliable Computing, 2013, 19(2). P. 215–228.

12. Panyukov A.V., Golodov V.A. Programmnaya realizatsii algoritma resheniya sistemy lineinykh algebraicheskikh uravnenii s interval'noi neopredelennost'yu v iskhodnykh dannykh [Software Engineering for Algorithm of Solving a Linear Equation Set Under Interval Uncertainty]. Upravlenie bol'shimi sistemami [Large-scale Systems Control], 2013, (43). P. 78–94. (in Russian).

13. Golodov V.A. Interval regularization approach to the Firodt method of the spectroscopic analysis of the nonseparated mixtures. In: 16th GAMM–IMACS International Symposium on Scientific Computing, Computer Arithmetic and Verified Numerical Computations(SCAN2014). Book of Abstracts. Department of Computer Science, Uniersity of Wurzburg, Germany, 2014. p. 58–59.

14. Golodov V.A., Panyukov A.V. Programma ExactISLAYSolutor 1.0 Svidetel'stvo o gosudarstvennoi registratsii programmy dlya EVM 2014610333 ot 09 yanvarya. [ExactISLAYSolutor 1.0 program] Programmy dlya EVM, bazy dannykh, topologii integral'nykh mikroskhem. Ofitsial'nyi byulleten' Rossiiskogo agentstva po patentam i tovarnym znakam. [Programs, Data Bases and Topologies of VLIS. Official bulletin of Russian Agensy of Patients and Trademarks] M.: FIPS, (2), 2014. (in Russian).

15. Shifrovanie GOST 28147-89 na x86- i GPUprotsessorakh [Encryption using algorithm GOST 28147-89 on x86 and GPU processors], April 16 2014. (http://securitycorp.org/infosecurity/crypto/26518-shifrovanie-gost-28147-89-na-h86-i-gpu-processorah.html ). (in Russian).

16. Panyukov A.V. Scalability of Algorithms for Arithmetic Operations in Radix Notation, Reliable Computing, 2015, 19(4). P. 417-434.

17. Golodov V.A., Panyukov A.V.. Masshtabiruemye algoritmy tselochislennoi arifmetiki i organizatsiya podderzhki ratsional'nykh vychislenii v geterogennykh sredakh. [Scalable Algorithms for The Integer Arithmetic and Rational Calculations in Heterogeneous Computation Environment] Vestnik Yuzhno Ural'skogo gosudarstvennogo universiteta Seriya: Vychislitel'naya matematika i informatika [South Ural State University Bulletin Series Computational Mathematics and Software Engineering], 2015, 4(2). pp. 71–88. (in Russian).

18. Panyukov A.V., Golodov V.A. Parallel Algorithms of Integer Arithmetic in Radix Notations for Heterogeneous Computation Systems with Massive Parallelism. South Ural State University Bulletin Series «Mathematical Modelling, Programming & Computer Software». 2015, 8(2). P. 117–126.