Multi-level model of parallel computing for linear algebra problems

A.V. Popov, O.V. Rudich, A.V. Chystiakov


The paper considers the use of a multilevel parallel computing model for solving linear algebra problems (systems of linear algebraic equations and an algebraic eigenvalue problem) on hybrid architecture computers and computers with multi-core processors Intel Xeon Phi x200 series. The basic principles of algorithm-software development of such computers are formulated. Examples of algorithms that use multilevel parallelism are given.

Problems in programming 2018; 2-3: 083-092


multilevel model of parallel computing; hybrid architecture computers; Intel Xeon Phi x200 processors; systems of linear algebraic equations; algebraic eigenvalue problem; parallel algorithms


TOP 500 – 2017(11).


Intel Math Kernel Library (MKL). Reference Manual/


OpenMP. V. 4.0.

A.N. Khimich, I.N. Molchanov, A V. Popov, T.V. Chistyakova, and M.F. Yakovlev, Parallel Algorithms for the Solving of Computational Mathematics Problems. [in Russian], Naukova Dumka, Kyiv (2008).

O.M. Khimich, A.Y. Baranov Hibrid Algorithm for Solving Linear Systems with Band Matrices by Direct Method. Computer Mathematics. Sb. sciences works. 2013. N 2. P. 80–87.

Popov O.V., Rudich O.V. On the Solving of Linear Systems on Hybrid-Architecture Computers. Mathematical and computer modeling. Series:

Physics and Mathematics: Sb. sciences works. 2017. N 15. P. 158–164.

I.N. Molchanov, A.V. Popov, A.N. Khimich Algorithm to Solve the Partial Eigenvalue Problem for Large Profile Matrices. Cybernetics and Systems Analysis. 1992. N 2. P. 141–147.

Khimich A.N., Popov A.V., Chistyakov A.V. Hybrid Algorithms for Solving the Algebraic Eigenvalue Problem with Sparse Matrices.

Cybernetics and Systems Analysis. 2017. N 6. P. 132–146.



  • There are currently no refbacks.