Open AccessRESEARCH OF IMPROVING THE PERFORMANCE OF EXPLICIT NUMERICAL METHODS ON THE X86 AND ARM CPU
Author(s) -
Vladislav Furgailo,
E. Elchinov,
N. Khohlov
Publication year - 2021
Publication title -
9th international conference "distributed computing and grid technologies in science and education"
Language(s) - English
Resource type - Conference proceedings
DOI - 10.54546/mlit.2021.79.83.001
Subject(s) - x86 , computer science , supercomputer , vectorization (mathematics) , parallel computing , task (project management) , computer architecture , metaprogramming , code (set theory) , program optimization , arm architecture , architecture , compiler , set (abstract data type) , software , operating system , programming language , engineering , art , systems engineering , visual arts
This paper is a continuation of the research of improving the computing performance of explicitnumerical methods on the CPU. We considered the computing possibility of such commoncomputing architectures as x86 and arm, for using optimizations on the data layer as vectorization andtiling. Other aspects of high-performance optimizations of explicit numerical methods have alsobeen explored - metaprogramming, code generation, and OpenMP technology. However, the noveltyof this research is the optimization of the arm architecture for the task of computing by the FDTDmethod and the assessment of the effectiveness of using the arm architecture for solving such a rangeof scientific problems. This paper considers a number of optimization algorithms, provides adescription of the algorithms, test calculations for various architectures.The results of the research andfurther directions of work on this topic are also presented.