The use of the simulation for the configure of scalable algorithms parameters in high-performance computing

Abstract

The main objective of this paper is development of new and efficient methods of simulation of super-scale software for supercomputers with peta and exza flops level of performance. Another objective is the development of the agent-oriented system for the simulation of algorithms for supercomputers. Several classes of parallel computing algorithms, for which the simulation with the use of multicore computing systems of the Siberian Supercomputer Center for collective use SB RAS (SSCC SB RAS) was carried out were considered.