About performance and intellectuality of supercomputer modeling

Abstract

The concept of supercomputer technologies is traditionally related to mapping algorithms onto the computer architecture, which, taking into account the explosive growth of computational capabilities, implies the necessity for an adequate increase in the performance of algorithms and programs. At the same time, it is well known that the rate of building “computer muscles” far exceeds the rate of increasing the labor productivity of software developers, which becomes a bottleneck of computer evolution. The only way to deal with this problem is to automate the construction of models, algorithms, and programs, which directly implies the revolutionary change in the level of artificial intelligence in supercomputer technologies. In this paper, it is from this standpoint that main computational stages of mathematical modeling of various processes and phenomena are discussed, some aspects of high logical complexity of modern high-performance methods for solving “large” applied problems are pointed out, and some intelligent solutions for various modeling problems are proposed.