UDC 004.891
METHODS AND ALGORITHMS FOR IDENTIFYING PROGRAM FRAGMENTS FOR MAKING RECOMMENDATIONS WITH THE AIM TO INCREASE THE SPEED OF SOFTWARE SYSTEMS
A. V. Gorchakov, assistant of the Department of Corporate Information Systems, Institute of Information Technologies, RTU MIREA, Moscow, Russia;
orcid.org/0000-0003-1977-8165, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
Innovations in server architecture made it possible to create heterogeneous computing platforms for solving specialized problems. There is a need to accelerate software systems based on the capabilities provided by the heterogeneous computing platform on which the software system is deployed and run, in order to achieve the best software performance when performing specialized calculations. The aim of this research is the development of methods and algorithms for identifying program fragments during the process of intelligent static analysis in order to make recommendations for reworking a software system in order to increase its performance. The results of the research are: a new method for converting programs into vector representations based on Markov chains constructed for abstract syntax trees and definition-use graphs; algorithm for searching fragments in abstract syntax trees by a program example based on the k-nearest neighbors method and the Jensen-Shannon distance function; a technique for identifying program fragments to make recommendations for improving the performance of software systems, based on the collection of a database of example programs and options for increasing their performance using accelerators available on a heterogeneous computing platform, followed by the use of the developed algorithm for fragments search in abstract syntax trees.
Key words: : static analysis, source code analysis, classification algorithm, Markov chains, abstract syntax trees.