Home         Authors   Papers   Year of conference   Themes   Organizations        To MES conference

Functional verification of microprocessors using machine-learning methods
Authors
	Grevtsev N.A.
Date of publication
	2022
DOI
	10.31114/2078-7707-2022-4-37-43
Abstract
	Nowadays, the major technology companies (such as ARM, Intel, IBM, and others) and universities are studying the practical applicability of machine learning techniques for microprocessor model functional verification. However, recent scientific research and experiments on machine learning are performed in the fields of formal verification, random test generation with CSP solvers, and symbolic execution and for post-silicon debugging or bug localization only. The novelty of the proposed approach comes from the usage of machine learning (ML) for user applications behavior imitation in random test generation flow. Its primary goal is to improve the quality of RTL-model verification by directed test generation. In this work, the basic concept of machine learning techniques for functional verification at the system level and base framework of the proposed methodology are presented. The key research findings are to provide the possibility of user applications behavior imitation on the executable machine code level to increase the coverage and automation of the hard-to-find bugs analysis process during the random test generation flow.
Keywords
	Functional verification, machine learning, coverage-driven test generation, executable machine code analysis, deep learning.
Library reference
	Grevtsev N.A. Functional verification of microprocessors using machine-learning methods // Problems of Perspective Micro- and Nanoelectronic Systems Development - 2022. Issue 4. P. 37-43. doi:10.31114/2078-7707-2022-4-37-43
URL of paper
	http://www.mes-conference.ru/data/year2022/pdf/D077.pdf