A Methodology for Opacity verification for Transactional Memory algorithms using Graph Transformation System

Abstract

With the constant research and development of Transactional Memory (TM) systems, various algorithms have been proposed, and their correctness is always an important aspect to take into account. When analyzing TM algorithms, one of the most commonly used correctness criterion is opacity, which infers that executions only observe consistent states of the shared memory. This paper proposes a formal definition to demonstrate that a given TM algorithm only generates opaque histories using a Graph Transformation System. The methodology introduced consists of translating an algorithm into production rules that manipulate the state of a graph. The proposed approach has demonstrated capability to deal with some of the complexity of TM algorithms and a case study has shown the working proof of opacity of the algorithm in question.