A new methodology of data dependence analysis for parallelizing C++

This paper details parallelism detection in NUAPC[4], an automatically parallelizing compiler based on C++. Two kinds of parallelism, i.e. the inter- and intra-object parallelism, are introduced in NUAPC. The paper presents the method for seeking these parallelism based on the key structures Conflict Set and Method Reference Vector, and proposes a new theoretical formation for expressing and performing data dependence analysis in object-oriented program paradigm. Data dependence involving inherited data and function members are also discussed.