Explorando a equivalência entre uma máquina química abstrata e computação dataflow

Abstract

Currently, the search for computational performance, both in scientific computing applications and general-purpose applications, points to parallel computing as an area of interest to overcome the existing difficulties in traditionally used computational paradigms. Among the parallel computational models where programs can be developed in a natural and transparent way, Gamma and Dataflow present a surprising similarity. However, the implementation of the Gamma computational paradigm presents several challenges with regard to the resulting scheduling to adapt to the available architectures. In this work, we demonstrate for the first time, the equivalence between Gamma and Dataflow computational models, where we present the similarity and the formal equivalence proofs between these two models. Due to this equivalence, we also propose the implementation of the first conversion tool between the Dataflow and Gamma models, called GFlow. Faced with the challenges related to Gamma implementations, we also propose GSink. It consists of the first implementation of a Gamma program execution environment that allows the parallel execution of instances of several reactions. For that, we use a mechanism based on scheduling by edges reversal of an acyclic directed graph. Through experimental results, we demonstrate the correctness of both the conversions proposed by GFlow and the results of GSink execution