A Graph-Based Semantics Workbench for Concurrent Asynchronous Programs

Abstract

Abstract. A number of novel programming languages and libraries have been proposed that offer simpler-to-use models of concurrency than threads. It is challenging, however, to devise execution models that successfully realise their abstractions without forfeiting performance or introducing unintended behaviours. This is exemplified by Scoop-a concurrent object-oriented message-passing language-which has seen multiple semantics proposed and implemented over its evolution. We propose a "semantics workbench" with fully and semi-automatic tools for Scoop, that can be used to analyse and compare programs with respect to different execution models. We demonstrate its use in checking the consistency of semantics by applying it to a set of representative programs, and highlighting a deadlock-related discrepancy between the principal execution models of the language. Our workbench is based on a modular and parameterisable graph transformation semantics implemented in the Groove tool. We discuss how graph transformations are leveraged to atomically model intricate language abstractions, and how the visual yet algebraic nature of the model can be used to ascertain soundness.

A semantics comparison workbench for a concurrent, asynchronous, distributed programming language

“…Acknowledgements. This work extends the research reported in our FASE 2016 paper [CHP16], which was partially funded by ERC Grant CME #291389.…”

“…We discuss how the visual yet algebraic nature of our Gts models can be used to ascertain soundness, and highlight how our approach could be applied to similar concurrent, asynchronous, distributed languages. This is a revised and extended version of our FASE 2016 paper, "A Graph-Based Semantics Workbench for Concurrent Asynchronous Programs" [CHP16] (itself based upon the preliminary modelling ideas in [HPCM15]), adding the following new content: (i) a new Gts semantics covering the distributed programming abstractions of D-Scoop, formalised orthogonally to the others by extending an existing semantics and not just replacing the components of one; (ii) a presentation of the underlying, compositional metamodel to which the family of Scoop and D-Scoop semantics all conform, including a discussion of the metamodel's genericity; (iii) an expanded evaluation that additionally explores the state spaces of our benchmarks in fully distributed contexts; and (iv) a significantly revised presentation, including new details, explanations, and examples throughout the paper.…”

An Interference-Free Programming Model for Network Objects

“…We will also investigate whether performance can be improved, by (safely) relaxing the requirement that one node communicates with another via a single connection. Finally, we want to formalize the D-SCOOP semantics using [5] to test extensions, and provide a formal proof that the protocol and algorithms correctly generalize the SCOOP guarantees.…”

“…We remark that earlier versions of the SCOOP runtime additionally provided timing guarantees by not allowing processes to enqueue requests concurrently [17]. A formal comparison with the current semantics is given in [5].…”