Search

Summary: This paper presents semi‐automatic software refactorings to introduce and tune structured parallelism in sequential Erlang code, as well as to generate code for running computations on GPUs and possibly other accelerators. Our refactorings are based on the lapedo framework for programming heterogeneous multi‐core systems in Erlang. lapedo is based on the PaRTE refactoring tool and also contains (1) a set of hybrid skeletons that target both CPU and GPU processors, (2) novel refactorings for introducing and tuning parallelism, and (3) a tool to generate the GPU offloading and scheduling code in Erlang, which is used as a component of hybrid skeletons. We demonstrate, on four realistic use‐case applications, that we are able to refactor sequential code and produce heterogeneous parallel versions that can achieve significant and scalable speedups of up to 220 over the original sequential Erlang program on a 24‐core machine with a GPU. [ABSTRACT FROM AUTHOR]

Metaheuristics are an effective and diverse class of optimization algorithms: a means of obtaining solutions of acceptable quality for otherwise intractable problems. The selection, construction, and configuration of a metaheuristic for a given problem has historically been a manually intensive process based on experience, experimentation, and reasoning by metaphor. More recently, there has been interest in automating the process of algorithm configuration. In this article, we identify shared state as an inhibitor of progress for such automation. To solve this problem, we introduce the Automated Open-Closed Principle (AOCP), which stipulates design requirements for unintrusive reuse of algorithm frameworks and automated assembly of algorithms from an extensible palette of components. We demonstrate how the AOCP enables a greater degree of automation than previously possible via an example implementation. [ABSTRACT FROM AUTHOR]

Evolutionary multi-agent systems (EMAS) play a critical role in many artificial intelligence applications that are in use today. In this paper, we present a new generic skeleton in Erlang for parallel EMAS computations. The skeleton enables us to capture a wide variety of concrete evolutionary computations that can exploit the same underlying parallel implementation. We demonstrate the use of our skeleton on two different evolutionary computing applications: (1) computing the minimum of the Rastrigin function; and (2) solving an urban traffic optimisation problem. We show that we can obtain very good speedups (up to 142.44 $$\times $$ the sequential performance) on a variety of different parallel hardware, while requiring very little parallelisation effort. [ABSTRACT FROM AUTHOR]

This paper considers how to use program shaping and algorithmic skeletons to parallelise a multi-agent system that is written in Erlang. Program shaping is the process of transforming a program to better enable the introduction of parallelism. Whilst algorithmic skeletons abstract away the low-level aspects of parallel programming that often plague traditional techniques, it is not always easy to introduce them into an arbitrary program, especially one that has not been written with parallelism in mind. Amongst other issues, data may not always be in a compatible format, function calls may need to be replicated to support alternative uses, sideeffects may need to be isolated, or there may be dependencies between functions and data that obstruct the introduction of parallelism. Program shaping can be used to transform such code to a form that allows skeletons to be more easily introduced. In this paper, we present a series of generic program shaping rewrite rules, provide their implementation as refactorings, and show how they can be used to parallelise an Evolutionary Multi-Agent System (MAS) written in Erlang. We show that we can significantly speed up this application, obtaining super-linear speedups of over 70 x the original sequential performance on a 64-core shared-memory machine. [ABSTRACT FROM AUTHOR]