Modular and non-invasive distributed memory parallelization

Abstract

This paper presents an aspect-oriented library to support parallelization of Java applications for distributed memory environments, using a message-passing approach. The library was implemented using AspectJ language, and aims to provide a set of mechanisms to make easier to parallelize applications, as well as to solve well known problems of parallelization, such as lack of modularity and reusability. We compare the advantages of this method over the traditional approach, and we discuss differences to recent approaches that address the same problem. Results show benefits over other approaches, and, in most of cases, a competitive performance.

Gaspar: a compositional aspect‐oriented approach for cluster applications

“…It has been previously shown that AOP can properly modularise concurrency and parallelism concerns . It also has been shown the feasibility of AOP to develop libraries that encapsulate common parallel programming abstractions . Two key properties of AOP are quantification and obliviousness .…”

“…The Gaspar framework includes a set of parallelism modules that provide OpenMP and MPI‐like programming constructs. Those modules are currently implemented in AspectJ and result from the redesign of two aspect libraries previously developed by the authors , with improved compositional properties and supporting the Gaspar data API. These modules use Java concurrency mechanisms (e.g.…”

“…Aspect‐oriented techniques were used previously to implement GPU‐specific execution concerns and to implement mappings of the base programs into Java threads and MPI . The Gaspar framework introduces aspect modules that implement mappings for GPU on top of Aparapi and Rootbeer and provides a unified approach where the same base program can be mapped into these platforms.…”