The ParaPhrase Project: Parallel Patterns for Adaptive Heterogeneous Multicore Systems

Abstract

Abstract. This paper describes the ParaPhrase project, a new 3-year targeted research project funded under EU Framework 7 Objective 3.4 (Computer Systems), starting in October 2011. ParaPhrase aims to follow a new approach to introducing parallelism using advanced refactoring techniques coupled with high-level parallel design patterns. The refactoring approach will use these design patterns to restructure programs defined as networks of software components into other forms that are more suited to parallel execution. The programmer will be aided by high-level cost information that will be integrated into the refactoring tools. The implementation of these patterns will then use a wellunderstood algorithmic skeleton approach to achieve good parallelism.A key ParaPhrase design goal is that parallel components are intended to match heterogeneous architectures, defined in terms of CPU/GPU combinations, for example. In order to achieve this, the ParaPhrase approach will map components at link time to the available hardware, and will then re-map them during program execution, taking account of multiple applications, changes in hardware resource availability, the desire to reduce communication costs etc. In this way, we aim to develop a new approach to programming that will be able to produce software that can adapt to dynamic changes in the system environment. Moreover, by using a strong component basis for parallelism, we can achieve potentially significant gains in terms of reducing sharing at a high level of abstraction, and so in reducing or even eliminating the costs that are usually associated with cache management, locking, and synchronisation.

A language-independent parallel refactoring framework

“…However this technique is not user-driven and can only be applied to very particular sets of problems, unlike our approach that will apply to a wide range of parallel problems. Hammond et al [5] gives a more detailed overview on refactoring and skeleton techniques for parallel programming. Finally, the application and derivation of the divide and conquer skeleton has been researched in detail, with key work by Smith [12] on applying divide and conquer to specific numerical problems.…”

Using Machine Learning Techniques to Detect Parallel Patterns of Multi-threaded Applications

“…-Selecting an optimal architecture: Parallel patterns have also been used for deciding the type of multicore architecture. For instance, in the ParaPhrase project [11,30], the authors decide the required resources in heterogeneous systems using the parallel pattern of a target application. In general, heterogeneous multicore architectures including larger and smaller CPU cores are suitable for divide and conquer and recursive data patterns.…”

“…Also, in [1], the authors use parallel design patterns to develop efficient, maintainable, and portable parallel applications. Designers can develop parallel applications using refactoring approaches based on well-understood high-level parallel design patterns as in the ParaPhrase project [30] and [32]. In the DiscoPoP project [18], different parallel patterns such as pipeline are detected to parallelize sequential programs.…”

Discovering parallel pattern candidates in Erlang

“…The EU ParaPhrase project [18] aims to produce new tools and techiques that can: i) (semi-)automatically locate suitable pattern candidates in Erlang programs; and ii) propose transformed versions of these pattern candidates that yield significant speedup on a given parallel architecture. The ParaPhrase Refactoring Tool for Erlang (PaRTE) identifies pattern candidates in Erlang programs, using static analyses to determine:…”