Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.654094
Title: Scalable support for process-oriented programming
Author: Ritson , Carl G.
Awarding Body: University of Kent
Current Institution: University of Kent
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Abstract:
Process-oriented programming is a method for applying a high degree of concurrency within software design while avoiding associated pitfalls such as deadlocks and race hazards. A process-oriented computer program contains multiple distinct software processes which execute concurrently. All interaction between processes, including information exchange, occurs via explicit communication and synchronisation mechanisms. The explicit nature of interaction in process-oriented programming underpins its ability to provide manageable concurrency. These interaction mechanisms represent both a potential overhead in the execution of process-oriented software and a point of mechanical sympathy with emerging multi-core computer architectures. This thesis details engineering to reduce the overheads associated with a process-oriented style of software design and evaluate its mechanical sympathy. The first half of this thesis provides an in-depth review of facilities for concurrent programming and their support in programming languages. Common concurrent programming facilities are defined and their relationship to process-oriented design established. It contains an analysis of the significance of mechanical sympathy in programming languages, trends in hardware and software design, and relates these to process-oriented programming. The latter part of this thesis describes techniques for the compilation and execution of process-oriented software on multi-core hardware so as to achieve the maximum utilisation of parallel computing resources with the minimum overhead from process-oriented interaction mechanisms. A new runtime kernel design for the occampi programming language is presented and evaluated. This design enables efficient cache-affine work-stealing scheduling of processes on multi-core hardware using waitfree and non-blocking algorithms. This is complemented by modern compilation techniques for occam-pi program code using machine independent assembly to improve performance and portability, and methods for debugging the execution of processoriented software using a virtual machine interpreter. Through application, these methods prove the mechanical sympathy and parallel execution potential of a process-oriented software.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.654094  DOI: Not available
Share: