Peter Welch

Professor of Parallel Computing

workTel:     +44 (0)1227 823629 faxFax:     +44 (0)1227 762811 Email: P.H.Welch@kent.ac.uk

Room S118 School of Computing University of Kent, School of Computing, University of Kent Canterbury Kent CT2 7NF England

Publications

My publications are available from the Computer Science department publications repository.

Research Interests

I belong to the following research groups:

• Programming Languages and Systems Group

Projects

Here are some of the projects I'm currently involved with:

The common theme in the above is lightweight and dynamic concurrency. I am interested in its theory, practice and education. Supporting the latter are materials for a short course on occam-pi.

Brief Bio

Peter Welch graduated in Mathematics from Warwick University (England) in 1969, taking his PhD in Computer Science from the same institution in 1975. His doctoral research was on semantic models for the lambda-calculus. Since 1972, he has been a Lecturer, Senior Lecturer and, from 1989, Professor of Parallel Computing at the University of Kent at Canterbury.

For the past 25 years, his main area of research and teaching has been in the field of concurrency and parallel computing. Applications of theory include a CSP model of Java thread synchronisation (that enables formal verification of Java multithreaded code) and CSP-based design rules for process network hierarchies (with proven safety properties such as the absence of deadlock, livelock and starvation). Long term research includes the design and development of tools supporting those rules, the design and compilation of parallel languages (occam-pi, ...), very lightweight CSP kernels (including efficient targeting of multicore) and the CSP class libraries for Java (JCSP).

He is Principal Investigator for the Kent part of the CoSMoS project (Google: "cosmos-research", 2007-2012). CoSMoS is researching patterns and frameworks for complex systems modelling and simulation, including the deliberate engineering of emergent behaviours. Typical models run to tens of millions of concurrent processes, with ever changing network topologies. Large models (needed for rich and interesting behaviours to emerge) rely on the lightweight concurrency support of occam-pi (for execution) and the CSP and pi-calculus process algebras (for formal, and informal, reasoning).