Refer Proceedings details

%T Virtual memory management for the transputer
%A P. J. Bakkes
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X During the last decade a number of microprocessors appeared
   on the market with operating system mechanisms designed into
   the firmware e.g. the Intel 80286 which supports segmented
   memory management and the 80386 supporting segmentation and
   paging. The idea was very attractive as protection amongst
   users could be enforced by the hardware. The support of
   multi\-users on one processor implied the conventional
   strategy of sharing the scarce processor resource amongst
   more than one user. The limitations in processing power of
   microprocessors did however limit their multi\-user
   exploitation.

%T CDL \-\- A distribution language for HELIOS
%A C. H. R. Grimsdale
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X Helios is an Operating System designed to exploit highly
   distributed systems. The central goal of such a distributed
   system, is not only to provide increased fault tolerance,
   but also increased performance. This necessitates adequate
   support for parallelism at the application level. It is
   proposed that instead of extending an existing language, or
   producing a completely new language, parallelism can be
   defined at the distribution phase. A high level Component
   Distribution Language has been designed, which allows simple
   sequential components to be combined by simple parallel
   constructors to form more complex parallel structures. The
   implementation language is irrelevant thus providing
   consistent support for standard sequential languages in a
   distributed environment.

%T A concurrent approach to the Towers of Hanoi
%A W. D. Crowe, P. E. D. Strain-Clark
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X The tower of Hanoi problem has long been known to have a
   closed (ie non\-recursive) solution. In this paper we
   analyse two approaches to this solution which involve
   concurrency. This is not an end in itself, but serves to
   introduce the main ideas and notations of CAP (Communicating
   Asynchronous Processes) \- a revision of CSP which the
   authors have used successfully to develop correct Occam
   programs. In turn, CAP is part of a wider development method
   (ODM) which is being studied with the intention of
   prototyping software tools to assist developers of
   concurrent software.

%T An OCCAM@ implementation of prolog and its preliminary performance
%A Kang Zhang
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X The paper presents an execution model for Prolog, which
   attempts to exploit the parallelism mainly at the argument
   level through the unification operation. The model,
   consisting of a number of virtual machine instructions, has
   been implemented in Occam2 on a Transputer Development
   System (TDS). The performance of the pure software
   implementation has been evaluated in real\-time. The speed,
   as assessed by running a few hand compiled benchmark
   programs on the TDS with a single transputer, ranges from
   7\-18 KLIPS. The paper gives some details of the
   performance, and then proposes a dataflow\-based
   functionally distributed configuration of the
   multitransputer system.

%T The meaning and implementation of PRI ALT in occam
%A Geoff Barrett, Michael Goldsmith, Geraint Jones, A. Kay
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X This paper describes what we believe to be an error in the
   implementation of PRI ALT on the Inmos transputer, and
   exhibits anomalous behaviour of that implementation. A
   correct implementation of comparable complexity is
   described, and some of the properties of the construct are
   described. Finally, an attempt is made to describe
   circumstances in which the behaviour of the existing
   implementation is adequate for the correctness of a program
   which uses it.

%T Simulation of gas pipeline networks
%A Minesh Patel, Paul Bentley, Clifton Hughes
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X As part of the Alvey ParSiFal project, Logica Cambridge has
   been developing a model of a significant gas distribution
   network, in order to explore the potential for improvements
   in modelling speed using large numbers of transputers.There
   are two basic approaches to continuous simulation \- the
   explicit and the implicit. The implicit approach models an
   entire system as a set of equations. Current implementations
   are well suited to conventional mono\-processor machines but
   it is difficult to parallelize the algorithm and
   consequently to fully exploit the potential of
   transputers.The explicit approach models components such as
   pipes as individual objects which can be directly mapped
   into parallel processes, making it ideally suited to the
   transputer/occam architecture. However, explicit simulation
   is inherently expensive in terms of computational resources.
   To enhance performance we are also working on improvements
   to the model.We have developed a system using an explicit
   algorithm and have simulated networks ranging from 20 to
   500+ objects. Preliminary results indicate that the explicit
   algorithm can efficiently exploit high degrees of
   parallelism.

%T An extension of the processor farm using a tree architecture
%A S. A. Green, Derek J. Paddon
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X For problems that require a small database a processor farm
   model was proposed by May and Sheppard, and has subsequently
   had much success. Here, we assume that the database is too
   large to store completely at each node of the farm and
   consequently, data must be communicated throughout the
   network during the execution of a problem. The processor
   farm model is extended to allow work tasks to be completed
   when data items not resident locally at a node can be
   obtained by making a request for data across the network.
   The farm model is further extended by configuring the system
   as a tree topology. The extended farm topology is evaluated
   by using a set of graphics benchmark problems. Benchmark
   results are given for a tree farm of upto fifteen worker
   processors.

%T A preprocessor to augment the description of occam processes for multitransputer machines
%A Hiromi Ohara, Hajime Iizuka
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X This paper describes a preprocessor which is designed to
   make the description of parallel processes of Occam easier.
   It takes a source program written in ordinary Occam language
   and a description of the configuration of the hardware on
   which it is intended to run. The programmer need not
   consider the particular hardware structure. The preprocessor
   assigns each process to an appropriate processor and
   produces the Occam program with process\-to\-processor
   allocation statements. Design philosophy, implementation
   details and some results are shown.

%T Randomised routing: "Hot potato" simulations
%A Xu Ming Qiang, Stephen J. Turner
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X In this paper, we describe experiments which simulate
   communication in a network of processors organised as a
   hypercube. A stochastic model o\[Po] communication has been
   implemented in which each processor chooses at random
   another processor in the network with which to communicate.
   Deterministic and randomized routing algorithms are compared
   with respect to both the time taken for packets to reach
   their destinations and the space required for packet queues.
   A mixed strategy is discussed and evaluated which combines
   some of the advantages of each method. Questions of fault
   tolerance are also discussed and possible solutions
   presented.

%T Mapping a process network onto a processor network
%A Francis C. M. Lau, K. M. Shea
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X This paper addresses the problem of mapping an Occam program
   onto a transputer network. This mapping is essentially a
   graph embedding (or related) problem: i.e., find a function
   f: N \-> { 1, 2.....| M |} such that certain
   performance criteria would be satisfied, where N is the set
   of processes in the Occam program (the graph) and M is the
   set of processors in the transputer network. Depending on
   the complexity of the graph as well as the target transputer
   network, many of these problems are NP\-complete [1,2]. This
   paper addresses not the mapping algorithm and how it may
   achieve optimality, which is a policy matter, but the
   mechanism of mapping \- that is, given a mapping algorithm,
   what exactly do we do to transform the program into an
   equivalent program that is ready to be downloaded onto the
   transputer network. Our answer is a precompiler which we
   have successfully implemented. We discuss several important
   problems that occurred in our construction of the mapping
   procedure, and briefly describe the actual implementation of
   the precompiler.

%T Support for occam channels via dynamic switching in multi\-transputer machines
%A Peter Jones, Alan Murta
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X The authors are motivated by the disparity between the
   arbitrarily large degree of inter\-process connectivity
   allowed in the occam model of computation (by means of occam
   channels) and the implementation of that model on the Inmos
   transputer (under the familiar four links per processor
   constraint). A communication technique has been developed
   and implemented on the ParSiFal machine which addresses this
   problem. The method involves a novel use of switching
   hardware. In normal use, links between transputers in a
   variable topology array such as the ParSiFal machine are
   arranged using switching hardware to form a static network
   before the application is loaded. This study explores the
   run\-time reconfiguration of the switch network under the
   direction of transputers executing user code, by means of a
   request\-resource / relinquish\-resource protocol.

%T The computing tower: A supercomputer for real\- time simulation of continuous systems
%A Patrick van Renterghem
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X Computer simulation of systems (e.g. magnetic fusion,
   nuclear power plants, weather forecasting, ...) will have a
   strong impact on theoretical and experimental research in
   the future, because experiments on real systems are
   impossible due to the danger, the cost or the duration of
   the experiment. Computer simulation of a system is therefore
   an attractive alternative. Generally, large and complicated
   systems are simulated, so this requires an enormous amount
   of computing power. In this paper, we describe the
   possibility to use a home\-built multitransputersystem, The
   Computing Tower, for system simulation and methods to
   implement a block\-structured simulation language on this
   system the most efficient way.

%T The application of transputers and occam to an industrial energy management system
%A Andy Sinclair, Paul Kelly
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X The work described here is part of a larger project to
   provide a knowledge\-based decision support system to offer
   advice on the control and utilisation of energy within a
   large integrated steel works. It is concerned with the
   electricity power generation subsystem of the steelworks.
   The aim of this work is to produce a decision support system
   that will give advice on the steam distribution required to
   maximise the power output of three turbines when viewed as a
   system. This paper is concerned with the practical details
   of implementing such a system in an industrial environment

%T Fast prototyping of architectural designs using transputers
%A David B. Skillicorn
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X The existence of a processor with on\-board communication
   facilities, such as the transputer, makes it possible to
   prototype new hardware designs by emulating their functional
   units on networks of transputers. This gives better
   information about performance and appearance than simulation
   and can be considerably easier to build. It also permits
   software to be built and executed with performance that
   approximates the final system\-hence, software prototyping
   is also possible. We describe our experiences with two
   different systems: an object\-based vision system and an
   implementation of the functional language Lucid.

%T GECKO: A graphical tool for the modelling and manipulation of occam software and transputer hardware topologies
%A Marc Stephenson, Olivier Boudillet
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X This paper outlines some of the work undertaken at the
   Polytechnic of Central London (PCL), within the framework of
   the Alvey supported Parsifal project. The aim of this work
   is to help the user with the problems of: conceptualising,
   designing, fine tunning and configuring occam applications.

%T A state\-of\-the\-art radar pulse deiterleaver
%A S. P. Turner, Rick D. Beton, C. Upstill
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X This paper discusses the conversion and development of a
   serial implementation of a state\-of\-the\-art radar
   deinterleaving algorithm into a full\-scale parallel
   implementation using occam2 and the transputer.

%T Techniques for rendering solid objects on a processor farm
%A Peter M. Dew, Nick Holliman, David Morris, Alan de Pennington
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X Collaborative research by Leeds University and the IBM UK
   Scientific Centre has resulted in an experimental parallel
   solid modeller called MISTRAL, based on the Constructive
   Solid Geometry (CSG) representation of solid objects. This
   paper gives an overview of the CSG rendering algorithms
   implemented in MISTRAL, a description of the MISTRAL
   modeller itself, and describes how several different kinds
   of parallelism may be exploited using a processor farm. It
   concludes by discussing some of the issues involved in
   writing complex parallel programs in OCCAM2.

%T A prototype simulator output movie system based on parallel processing technology
%A N. Carmichael, D. Hewson, J. van der Vorst
%E Charlie Askew
%B OUG\-9: Occam and the Transputer \-\- Research and Applications
%X New parallel processing technologies potentially offer
   immense computing power at a very reasonable price. However
   it is not yet clear how best this potential may be realised
   for the large class of applications of concern to the Shell
   computing community. Shell U.K. (Expro) Ltd., as one of that
   company\[rs]s ninth licensing round initiatives, decided to
   purchase a parallel processing machine (a member of the
   "computing surface" range from Meiko
   Ltd.,a Bristol ,U.K., based company). In order to test the
   abilities of the machine and its suppliers, dynamic
   "movie style" display of oil or gas
   reservoir simulator output (ie. changing fluid distributions
   within the reservoir) was chosen as an application example.
   The presentation will give a brief description of the
   application and of how it is placed on the machine. These
   devices are now being routinely used by reservoir
   engineers.Operational usage of the system will be discussed
   (a video of an interactive session will be available) and
   some more general remarks about the placement of such
   technology within an extensive heterogeneous computing
   environment will be made.

If you have any comments on this database, including inaccuracies, requests to remove or add information, or suggestions for improvement, the WoTUG web team are happy to hear of them. We will do our best to resolve problems to everyone's satisfaction.

Copyright for the papers presented in this database normally resides with the authors; please contact them directly for more information. Addresses are normally presented in the full paper.