WoTUG - The place for concurrent processes

BibTeX Proceedings details

BibTex html text dump of PaperDB database.

PaperDB is GPL Software, see: http://paperdb.sourceforge.net.

  title = "{T}he advancements of transputers and occam",
  author= "Edwards, Janet and Lawson, Philip",
  editor= "Edwards, Janet",
  pages = "1--12",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "In our eagerness to promote Occam and Transputers on the
     World stage, we must not forget that the now common place
     jargon of the Occam User Group may be very new to those
     practitioners that are entering this realm for the first
     time. For the benefit of students, old and young alike, who
     have missed out on the hype that surrounded the transputer
     in the 1980s, what follows should provide a brief insight
     into this complex field. Beyond introducing the current
     transputer hardware and Occam software, a glimpse of things
     to come is provided by a pre-release view of the IMS T9000
     transputer and the next generation of the Occam language.
     This overview concludes with a short pr\`{e}cis of each of
     the papers that contributed to this, the 14th World Occam
     and Transputer User Group technical meeting."
  title = "{A}spects of database machine design using the {H}1, {C}104 and {O}ccam91",
  author= "Kerridge, Jon and Oates, Richard J.",
  editor= "Edwards, Janet",
  pages = "13--27",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "In this paper we explore the design of a database machine
     that is already being implemented using T series transputers
     to determine the benefits of using the T9000 series
     transputer togeher eith its associated routing chip, the
     C104, and occam91. The language occam91, a superset of
     occam2, contains language features that directly support the
     exploitation of hardware features that have been
     incorporated into the T9000 transputer. We briefly introduce
     the underlying design of the database machine. The way in
     which the C104 routing chips can be used is then described
     in conjunction with a planar approach to the interconnection
     of nodes in the machine. Finally, we give fragments of
     occam91 that demonstrated how its features can be used in
     two different parts of the database machine. The first
     application considers the problems of allocating resources
     to a user query and the second looks at how recovery
     features can be specified."
  title = "{P}erformance of post-game analysis on transputers",
  author= "Sunter, Johan P. E. and Bakkers, Andr\`{e} W. P.",
  editor= "Edwards, Janet",
  pages = "28--38",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "In this paper the performance of a Post-Game Analysis system
     is studied. For this purpose several simple cases have been
     designed. These cases consist of a number of processes which
     have to be distributed over a network of transputers. The
     final distributions for these cases are compared to the
     optimal ones, which can be easily derived for these simple
     cases. Performance measures considered include the number of
     iterations required to reach the final distribution, and the
     overhead caused by monitoring the program behaviour."
  title = "{S}trategies for workload distribution",
  author= "Phillips, Iain and Capon, Peter C.",
  editor= "Edwards, Janet",
  pages = "39--51",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "If large distributed systems, such as transputer networks,
     are to be exploited fully, effective workload distribution
     strategies are required. For such systems to be successfully
     scalable, decision making in such strategies will need to be
     both dynamic and distributed. The choice of strategy depends
     on many factors, including the size and topology of the
     network together with the nature of the application. The
     success of any particular strategy is measured by comparing
     the time taken to execute the application on many processors
     with the time taken on few processors."
  title = "{A} process migration harness for dynamic load balancing",
  author= "Baker, S. A. and Milner, K. R.",
  editor= "Edwards, Janet",
  pages = "52--61",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "In order to attain high efficiency and maximum throughput
     from MIMD architectures, it is necessary to have a balanced
     load with all processors being utilised. If the optimum
     program configuration is data-dependent, the process
     distribution will have to be performed at run time, i.e.
     dynamically. A message-passing harness has been developed
     with a process migration capability for dynamic
     loca-balancing on transputer-based machines. This paper
     describes the process migration harness and the way in which
     it is integrated with the user's application, gives a brief
     survey of the types of dynamic load balancing algorithm
     currently under investigation and concludes with some
     preliminary performance results."
  title = "{A} simple parallel algebraic multigrid",
  author= "Robinson, Guy",
  editor= "Edwards, Janet",
  pages = "62--75",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "This paper describes an \"algebraic multigrid\"
     scheme which can be applied to a wide range of matrix based
     problems. Multigrid schemed offer significant gains in both
     numerical performance and runtimes compared to conventional
     solvers. The equations for the hierachy of grids are
     generated solely from the equation for the fine mesh without
     generating the intermediate grids or relying on geometrical
     features of the fine mesh. The development of the code for
     distributed memory Multiple Instruction Multiple Data
     architectures is detailed. The numerical and run time
     performance is described for simple linear equation sets and
     as a linear solver for coupled equations as part a 3D
     computational fluid dynamics code."
  title = "{F}ast fourier transform on transputers",
  author= "Khan, Aman and Stephens, Nelson",
  editor= "Edwards, Janet",
  pages = "76--84",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "The fast evaluation of the Discrete Fourier Transform on a
     system with a large number of transputers is considered. The
     implementation uses a configuration of the transputers which
     leads to a very high performance. The design and
     implementation incorporate several new features. Tables of
     this performance, in terms of actual times and speed-up as
     the number of data points and transputers vary, are
     presented. Implementation aspects of the radix-2 and higher
     radix DFT in one and two dimensions are considered. The
     timings are compared with those for other computers and
     found to be favourable."
  title = "{A} new adaptive algorithm for the solution of systems of linear equations",
  author= "Cunha, Rudnei Dias da and Hopkins, Tim",
  editor= "Edwards, Janet",
  pages = "85--99",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "We present a comparison between serial and parallel
     implementations of some iterative methods to solve systems
     of linear equations. The basic vector arithmetic operations
     used in the implementations are discussed with respect to
     its parallelization. The iterative methods considered are
     the Adaptive SOR (A-SOR), the Steepest-descent (G), an
     adaptive version of the steepest-descent method (A-G), the
     Richardson's Optimum-Extrapolated (RF-OE), and the Conjugate
     Gradient (CG)."
  title = "{S}imulation of optical systolic and neural network using occam",
  author= "Evans, D. J. and Margaritis, K. G.",
  editor= "Edwards, Janet",
  pages = "100--110",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "This paper describes the simulation of optical processors
     for performing optical systolic and neural net algorithms.
     Occam is used as a simulation langugage and each component
     of the optical processor is defined as an occam process."
  title = "{D}evelopement methods and occam",
  author= "Gee, David M. and Worrall, Barry P. and Henderson, W. D.",
  editor= "Edwards, Janet",
  pages = "111--122",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "Existing structured methods for real-time systems are
     critically reviewed with respect to their suitability for
     direct implementation in occam without transformation. All
     are found to have weaknesses in some respects; consequently
     new notations are proposed. Use of these within an
     appropriate tool framework could assist in the automatic
     generation of occam code."
  title = "{A} general-purpose parallel programming environment",
  author= "Debbage, Mark and Hill, Mark and Nicole, Denis A.",
  editor= "Edwards, Janet",
  pages = "123--132",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "A parallel programming environment has been developed which
     is superior to many available systems on current generation
     transputers. The core of the package is a packet router
     which delivers asynchronous datagrams around arbitrary
     networks. The router is guaranteed not to deadlock provided
     that higher levels do not violate the eager-readership
     edict. This eager0readership can be guaranteed by the use of
     communication protocols which ensure that packets are never
     sent to a destination which has insufficient buffer space
     available. The router has been implemented in C and is
     applicable to a wide number of loosely coupled
     multiprocessing architectures. The current implementation
     has been optimised for transputer networks, on which it
     achieves impressive communications performance. The
     implementation of a virtual channel protocol within occam
     semantics on such a router is discussed. This forms the
     basis of version 2.0 of the popular Virtual Channel Router
     (VCR) package. The user interface to these channels is the
     conventional occam syntax for communications but with the
     configuration restriction of four channel-pairs per
     processor eliminated. The virtual channels provided by this
     package can be exploited from other languages through the
     appropriate Inmos Toolset."
  title = "{R}epeatable execution of occam programs",
  author= "Villano, Umberto",
  editor= "Edwards, Janet",
  pages = "133--142",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "Parallel programs are intrinsically non-deterministic, and
     therefore the techniques of cyclical debugging that are
     commonly used for sequential programs are not suitable for
     parallel ones. This paper proposes a method for guaranteeing
     the repeatability of occam program behaviour. Saving
     information on the ALT guards selected at run-time allows
     program replay. i.e. makes it possible to re-execute the
     program following the same instruction path whatever the
     speed at which each component process is executed. This
     enables the software developer to use tools such as
     debuggers and intrusive monitors to help identify program
     errors. After a discussion on the possible implementations
     of the proposed technique, a prototype tools that allows the
     replat of occam programs within TDS is briefly described."
  title = "{A} {S}oftware {D}evelopement {E}nvironment for {P}arallel {I}mage {P}rocessing: {I}mplementation techniques and issues",
  author= "Crookes, D. and Morrow, Philip J. and McClatchey, I. and Rafferty, T.",
  editor= "Edwards, Janet",
  pages = "143--149",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "A program development environment has been designed for
     building transputer based image processing systems. The
     system has two main components: a library of (parallelised)
     image processing operations, and a graphical user interface.
     After introducing the user interface, the paper concentrates
     on the implementation techniques and issues involved in
     constructing the environment. The main implementation areas
     considered are: a flexible and highly parameterised library
     of image processing routines; the provision of a
     communications shell which utilises the underlying
     parallelism without the user having to be concerned about
     it; and the extensibility of the environment by adding new
     components to the library."
  title = "{I}ntegration of {C}lassification and {E}valuation {P}rocedures in the {I}mplementation of {P}arallel {I}mage {A}nalysis {A}lgorithms",
  author= "Brittan, P. and Fairhurst, M. C.",
  editor= "Edwards, Janet",
  pages = "150--158",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "The requirements of adaptability and high performance often
     dictate a parallel processing approach for cost-effective
     algorithm implementation in many real-time applications of
     image classification. The aim of this paper is twofold.
     First it seeks to illustrate that by giving careful
     consideration to the conceptual architecture of the image
     classifier and its implementation base levels of performance
     can be achieved. Second, it will discuss the principal
     features of implementing multi-layer classifier algorithms,
     performance evaluation processes, and high-speed data
     handling required for evaluation, over an array of
  title = "{O}n the serialisation of parallel programs",
  author= "Welch, Peter H. and Justo, G. R. Ribeiro",
  editor= "Edwards, Janet",
  pages = "159--180",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "This paper argues that one of the key techniques for making
     the most efective use of multi-processor architectures is
     the serialisation of parallel code! Parallel algorithms are
     presented as having strong engineering merits that will form
     the natural basis for systems design in the future.
     Parallelisation of serial code is regarded as having only
     short-term value (for \"dusty-decks\", whose
     correctness cannot be verified) as well as being
     mathematically intractable. Serialisation, on the other
     hand, is much easier to automate and can be profitably
     employed today. Several serialising transforms for occam
     processes are presented and applied to various simulation
     and image compression tasks."
  title = "{A}pplication of occam to biological sequence comparisons",
  author= "Sturrock, Shane S. and Salmon, Ian",
  editor= "Edwards, Janet",
  pages = "181--190",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "One of the major uses of computers by molecular biologists
     is the alignment of protein or DNA sequences. Since the
     structure and function of proteins and DNA cannot be
     predicted directly from sequence data alone, the most useful
     procedure is to compare unknown sequences against a database
     of known sequences. Those with a high degree of homology are
     likely to have similar properties. At present, the databases
     are growing exponentially with doubling in around 20 months.
     Consequently, comparisons are taking proportionately longer
     to complete. We have applied various sequential comparative
     algorithms within a farming harness constructed in occam
     from a pipeline of transputers. The design of the farming
     harness for optimum link communication is discussed. The
     advantage our method offers is scalability; an important
     consideration as the databases multiply in content. The
     ability to tailor the algorithm to work on various size
     networks of transputers also allows the user to weigh time
     considerations against the total cost of the system. Our
     present implementation can perform around 600 comparisons
     per second between a query sequence of 50 residues and a
     database using 64 worker processors."
  title = "{I}mplementing an {A}ctive {C}hart {P}arser on a {T}ransputer {N}etwork",
  author= "Edwards, Janet and Connolly, John H.",
  editor= "Edwards, Janet",
  pages = "191--200",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "Active chart parsing is a well-known technique for the
     syntactic analysis of natural language. This paper is
     concerned with the parallel implementation of the relevant
     algorithm in Occam-2 on a Transputer network. Different
     parallelisation strategies are compared, and it is concluded
     that better results are obtained by splitting the grammar
     associated with the parser than by decomposing the algorithm
  title = "{A} {S}calable {C}ommunication {N}etwork for a {P}arallel {D}atabase {M}achine",
  author= "Walter, David and Kerridge, Jon",
  editor= "Edwards, Janet",
  pages = "201--216",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "The IDIOMS parallel database machine has at its kernel a
     scalable communications structure. The scalability is
     fundamental to the database machine because it is intended
     that the parts which are joined together by the
     communications network can be added to as demands on the
     database machine change. The network must be scalable both
     in terms og the number of user processes which can be
     supported, and, in order to meet the performance
     requirements of large systems, it should be possible to
     scale up the bandwidth of the network by adding further
     communication processors. The structure of a node in the
     network is described in terms of itsccam behavior which
     shows how node control is achieved by a control process that
     manipulates a set of flags by means of channel
     communications. The network is shown to be free from the
     possibility of deadlock. Finally, the performance and
     operational characteristics of the network are discussed and
     it is shown that the network is ideally suited for parallel
     database operation."
  title = "{P}arallel {S}can {L}ine algorithm for {H}idden {S}urface {E}limination",
  author= "Highfield, Julian C.",
  editor= "Edwards, Janet",
  pages = "217--224",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "With the general availability of general purpose parallel
     computers, there is a need to reconsider scan conversion
     algorithms with respect to their parallel implementation.
     This paper considers MIMD parallel implementation of two
     versions of the scan conversion algorithm, one the common
     edge table optimisation, and one not. Their suitability for
     parallel implementation is investigated and their relative
     performance in multi-processor systems is measured using
     polygonal scene descriptions of between 150 and 2600
     polygons. Dependence upon the size of scene description is
     measured and results are extrapolated to larger scene
     descriptions. It is shown that scan conversion algorithms
     may be efficiently parallelised. It is also shown that the
     edge table optimisation, while appropriate to the single
     processor case, becomes useless at around twenty processors,
     and would actually be a disadvantage in the limiting case of
     one processor per scan line."
  title = "{P}rocessor {I}ndependant and {E}xtendable {R}outing {S}ystem using a {C}yclic {R}outing {A}lgorithm",
  author= "Shallow, P. A.",
  editor= "Edwards, Janet",
  pages = "225--233",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "PIERS is a deadlock free virtual routing system developed
     for distributed concurrent applications. It deliberately
     exploits the advantages gained by allowing cycles to occur
     in the channel dependency graph and incorporates the cycles
     into the routing algorithm. It directly interfaces with the
     user's application and uses the existing OCCAM channel
     primitives. The paper describes how this cyclic
     routingalgorithm operates and how deadlock is averted. It
     describes how the routing system is implemented, presents
     the performance figures obtained and summarises the network
     configuration utility written to integrate the communication
     harness with the user's application."
  title = "{D}eterministic {M}essage {R}outing for {S}afety-{C}ritical {A}pplications",
  author= "Croll, Peter R.",
  editor= "Edwards, Janet",
  pages = "234--246",
  booktitle= "{P}roceedings of {W}o{TUG}-14: {O}ccam and the {T}ransputer-{C}urrent {D}evelopments",
  isbn= "90 5199 063 4",
  year= "1991",
  month= "sep",
  abstract= "This paper considers a technique of message passing which
     can be applied in the development of parallel programs for
     safety-critical applications. The routing algorithm used to
     ensure that messages will always be able to meet hard
     real-time constraints and yet cope with some degree of
     hardware failure. This paper will firstly introduce the
     routing algorithm, it will indicate what properties can be
     proved about deterministic message passing and describe how
     the algorithms cope with hardware failure. From this, the
     details of possible solutions that the new T9000 family can
     offer will be presented."

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.

Pages © WoTUG, or the indicated author. All Rights Reserved.
Comments on these web pages should be addressed to: www at wotug.org

Valid HTML 4.01!