Publications on Paralelism by Gene Cooperman

This page is still being developed.

The papers on this page are concerned with my model for parallelism in distributed memory, I have incrementally develop the TOP-C (Task-Oriented Parallel C) model. A simple model requires that the performance issues concerned with when a message is sent or arrives should be hidden from the programmer, while not greatly impacting overal efficiency. Hence, I felt that the key issue in maintaining good efficiency under these conditions was latency tolerance. This led me to the three abstractions of this model.

I assume a master-slave architecture. The three abstractions are (1) the task, (2) the environment, and (3) the action. The environment is the globally shared data. The task is a procedure executed on a slave process that takes an input, and the environment, and produces a unique output. The action is the decision taken by the master process when it is presented with a task input-output pair. Example actions are NO_ACTION, REDO, UPDATE_ENVIRONMENT, and CONTINUATION(), and various hybrid strategies are possible. Most of the issues of efficient parallelization can then be reduced to the choice of definition of a taks, and what action to take for various input-output pairs.

Some of the features that come with this abstraction are easy reuse of "legacy" sequential code, easy checkpointing (sinc This parallel model has proved quite successful in a variety of examples of largest applications of their kind. At this time, my research issues are: the scalability of CPU's (> 100); scalability for large, distributed memory (> 1 GB); and extension of the model to SMP/DSM (hopefully with minimal rewriting of application code).

In addition to the papers below, the software implementations are available for C (TOP-C), GAP (ParGAP or Parallel Groups, Algorithms and Programming) and GCL (Parallel GNU LISP). Descriptions of applications done with these systems are available here.

o

``Parallel GAP: Mature Interactive Parallel'', G.~Cooperman, {\sl Proceedings of Conference on Groups and Computation}, DeGruyter Publishers, 2000, to appear.

o

``TOP-C: Task-Oriented Parallel C for Distributed and Shared Memory'', G. Cooperman, Workshop on Wide Area Networks and High Performance Computing, Lecture Notes in Control and Information Sciences 249, 1999, Springer Verlag, pp. 109-118.

o

``TOP-C: Task-Oriented Parallel C for Distributed and Shared Memory'', G. Cooperman, Lecture Notes in Control and Information Sciences 249, 1999, Springer Verlag

o

``TOP-WEB: Task-Oriented Metacomputing on the WEB'', G. Cooperman and V. Grinberg, International Journal of Parallel and Distributed Systems and Networks 1, 1998, pp. 184-192

o

``TOP-WEB: Task-Oriented Metacomputing on the Web'', G. Cooperman and V. Grinberg, Proceedings of Ninth IASTED International Conference on Parallel and Distributed Computing and Systems (PDCS-97), IASTED/Acta Press, Anaheim, 1997, pp. 279-286

o

``Practical Task-Oriented Parallelism for Gaussian Elimination in Distributed Memory'', G. Cooperman, Linear Algebra and its Applications 275-276, 1998, pp. 107-120

o

``Using Tadpoles to Reduce Memory and Communication Requirements for Exhaustive, Breadth-First Search Using Distributed Computers'', G. Cooperman and M. Tselman, Proc. of ACM Symposium on Parallel Architectures and Algorithms (SPAA-97), ACM Press, 1997, pp. 231-238; also N.U. Tech. Rept. NU-CCS-97-02, 1997

o

``Practical Parallel Coset Enumeration'', G. Cooperman and G. Havas, Proc. of Workshop on High Performance Computation and Gigabit Local Area Networks, Lecture notes in control and information sciences 226, Springer Verlag, 1997, pp. 15-27

o

``GAP/MPI: Facilitating Parallelism'', G. Cooperman, Proc. of DIMACS Workshop on Groups and Computation II 28, DIMACS Series in Discrete Mathematics and Theoretical Computer Science, AMS, Providence, RI, 1997, pp. 69-84.

o

``New Sequential and Parallel Algorithms for Generating High Dimension Hecke Algebras using the Condensation Technique'', G. Cooperman and M. Tselman, Proc. of International Symposium on Symbolic and Algebraic Computation (ISSAC '96), ACM Press, 1996, pp. 155-160

o

``TOP-C: A Task-Oriented Parallel C Interface'', G. Cooperman, 5^th International Symposium on High Performance Distributed Computing (HPDC-5), IEEE Press, 1996, pp. 141-150

o

``Computing with Matrix Groups using Permutation Representations'', G. Cooperman, L. Finkelstein and M. Tselman, Proc. of International Symposium on Symbolic and Algebraic Computation (ISSAC '95), ACM Press, 1995, pp. 259-264

o

``STAR/MPI: Binding a Parallel Library to Interactive Symbolic Algebra Systems'', G. Cooperman, Proc. of International Symposium on Symbolic and Algebraic Computation (ISSAC '95), ACM Press, 1995, pp. 126-132

o

``Parallel Implementations of Group Membership and the Method of Random Subproducts'', G. Cooperman, L. Finkelstein, and B. York, Proc. of 1992 Dartmouth Institute for Advanced Graduate Studies in Parallel Computation Symposium (DAGS'92), D. Johnson, F. Makedon, and P. Metaxas (eds.), Department of Mathematics and Computer Science, Dartmouth College, pp. 94-100, June 17-20, 1992