SCALABLE HIGH PERFORMANCE COMPUTING CONFERENCE SHPCC '92 Sponsored by IEEE Computer Society April 27-29, 1992 Williamsburg Hilton and National Conference Center Williamsburg, Virginia ADVANCE PROGRAM SUNDAY 6:00-9:30 Registration 6:30-9:00 Cocktail Reception hosted by Intel Supercomputer Systems Div. MONDAY 8:30-9:30 Scalability of Data Transport, Harry Jordan, Univ. of Colorado 9:30-10:30 APPLICATIONS PERFORMANCE Image Algebra: An Object Predicting the Performance of Large Oriented Approach to Transparently Programs on Scalable Multicomputers, Concurrent Image Processing, I. B. Stramm and F. Berman, UC-San Diego Angus, Northrop Parallel Volume Rendering for Using Atomic Data Structures for Curvilinear Volumes, J. Challinger, Parallel Simulation, P. Barth, MIT UC-Santa Cruz 10:30-11:00 Break 11:00-12:00 APPLICATIONS LANGUAGES Abstractions for Parallel Vienna Fortran 90, S. Benkner, B. N-Body Simulations, S. Bhatt, M. Chapman, and H. Zima, Univ. of Vienna Chen, C.-Y. Lin, and P. Liu, Yale Univ. Real-Time Processing of Program Analysis and Transformations Spanning Trees, B. Lee, Stanford, for Message-Passing Programs, M. Univ. Gerndt, Univ. of Vienna 12:00-1:30 Lunch 1:30-2:30 Interleaved Mass Storage: Parallel Processing in Secondary and Tertiary Memory, Randy Katz, UC-Berkely 2:30-3:30 ALGORITHMS SYSTEMS SUPPORT FOR LANGUAGES Parallel Preconditioning and Improving Application Performance by Approximate Inverses on the Multithreading, E. Felten and D. Connection Machine, M. Grote, McNamee, Univ. of Washington Stanford Univ., and H. Simon, Computer Sciences Corp. Iterative Solution of Large Sparse Programming Distributed Memory Unsymmetric Linear Systems Parallel Computers without Explicit Resulting from CFD Problems on Message Passing, F. Andre and T. Massively Parallel Computers, A. Priol, IRISA Yeremin, Steklov's Mathematical Institute 3:30-4:00 Break 4:00-5:30 MOLECULAR DYNAMICS TOOLS Evaluating Parallel Languages for The 'Multicomputer Toolbox' Approach Molecular Dynamics Computations, Concurrent BLAS and Linear Algebra T. Clark, Univ. of Houston, R. Communication Subprograms, B. Still von Hanxleden, K. Kennedy, C. and T. Skjellum, Lawrence Livermore Koelbel, Rice Univ., and R. Scott, Univ. of Houston Parallel Molecular Dynamics on a HeNCE: Graphical Development Tools Torus Network for Surfactants for Network-Based Concurrent and Polymer Chains, K. Esselink Computing, A. Beguelin and J. Dongarra, and P. Hilbers, Shell-Amsterdam Oak Ridge and Univ. of Tennessee, A. Geist, Oak Ridge, R. Manchek, K. Moore, and R. Wade, Univ. of Tennessee Molecular Dynamics Simulation on Data Remapping for Distributed- an AP1000 Distributed Memory Memory Multicomputers, C. Chase and Parallel Computer, H. Sato and Y. A. Reeves, Cornell Univ. Tanaka, Fujitsu, H. Iwama, S. Kawakika, M. Saito, K. Morikami, and T. Yao, Protein Engineering Research Institute, S. Tsutsumi and H. Yoshijima, Fujitsu 6:30-8:30 Buffet Dinner and Poster Session Data Storage, Concurrency, and Portability: An Object Oriented Approach to Boundary Conditions in Fluid Mechanics, I. Angus, Northrop Problem-Specific Environments for Parallel Computing, L. Auvil, C. C. Ribbens, and L. Watson, VPI&SU Applications of a Parallel Pressure-Correction Algorithm to 3D Turbomachinery Flows, M. Braaten, General Electric Load Balancing and Parallel Implementation of Iterative Algorithms for Markov Chains, M. Colajanni and M. Angelaccio, Univ. of Rome Technical Challenges in the Monitoring & Visualization of Parallel Programs, C. Fineman, Sterling Federal Systems, and P. Hontalas, NASA Ames The RSCG Algorithm on the Intel iPSC/860, L. Freitag, Univ. of Virginia PFP, A Scalable Parallel Programming Model, B. Gorda and K. Warren, Lawrence Livermore Load Information Distribution via Active Interconnection Networks, D. Grunwald, Univ. of Colorado Compiler Optimizations for Distributed-Memory Programs, R. Gupta, Univ. of Pittsburgh Applications of Forall-Formed Computations in Large Scale Stochastic Dynamic Programming, F. Hanson and H. Xu, Univ. of Illinois at Chicago Distributed Hierarchical Scheduling with Explicit Grain Size Control, R. Hofman and W. Vree, Univ. of Amsterdam The Algorithm of Choice for Matrix Multiplication, C. Lin and L. Snyder, Univ. of Washington METRICS: A Tool for the Display and Analysis of Mappings in Message-Passing Multicomputers, V. Lo, K. Windisch, and R. Datta, Univ. of Oregon Debugging Mapped Parallel Programs, J. May and F. Berman, UC-San Diego Parameterized Memory/Processor Optimizing FORTRAN Comiler for Parallel Computers, D. Nosenchuck, Princeton Univ. An Expressive Annotation Model for Generating SPMD Programs, E. Paalvast, Institute of Applied Computer Science, and L. Breebaart, Delft Univ. Monte Carlo Particle Simulation of Low-Density Fluid Flow on MIMD Supercomputers, S. Plimpton and T. Bartel, Sandia-Albuquerque A Runtime Data Mapping Scheme for Irregular Problems, R. Ponnusamy, Syracuse Univ., J. Saltz, ICASE, R. Das, ICASE, C. Koelbel, Rice Univ., and A. Choudhary, Syracuse Univ. A Global Synchronization Algorithm for the Intel iPSC/860, S. Seidel and M. Davis, Michigan Technological Univ. Application of Linda to Molecular Modeling, T. Mattson and M. Shifman, Yale Univ. Synthesizing Scalable Computations from Sequential Programs, R. Govindaraju and B. Szymanski, RPI HYPERGEN - A Distributed Genetic Algorithm on a Hypercube, L. Knight and R. Wainwright, Univ. of Tulsa Programming an Astrophysics Application in an Object-Oriented Parallel Language, S. Yang, S. Narayana, J. Lee, and D. Gannon, Indiana Univ. 8:30-10:00 Panel on Educational Issues in Parallel Computing organized by R. Pargas, Clemson Univ. TUESDAY 8:30-9:30 Parallel Methods and Applications for Macromolecular Simulations, Bernard Brooks, National Institutes of Health 9:30-10:30 MOLECULAR DYNAMICS SUPPORT FOR IRREGULAR PROBLEMS A Parallel Scalable Approach to Automatic Mapping and Load Balancing Short-Range Molecular Dynamics on of Pointer-Based Dynamic Data the CM-5, R. Giles, Boston Univ. Structures on Distributed Memory and P. Tamayo, Boston Univ. and Machines, R. Weaver and R. Schnabel, Thinking Machines Univ. of Colorado Scalable Parallel Molecular Sparse Data Representation for Dynamics on MIMD Supercomputers, Data-Parallel Dense Computation, A. S. Plimpton, Sandia-Albuquerque Cheung and A. Reeves, Cornell Univ. 10:30-11:00 Break 11:00-12:00 COMPUTATIONAL FLUID DYNAMICS SYSTEMS ISSUES Scalability Issues for a Class of On the Influence of Programming CFD Applications, V. Naik, T.J. Models on Shared Memory Computer Watson Research Center Performance, T. Ngo and L. Snyder, Univ. of Washington Preliminary Experience in Devel- Communication Efficient Global Load oping a Parallel Thin-Layer Navier Balancing, D. Nicol, College of Stokes Code and Implications William & Mary for Parallel Language Design, D. Olander and R. Schnabel, Univ. of Colorado 12:00-1:30 Lunch 1:30-2:30 Computational Aerodynamics with Unstructured Meshes, Dimitri Mavriplis, ICASE 2:30-3:30 COMMUNICATION TOOLS Complete Exchange on a Circuit SUPERB Support for Irregular Switched Mesh, S. Bokhari, Univ. Scientific Computation, P. Brezany, of Engineering & Technology, and M. Gerndt, V. Sipkova, and H. Zima, S. Berryman, Yale Univ. Univ. of Vienna Intercube Communication for Phase Modeling of a Parallel the iPSC/860, E. Barszcz, NASA Scientific Code, P. Worley, Oak Ames Ridge 3:30-4:00 Break 4:00-5:30 COMPUTATIONAL FLUID DYNAMICS COMPILERS Massively Parallel MIMD Solution A Compile Time Scheduling and Code of the Parabolized Navier-Stokes Generation Tool for Distributed Equations, A. Stagg and G. Carey, Memory Multiprocessors, T. Yang and Univ. of Texas, J. Shadid and D. A. Gerasoulis, Rutgers Univ. Cline, Sandia-Albuquerque Unsteady Flow Simulation using Balancing Interprocessor Communica- an MIMD Computer, S. Chakravarthy tion and Computation on Two-Dimen- and S. Palaniswamy, Rockwell sional, Mesh-Connected Processors Running Compiler-Parallelized Code, M. Annaratone and R. Ruhl, ETH A Parallel Implementation of the Data Alignment: Transformations to Chemically Reacting CFD Code, Reduce Communication on Distributed SPARK, J. Otto, NASA Langley Memory Architectures, M. O'Boyle and G. Hedayat, Univ. of Manchester 6:30-8:30 Dinner hosted by Intel Supercomputer Systems Div. Location to be announced. 9:00- ?? Technical Vendor Presentations WEDNESDAY 8:30-9:30 Performance Animation, Dennis Gannon, Indiana Univ. 9:30-10:30 APPLICATIONS LANGUAGES A Look at Some Scalable Algorithms Towards a Distributed Memory from Linear Algebra for Implementation of Sisal, M. Haines Distributed Memory Architectures, and W. Bohm, Colarado State Univ. J. Dongarra, Oak Ridge and Univ. of Tennessee, and R. van de Geijm, Univ. of Tennessee Solving Equality Constrained A Test Suite Approach for Fortran Least Squares Problems, U. 90D Compilers on MIMD Distributed Vemulapati, Univ. of Central Memory Parallel Computers, M.-Y. Wu, Florida SUNY-Buffalo, and G. Fox, Syracuse Univ. 10:30-11:00 Break 11:00-12:30 ALGORITHMS PERFORMANCE Incremental Mapping for A Methodology for Visualizing the Solution-Adaptive Multigrid Performance of Loosely Synchronous Hierarchies, J. Keyser and D. Programs, S. Sarukkai, Indiana Univ., Roose, Catholic Univ. of Leuven D. Kimelman, and L. Rudolph, T.J. Watson Research Center Adaptive Methods and Rectangular Visual-Aural Representations of Workload Partitioning Problem, C. Performance for a Scalable Ozturan, B. Szymanski, and J. Application Program, J. Francioni, Flaherty, RPI Univ. of Southwestern Louisiana, and D. Rover, Michigan State Univ. Portable Parallel Level-3 BLAS in Portable Execution Traces for Linda: Performance and Use, B. Parallel Program Debugging and Ghosh and M. Schultz, Yale Univ. Performance Visualization, A. Couch and D. Krumme, Tufts Univ. 12:30 Conference Ends TUTORIALS April 26, 1992 Computational Fluid Dynamics on Parallel Machines - Algorithms and Applications 9:00-4:30 Ramesh Agarwal, McDonnell Douglas Research Laboratories Greater computational power is needed for solving computational fluid dynamics problems of interest in engineering design. This tutorial is mainly concerned with the implementation of algorithms to solve the Euler and Navier-Stokes equations on concurrent computers with grain sizes ranging from fine to coarse. The tutorial will begin with an introduction to the terminology of parallel processing, and a discussion of performance evaluation parameters. An overview of commercially available parallel machines will also be given. Modern numerical algorithms for the solution of the Euler and Navier-Stokes equations will be discussed, and their implementation on machines such as the Cray X-MP/48, IBM 3090-E, Ardent Titan, BBN Butterfly, INTEL iPSC/2, Myrias, Symult 2010, and the Connection Machine CM-5 will be described. Particular attention will be paid to differences in implementation on SIMD and MIMD architectures. Using Massively Parallel Supercomputers: An Applications Perspective 9:00-12:00 Horst Simon, Computer Sciences Corporation, NASA Ames Research Center Because of their considerable potential computational power, parallel supercomputers are increasingly considered as an alternative to the more conventional supercomputers based on a small number of powerful vector processors. Even though many research issues concerning their effective use and their integration into a large scale production facility are still unresolved, parallel supercomputers are already used for production computing, although mostly in a single application or research mode. In this tutorial, we will summarize the experience made with massively parallel supercomputers such as the Connection Machine from Thinking Machines Corporation and the iPSC/860 (Touchstone Gamma Machine) from Intel at NASA Ames Research Center. We will survey current hardware and software trends for massively parallel supercomputers from the perspective of applications users, and discuss in case studies the lessons learned at NASA Ames. Compilers for Scalable Architectures 1:30-4:30 Ken Kennedy, Rice University and Joel Saltz, ICASE This tutorial will describe how to design distributed memory compilers capable of generating efficient code for scalable architectures. Directives that specify distribution of data are added to a sequential source program, this is translated to a parallel program for a scalable multiprocessor. This tutorial will focus on Fortran D and High Performance Fortran as source languages. Fortran D will first be described. We will then cover the techniques used to build compilers able to handle structured scientific problems. This will include a description of the types of analyses and optimizations used in these problems. Following this will be a description of how to build compilers able to handle various types of irregular and adaptive problems. We will give examples from the areas of computational fluid dynamics (unstructured and block structured codes), molecular dynamics and structures calculations. Morning and afternoon breaks will be provided with all tutorials. In addition, lunch will be provided for those signing up for the full day tutorial or for both of the half day tutorials. GENERAL INFORMATION LOCATION: The technical sessions will be held in the National Conference Center of the Williamsburg Hilton. The hotel is less than two miles from Colonial Williamsburg which forms one third of Virginia's Historical Triangle along with Jamestown and Yorktown. The restoration of Colonial Williamsburg has set a standard of excellence recognized througout the world. A variety of tours and admission packages are available; in addition, several restaurants provide dining in the spirit of colonial times. For further information write to Colonial Williamsburg, P.O. Box 1776, Williamsburg, VA, 23185, or call 1-800-HISTORY. Limited complementary transportation is available from the hotel. The weather in Williamsburg ranges from 65-75 degrees in April and many flowers and flowering trees will be in bloom. TRANSPORTATION: Williamsburg, Virginia is 15 miles from Newport News/Williamsburg International Airport in Newport News, Virginia and 45 miles from Byrd International Airport in Richmond and Norfolk International Airport. For ground transportation from the airport to the hotel the following rental car and limousine services are available: Byrd Airport (Richmond): Groome Transportation, (804) 857-1231. Leaves every hour on the hour; 20/one person; $13/two more. Rental Car Companies: Thrifty, Hertz, National, Alamo and Budget. Norfolk Airport: Groome Transportation, (804) 857-1231. Leaves every hour on the half hour; $21/one person; $14/two or more. Rental Car Companies: Thrifty, Avis, Budget, Dollar, Hertz and National. Newport News/Williamsburg Airport: Williamsburg Limousine (804) 877)0279; $15 per person. Rental Car Companies: Avis, Hertz, Budget, National and Thrifty. Conference rates at the Hilton are available beginning on Saturday, April 25 in order to take advantage of the location and the cheaper airfares involving a Saturday night stayover. MEALS: A continental breakfast and lunch will be provided for tutorial participants. A reception hosted by Intel Supercomputer Systems Division is planned for Sunday evening. During the conference, a continental breakfast will be provided along with morning and afternoon coffee breaks. Monday evening during the Poster Session a buffet of international dishes will be provided and a cash bar will be available. Tuesday evening Intel Supercomputer Systems Division will host a dinner for all attendees. REGISTRATION INFORMATION Mail To: SHPCC '92 Registration IEEE Computer Society 1730 Massachusetts Ave., N.W. Washington, DC 20036-1903 (202) 371-1013 FAX (202) 728-0884 Please Print or Type Name_____________________________________________ Last First Middle Initial Badge Name_______________________________________ Company__________________________________________ Address__________________________________________ City/State/Zip/Country___________________________ Daytime Phone Number_____________________________ FAX Number_______________________________________ E-Mail Address___________________________________ IEEE/CS Membership Number________________________ Registration fees include Conference and/or Tutorials attendance, one copy of the proceedings available at registration, and attendance at the social functions described elsewhere in this program. Tutorial Registration (Fee per tutorial) Full Day Half Day Advance Late Advance Late Member 0 $300 0 $350 0 $200 0 $240 Nonmember 0 $360 0 $410 0 $250 0 $295 Student Any Combination 0$95 FULL DAY: 0 #1 Agarwal HALF DAY: 0 #2 Simon 0 #3 Kennedy/Saltz Conference Registration Advance Late/On Site Member 0 $190 0 $230 Nonmember 0 $225 0 $275 Student 0 $95 0 $125 Total Payment (Must Be Enclosed) $_________ Make checks payable to IEEE Computer Society. Payment must be in U.S. dollars, drawn on U.S. banks. Advance registration must be postmarked by APRIL 6. Method of Payment 0 Check 0 Visa 0 MasterCard 0 American Express 0 Purchase Order Cardholder Name_______________________________ Signature_____________________________________ Card Number___________________________________ Expiration Date_______________________________ Written requests for refunds must be received by the IEEE Computer Society office no later than April 6, 1992. Refunds are subject to a $50 processing fee. All no-show registrants will be billed in full. Registration after April 14 will be accepted on-site only. HOTEL RESERVATIONS Room reservations should be made by contacting the Williamsburg Hilton at (804) 220-2500. The conference rate is $68, single, $74, double, both subject to a room tax of 6.5%; to assure these rates, be sure to refer to the conference. The hotel will also honor the rate for April 25 through April 30 to facilitate taking advantage of the area and lower airfares for a Saturday night stayover. The reservation deadline is March 20. For a hard copy of the Advance Program including a hotel reservation card, contact emily@icase.edu.