Tutorial 1

Note: This tutorial is a joint tutorial between the ACM JavaGrande/ISCOPE conference and the SUN JavaOne conference.

Participants that register for this tutorial through ACM as part of the JavaGrande conference registration are entitled to visit the tutorial at the JavaOne location. To attend other events at the JavaOne conference you have to register with the JavaOne conference.
The tutorial 1 meeting location for this totorial will be in downtown San Fransisco, in the Palace Hotel. No transportation will be provided from Palo Alto. Attendees should allow at least two hours for travel and parking. No lunch will be provided.
Space for JavaGrande participants is limited to 50 seats requirening early registration by May 11th. The seats are given out on a first come first serve basis.
In case you have registered for this tutorial but intend to cancel it a penalty of $125 is required as to cover the costs ACM pays to JavaUniversity for the seat reservation.

Joint JavaGrande and JavaOne tutorial on:
Java^TM for High Performance Computing

Instructors:

Dr. José E. Moreira
Research Staff Member
IBM Thomas J. Watson Research Center

Dr. Roldan Pozo
Leader, Mathematical Software Group
National Institute of Standards and Technology

Dr. Vladimir Getov
School of Computer Science, Westminster,
UK

Dr. Gregor von Laszewski
Argonne National Laboratory, U.S.A.

others TBD

Who Should Attend:

Skilled Java technology-based software developers, engineers, programmers, analysts, architects, and computational scientists

Prerequisites:

Hands-on Java programming language experience is required. Understanding of distributed computing and object-oriented programming is required.

Benefits:

Highlights current trends in the use of Java for high-performance computing
Introduces Java numeric issues, including
- efficient multi-arrays and complex numbers
- efficient numerical calculations in Java
- Java numerical libraries
Compares the performance of Java with C/C++ and Fortran
Discusses how Java can be used for distributed computing, including
- use of message-passing libraries
- distributed computing with remote objects
- efficient access to computational Grids

Outline:

full day

INTRODUCTION TO HIGH-PERFORMANCE COMPUTING
- History
- Parallel programming paradigms
- High-performance and distributed computing challenges
- Current trends in the high-performance computing community
JAVA NUMERIC ISSUES
- Introduction to Java for numerical computing
- Bit-wise reproducibility of floating-point results
- Multi-arrays
- Complex numbers
- Transcendental functions
- Numerical libraries
- Performance analysis
  - Comparison with C/C++ and Fortran
  - Optimizing BLAS kernels
  - Java numeric benchmarks
  - General performance improvements
OPTIMIZING COMPILER TECHNOLOGY
- Java vs. Fortran array data structures
- Optimizing compilers
- Arrays
- Performance results
JAVA FOR DISTRIBUTED COMPUTING
- a. Distributed programming with message based libraries
- MPI, openMP, JMS
- FAST RMI, CORBA, and Jini
- Distributed programming in computational Grids
  - Introduction to computational Grids
  - Secure access to remote resources in Grids
  - Remote job submission and monitoring
  - Distributed Grid information management an remote data access
  - Designing graphical component to access Grids

Abstract:

This course will discuss in detail how to use Java for tasks requiring high-performance computing. The course is intended for developers and computational scientists experienced in using Java technology. The course will cover three main areas:

How to use Java in applications that require enhanced numerics
How to optimize compilers to achieve high performance
How to design distributed high-performance applications that use well-established programming paradigms

Acknowledgments:

This course is developed with the input of the following researchers (in alphabetical order):

Boisvert, Ron, NIST
Fox, Geoffrey C., Florida State University
Getov, Vladimir, University of Westminster, UK
Gupta, Manish, IBM T.J. Watson Research Center
Laszewski, Gregor von, Argonne National Laboratory
Midkiff, Samuel, IBM T.J. Watson Research Center
Moreira, Jose, IBM T.J. Watson Research Center
Philippsen, Michael, University of Karlsruhe, Germany
Pozo, Roland, NIST
Thiruvathukal, George K., Layola University

This tutorial is also announced through the JavaOne conference.