IPDPS 2003 TUTORIAL
IPDPS 2003 is presenting three plenary tutorials, two on Tuesday and one on Saturday. Registration for IPDPS 2003 tutorials is in addition to registration for the symposium. They are not included in the technical program registration and require separate registration for each tutorial. See the IPDPS 2003 registration form. If you have already registered for the symposium, use the same form to register for tutorials only. The tutorials are held on the same days as workshops and there will be morning and afternoon refreshment breaks for all participants.

QUICK CLICK  [ Tutorial 1 ]  [ Tutorial 2 ]  [ Tutorial 3 ]  [ Download PDF ]

Tutorial 1 - Tuesday Morning, April 22, 2003

Introduction to Performance Tools Development: Build Your Own Tools!
 · Luiz DeRose, IBM T.J. Watson Research Center, ACTC
 · Bernd Mohr, Research Centre Julich, ZAM

%Introductory-20 · %Intermediate-60 · %Advanced-20

Description: Application developers are facing new and more intricate performance tuning and optimization problems as parallel architectures become more complex. Hence, users of high-performance computing systems normally complain that it is difficult to achieve a good performance for their application and are constantly asking for more and better performance analysis tools. However, the availability of tools development resources across the industry is shrinking rapidly because development of performance tools is complex and generally not profitable. In summary, more performance tools are needed, but fewer tools can be created by the industry. In this tutorial we present an overview of the major issues, techniques, and re-sources in performance tools development. Our goals are twofold: first, we will provide enough information, such that users who need specialized tools could attempt to do in-house development, in order to fulfill their needs. Second, we will discuss open problems in the area for researchers and students interested in working in the field of performance tools. Areas covered will include instrumentation, performance measurement, performance data representation, analysis, and visualization techniques. For more information see http://www.fz-juelich.de/zam/tools101/

QUICK CLICK  [ Tutorial 1 ]  [ Tutorial 2 ]  [ Tutorial 3 ]  [ Download PDF ]

Tutorial 2 - Tuesday Afternoon, April 22, 2003

          Object-Oriented Middleware and Components for the Grid: 
          Java, Corba Techniques and Tools
           · Denis Caromel, University Nice Sophia Antipolis, INRIA-CNRS-IUF
           · Christian Perez, IRISA INRIA

%Introductory-20 · %Intermediate-60 · %Advanced-20

Description: Object technology and middleware are increasingly used within parallel and distributed computing. At the same time, components are becoming a very effective tool for the composition and deployment of business applications. While the difficulties to actually program and deploy on the Grid have been, to a great extend, acknowledged, components technology is in the process of bringing many advances for such parallel programming. The aim of this course is to explain and detail this evolution.

The course objectives are:

­ to explain the main principles of component technology,
­ to explain how object-oriented middleware can be used for parallel and distributed programming,
­ to state and to detail how object-oriented middleware, together with components turn out to be very effective for the Grid.

The course is based on widespread, industrial languages and frameworks, such as Java, Corba, CCM.

QUICK CLICK  [ Tutorial 1 ]  [ Tutorial 2 ]  [ Tutorial 3 ]  [ Download PDF ]

Tutorial 3 - Saturday, Full Day, April 26, 2003

Programming in the Distributed Shared-memory Model
 · Tarek El-Ghazawi, George Washington University
 · Bob Numrich, Minnesota Supercomputing Institute
 · Dan Bonachea, University of California, Berkeley

%Introductory-30 · %Intermediate-50 · %Advanced-20

Description: The distributed shared-memory programming paradigm has been lately getting rising attention. Recent developments have resulted in viable distributed shared memory languages that are gaining vendors' support, and several early compilers have been developed. This programming model has the potential of achieving a balance between ease-of-programming and performance. As in the shared-memory model, programmers need not to explicitly specify whether accesses are local or remote. Meanwhile, programmers can exploit data locality in distributed memory systems using an abstract model that can enable program portability.

In this tutorial, we present the fundamental concepts associated with the distributed shared-memory programming model. These will include execution models, synchronization, workload distribution, and memory consistency. We then introduce the syntax and semantics of three parallel programming language instances with growing interest. These are the Unified Parallel C or UPC, which is developed by a consortium of academia, industry, and government; Co-Array FORTRAN, which is mainly developed at Cray Inc.; and Titanium, a JAVA implementation from UCB. It will be shown through experimental case studies that optimized distributed shared memory programs can perform at least as good as message passing codes, without much departure from the ease of programming of the shared memory model. For more information see
ftp://ftp.seas.gwu.edu/pub/upc/downloads/upcsc01.pdf