3rd UKHEC Seminar Programme

Monday 9th December, 2002

timespeaker or activity
10:30-11:00registration and coffee
11:00-11:30Mike Ashworth
11:30-12:30Martyn Guest
12:30-13:30John Hague
13:30-14:30buffet lunch
14:30-15:305Deborah Salmond
15:30-16:00tea
16:00-17:00Jan Boerhout
17:00-18:00Jennifer Scott
17:30-18:00free time
18:00-20:00wine reception

Tuesday 10th December, 2002

timespeaker or activity
9:15-10:30Rick Kufrin
10:30-11:00coffee
11:00-12:15Gernot Hoyler
12:15-13:30Bob Carruthers
13:30-14:30buffet lunch
14:30-15:45Werner Krotz-Vogel
15:45-16:15tea
16:15-17:30Judit Gimenez

Titles and Abstracts of Talks follow:

Mike Ashworth

(CLRC) - The UKHEC Collaboration
Abstract: The aim of the UKHEC initiative is to investigate emerging areas of computing and to inform and provide advice to the user community in hardware and software developments, in new tools, in best practice code development and in data management.

There are three centres supported under the EPSRC co-ordinated High End Computing (HEC) programme. The centres are:

In this presentation I shall summarise UKHEC activities, present examples from recent work and look to the future of HEC research in the UK.
Presentation slides [PDF]

Jan Boerhout

(NEC ESS) - High Performance through Vector-Parallel Processing
Presentation slides [PDF]

Bob Carruthers

(Cray UK) - Optimisation Strategy for exploiting the Cray X1 Architecture
Presentation slides [PDF]

Judit Gimenez

(CEPBA, Barcelona) - Obtaining useful information from raw performance data
Abstract: Performance tuning of parallel programs presents great difficulty due to the huge number of factors that influence performance. To be able to properly understand the behavior of a program it is of key importance to have very flexible tools that can be used to get responses to the hypotheses the analyst may wish to test.
In this talk we will describe Paraver, a trace-driven tool that lets an analyst fully extract the huge amount of information that is actually captured by a single trace. We will show some examples of how the tool has been used to better understand the behaviour of a wide range of codes and systems and the type of information obtained.
The presentation will finish with a live demo of the tool.

Martyn Guest

(CLRC, Daresbury) - Benchmark Performance of Current Processors

John Hague

(IBM UK) - Optimisation for IBM's p690 Power4 system.
Abstract: With the very advanced hardware and compiler technology that is now available, the emphasis in application optimisation for highly parallel progams has changed somewhat. Many of the techniques that have been used previously, have been overtaken by a match between compiler and hardware . This is not to say there is nothing left for the user to optimise!. However, the focus has changed a little.
Presentation slides [PDF]

Gernot Hoyler

(Technical Marketing Engineer, Intel, EMEA) - Getting the most out of the Intel(r) Itanium(r) Architecture
Presentation slides [PDF]

Werner Krotz-Vogel

(Pallas) - Software Development Tools for Getting the Most from your Processor
Presentation slides [PDF]

Rick Kufrin

(Performance Engineering and Computational Methods Group (PECM), National Computational Science Alliance (NCSA)) - Experiences with First-Generation Itanium at the National Center for Supercomputing Applications.
Abstract: The National Center for Supercomputing Applications (NCSA), located on the campus of the University of Illinois at Urbana-Champaign, is the leading edge site for the National Computational Science Alliance and is a part of a team of institutions involved in a multi-year effort to build and deploy the world's largest, fastest, distributed infrastucture for open scientific research. When completed, the TeraGrid will include 20 teraflops of computing power distributed at five sites, facilities capable of managing and storing nearly 1 petabyte of data, high-resolution visualization environments, and toolkits for grid computing. These components will be tightly integrated and connected through a network that will operate at 40 gigabits per second - the fastest research network on the planet.
The bulk of the computing power behind the TeraGrid will be housed at NCSA and will consist of 10 teraflops of computing capacity in IBM Linux clusters powered by Intel (R) Itanium (TM) 2 processors, the second generation of the EPIC/IA-64 architecture developed jointly by Hewlett-Packard Company and Intel Corporation. The process of porting applications to and optimizing them for the IA-64 architecture has been underway since the installation of a first-generation Itanium-based cluster at NCSA in late 2000 which has been successfully operating in production mode for scientific research since April 2001.
This talk will describe some of our experiences with bringing applications to the IA-64 platform and lessons learned along the way. We will also discuss performance analysis tools that have proven to be helpful in locating performance bottlenecks on this architecture and will provide examples of performance improvements as well as outstanding issues and difficulties that we have encountered as we begin to lay the foundation for the TeraGrid.
Presentation slides [PDF]

Deborah Salmond

(ECMWF) - Implementation of a global weather forecasting system on an IBM highly parallel scalar system with 960 Power4 processors.
Presentation slides [PDF]

Jennifer Scott

(CLRC) - Numerical libraries, HSL and large sparse systems
Abstract: For developers of computational software, there are significant advantages in exploiting mathematical software libraries. One such library is HSL (formerly the Harwell Subroutine Library). HSL is a collection of ISO Fortran packages for large scale scientific computation. Each package is a collection of subprograms that perform a basic numerical task. Started in 1963, the Library is written and developed primarily by members of the Numerical Analysis Group at the Rutherford Appleton Laboratory. Over the past forty years, HSL has built up a worldwide reputation as a prime source of numerical codes, with some of its best known packages being for sparse linear algebra and optimization.
The first part of this talk looks at why programmers should consider using numerical libraries, both for the development of commercial codes and for academic research. We then present a general introduction to HSL, including how to access the Library plus an overview of its main features. We discuss the design of software in HSL and how we achieve our goals of portability, efficiency, reliability, ease of use and flexibility. Finally, we provide a brief introduction to the HSL packages for sparse linear systems of equations.
R.J.Allan
Last modified: Monday November 25 13:48:50 GMT 2002