Newsgroups: comp.parallel
From: Thierry Cornu <cornu@mantrasun5.epfl.ch>
Subject: Call for participation, JIP'95
Organization: Ecole Polytechnique Federale de Lausanne
Date: 11 Sep 1995 15:31:20 GMT
Message-ID: <431ko8$2k7@usenet.srv.cis.pitt.edu>

	    ===============================================
	          #   ###   ######    ###    #####  #######
	          #    #    #     #   ###   #     # #
	          #    #    #     #    #    #     # #
	          #    #    ######    #      ###### ######
	    #     #    #    #                     #       #
	    #     #    #    #               #     # #     #
	     #####    ###   #                #####   #####
	    ===============================================


			Call For Participation

			       JIP' 95
	       (Journees Industrielles du Parallelisme)

		    PARALLEL COMPUTING IN INDUSTRY

		 ------------------------------------
		 Lausanne, CH, 28-29th September 1995
		 ------------------------------------

			     Organized by


				 GRIP
	    Groupe de Recherche en Informatique Parallele
		Swiss Federal Institute of Technology
		     Computer Science Department

				 and

			       COSMASE
	  Computation in Sciences, Methods and Algorithms on
		    Supercomputing for Engineering




PARALLEL COMPUTING IN INDUSTRY 7 AIM OF THE COURSE
--------------------------------------------------
The field of parallel computing and its applications is getting
increasingly vital for both industrial and scientific
activities. Parallel programming on clusters of workstations makes
high performance computing available at affordable price. Also, in the
domain of super-computing, several major vendors gradually introduce
massively parallel processing as a complement to or a replacement of
their more traditional vector computing architectures.  However,
industrial use of parallel computing is often hindered by a number of
difficulties: choice of a programming method, adaptation of existing
code, performance optimising, limited code portability, etc.  The goal
of the JIPs is to make the audience more familiar with the problems
related to the use of parallel computing systems. This year, there
will be two editions of the JIP. The first one was held in Bordeaux,
France in May 1995.  The second one will take place at EPFL in
Lausanne, SWITZERLAND in September 1995.

The JIP'95 in Lausanne will take advantage of the presence at EPFL of
the Parallel Application Technology Programme, a joint research
programme between EPFL and Cray Research. A first day will be devoted
to a series of conferences, by speakers from the industry as well as
from the academic field, presenting their experience with parallel
computing.  The second day will be devoted to a tutorial that will
make the participants more familiar with the different programming
standards currently available.  The participants will have the
opportunity to program on the Cray T3D massively parallel
supercomputer.



LECTURE ABSTRACTS
-----------------

Evolutionary Developments in the Field of Parallel Computing, R. PERROT 

In sequential computing a single model of computation has been
available for application programmers, language designers, compiler
writers and compiler architects; this has enabled the rapid
development of the field. The lack of a single agreed model has
inhibited the development and spread of parallel computing. This
presentation traces the major developments in hardware and software
which has influenced parallel computing and considers their
contribution towards the advancement of parallel computing and a
single model of computation.

Architectures and Programming Models for High Performance Parallel Computers, M. L. SAWLEY

This presentation is intended to address a number of questions
associated with the often difficult choice of parallel hardware and
programming models. The factors affecting this choice - including code
performance and portability - are discussed, based on experience
gained by porting a numerical simulation code to a number of different
parallel computer systems.  The use of programming standards, such as
MPI and HPF, to alleviate some of the potential difficulties of high
performance parallel computing will also be discussed.

Parallel CFD Frameworks: the Experience with PARAGRID and FLO67P, M. VITALETTI 

PARAGRID is a parallel multiblock framework developed at IBM ECSEC,
where CFD algorithms designed for single-block structured grid can be
integrated as subdomain solvers. Applications can also run very
efficiently in parallel mode by exploiting block-level
parallelism. This work reports on the joint development effort by IBM
ECSEC and Aermacchi in the development of FLO67P, a parallel CFD code
based on the Jameson FLO67 Euler solver which exploits the PARAGRID
framework.

Code for use in Gas Turbine CFD, I.K. JENNIONS / C. HUSTAD 

The presentation will describe implementation and validation of an
unstructured 3D Navier-Stokes code for solving complex geometry
internal flows in an industrial turbomachinery environment. The code
has been implemented using both standard PVM for workstation clusters
and specific shared memory routines for use on the Cray T3D.

Impact of Massively Parallel Computing Techniques and Systems to
perform Interactive 3D Seismic Structural Imaging: "a Commercial
Success", C. ROULET

In a partnership with MPP suppliers, powerful and efficient tools have
been made available to geophysicists to solve two key issues in
seismic data analysis : uneconomical turn around time of 3D data
processing and impractical iterative 3D structural and stratigraphic
inversion. It is now technically feasible to commercially provide oil
companies with " weeks" turn around time in order to process large 3D
seismic survey. An example will be described which depicts the
commercial use of a distributed, heterogeneous computing environment,
the heart of which is a powerful 512 nodes CM5.

RISK 2001: A Parallel Application for Banks, J. F. BALLIF 

RISK 2001 aims to calculate the global risk for banking institution in
the field of: credit, market, liquidity, interest, exchange, index &
stock risk. Using efficient simulation tools based on probabilistic
methods RISK 2001 computes a simulate global risk by variation on
relevant market parameters such as exchange rate, devise etc. The talk
will present the objectives of RISK 2001, its parallel processing
aspects and finally some results and bilan.

Dosimetric Calculation for Radiotherapy, M. STUDER.

A parallelized Monte Carlo Method for Electron Beam Dose
Calculation. In a joint project of the Dept. of Medical Radiation
Physics of the Berne University and TnTech Parallel Computing Support
Ltd. - embedded in the European AIM- project "MILORD" - the task was
given to accurately and fast calculate the absorbed dose in irradiated
organs during radiation therapy. This is conducted with a Macro Monte
Carlo Method on a parallel computer system. The MMC code was therefore
parallelized and implemented on the PowerPC based Parsytec
PowerXplorer and Parsytec GC PowerPlus parallel computers.

Address Space Programming Models for Distributed Memory Systems, A. MELTZER

This talk present an overview of global address space programming
models for distributed memory parallel architectures, discussing the
motivations behind the creation of these languages and the problems
faced by language designers. Many programming models such as HPF,
Vienna Fortran, Fortran-D have their own strengths and weaknesses;
many of these languages also struggle with a tension between
expressibility and performance and a tension between automatic and
user-specified processing. These issues will be discussed in the
context of the above languages as well as Cray's approach to solving
these problems with the CRAFT parallel programming model.

Message Passing Paradigms for Programming Massively Parallel Processors, H. POXON. 

Different message passing methods have been used for programming
different massively parallel architectures or cluster of workstations.
The most widely used of them: PVM (Parallel Virtual Machine), allows a
certain degree of portability of programs between different
architectures. Recently a new standard has emerged, the Message
Passing Interface (MPI), which is fast gaining a wide popularity. The
talk will compare the relative merits from a users point of view of
PVM and MPI implentations on the Cray T3D massively parallel
machine. Their relative performance characteristics will be discussed
as well.



SPEAKERS
--------
Dr. J.F. BALLIF is the Director of the Research & Technology
Department at LinkVest SA. He is mainly involved in the development of
software for banks. He is currently working on the project RISK 2001,
a parallel application for the computation of the global risk for a
financial company.

Dr. C.-W. HUSTAD is a Research Scientist in the Laboratory of Applied
Thermodynamics and Turbomachinery of the EPFL. He is responsible for a
collaborative project with ABB Power Generation Ltd, involving the
parallel computation of three-dimensional transonic turbomachinery
flows using unstructured grids.

Dr. I.K. JENNIONS has over 20 years of experience in the research,
development and introduction of advanced CFD tools into aerodynamic
design systems. He has worked for British Aerospace, UK Ministry of
Defence, Rolls Royce, General Electric Aircraft Engines and ABB Power
Generation. He obtained his BSc and PhD degrees from Imperial College,
London. Work for the PhD was sponsored by the Ministry of Defence and
involved CFD modelling of missile exhaust plumes. Dr. Jennions is
currently responsible for CFD Technology at ABB Power Generation.

Dr. A. MELTZER obtained his Masters of Science degree in Computer
Science of Cornell University. He has worked on a number of parallel
and massively parallel systems such as Thinking Machines CM1, CM2, CM
200, MasPar MP1, MP2, etc.; he has written a compiler for the Suprenum
compiler project; has done early compiler work for ACRI and has worked
on numerous designs of proprietary languages. Currently an employee of
Cray Research, Inc., he is co-designer of CRAFT, Cray's shared memory
model and the HPF representative for Cray Research. He is working in
Cray T3D/T3E compiler design and development.

Prof. R. PERROTT has been Professor of Software Engineering at the
Computer Science Department at the Queen's University of Belfast since
1985. He has been carrying out research into parallel computing for
more than twenty years particularly at NASA Ames research Centers and
CERN. He has published five books, two of them on parallel programming
and more than 80 articles. He is currently an invited professor at the
Computer Science Department of the EPFL.

Dr. H. POXON is the leading member of the PVM and MPI implementation
team at Cray Research, Inc. She began her involvement with the Cray
T3D and MPP programming methods with the Cray T3D Emulator, a program
that allowed users to emulate the different programming models
available on the Cray T3D. Prior to this, she worked as a member of
Cray's parallel processing software team, where she focussed on
software performance improvements for emerging Cray parallel vector
architectures. Her wealth of parallel processing experience is being
used to provide performance enhancements to message passing software
on Cray architectures.

Dr. M. VITALETTI received his degree in physics from the University of
Rome in 1981. After joining IBM in 1984 he participated to the
development of mathematical software for IBM vector computers,
including sparse iterative solvers for the IBM ESSL library. His
interests as a research scientist of IBM ECSEC focused on numerical
algorithms for computational fluid dynamics, including unfactored
implicit schemes for compressible flows. He was also involved in
cooperations with various European Aerospace industries on parallel
multi-block codes. This experience led to the design and development
of PARAGRID, a parallel multi-block framework designed to help the
porting of CFD codes to distributed memory systems.

Dr. C. ROULET is the Vice-President Technology for Schlumberger
Geco-Prakla. He graduated from "Ecole Normale Supirieure, Paris" ,
holds a Ph.D. in physics and has lead the Schlumberger seismic product
development group since 1991. His group has developed and
commercialised highly efficient 3D seismic structural imaging services
using massively parallel systems, operating in a distributed and
heterogeneous computing environment.

Dr. M. L. SAWLEY is a Research Scientist at the Fluid Mechanics
Laboratory of the EPFL. His present interests include the use of
different parallel programming models and computer systems for CFD
applications.

Dr. M. STUDER is the Managing Director of TnTech Parallel Computing
Support AG, a Swiss company founded in 1989 and domiciled in the Berne
Technopark. TnTech is specialised on and committed to parallel and
high performance computing. TnTech is the Swiss distributor and
representative of Parsytec Industrial Systems GmbH with desktop and
high end supercomputers. TnTech's courses are normally conducted
through NDIT/FPIT (a post- graduate educational program) or FORMITT (a
COMETT programme). TnTech has close relationships to various
educational and industrial institutions performing joint activities in
several national and European R&D- programs.



SCHEDULE
--------
Thursday, 28th September 1995, CM 4


08 h 00 - 08 h 30       Registration

08 h 30 - 09 h 00       Welcome and Introduction

09 h 00 - 10 h 00       Evolution in the Field of Parallel Computing, 
                        R. PERROTT

10 h 00 - 10 h 30       Coffee Break

10 h 30 - 11 h 30       Architectures and Programming Models for High 
                        Performance Parallel Computers, 
                        M. L. SAWLEY

11 h 30 - 12 h 15       Parallel CFD Frameworks: the Experience with 
                        PARAGRID and FLO67P, 
                        M. VITALETTI

12 h 15 - 14 h 00       Lunch

14 h 00 - 14 h 45       Code for use in Gas Turbine CFD, 
                        I.K. JENNIONS / C. HUSTAD

14 h 45 - 15 h 30       Impact of Massively Parallel Computing Techniques 
                        and Systems to perform Interactive 3D Seismic 
                        Structural Imaging: "a Commercial Success", 
                        C. ROULET

15 h 30 - 16 h 00       Coffee Break

16 h 00 - 16 h 45       RISK 2001: A Parallel Application for Banks, 
                        J. F. BALLIF

16 h 45 - 17 h 30       Application in the Medical Field, 
                        M. STUDER

17 h 30 - 18 h 00       Discussion 

20 h 00 - 23 h 00       JIP'95 Dinner

Friday, 29th September 1995, CM 4 and CM 103 rooms


08 h 30 - 10 h 00       Address Space Programming Models for Distributed 
                        Memory Systems, 
                        A. MELTZER

10 h 00 - 10 h 30       Coffee Break

10 h 30 - 12 h 00       Message Passing Paradigms for Programming 
                        Massively Parallel Processors, 
                        H. POXON.

12 h 00 - 13 h 30       Lunch

13 h 30 - 16 h 00       Hands-on Training Session on Cray T3D

16 h 00 - 16 h 30       Coffee Break

16 h 30 - 18 h 00       Hands-on Training Session on Cray T3D

        
        Friday, 29th September 1995, CM 113 room

        Afternoon, presentation of commercial products and posters: 
        Cray Research, IBM, Parsytec, AF-CSE (Computer Science Engineering)



ACCOMODATION FORM
-----------------
FEES

The enrolment fee for this course is: 

       Industry 800 CHF 
       University 500 CHF 

which includes lectures, course material, lunches and coffee breaks. 


PAYMENT

       Upon registration, we will send you directly an invoice for the registration fee. 
       You can pay by bank transfer or in cash the first day of the course. 
       Delayed registrations will increase the registration fees by 10%. 



ACCOMODATION

A block of rooms has been reserved at: 

ALPHAHOTEL, 34 rue du Petit-Chjne, (near the main railway station) 

Prices in Swiss Francs per room, per night and taxes included, but without breakfast 

       Single room with shower     CHF 90 
       Double room with shower     CHF 130 
       Breakfast at the restaurant CHF 8 

Please note: 

       The participants is advised to return the form as soon as
       possible in order that we may satisfy his/her choice 

       Subsequent changes in the reservation should be notified immediately 
       The registration form must be returned: by September 18th, 1995 



REGISTRATION FORM 
-----------------
(also available by WWW at URL: 	
	http://imhefwww.epfl.ch/COSMASE/JIP)

to be returned to:

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE
IMHEF / DGM
Mrs Christina BERG
CH - 1015 LAUSANNE, SWITZERLAND

-----------------------------------------------------------------------------                                                               
Course Registration Form - JIP'95 - September 28-29th, 1995

This registration form must be returned: by September 18th, 1995 
                                                               


Industry O                              University O

Title (Prof., Dr, Mr, Mrs)

Name                                    First Name

Position or Job Title

Company Address

Phone                   Fax                     E-Mail

I wish to reserve single room (s) for           nights

I wish to reserve double room (s) for           nights

Accompanying Person (s)

Name                            First Name              


Date of                 Date of                 Arrival time at 
Arrival                 Departure               the hotel        


I will participate to the gala dinner           yes     O        no     O

Date                    Signature and Stamp of Institution







-----------------------------------------------------------------------------                                       

                        
FURTHER INFORMATION
-------------------

JIP'95 WWW PAGE

	http://imhefwww.epfl.ch/COSMASE/JIP


GENERAL COORDINATOR                     SECRETARIAT

Dr. Alain DROTZ                         Mrs Christina BERG
EPFL                                    EPFL
IMHEF / DGM                             IMHEF / DGM
CH-1015 LAUSANNE                        CH-1015 LAUSANNE
Phone   ++ 41 21 693 35 08              Phone   ++ 41 21 693 35 04
Fax     ++ 41 21 693 36 46              Fax     ++ 41 21 693 36 46
E-Mail   drotz@dgm.epfl.ch              E-Mail  berg@dgm.epfl.ch

SCIENTIFIC COORDINATORS                 OR

Dr. Pierre Kuonen                       Dr. Thierry Cornu                       
EPFL                                    EPFL                                    
DI / LITH                               DI / LAMI                               
Phone   ++41 21 693 52 83               Phone   ++41 21 693 66 92               
Fax     ++41 21 693 52 63               Fax     ++41 21 693 52 63
E-Mail  grip@di.epfl.ch                 E-Mail  grip@di.epfl.ch

