Industrial Applications of High-Performance Computing

195

Chapter 13

Industrial Letters

of Testimony to

the U.S. National

Science Foundation

Merle Giles

13.1 INTRODUCTION

e following excerpts are taken from the archives at the National Center

for Supercomputing Applications (NCSA) at the University of Illinois at

Urbana-Champaign. ey are letters submitted by NCSA industrial partners

to the Task Force on the Future of the NSF Supercomputer Centers Program.

e NSF-supported Supercomputer Centers played a major role in U.S.

CONTENTS

13.1 Introduction 195

13.2 Eli Lilly and Company 196

13.3 AT&T 198

13.4 Caterpillar Inc. 199

13.5 Dow Chemical Company 199

13.6 Eastman Kodak Company 200

13.7 FMC Corporation 200

13.8 Morgan Guarantee Trust Company of New York 201

13.9 Motorola 201

13.10 Phillips Petroleum Company 202

13.11 Tribune Interactive Network Services 203

Endnotes 203

196 ◾ Industrial Applications of High-Performance Computing

advancement of science and engineering research, and the Task Force Report

was prepared as the Centers Program approached its ten-year anniversary.

ese letters represent a rare compendium of industrial value that

comes from supercomputing.

1

Do note the comments that state value in

terms other than economic.

All content is in the form of quotes from the named authors, answering

the following questions:

1. Does the existence of the Centers benet your organization either

as a resource for technical help and development, or by providing

access to very high-end computational resources? If so, how?

2. Are there any quantitative measures?

NSF committee members were

1. Edward Hayes (Chairman), e Ohio State University

2. Arden Bement Jr., Purdue University

3. John Hennessy, Stanford University

4. John Ingram, Schlumberger, Austin, Texas

5. Peter Kollman, University of California, San Francisco

6. Mary Vernon, University of Wisconsin

7. Andrew White Jr., Los Alamos National Laboratory

8. William Wulf, University of Virginia

9. Nathaniel Pitts, Paul Young and Robert Voigt, NSF

13.2 ELI LILLY AND COMPANY

R.F. Abdulla, PhD, Director, Information Services (Scientic),

Eli Lilly and Company, Indianapolis, Indiana

We are currently partners at the NCSA at the University of Illinois at

Urbana-Champaign and are in the nal stages of negotiation for our

eighth consecutive year of membership. In general we have gained at Lilly

through this partnership in the following areas:

• High Performance Computing Applications in Drug Discovery

• e dissemination of decision support algorithms

Industrial Letters of Testimony ◾ 197

• Computational uid dynamics and manufacturing plant modeling

• Scientic visualization and Virtual Environments

• e institutionalization of structure-based drug design

• e institutionalization of molecular bioinformatics

• e creation of an expert systems group within the HPCC discipline

in Scientic Information Systems

e attempt to “quantify” the impact above, via outdated economic

precepts and tools, does both the analysis as well as the disciplines

represented above a disservice. I would state that the documented evidence

suggests that the technologies represented above have added unique value

to thisand other companies. Much of the added value remains in our

proprietary domains but a search of the literature will easily identify the

added value we, at Lilly, have chosen to share.

It is my opinion that the paradigm of high-performance computing

has changed since the creation of the NSF Supercomputing Centers

in 1986. e drive to produce faster chips … has met with both cost

and physics limits. Recent evidence … has suggested the dawn of a

“new” paradigm in … parallel computing. e literature indicates,

however, that in Europe, widely conceded as a theater with second tier

chip design capability when compared to Japan and American (sic),

this “new” high-performance computing paradigm was conceived of

necessity in the 1980s. I refer, ofcourse, to parallel computing in all

its manifestations ranging from tightly coupled printed circuit boards

and domain decomposition algorithms running across those boards,

to the more sophisticated machines being built in the United States of

various congurations.

2

Technically, these systems will be capable of

running a few optimized algorithms at teraop speeds before the end

of the decade. Algorithm development is lagging seriously, and will

continue to be a seriously limited factor in our capability to optimally

utilize these systems.

If the preceding paragraph encapsulates a problem, it also denes

a solution which could be invoked to solve the problem. e solution is

one which we continue to follow in America, the successful model of the

tripartite collaboration between the National Government, the Universities

and the Industrial Corporations represented by the NSF Centers. It is

evident that the mechanism for this to occur in the United States exists