Critical Problems in Very Large Scale Computer Systems
Abstract
This is the first semiannual report on this contract. The purpose of the present contract is to investigate limiting technologies for a very large computer system, one which, if built during the mid 1990's, would be so large that the nation could only afford one or two of them. The purpose of this contract is to investigate plausibility, which is defined as the step before feasibility. Once plausibility is demonstrated, then separate efforts at demonstrating feasibility and then design and manufacture would be appropriate. It is estimated that the cost of such 'building-size computer' would be in excess of $1,000,000.000. Six critical areas were identified, and work is in progress in each of these areas. The six areas, and the MIT faculty members who are working in each of the areas, are listed in the chart. The format of this report is based on this table. The work in circuits is a combination of improved techniques in waveform bounding and new ideas in LU factorization and relaxation methods. It has been found that in the context of highly parallel computation, the Gauss-Jacobi technique is never inferior to the Gauss-Seidel technique. Prototype circuits for a proposed processing element have been fabricated, and are under evaluation. The investigation of communications topology and related architectural concerns centers around two projects. In one of them, the Message- Driven Processor is the object of study.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 31, 1988
- Accession Number
- ADA203299
Entities
People
- Anant Agarwal
- Bill Dally
- Charles E. Leiserson
- F. T. Leighton
- Jacob K. White
- John L. Wyatt Jr.
- Lance A. Glasser
- Paul Penfield Jr.
- Thomas F. Knight Jr.
Organizations
- Massachusetts Institute of Technology