TWO ATTRIBUTES OF THE CURRENT SUPERCOMPUTER MILIEU REVEAL AN IMPORTANT NATIONAL NEED ADDRESSED IN THIS PROPOSAL: A) THE KNOWN DISTRIBUTION OF SUPERCOMPUTER RESOURCES IN THE AMERICA AT OTHER THAN UNIVERSITY CENTERS OF EXCELLENCE, AND, B) THE RESULTS OF

TWO ATTRIBUTES OF THE CURRENT SUPERCOMPUTER MILIEU REVEAL AN IMPORTANT NATIONAL NEED ADDRESSED IN THIS PROPOSAL: A) THE KNOWN DISTRIBUTION OF SUPERCOMPUTER RESOURCES IN THE AMERICA AT OTHER THAN UNIVERSITY CENTERS OF EXCELLENCE, AND, B) THE RESULTS OF RECENT STATISTICAL ANALYSES, DISCLOSED IN THIS PROPOSAL, THAT IDENTIFY IMPORTANT (TYPICAL) PATTERNS OF SUPERCOMPUTER USAGE BY RESEARCHERS. IT IS THE POSITION OF THIS PROPOSAL THAT THESE DATA INDICATE THAT GREATER NATIONAL PROGRESS COULD BE MADE IN SCIENCE AND TECHNOLOGY TODAY THROUGH THE READY AVAILABILITY OF A LOW COST MINI-SUPERCOMPUTER RATHER THAN ANOTHER FIFTH GENERATION SUPERCOMPUTER. DATA IN THIS PROPOSAL WILL SHOW THAT A MODEST-PERFORMANCE, STAND-ALONE, 64-BIT, VECTOR-PROCESSING MINI-SUPERCOMPUTER SYSTEM, WITH A COMPATIBLE REGISTER ARCHITECTURE, AND A COMPATIBLE INSTRUCTION SET, THAT IS MICROCODED TO EXECUTE THE IDENTICAL OBJECT CODE OF THE MAINLINE SUPERCOMPUTER, COULD: A) SUBSTANTIALLY IMPROVE RESEARCHER SOFTWARE DEVELOPMENT PRODUCTIVITY, B) INCREASE THE EFFECTIVENESS OF EXISTING SUPERCOMPUTER FACILITIES BY RELEASING SUPERCOMPUTER RESOURCES TO WORK MORE EXCLUSIVELY ON COMPUTE INTENSIVE PROBLEMS, C) PROVIDE MORE COMPETITIVE RESOURCES FOR SIMULATION AND NUMERICAL ANALYSIS ALGORITHM DEVELOPMENT, AND, D) IMPROVE THE QUALITY OF TRAINING RECEIVED BY RESEARCH SCIENTISTS WHO COULD BE TRAINED, RATHER THAN RETRAINED, TO USE SCIENTIFIC VECTOR PROCESSORS IN PURSUIT OF THE COMPUTATIONAL EXPERIMENTS ADVANCED RESEARCH.