Grouped Multiprocessor Architecture
Original Publication Date: 1985-May-01
Included in the Prior Art Database: 2005-Feb-18
This article describes a system architecture, the grouped multiprocessor, which substantially increases the cost/performance of a highly-parallel multiprocessor. Highly-parallel multiprocessors, such as the NYU Ultracomputer 1, suffer substantial performance degradation because of long primary memory latency, even when a cache memory is associated with each processor.