Polymorphous Computing Architecture (PCA) Kernel Benchmark Measurements on the MIT Raw Microprocessor

摘要

The DARPA Polymorphous Computing Architecture (PCA) program is a research initiative aimed at developing new computer architectures with a high degree of flexibility. Unlike current computer architectures that are rigid in nature, PCAs will have the capability to adapt ('morph') to match the problem being solved. This flexibility will allow higher overall system performance in a broad range of applications. MIT Lincoln Laboratory has defined a set of kernel benchmarks for the PCA program. The kernel-level benchmarks have been chosen to stress both computation and communication aspects of the architecture. The particular benchmarks chosen are based on the frequency of their use in current and future applications. They are drawn from the areas of signal and image processing, communication, and information and knowledge processing. Each of these areas imposes different processing requirements on the architecture in terms of operations performed and memory bandwidth required. This document describes a set of measurements of the PCA kernel benchmarks on a prototype PCA chip.