A New Computer Architecture Using a New Program Driving Method
Xiaobo Li, Xiangdong Cui, Xiaoqiang Ni, Chiyuan Ma, Jingfei Jiang, Zhou Huiping, Zhou Zhou, Ke Luo, Jiangsha Deng · 2006
Much dependency among the instructions of one program often limits the parallelism that can be exploited. In the other hand, the dependency among the instructions from different programs is much smaller, and the parallelism is more easily achieved. However, traditional processors are all using "hardware PC+" to drive a program to run, which is unable to run more programs in a single kernel processor concurrently and is therefore unable to exploit the parallelism of instructions of programs. This paper proposes a new type of computer processor (called concurrent multiple program processor CMPP) architecture using a new program driving method which can run more than one program in a single kernel processor concurrently without using interrupt technique. The main idea of the new driving method is to pick up all the program/instruction driving elements/factors to form a driving vector, called program driving vector (PDV) or program driver (PD) for short, so that many PDs can drive many programs to run concurrently