RISC微处理器设计中流水线数据相关问题研究与实现

The Research and Implementation of Pipeline Data Dependency in RISC Microprocessor Design

摘要: 处理器流水线中的数据相关问题，通常会影响指令执行结果的正确性，并限制处理器性能的发挥。本文首先对流水线数据相关问题的本质及成因进行剖析，并提出解决数据相关的基本原理。依据此原理，并结合两种常用解决方法，流水线暂停法和数据前推法的优点，提出新的解决方案——“最小指令周期写回法”，该方法既灵活又充分节省了硬件资源的消耗，不失一般性，该方法可为解决用于嵌入式设备的RISC微处理器流水线中的数据相关提供参考。

Abstract: Data dependency in the processor pipeline usually affects the correctness of the results of the instruction execution, and limits the performance of the processor. In this paper, we first analyze the nature and causes of pipeline data dependency, and put forward the basic principle of data dependency. Then combining this principle with the advantages that both of the suspended pipeline method and the data push forward method, we propose a new solution “a minimum instruction cycle writing back method”. The method is flexible and fully saves the consumption of hardware resources. Without loss of generality, the method also can be used for solving the pipeline data dependency problem of the embedded RISC microprocessor.