Method for realizing instruction prefetching in front-end assembly line by advanced pointer method

Abstract

The invention discloses a method for realizing instruction prefetching in a front-end assembly line by an advanced pointer method in the technical field of front-end assembly lines. The method for realizing instruction prefetching in the front-end assembly line by using the lead pointer method comprises the following steps of: pre-knowing cache miss in advance, pre-knowing the specific position ofdata to be accessed by pc (Personal Computer) of a fetch in data sram and pre-knowing branch skip in advance by pre-fetching tag logic of a pointer in advance in bottom hardware; therefore, the dataaccessed by the pipeline in the icache module due to branch skip can be reduced; the power consumption of the icache is further reduced; the flow series of the front-end assembly line is shortened; after the icache miss is foreseen, the data L2 can be scheduled back in advance before the real pc points to the cache line of the miss; therefore, assembly line bubles caused by icache miss on a front-end assembly line are reduced to a great extent, and the execution efficiency of the front-end assembly line is improved.

Description

technical field [0001] The invention discloses a method for realizing instruction prefetching in a front-end assembly line by an advanced pointer method, and specifically relates to the technical field of the front-end assembly line. Background technique [0002] Due to most of the current CPU architecture design, the front-end pipeline of the CPU uses icache and branch prediction at the same time, and icache / branch prediction usually requires a delay of 2-3 clocks regardless of whether serial or parallel access is used. , that is to say, in the case of a branch instruction jump, the result of the instruction jump needs to be known after accessing the icache and the branch predictor. If the jump needs to be erased at this time, the third level of the icache accessed before the jump instruction Pipeline, this will inevitably waste the dynamic power consumption during the period. In addition, when a cache miss occurs in the icache, the L2cache is accessed. At this time, the co...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a method for realizing instruction prefetching in the front-end pipeline by the advanced pointer method, so as to solve the problem that the traditional pc access mode icache proposed in the above background technology will cause the internal instruction pipeline of icache to be erased due to the jump of the branch instruction, and branch The cost caused by prediction errors is too high, and cache misses cannot be predicted, so icache misses will send a large number of bubbles into the front-end pipeline, affecting the execution efficiency of the front-end pipeline

Images