A Fast Extraction Method of Expected Stack Distance Based on Statistical Deduction

Abstract

The invention discloses a method for rapidly extracting expected stack distance based on statistics and deduction. The method comprises the following steps: acquiring access Trace flow of a target program by way of Gem5, and marking each access request according to an execution order; using the red-black tree index structure to record access addresses and mark; when a reuse occurs, acquiring a reuse distance by subtracting current request execution mark and a previous access mark recorded in the red-black tree. The method of the invention finds out a conversion relationship between the reuse distance and the reuse distance based on a Cache set associative, and derives a conversion relationship between a reuse distance distribution and an expected stack distance. Since the time cost of reuse distance extraction is low, and the expected stack distance distribution based on the Cache set associative can be obtained from the aforementioned two conversion process, the method of the invention substantially increases the velocity of evaluating LRU-Cache access behavior by using the stack distance.

Description

technical field [0001] The invention relates to the field of software-hardware collaborative design, in particular to a method for quickly extracting expected stack distances based on statistical deduction, which is used for evaluating processor Cache memory access performance. Background technique [0002] With the development of semiconductor technology, the computing power of processors has been significantly improved. But the "memory wall" problem seriously affects the computing performance of the processor. As an on-chip cache, Cache can reduce the number of off-chip memory accesses and improve system memory access performance. [0003] In the embedded field, in addition to computing performance, area and power consumption are also important design indicators for system chips. Limited by product launch time and chip design costs, the industry needs to quickly obtain correct hardware design solutions to reduce product design cycles and design costs. Therefore, at the ...

AI Technical Summary

Problems solved by technology

[0006] Purpose of the invention: In order to overcome the problem in the prior art that the time complexity of the stack distance extraction algorithm that occurs in the evaluation of LRU-Cache memory access behavior is too high, the present invention provides a method for quickly extracting the expected stack distance based on statistical deduction. On the basis of quickly obtaining the memory access reuse distance, the design formula derives the expected stack distance distribution, which is used to quickly evaluate the LRU-Cache memory access behavior

Images