Dynamical Cache Partitioning Based On Cache Missing Ratio And Process Priority

Presently speaking, the Multi-Core system has been one mature technology abroad, however, the related research about the Multi-Core system internally start just in recent two years. For the reason of the closely relationship with the hardware of the Multi-Core, the less extension of the theory presently can be utilized to it and make it better performance. So, there is a little academic achievements which can obviously and advanced promote the performance of the Multi-Core system so far.The emergence of the Multi-Core system is one way for people who pursuits more advanced performance of the computer. This idea gives us the enlightenment to make the ideal destination come true. Meanwhile, it brings off many problems which should be resolved later. Of course, adding more computing core and making it integrated to one chip will strengthen the ability of the computing, anyhow, the emergence of the resource distributing, the task scheduling, the temperature control of the chip and the how to effectively control the power of each core and so forth are the topics awaited after introducing the Multi-Core system. This paper based on the research of the pre-research on this object, simply interprets the hot topics in current, its method and achievement, then detail elaborates our major, and progress.The research about this paper is in order to resolve and promote the performance of the system. A great many data show that, not more processor unit integrated into one chip will make faster of the computer. Because of this, we choose the parameters which are the affect the system performance directly and sensitively to analysis the relationship related, after this, we give rise to this algorithm according to the actual need.This paper studies the fairness model in the sharing cache between the cores in a chip Multi-processor (CMP) architecture, and makes two contributions as follows:firstly, this paper discussed the performance of the Multi-Core system based on the priority and the missing ratio, in contrast with the relative researches focus on the process priority, thread missing ratio or the relation between the process priority and the throughput respectively in past. Secondly, this paper proposed the new metrics to measure the fairness of the sharing cache partitioning, and the algorithm is under the guideline of this theory of the fairness model to partition the cache in order to maximize the ability of the processing of the core and on the whole. The experiments proved that this algorithm is effective and it improves the CPU's throughput.As for the reality of the algorithm, we adopt the common and general system performance detection tool Multi2Sim, and the CPU system performance test standard benchmarks to test the effectiveness of our algorithm.Experiment result indicates that, this algorithm promotes the throughput of the CPU. The experiment is tested under the two different conditions, and then making the comparison. Firstly, it distributes the resource depending on the priority purely recording the computing time lost. This time is compared with the second scheme which attributes the resource by the control of two conditions which are the priority and the missing ratio. The computing standard is that, the time cost under the equal quantity of the task is less than another which shows that the performance of the system performed with the first algorithm is better than another. The experiment shows that the system throughput is promoted obviously.The meaning of this algorithm is that, it improves the performance by more consideration of the bound variable which affects performance of the system. The better controlling appropriately decreases the missing of the thread to increase the throughput, and to reach the target of decreasing the overhead of time.