Research On Dynamic Percolation Dispatch Method Of Data On Processor Chip

With the increase demand for high-performance and high-throughput computing in the computer field,the requirements for computer performance in academia and industry are becoming more and more stringent.As the main bottleneck that has plagued processor performance in recent decades,the "memory wall",there are three main directions to alleviate this problem: delay avoidance,delay reduction,and delay tolerance.Due to the gradual failure of dennard's scaling law and Moore's law,it is difficult to make further research progress in the solution of delay avoidance and delay reduction,which are implemented mainly by means of industrial integrated.This paper adopts software method,from the perspective of delay tolerance,based on a new type of percolation cache,through research in structure of cache and timely locality of data group,this paper proposes a dynamic on-chip data percolate dynamic deployment method.This method complements the research shortcomings and related research theories of dynamic allocation mode in percolate mechanism.As a medium to store data,the difference in cache's function and structure has a direct impact on its own performance,it also has an indirect impact on the performance of the processor.Therefore,rationally devise the cache structure and adopt a proper cache data deployment method can help improve the performance of cache and processor.Due to the local nature of data which accessed during processor operating,that is,within a certain time range,the data accessed by the processor is likely to be accessed again by the processor in the near future;within a certain spatial range,the data adjacent to the accessed data may also be accessed by the processor in the near future.Therefore,with the advancement of program sequential execution,by analyze the local characteristics of historical fetched data,we can find a cache structure that is more conducive to maintain characteristics of the current data in timely and locality,which can improve cache hit rate and improve cache performance,and alleviate the memory wall problem.In order to further explore the structural advantages of percolation cache,this paper proposes the concept of dynamic timely locality group and dynamic percolate cache structure.Under the respective matching cache data deployment mode,it realizes the adaptive processing of different programs and different segments of the same program,and finally obtains the percolation cache configuration structure more suitable for the current running program segment.In this paper,Verilog and Modelsim are used for simulation.By compare the dynamic percolate and composite dynamic percolate with the traditional cache and static percolate,this paper comprehensively demonstrates the functional advantages of dynamic deployment mode,dynamic timely locality of data group,and dynamic percolation cache structure of data percolate migration on processor chip.Simulation results show that dynamic percolate and composite dynamic percolate can not only improve cache hit rate and processor performance,but also indirectly alleviate the memory wall problem,which is a new way to solve the memory wall problem from the computer architecture level.