Research On The Method Of On-Chip Data Percolation For Dual-Core Processors

The inconsistency between the slow growth of memory access speed and the rapid increase of processor computing speed causes memory performance to severely limit the performance of the processor.The proposal of the percolation delay tolerance mechanism provides a new way to solve this problem.The percolation cache structure model is constructed by imitating the flow mode of spring water,which is mainly composed of spring suction and spring gushing.The just-in-time block constructed according to the principle of data locality is used as percolation data.The tuple is used for data percolation operation,and the simulation experiment shows that this mechanism can effectively improve the memory access performance,so as to achieve the purpose of improving the performance of the processor.However,there are still some shortcomings in this research:the current research is carried out on a single processor,the multicore percolation cache structure and data allocation methods have not been deeply studied,the difference between the single-core percolation structure and the multicore percolation structure has an impact on the cache hit rate There is a lack of corresponding research on the impact of processor performance and whether single-core fixed just-in-time block data allocation can meet the needs of multicore data allocation.In view of the above shortcomings,the research contents of this paper are as follows:(1)A dual-core on-chip cache data penetration method based on the basic idea of partial assignment is proposed.In the dual-core structure,each processor has its own a first-level cache and a second-level cache.Try to analyze the change of the processor’s demand for the local percolation cache capacity,so as to allocate the local percolation cache to the processor with greater demand for the cache capacity,so as to realize the partial assignment of the local percolation cache,and improve the effective utilization of the cache resources.(2)The dynamic adjustment strategy for the radius of just-in-time block is proposed.In-depth study of the performance of timely locality in permeating data migration state,permeating cache state,etc.,based on this,infer what requirements the current processor has on the timely locality of data,and realize the elastic change of the just-in-time block radius according to this requirement.To achieve the purpose of improving the percolation cache utilization and hit rate.(3)Joint scheduling of dual-core penetration cache is proposed.n the basis of partial assignment of multicore percolation cache,due to the change of cache capacity,the requirement for locality of data in time is different from that of fixed cache capacity.Therefore,joint scheduling is to dynamically adjust the radius of each processor’s own just-in-time block based on the partial assignment of the percolation cache,according to the change of the cache capacity,so as to improve the cache hit rate.(4)The simulation platform of dual-core penetration cache is designed and implemented,and four groups of comparative simulation experiments are designed and implemented based on the platform.The experimental results prove the feasibility of the local assignment idea,and verify that the dynamic adjustment of just-in-time block can improve the cache hit ratio and thus achieve the purpose of improving processor performance.In addition,the joint scheduling experiment shows that the dynamic adjustment of the just-in-time block according to the variation of cache capacity can effectively improve the cache hit ratio.