Research on Fine-Grain Cache Assignment Scheduling Algorithm for Multi-Core Processors

Currently, multi-core processors are becoming mainstream due to their superior performance and power characteristics. The problem of shared L2 cache allocation is critical to the effective utilization of multi-core processors. When running under the conventional LRU cache replacement algorithm, the cache is allocated to an application is proportional to the frequency of cache requests. But there is no requirement that the application uses this storage efficiently. Unbalanced cache allocation will happen, and this situation can easily leads to serious problems such as thread starvation and priority inversion, which threatens to processor´s utilization ratio and system performance. Aiming at this problem, a cache assignment scheduling algorithm which allocates suitable cache to running threads from fine-grain level is proposed. Experiment results show that the fine-grain cache assignment scheduling algorithm achieves higher performance and fairer cache assignment than that adopting conventional LRU policy. As a result, fine-grain cache assignment scheduling algorithm improves the operating system performance and utilization effectively.