A Multi-time Sharing Scheme in Optimal Memory Time Partitioning for Chip-Multiprocessors

When there are several concurrently executing applications in Chip-Multiprocessor (CMP), it´s a problem to allocate cache resource between these applications. Cache partitioning is commonly used to solve this problem. Most of the existing partitioning schemes is to reduce the miss rate of concurrently executing applications, but the optimal miss rate does not mean the best performance. This paper presents a Multi-Time Sharing (MTS) scheme based on Optimal Memory Time cache Partitioning (OMTP) to improve the overall performance and the fairness metrics of concurrently executing applications. the MTS scheme uses dynamic weighted values to implement multiple time sharing. the experiment results show that OMTP has higher miss rate compared with utility-based cache partitioning (UCP) but has better IPC throughput and weighted speedup, the MTS scheme used in OMTP can improve both fairness and performance.