IPC-Based Cache Partitioning: An IPC-Oriented Dynamic Shared Cache Partitioning Mechanism

In a chip-multiprocessor with a shared cache structure, the last level cache is shared by multiple applications executing simultaneously. The competing accesses from different applications degrade the system performance, resulting in non-predicting executing time. Cache partitioning techniques partition the shared cache for multiple applications. The aim of traditional cache partitioning mechanism, Utility-based Cache Partition (UCP) for example, is to lower the overall miss rate of shared cache. But the lowest miss rate doesn't mean the highest performance (IPC). This paper investigates IPC-based Cache Partitioning (IPC-CP), an IPC performance oriented dynamic cache partitioning method. We design a Miss Rate Monitor to collect miss rate information of competing applications at run-time. Then the information collected is inputted to a Miss-Rate to IPC model to get the corresponding IPC performance. Lastly, we get the optimal cache partitioning based on IPC optimum objective function. Our evaluation, on top of a four cores CMP processor with 20 multi-programmed workloads shows that IPC-CP improves throughput by up to 53% and on average 9% over UCP.