Dynamically adaptive multi-core architectures have been proposed as aneffective solution to optimize performance for peak power constrainedprocessors. In processors, the micro-architectural parameters orvoltage/frequency of each core to be changed at run-time, thus providing arange of power/performance operating points for each core. In this paper, wepropose Thread Progress Equalization (TPEq), a run-time mechanism for powerconstrained performance maximization of multithreaded applications running ondynamically adaptive multicore processors. Compared to existing approaches,TPEq (i) identifies and addresses two primary sources of inter-threadheterogeneity in multithreaded applications, (ii) determines the optimal coreconfigurations in polynomial time with respect to the number of cores andconfigurations, and (iii) requires no modifications in the user-level sourcecode. Our experimental evaluations demonstrate that TPEq outperformsstate-of-the-art run-time power/performance optimization techniques proposed inliterature for dynamically adaptive multicores by up to 23%.
展开▼
