Evaluation of relaxed memory consistency models for multithreaded multiprocessors

Abstract

Stochastic timed Petri nets are developed to evaluate the relative performance of distributed shared memory models for scalable multithreaded multiprocessors. The shared memory models evaluated include the Sequential Consistency (SC), the Weak Consistency (WC), the Processor Consistency (PC) and the Release Consistency (RC) models. Under saturated conditions, we found that multithreading contributes more than 50% of the performance improvement, while the improvement from memory consistency models varies between 20% to 40% of the total performance gain. Our analytical results reveal the lowest performance of the SC model. The PC model requires to use larger write buffers and may perform even lower than the SC model if a small buffer was used. The performance of the WC model depends heavily on the synchronization rate in user code. For a low synchronization rate, the WC model performs as well as the RC model. With sufficient multithreading and network bandwidth, the RC model shows the best performance among the four models.