An Infrastructure for Efficient Synchronization of Asymmetric Threads on Hyper-Threaded Processors

So far, the privileged instructions MONITOR and MWAIT introduced with Intel Prescott core, have been used mostly for inter-thread synchronization in operating systems code. In a hyperthreaded processor, these instructions offer a “performance-optimized” way for threads involved in synchronization events to wait on a condition. In this work, we explore the potential of using these instructions for synchronizing application threads on hyper-threaded processors which are characterized by workload imbalance. Initially, we propose a framework through which one can use MONITOR/MWAIT to build condition wait and notification primitives, with minimal kernel involvement. Then, we evaluate the efficiency of these primitives in an incremental manner: at first, we quantify certain performance aspects of the primitives that reflect the considered execution model, such as resource consumption and responsiveness. As a further step, we use our primitives to build synchronization barriers. We examine the same performance issues as before, and we evaluate also the efficiency of our implementation for fine-grained inter-thread synchronization. Finally, we test our barriers in a real-world scenario, and specifically, in applying thread-level Speculative Precomputation. For the application model under consideration, our proposed primitives is the option that best balances low resource consumption with high responsiveness and reduced call overhead.