Preliminary evaluation of eco-threading framework for commodity platform

We are developing a framework to enhance performance and efficiency of multi-thread applications. The framework consists of a kernel-level thread scheduler, an application programming interface (API) for the scheduler, and a helper thread to control the behavior of the scheduler through the API. We exploit the affinity of sibling threads, which have the same parent process and share the context, so that we can reduce the occurrences of avoidable events such as cache misses and enhance the efficiency of multi-thread programs. The API enables us to adjust the control parameter to make a balance between the affinity and other issues in scheduling such as fairness. The helper thread monitors performance counters and automatically adjusts the behavior of the scheduler through the API to try to maximize the effectiveness of the scheduler. According to the results of the preliminary evaluation, our framework is promising to increase the energy efficiency as well as to reduce the execution time.