Using the parallelism viewpoint to optimize the use of threads in parallelism-intensive software systems

The use of multithreading can enhance the performance of a software system. However, its excessive use can degrade the performance. For example, a thread-per-job approach might lead to a large amount of threads with increased associated overheads. In this paper we explore the use of the Parallelism Viewpoint to support one possible strategy to reduce the number of threads, namely finding candidate threads that can be replaced by thread pooling. Thread pooling reduces the large number of threads by reusing threads from an existing pool. As an example we analyze the threads of a precision critical parallelism-intensive electron microscope software system. Results show that the viewpoint provides a profound insight into the threading structure of the system, which helps in reducing the number of threads in a cost-effective way. And, the total time gain along with such reduction is encouraging.status: publishe