Dynamic adaptation in HPX - A task-based parallel runtime system

As parallel computation enters the exascale era where applications may run on millions to billions of processors concurrently, all aspects of the computational model need to undergo a transformation to meet the challenges of scaling impaired applications. One class of models aimed towards exascale computation is the task-based parallel computational model. Task-based execution models and their implementations aim to support parallelism through massive multi-threading where an application is split into numerous tasks of varying size that execute concurrently. Thread scheduling mechanisms used to manage application level tasks are a fundamental part of the task-based parallel computational model.;In task-based systems, scheduling threads onto resources can incur large overheads that vary with the underlying hardware. In this work, our goal is to dynamically control task grain size to minimize these overheads. We use performance studies to determine measurable events and metrics derived from them that indicate how tuning task granularity will improve performance. We aim to build a closed loop system that measures pertinent events and dynamically tunes task grain size to improve performance of parallel applications. High Performance ParalleX (HPX), the first implementation of the ParalleX execution model, is a runtime system that employs asynchronous fine-grained tasks and asynchronous communication for improved scaling of parallel applications. HPX is a modular system that has a dynamic performance modeling capability and a variety of thread scheduling policies and queuing models for work stealing and load balancing. It provides the ideal framework for studying parallel applications with the ability to make dynamic performance measurements and implement adaptive mechanisms. Therefore, dynamic tuning of task granularity is developed within the HPX framework.