HPX

HPX is an open-source C++ runtime system and library for parallel and distributed applications, implementing the ParalleX execution model to enable scalable, fine-grained parallelism on heterogeneous hardware and large clusters.

Its core concepts include the Active Global Address Space (AGAS), which provides a globally addressable object space; a message-driven runtime; and a work-stealing scheduler that manages lightweight user-level threads. Tasks, or HPX work items, execute asynchronously and can be composed with futures and dataflow constructs; actions are remote procedures that operate on localities.

Programming model: HPX exposes an API that parallels the C++ standard library and parallel algorithms, while adding asynchronous actions, futures, and executors. Developers can write portable C++ that scales across nodes, and HPX can interoperate with existing C++ code and MPI-based codes. It supports various network backends and can run on clusters, campus HPC systems, and supercomputers.

Usage and ecosystem: HPX is used in research and industry to explore parallelism models and to scale applications from small to large scales. The project includes tutorials, examples, and a growing set of components, including numerical algorithms, I/O facilities, and caching.

History and licensing: HPX originated from the ParalleX project to provide an implementation of its execution model. It is maintained by a community of researchers and developers and is open-source under a permissive license.