HPX - A C++ Runtime System for Parallel and Distributed Computing
Explore a C++ runtime system for parallel and distributed computing in this comprehensive conference talk from CppNow. Dive into HPX, a new runtime system based on the ParalleX computation model, designed to address challenges in multi-core, multi-threaded environments. Learn how HPX leverages asynchrony between computation and communication, implements remote asynchronous functions called Actions, and utilizes data constructs like Futures and Dataflow to support task execution asynchrony. Discover the Active Global Address Space (AGAS) addressing scheme, which enables temporary suspension of remote thread objects and task migration across resources. Gain insights into how HPX conforms to the latest C++ standards and Boost Libraries, providing an easy-to-understand interface for seamless integration of multi-core, multi-threaded heterogeneous architectures with user-level software applications. Understand how HPX aims to improve resource utilization and scalability in high-performance computing systems, addressing challenges such as starvation, latency, overheads, and contention resolution. Read more