Dominik Charousset: Towards Actor Programming for High-Performance Computing

The actor model of computation has gained significant popularity over the last decade. Its high level of abstraction combined with its flexibility and efficiency makes it appealing for a wide range of applications. The actor model addresses concurrency as well as distribution and allows developers to scale from one core to many cores and from one node to many nodes.

In this presentation, I present our work on CAF, the C++ Actor Framework. CAF targets at providing an extremely scalable native environment for building fast and efficient software. CAF provides building blocks for infrastructure software and can even scale down to the IoT. However, the native C++ runtime of CAF also makes it attractive for the HPC domain. Scientific software could benefit from a higher level of abstraction to make reasoning about distributed software easier and to lower overall system complexity and ultimately development time. After introducing CAF, I will compare it to traditional, MPI-based HPC programming. Further, I will outline the upcoming challenges we face to scale CAF up to HPC demands.

Slides.