Does my C++ Object Model Work with a GPU, and Can I Make It “Safe”?

Speaker: Erik Tomusk

Audience level: [Beginner | Intermediate]

The object model is one of the defining features of C++, underpinning RAII, value semantics, automatic variable lifetimes, etc., and contributing to developer productivity, code reusability, and higher-level abstractions. Current industry trends include the explosive growth of GPU-based compute and an ever-increasing demand for safety in its various forms, and it’s often unclear if and how C++ is able to address these emerging requirements. Can C++ continue to deliver value, or is it just a liability at this point?

This talk will be a deep dive into the intersection of C++ and accelerated compute, exploring the benefits of C++ as well as points of tension. The discussion will include code examples, accelerator API design, and plenty of hardware details. There will be a particular focus on functional safety and its relationship to other "safeties."

The talk will be particularly relevant to those with an interest in real-time and safety-critical systems.