Concurrency and Parallelism for Game Programming – A Deep Dive (Jason Gregory - Naughty Dog)

Modern gaming platforms achieve their blazing speeds via highly parallelized hardware architectures: Multicore CPUs operate in tandem with GPUs containing thousands of SIMD cores. Take an in-depth guided tour through the fascinating, mysterious, and often vexing landscape of multithreading and concurrent programming on parallelized hardware within the context of high-performance real-time game programming. Join programming lead Jason Gregory of Naughty Dog for a deep dive into the intricacies of building a game that maximizes the benefits of parallel hardware. Through lectures, interactive exercises, group discussion and real-world examples, attendees will learn how to write reliable performant programs that take full advantage of multicore CPUs and the GPU. The course does a deep dive into how multithreaded synchronization works under the hood. We’ll discuss how to implement custom locks and the basics of lock-free programming. Participants will also learn how to apply their knowledge to build a job system, and how to build engine systems that fully utilize the hardware. We’ll also cover optimization via SIMD and GPU programming. The course concludes with a discussion of best practices, and how to think in a concurrent, data-oriented manner. October 13th in Bordeaux - October 15th in Paris - October 17th in Lyon Intro: Motivation, The Problem We’re Solving, Agenda I - Hardware and OS for Concurrent Programming II - Race Conditions and Locking III - Common Gotchas in Concurrent Programming IV - Under the Hood: Atomics, Custom Locks V - Concurrent Game Engines and Job Systems VI - Intro to Lock-Free Programming VII - SIMD and GPU Programming VIII - Best Practices and Thinking Concurrently