James Spooner explains how Data Flow Parallelism works and how it helps to design efficient parallel algorithms. Also: OOP vs. Parallelism.
John Nolan shows the state of hardware acceleration with GPUs and FPGAs, why it's hard to write efficient code for them, and why to favor polymorphism over if statements for performance.
Mike Williams, co-creator of Erlang discusses the history of and influences on Erlang as well as languages and paradigms used at Ericsson for large scale development and embedded programming.