Timothy Lochner

Writing a blog post on Data Oriented Design. Any skeptics want to share their worries/criticisms? Looking to address some of the most common ones.

Not skepticism but I’d love to see ways for defining scheduled work. The ones I’ve seen haven’t felt that great so far. BTW, in rendering, the concept of render graphs / frame graphs is really similar to dod but for graphics/gpu. It’s uncannily similar, very weird.

Continuing: What do you do about task thread affinities (to avoid trashing the cache)? What do you do about thread priorities? Do you have thread group based scheduling? And the biggest question of them all. How do you guarantee thread execution order independence?

I work on the Mana engine with Tim. Execution Order Independence is guaranteed by having each system declare data dependencies, and then building a static schedule that satisfies those dependencies. ( A before B ).

It's possible for the order to shift around a bit at runtime, but only between systems that have no dependencies on each other.

We check the dependency graph for deadlocks at startup, and prevent such a schedule from running in the first place.