Timothy Lochner

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.

To be a devil's advocate... "some developers don't understand multithreading, and [with this system] they don't have to" - sounds just like the reasoning that lead to introduction of garbage collection. Not saying DoD is inherently bad. Just that such a goal is questionable.

Nothing wrong with garbage collection as long as it’s done “right” according to your needs and purposes. In a way, garbage collection is also in Mana in several ways. The Gfx API does “GC” to frame data when it’s done being used.

But I think you’re right to be skeptical. I have another post later on illustrating why it’s useful. You can implement something “single threaded” and then if you need to, make MT improvements later.

This reflects the realities of development. You often work on a team, where each member has different skills. Initial implementation of a feature can be “single threaded”, where safety is guaranteed and concurrent work will still be scheduled.

If performance demands it, you can later make MT improvements. The DoD design makes that an easy refactor in many cases.

As a final note: on days when my brain is deep in some other subject, *I* am bad at multithreading. Being able to tackle the actual problem at and and leave the MT for another day is a boon.