russell-taylor
commented
7 years ago Jeremy notes: I thought they used two contexts, one high priority context for the ATW thread and a lower priority one for the render thread, so the ATW thread could preempt the render thread. hmm, they shouldn't be then. the whole point is to use one high priority device and a separate low priority device. if they're sharing the same device, maybe ATW isn't working so well.
We only need a shared mutex if two threads are going to render to the same buffer, not if one thread writes it and another reads it.
Verify this again before committing, but the HighPoly example ran fine when the mutex code was removed.