Mir

Code review comment for lp:~vanvugt/mir/predictive-bypass-2

Revision history for this message
Andreas Pokorny (andreas-pokorny) wrote :

> "In the meantime, why is this done on the mg{a,m}::DisplayBuffer level?"
> Again, to ensure safety and that we are never guessing about render times, the
> optimization is only applied when there is no render time -- bypass/overlay
> mode. This is good for Unity8 because it's always bypassed/overlayed.

Ok I complained about the - "at the end of post()" mostly because of making the decision inside DisplayBuffer instead of inside the rendering loop.. just as kdub stated. I think this behavior makes also sense if you have to render and if the render time is low enough. Of course then one has to deal with nesting. For a nested-bypass to make sense it should submit its updates as soon as possible. (Or as late as possible while ensuring that the host compositor still has enough time to redraw or bypass/overlay).

« Back to merge proposal