This is the end game for OMTC - where we don't have any main thread compositing. So far we have OMTC on for Android, b2g, and Mac where we have HWA. To do: d3d9 and 11 for Windows. Open GL for Linux (low priority because we don't properly support this platform), and basic compositor as a fallback for everything and a primary solution for Linux and older Mac OS.

When we get here, we should remove all the main thread compositing code. That is all the layer managers (and their supporting classes) other then client and composite. We can then probably do a whole bunch of pleasant refactoring.

Is there any news on how this is coming along? The UX team is currently looking at a couple of ways to improve the perceived performance of Firefox and having reliable OMTC (and OMTA) would be extremely helpful.

Track the dependent bugs. What platforms are you interested in? OMTC is on by default on OS X, in progress on Windows (see the dependent bugs) and getting accelerated (aiming for train 34) for Linux/software. What exactly is UX looking for, and how is OMTC helping?

(In reply to Milan Sreckovic [:milan] from comment #3) > Track the dependent bugs. What platforms are you interested in? OMTC is on > by default on OS X, in progress on Windows (see the dependent bugs) and > getting accelerated (aiming for train 34) for Linux/software. What exactly > is UX looking for, and how is OMTC helping? Great, that was exactly the information I was looking for :) There's an etherpad with ideas around perceived performance here: https://etherpad.mozilla.org/PerceivedPerformance A lot of it depends on smooth animations even if the application is slow, which is where OMTA should come to the rescue.

Oh hey, we finished this.