Works well on Windows now and should work fine when the basic compositor is ready.

Comment on attachment 8365375 [details] [diff] [review] patch Review of attachment 8365375 [details] [diff] [review]: ----------------------------------------------------------------- ::: gfx/layers/composite/LayerManagerComposite.cpp @@ +316,5 @@ > + int aOffsetX, int aOffsetY, > + Compositor* aCompositor, > + EffectChain& aEffectChain) > +{ > + unsigned int divisor = 100; NOTE TO SELF: > 1000 -> 999

see also bug 874781

Yeah, I keep meaning to get back to bug 874781 :/

Comment on attachment 8365375 [details] [diff] [review] patch Vlad are you happy with going with this version instead of yours? Our approach are very different. I'm not sure how they compare in performance but my patch is simpler and the performance is good enough. We could speed it up by extending Compositor API to support a DrawQuad API and possible add a way to retain an abtraction over VBO for accelerated backends.

Comment on attachment 8365375 [details] [diff] [review] patch Review of attachment 8365375 [details] [diff] [review]: ----------------------------------------------------------------- Things that make things more cross-platform are generally good... ::: gfx/layers/composite/LayerManagerComposite.cpp @@ +440,5 @@ > /** Our more efficient but less powerful alter ego, if one is available. */ > nsRefPtr<Composer2D> composer2D = mCompositor->GetWidget()->GetComposer2D(); > > if (composer2D && composer2D->TryRender(mRoot, mWorldMatrix)) { > + double fps = mFPS->mCompositionFps.AddFrameAndGetFps(TimeStamp::Now()); shouldn't this only be called if mFPS != nullptr?

Yes, another other review comments?

BenWa, the following code causes the regression as not to use HWComposer at all. Can you fix it? ------------------------------------------------ /** Our more efficient but less powerful alter ego, if one is available. */ nsRefPtr<Composer2D> composer2D = mCompositor->GetWidget()->GetComposer2D(); - if (composer2D && composer2D->TryRender(mRoot, mWorldMatrix)) { + if (mFPS && composer2D && composer2D->TryRender(mRoot, mWorldMatrix)) { + double fps = mFPS->mCompositionFps.AddFrameAndGetFps(TimeStamp::Now()); + if (gfxPlatform::GetPrefLayersDrawFPS()) { + printf_stderr("HWComposer: FPS is %g\n", fps); + } mCompositor->EndFrameForExternalComposition(mWorldMatrix); return; }

(In reply to Benoit Girard (:BenWa) from comment #4) > Comment on attachment 8365375 [details] [diff] [review] > patch > > Vlad are you happy with going with this version instead of yours? > > Our approach are very different. I'm not sure how they compare in > performance but my patch is simpler and the performance is good enough. We > could speed it up by extending Compositor API to support a DrawQuad API and > possible add a way to retain an abtraction over VBO for accelerated backends. Oh yeah, go nuts. Getting something in is way better than blocking; the world has changed quite a bit since my patches too. It shouldn't hard to get any of the interesting bits from my patches back in again at some point.

Opps, filed bug 971875 to fix the regression. It's really bad that we don't have any test for HWC on checkin :(