[Webkit-unassigned] [Bug 39123] Unneeded fillRect call in RenderBoxModelObject::paintFillLayerExtended

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=39123





--- Comment #13 from Pat Galizia <pgalizia at codeaurora.org>  2010-05-26 10:22:30 PST ---
(In reply to comment #12)
> (In reply to comment #10)
> > (In reply to comment #9)
> > > This patch isn't avoiding doing a fillRect for a fully transparent color. It's checking to see if a later fillRect will fully obscure an earlier one, which I think is a fine optimization.
> > 
> > Sorry, I was totally confused. Now that I know what this is, I can make a different comment.
> > 
> > We have tried this optimization multiple times in the past and have never need able to actually squeeze a speed-up out of it. I’m excited that things are different this time, and also a little skeptical because of the previous attempts.
> 
> The improvement may be due to our particular device configuration.  I'll look into applying this to a different device and see if we end up losing cycles with that change there.  If so, we can determine what to do next, since it makes no sense to apply this for one type of device if it regresses others.

I reran the patch on a higher-end x86_64 Linux system.  I don't see a difference one way or the other at a benchmark level, nor with instrumentation in the code.  From what I'm seeing, it looks like this patch may help lower-end devices, but won't do anything for mid-to-higher-end systems.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.