Andreas Gal :gal Reporter
	Description • 11 years ago

Cache surfaces during allocation spikes to avoid allocating and memset-ing the surfaces constantly.

	Jeff Muizelaar [:jrmuizel]
	Comment 3 • 11 years ago

Comment on attachment 772877 [details] [diff] [review]
Recycle memory surface

I didn't review the new code very thoroughly as I assume that matt has already reviewed that.

	Matt Woodrow (:mattwoodrow)
	Comment 4 • 11 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/b61b7f2e0d0b

	Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)
	Comment 5 • 11 years ago

Backed out under suspicion of causing m-oth leaks like https://tbpl.mozilla.org/php/getParsedLog.php?id=25097117&tree=Mozilla-Inbound on OSX debug builds:
https://hg.mozilla.org/integration/mozilla-inbound/rev/f8257d93273a

	Daniel Holbert [:dholbert]
	Comment 6 • 11 years ago

This also caused:
 <Regression> Mozilla-Inbound - Tp5 Optimized (Main RSS) - MacOSX 10.7 - 15.6% 
which went away when KWierso backed it out.

(dev-tree-management bot incorrectly blamed the previous push, but if you look at the graph it's abundantly clear that it started on this cset's push.)

More info here:
https://groups.google.com/forum/#!topic/mozilla.dev.tree-management/BwXpmvAQzng

	Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)
	Comment 7 • 11 years ago

Well, my backout had that same conformance test failure that I backed the push out for: https://tbpl.mozilla.org/php/getParsedLog.php?id=25106631&tree=Mozilla-Inbound

Though the failure didn't show up until the following push, so maybe this won't be fixed immediately by the backout, either?

	Florian Bender
	Comment 8 • 11 years ago

Wes, Jeff, can you please revisit this bug? It's one of the last OMTC issues on Mac.

	Florian Bender
	Comment 9 • 11 years ago

… I actually should have meant Andreas. Sorry!

	Andreas Gal :gal Reporter
	Comment 10 • 11 years ago

Milan can you find an assignee?

	Milan Sreckovic [:milan] (needinfo for best results)
	Comment 11 • 10 years ago

With the demise of MemoryImage with bug 991028, this is not relevant anymore, right?