Confirmed users
856
edits
Platform/GFX/2012-February-27: Difference between revisions
Platform/GFX/2012-February-27 (view source)
Revision as of 23:34, 27 February 2012
, 27 February 2012→Agenda
(→Agenda) |
(→Agenda) |
||
| Line 7: | Line 7: | ||
=Agenda= | =Agenda= | ||
* Fennec status update (joe, jrmuizel) | * Fennec status update (joe, jrmuizel) | ||
** | ** Checkerboardiness | ||
*** better breakdown of spending our time | *** Getting better. | ||
*** | *** We have a better breakdown of where we're spending our time, and fixing them as we're going along. | ||
*** Drawing frames seems to have a very high (~10ms) fixed cost; not a big difference between complex and simple pages. | |||
*** JSON string building (JS->Java) isn't helping | |||
*** Dalvik JSON is known to be very very slow. | |||
*** kats has filed a bug, & blassey attached a patch, to kill JSON use when updating the viewport on mobile {{bug|727352}} | |||
*** There are still unknown issues. We expect lots of multi-millisecond costs. | |||
*** From the beginning of OnDraw until the time we actually start drawing layers can be high (~12 ms) | |||
*** Kats' patches seem to have stopped the multi-second checkerboarding, though. | |||
*** jrmuizel has not found any urgent invalidation issues | |||
*** Cwiiis has been looking at buffer rotation | |||
** Correctness | |||
*** joe, gw280, ajuma have been fixing bugs | |||
*** jrmuizel knows of a bug in which things stop drawing: you scroll, get checkerboarding, and then it never draws | |||
*** BenWa notes that the Adreno has a limited number of GL contexts available; possible minefield ahead. | |||
** mozilla-central merge | |||
*** multi-second checkerboarding was our biggest problem | |||
*** what do we need to land? how long will it take? | |||
*** head-to-head comparison with nightly | |||
*** won't block on flash on ICS | |||
*** upload costs seem to be causing lots of jank, that pcwalton is working on (EGLImage) | |||
* {{bug|711656}}, the blacklisting bug (Doug working on it) | * {{bug|711656}}, the blacklisting bug (Doug working on it) | ||
** Lots of crashes on drivers that are explicitly blacklisted | |||
*** It's possible but unlikely that they are force-enabling acceleration. | |||
** They are mostly on a particular integrated Intel graphics card. | |||
** Doug's going to make a patch to reimplement blacklisting again because we think the current version doesn't work | |||
* {{bug|717658}}, #4 mobile topcrasher at some point, the weird layout bug that feeds us with stupid huge sizes. | * {{bug|717658}}, #4 mobile topcrasher at some point, the weird layout bug that feeds us with stupid huge sizes. | ||
** OOM crash | |||
** absurdly large, like 50kx50k | |||
* WebGL conformance update (bjacob, jgilbert) | * WebGL conformance update (bjacob, jgilbert) | ||
** Ms2ger has a complete patchset for ImageData, pending review | ** Ms2ger has a complete patchset for ImageData, pending review | ||
** ANGLE identifier mapping patchset complete, pending review | ** ANGLE identifier mapping patchset complete, pending review | ||
** need status update on: alpha premultiplication (jgilbert, jrmuizel) | ** need status update on: alpha premultiplication (jgilbert, jrmuizel) | ||
*** {{bug|730157}} blocks WebGL alpha-premult, jgilbert to put a week into it | |||
** bjacob now working on: getExtension object persistence | ** bjacob now working on: getExtension object persistence | ||
* Tracking 11 Update | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=723444 | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=711656 | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=715401 | |||
*** https://bugzilla.mozilla.org/show_bug.cgi?id=729116 | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=726797 | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=713305 | |||
** https://bugzilla.mozilla.org/show_bug.cgi?id=668344 | |||
** We made sure all the bugs were owned/understood. | |||