cc: Avoid power of two textures.

cc: Avoid power of two textures.

After rounding up to 64, it becomes more likely to hit
power-of-two textures which are also bad for IMG. So, this
avoids powers-of-two for IMG, through a fast-path extension.

BUG=175319
No try, since it only failed win_rel full-screen tests, but
this only has any effect on Android Imagination devices.
NOTRY=true

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=182087

[epenner, 2013-02-11 19:19:33]

Ptal.

[danakj, 2013-02-11 19:28:04]

I'm not sure if enne/nduca will agree, but I am wondering if this should be
LayerTreeSettings::avoidPowerOfTwoTextureSizes. I'm a bit hesitant about
non-obvious side effects from choosing different default tile sizes.

[enne (OOO), 2013-02-11 19:46:15]

Yeah, I have the same thought as danakj.  It's a little unexpected to have this
behavior be implicit in the choice of tile size.

[nduca, 2013-02-11 22:21:06]

yesplease, though not settings...

this sounds like a renderer capabilities thing

[epenner, 2013-02-11 23:07:02]

Okay renderer capabilities it is. Multiple-of-eight was also added with no flag,
but I'm liking that being always enabled.

[nduca, 2013-02-12 00:57:53]

I think the difference is that mo8 is pretty agreeable everywhere, whereas most
gpus can handle 256x256 tiles, and even prefer them.

[epenner, 2013-02-12 05:43:48]

Ptal. So while it was 'neat' to find that our GL_VENDOR is 'Chromium', it makes
doing this the 'right way' pretty difficult. So I'm hoping this is acceptable.

[nduca, 2013-02-12 05:45:49]

How about just exposing an extension via feature_info.cc and then testing for
its presence? Its a clean path from gpu process which knows this up to cc.

[epenner, 2013-02-12 05:50:55]

Yep. okay. I'd like to call it the "IMG" feature, but I don't get into it ;)

[epenner, 2013-02-12 06:22:21]

+Gman for new extension review. Ptal.

[nduca, 2013-02-12 06:42:28]

Hahah nice. GL_CHROMIUM_your_driver_is_hopelessly_bizarre_give_up_all_hope_1

[nduca, 2013-02-12 06:42:36]

parenthetically, lgtm

[greggman, 2013-02-12 17:50:43]

This seems like abuse of the extensions system. The extension string is suppose
to tell you what features are added, not what features are missing or broken

I'll lgtm it but my personal preference would be a positive attribute. In other
words, add an GL_CHROMIUM_fast_uploads_for_all_sizes for non IMG drivers or
something along those lines.

[epenner, 2013-02-12 18:47:49]

Yeah I agree. It's really gross, which is why I think everyone has avoided this
change. But it's just the reality we are in at the moment.

I'll try to make it a positive extension like you say, at least.

[epenner, 2013-02-12 20:40:58]

Okay I advertised a special fast path extension, which I think fits the model a
bit more. It's very limited, but so is IMG ;) If we can speed up an other cases
such that they don't cause dropped frames, I'll remove this extension.

[commit-bot: I haz the power, 2013-02-12 20:44:35]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/12217102/13001

[commit-bot: I haz the power, 2013-02-12 21:32:10]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/12217102/16002

[commit-bot: I haz the power, 2013-02-13 01:14:00]

Retried try job too often on win_rel for step(s) interactive_ui_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...

[commit-bot: I haz the power, 2013-02-13 01:45:33]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/12217102/16002

[commit-bot: I haz the power, 2013-02-13 02:02:54]

Change committed as 182087