Disable accelerated 2D canvas on DX9 by not initializing GraphicsContext3D if it can be lost.

Disable accelerated 2D canvas on DX9 by not initializing GraphicsContext3D if it can be lost.

BUG=52689
TEST=http://www.corp.google.com/~scheib/no_crawl/canvas2dtests/minupload1.html. Jitters/popping in texture are a side-effect of NON GPU accelerated rendering. Smooth rotation will occur only with GPU acceleration. On Windows XP even with --enable-accelerated-2d-canvas, only software rendering should appear. On Vista and later the flag should result in smooth rotation.

[scheib, 2010-11-10 21:19:27]

James, this will enable the feature of canvas detecting XP / DX9. The test is
manual still, but we don't currently have infrastructure to regression test
this. I'll leave the bug open until a regression test is completed.

[jamesr, 2010-11-10 21:21:54]

On 2010/11/10 21:19:27, scheib wrote:
> James, this will enable the feature of canvas detecting XP / DX9. The test is
> manual still, but we don't currently have infrastructure to regression test
> this. I'll leave the bug open until a regression test is completed.
Sounds like a plan.

LGTM

[jamesr, 2010-11-10 21:22:01]

http://codereview.chromium.org/4320003/diff/1/chrome/renderer/webgraphicscont...
File chrome/renderer/webgraphicscontext3d_command_buffer_impl.cc (right):

http://codereview.chromium.org/4320003/diff/1/chrome/renderer/webgraphicscont...
chrome/renderer/webgraphicscontext3d_command_buffer_impl.cc:90: if
(attributes.canRecoverFromContextLoss == false) {
nit: if (!attributes.canRecoverFromContextLoss)

[Svyatpro, 2010-11-15 13:45:42]

On 2010/11/10 21:22:01, jamesr wrote:
>
http://codereview.chromium.org/4320003/diff/1/chrome/renderer/webgraphicscont...
> File chrome/renderer/webgraphicscontext3d_command_buffer_impl.cc (right):
> 
>
http://codereview.chromium.org/4320003/diff/1/chrome/renderer/webgraphicscont...
> chrome/renderer/webgraphicscontext3d_command_buffer_impl.cc:90: if
> (attributes.canRecoverFromContextLoss == false) {
> nit: if (!attributes.canRecoverFromContextLoss)

I used lastest Chromium builds on XP/2003, and HW acceleration worked as well,
and now from this (65733) build "IE9 Platform demos" works very slow!
Why is this build so poor?

[scheib, 2010-11-15 18:11:28]

> I used lastest Chromium builds on XP/2003, and HW acceleration worked as well,
> and now from this (65733) build "IE9 Platform demos" works very slow!
> Why is this build so poor?

As the bug http://code.google.com/p/chromium/issues/detail?id=52689 notes, there
is no good way to handle a lost context on DX9 - and we must offer correct
behavior even in the event of lost context.

[Svyatpro, 2010-11-15 19:29:06]

On 2010/11/15 18:11:28, scheib wrote:
> > I used lastest Chromium builds on XP/2003, and HW acceleration worked as
well,
> > and now from this (65733) build "IE9 Platform demos" works very slow!
> > Why is this build so poor?
> 
> As the bug http://code.google.com/p/chromium/issues/detail?id=52689 notes,
there
> is no good way to handle a lost context on DX9 - and we must offer correct
> behavior even in the event of lost context.

In this case Chrome is not very interesting, Because being on NT6+ it's simply
to use IE9 or FireFox 4, the main advantage of Chrome was hardware acceleration
of 2D Canvas under XP/2003, I was always using Opera. Only HW accel. of Chrome
made me to move on Chrome. But now, I should to back to Opera again because it
is more pleasant to use.