Add device-independent rendering of ovals, take two.

Add device-independent rendering of ovals, take two.

This permits GPU support for arbitrary matrices. The only exception is 
not all stroked ovals are supported, as thin ovals + fat strokes do not
produce elliptical borders.

Committed: http://code.google.com/p/skia/source/detail?r=11115

[jvanverth1, 2013-09-04 20:40:06]

Second shot at this -- fixes all the issues from this morning. The interesting
bits happen in Patch 2 and Patch 3. Patch 3 is a simplification of 2 which takes
us back to one attribute.

[bsalomon, 2013-09-04 20:51:16]

lgtm w/ one q

https://codereview.chromium.org/23701013/diff/13001/src/gpu/GrOvalRenderer.cpp
File src/gpu/GrOvalRenderer.cpp (right):

https://codereview.chromium.org/23701013/diff/13001/src/gpu/GrOvalRenderer.cp...
src/gpu/GrOvalRenderer.cpp:471: } else if
(target->caps()->shaderDerivativeSupport()) {
Is it better to prefer the circle renderer or should we always DIE if we can?

[jvanverth1, 2013-09-04 21:00:23]

On 2013/09/04 20:51:16, bsalomon wrote:
> lgtm w/ one q
> 
> https://codereview.chromium.org/23701013/diff/13001/src/gpu/GrOvalRenderer.cpp
> File src/gpu/GrOvalRenderer.cpp (right):
> 
>
https://codereview.chromium.org/23701013/diff/13001/src/gpu/GrOvalRenderer.cp...
> src/gpu/GrOvalRenderer.cpp:471: } else if
> (target->caps()->shaderDerivativeSupport()) {
> Is it better to prefer the circle renderer or should we always DIE if we can?

Given that 90+% of the ovals we see will be circles, I'd say its probably better
to default to the faster circle renderer (unless you have another reason in
mind). There isn't much difference in speed between the DIE renderer and the old
ellipse one; the old one is just there for the legacy hardware.

[commit-bot: I haz the power, 2013-09-05 18:35:02]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/jvanverth@google.com/23701013/13001

[commit-bot: I haz the power, 2013-09-05 19:26:54]

Change committed as 11115