Add direct bezier cubic support for GPU shaders

Add direct bezier cubic support for GPU shaders

BUG=

Committed: http://code.google.com/p/skia/source/detail?r=10814

[egdaniel, 2013-08-16 17:12:47]

So here is the first round of cubic support. Most of the math and KLM
calculations have been moved to GrPathUtils. 

In terms of rendering, the cubics are still off since correct clipping is not
happening (right now it is just a box around control points). For this reason
should I revert the HairLinePathRenderer back to its current patched in state
(using quads for cubics) since my version will pretty much fail all hairline gpu
GM's?

[bsalomon, 2013-08-16 17:47:00]

This looks really nicely written.

I'd say lets revert most of the hairline code except the new effect. Then we
should move all the effects out of the hairline renderer (not as part of this
change).

You should probably use SkASSERT instead of GrAssert, we're in the process of
removing GrAssert.

I'll try to investigate writing a GM that directly invokes an effect over the
weekend. It'd be great to have a way of testing what you've written without
having to hack in code path through the Skia API.

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp
File src/gpu/GrPathUtils.cpp (right):

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:482: enum CubicType {
style nit:

kSerpentine_CubicType,
kCusp_CubicType
...
kLine_CubicType
...

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:507: return kLoop;
indent

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:577: k[i] *= -1.f;
I'd say k[i] = -k[i]

maybe not faster but seems like it can't be slower.

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:611: k[i] *= -1.f;
ditto

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:664: k[i] *= -1.f;
ditto

https://codereview.chromium.org/22900007/diff/3001/src/gpu/GrPathUtils.cpp#ne...
src/gpu/GrPathUtils.cpp:810: 
style-nit: a few extra \ns here

[egdaniel, 2013-08-16 19:13:53]

Changes made.

My plan after this CL was to make a quick cl to take out the conic (and quad if
there is any) util code and move into pathUtils, and then make a 3rd one where I
actually move the effects out.

[bsalomon, 2013-08-19 13:34:24]

On 2013/08/16 19:13:53, egdaniel wrote:
> Changes made.
> 
> My plan after this CL was to make a quick cl to take out the conic (and quad
if
> there is any) util code and move into pathUtils, and then make a 3rd one where
I
> actually move the effects out.

Sounds good to me.

I think Jim should look at this CL since it's up his alley and he has to write a
review for you :)

[jvanverth1, 2013-08-19 15:12:06]

LGTM, just one potential f.p. issue:

https://codereview.chromium.org/22900007/diff/11001/src/gpu/GrPathUtils.cpp
File src/gpu/GrPathUtils.cpp (right):

https://codereview.chromium.org/22900007/diff/11001/src/gpu/GrPathUtils.cpp#n...
src/gpu/GrPathUtils.cpp:503: const SkScalar discr = d[0] * d[0] * (3.f * d[1] *
d[1] - 4.f * d[0] * d[2]);
Might want to make these checks against SK_ScalarNearlyZero.

[commit-bot: I haz the power, 2013-08-20 14:34:25]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/egdaniel@google.com/22900007/32001

[commit-bot: I haz the power, 2013-08-20 14:45:50]

Change committed as 10814