Add Loop Blinn hairline conics to GPU

Add Loop Blinn hairline conics to GPU

BUG=
R=bsalomon@google.com

Committed: https://code.google.com/p/skia/source/detail?r=10487

[egdaniel, 2013-07-31 18:26:02]

Here is the Loop Blinn conic shader.

For conics the bench shows that it is faster then the implicit version. It is
slightly slower than the current implicit version that is currently checked in,
but that is only because the current one only makes one chop at the quad max
curvature which is not enough. So the test I ran is using the implicit with two
iterations of chopping.

We should also see improvements for AA cubic, quad, and line rendering on
devices the mac mini since the vertex size has been reduced back down compared
to what is currently checked in.

[jvanverth1, 2013-07-31 18:41:39]

Just some typos I noticed:

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
File src/gpu/GrAAHairLinePathRenderer.cpp (right):

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
src/gpu/GrAAHairLinePathRenderer.cpp:168: // Most cases will result in etiher 1
conic (chop point is not within t range)
Typos

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
src/gpu/GrAAHairLinePathRenderer.cpp:721: * When comparing the second order
approximations to the first order the following
More typos.

[bsalomon, 2013-07-31 18:48:33]

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
File src/gpu/GrAAHairLinePathRenderer.cpp (right):

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
src/gpu/GrAAHairLinePathRenderer.cpp:1116: static const uint32_t gReqDerivMask =
SkPath::kCubic_SegmentMask |
I think this needs to be updated. Perhaps we should invert the logic.

[egdaniel, 2013-07-31 19:58:32]

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
File src/gpu/GrAAHairLinePathRenderer.cpp (right):

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
src/gpu/GrAAHairLinePathRenderer.cpp:1116: static const uint32_t gReqDerivMask =
SkPath::kCubic_SegmentMask |
On 2013/07/31 18:48:33, bsalomon wrote:
> I think this needs to be updated. Perhaps we should invert the logic.

So we definitely needed to add the Conic mask (which I've done). As for
inverting are you referring to the if statement and making it:
if(hasDervSupport || !(needDervSupport))
?

[egdaniel, 2013-07-31 19:58:32]

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
File src/gpu/GrAAHairLinePathRenderer.cpp (right):

https://codereview.chromium.org/21036008/diff/1/src/gpu/GrAAHairLinePathRende...
src/gpu/GrAAHairLinePathRenderer.cpp:1116: static const uint32_t gReqDerivMask =
SkPath::kCubic_SegmentMask |
On 2013/07/31 18:48:33, bsalomon wrote:
> I think this needs to be updated. Perhaps we should invert the logic.

So we definitely needed to add the Conic mask (which I've done). As for
inverting are you referring to the if statement and making it:
if(hasDervSupport || !(needDervSupport))
?

[egdaniel, 2013-08-01 13:52:50]

Changed the canDrawPath to look specifically for if line or has deriv support.

Hopefully fixed all the typos, which as a side note has made me discover vim's
built in spell checker which very well my be life changing for me.

[bsalomon, 2013-08-01 13:58:48]

lgtm

I had a spell checker installed in VS 2010 and realized that I can neither type
nor spell. Actually Rob already made that pretty clear. :)

[egdaniel, 2013-08-01 17:09:19]

Committed patchset #3 manually as r10487 (presubmit successful).