Specialize GrConvexPolyEffect for AA rects, use for AA clip rects.

Specialize GrConvexPolyEffect for AA rects, use for AA clip rects.

BUG=skia:2051
Committed: http://code.google.com/p/skia/source/detail?r=13375

[robertphillips, 2014-02-04 19:38:50]

lgtm + nits, questions & suggestions

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp
File gm/convexpolyeffect.cpp (right):

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:91: // half-integer edges
Seems like second # should be 1.5

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:99: // small in x and y
Seems like this the last # should be .95

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:101: // inverted in x and y
Seems like first number should be 100.5

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:217: 
// Draw AA rect using normal API for reference?

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.cpp (right):

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:100: const GrDrawEffect& drawEffect)
tab here?

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:119: // respectively.
Maybe a more expansive comment about how coverage is being computed?

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:133: const SkRect& rect = aare.getRect();
Do you think this caching will actually pay off?

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:270: } else {
It seems like we would never want to take this path?

[bsalomon, 2014-02-08 19:18:01]

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp
File gm/convexpolyeffect.cpp (right):

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:91: // half-integer edges
On 2014/02/04 19:38:50, robertphillips wrote:
> Seems like second # should be 1.5
I don't really understand any of these comments about the rect edges. The exact
numbers are arbitrary. I just wanted to catch the 50% case, cases where the
pixel center isn't covered, a subpixel case in x and y that missed the pixel
center, and an inverted rectangle.

https://codereview.chromium.org/140093004/diff/60002/gm/convexpolyeffect.cpp#...
gm/convexpolyeffect.cpp:217: 
On 2014/02/04 19:38:50, robertphillips wrote:
> // Draw AA rect using normal API for reference?

Done.

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
File src/gpu/effects/GrConvexPolyEffect.cpp (right):

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:100: const GrDrawEffect& drawEffect)
On 2014/02/04 19:38:50, robertphillips wrote:
> tab here?

Done.

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:119: // respectively.
On 2014/02/04 19:38:50, robertphillips wrote:
> Maybe a more expansive comment about how coverage is being computed?

Done.

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:133: const SkRect& rect = aare.getRect();
On 2014/02/04 19:38:50, robertphillips wrote:
> Do you think this caching will actually pay off?

We can have the same clip active for a lot of draws and I'm concerned about
flooding uniform uploads, esp on the command buffer. Maybe it doesn't matter
much, but I don't think it hurts.

https://codereview.chromium.org/140093004/diff/60002/src/gpu/effects/GrConvex...
src/gpu/effects/GrConvexPolyEffect.cpp:270: } else {
On 2014/02/04 19:38:50, robertphillips wrote:
> It seems like we would never want to take this path?

True. I changed the function to CreateAAFillRect and removed the type param.

[bsalomon, 2014-02-08 19:18:05]

The CQ bit was checked by bsalomon@google.com

[commit-bot: I haz the power, 2014-02-08 19:18:11]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/bsalomon@google.com/140093004/120001

[commit-bot: I haz the power, 2014-02-08 19:31:08]

Change committed as 13375