Oval and stroke AA rect now batch

Oval and stroke AA rect now batch

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/bc54fab1a4e5b51628a8c1557c62bc74e078870d

Committed: https://skia.googlesource.com/skia/+/18055afb838a278b5a8436cd51dbfbb688e1e0a0

Committed: https://skia.googlesource.com/skia/+/5ead6da4abaa3d8fb479fd84631138f83b7aa5b6

[joshualitt, 2014-10-20 19:25:53]

joshualitt@google.com changed reviewers:
+ joshualitt@google.com, robertphillips@google.com

[joshualitt, 2014-10-21 18:49:19]

On 2014/10/20 19:25:53, joshualitt wrote:

Some perf benchmarks, there were a few others which seemed to show an
improvement but for simplicity:
TOT
  98M	3781	1.33µs	1.33µs	1.33µs	1.34µs	0%	▇▂▄▂▄▅█▁▄▂	gpu	draw_stroke_oval_miter	
  98M	3743	1.33µs	1.34µs	1.33µs	1.34µs	0%	▄▆▃█▁▇▁█▇▃	gpu	draw_stroke_oval_bevel	
  98M	3765	1.33µs	1.33µs	1.33µs	1.34µs	0%	▁██▁▅▆▃▃▄▃	gpu	draw_stroke_oval_round	
  98M	3095	1.58µs	1.61µs	1.6µs	1.62µs	1%	▅▅█▆▆▃▁▅▇▆	gpu	draw_stroke_rrect_miter	
 
98M	3091	1.62µs	1.62µs	1.62µs	1.63µs	0%	▅█▅▄▃▁▁▁▁▃	gpu	draw_stroke_rrect_bevel	
 
98M	3109	1.56µs	1.58µs	1.58µs	1.63µs	1%	█▄▃▂▃▃▁▃▃▃	gpu	draw_stroke_rrect_round	
  98M	3644	1.35µs	1.39µs	1.38µs	1.42µs	2%	█▆▄▂▇▂▆▁▁▂	gpu	draw_stroke_rect_miter	
  98M	3319	1.47µs	1.49µs	1.48µs	1.5µs	1%	▄▁▄▂▄▂█▄▅▃	gpu	draw_stroke_rect_bevel	

New code
  98M	8427	597ns	598ns	599ns	606ns	0%	▂▃▂▂▂▁▁▂█▁	gpu	draw_stroke_oval_miter
  98M	8405	599ns	601ns	601ns	607ns	0%	▁▃▂▂█▃▁▁▃▄	gpu	draw_stroke_oval_bevel
  98M	8398	597ns	599ns	599ns	603ns	0%	▄▁▃▃▁▂▄▄█▂	gpu	draw_stroke_oval_round
  98M	5706	871ns	886ns	883ns	894ns	1%	▃▂▃▆█▂▆▇█▁	gpu	draw_stroke_rrect_miter
  98M	5761	867ns	873ns	873ns	890ns	1%	▃▁▃▂▁▁▄▁█▃	gpu	draw_stroke_rrect_bevel
  98M	5663	877ns	887ns	885ns	893ns	1%	█▃▃▆▅▁▂█▇▃	gpu	draw_stroke_rrect_round
  98M	6834	725ns	739ns	738ns	758ns	1%	▂▁█▅▄▄▃▁▄▄	gpu	draw_stroke_rect_miter
  98M	6020	810ns	822ns	821ns	829ns	1%	▅█▅▄▄▆▆▄█▁	gpu	draw_stroke_rect_bevel

[robertphillips, 2014-10-21 19:12:38]

lgtm. One somewhat minor concern I have is if the kNum*IndexBuffer numbers are
too large? It might make sense to churn through the skps and spit out the
maximum occupancy of the buffers.

https://codereview.chromium.org/664193002/diff/30001/src/gpu/GrGpu.cpp
File src/gpu/GrGpu.cpp (right):

https://codereview.chromium.org/664193002/diff/30001/src/gpu/GrGpu.cpp#newcod...
src/gpu/GrGpu.cpp:180: for (int i = 0; i < reps; ++i) {
Fix comment ?

[joshualitt, 2014-10-22 13:38:53]

joshualitt@google.com changed reviewers:
+ bsalomon@google.com

[joshualitt, 2014-10-22 13:38:54]

adding Brian to review.  Greg mentioned I might be able to add per vertex
colors.  I think that is a good idea.  Is there any reason I can't do that with
these shapes?

[bsalomon, 2014-10-22 17:51:50]

https://codereview.chromium.org/664193002/diff/30001/src/gpu/GrGpu.h
File src/gpu/GrGpu.h (right):

https://codereview.chromium.org/664193002/diff/30001/src/gpu/GrGpu.h#newcode146
src/gpu/GrGpu.h:146: GrIndexBuffer* createIndexBufferPattern(const uint16_t*
pattern,
I like this... but maybe we should call it createInstancedIndexBuffer or
something similar to tie its name to drawIndexedInstances?

Also, can we make the existing instance ibs use this?

[joshualitt, 2014-10-22 18:56:07]

feedback inc

[bsalomon, 2014-10-22 20:23:11]

lgtm with a couple nits below.

https://codereview.chromium.org/664193002/diff/90001/src/gpu/GrAAHairLinePath...
File src/gpu/GrAAHairLinePathRenderer.cpp (right):

https://codereview.chromium.org/664193002/diff/90001/src/gpu/GrAAHairLinePath...
src/gpu/GrAAHairLinePathRenderer.cpp:890: kLineSegNumVertices*lines,     //
startV
can you realign these comments?

https://codereview.chromium.org/664193002/diff/90001/src/gpu/GrGpu.cpp
File src/gpu/GrGpu.cpp (right):

https://codereview.chromium.org/664193002/diff/90001/src/gpu/GrGpu.cpp#newcod...
src/gpu/GrGpu.cpp:167: GrIndexBuffer* GrGpu::createInstancedIndexBuffer(const
uint16_t* pattern,
I take it that the const change is to support the quad index buffer? I think
that was a dubious call to make getQIB const.. can we confine the const cast to
that function?

https://codereview.chromium.org/664193002/diff/90001/src/gpu/GrGpu.cpp#newcod...
src/gpu/GrGpu.cpp:182: // Each AA filled rect is drawn with 8 vertices and 10
triangles (8 around
this comment should go

[joshualitt, 2014-10-22 20:57:26]

The CQ bit was checked by joshualitt@google.com

[commit-bot: I haz the power, 2014-10-22 20:59:02]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/664193002/110001

[joshualitt, 2014-10-22 20:59:13]

The CQ bit was unchecked by joshualitt@google.com

[joshualitt, 2014-10-22 21:40:55]

The CQ bit was checked by joshualitt@google.com

[commit-bot: I haz the power, 2014-10-22 21:41:31]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/664193002/110001

[commit-bot: I haz the power, 2014-10-22 21:41:50]

Committed patchset #7 (id:110001) as bc54fab1a4e5b51628a8c1557c62bc74e078870d

[mtklein, 2014-10-22 22:22:26]

A revert of this CL (patchset #7 id:110001) has been created in
https://codereview.chromium.org/671993003/ by mtklein@google.com.

The reason for reverting is: Many GMs failing, e.g.
http://build.chromium.org/p/client.skia.android/builders/Test-Android-Nexus5-....

[joshualitt, 2014-10-22 22:47:38]

The CQ bit was checked by joshualitt@google.com

[commit-bot: I haz the power, 2014-10-22 22:48:36]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/664193002/110001

[joshualitt, 2014-10-22 22:48:47]

The CQ bit was unchecked by joshualitt@google.com

[commit-bot: I haz the power, 2014-10-22 22:48:53]

Committed patchset #7 (id:110001) as 18055afb838a278b5a8436cd51dbfbb688e1e0a0

[joshualitt, 2014-10-22 22:51:26]

A revert of this CL (patchset #7 id:110001) has been created in
https://codereview.chromium.org/675623002/ by joshualitt@google.com.

The reason for reverting is: whoops, accidental commit.

[joshualitt, 2014-10-22 22:52:19]

The CQ bit was checked by joshualitt@google.com

[commit-bot: I haz the power, 2014-10-22 22:52:41]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/664193002/130001

[commit-bot: I haz the power, 2014-10-22 23:00:37]

Committed patchset #8 (id:130001) as 5ead6da4abaa3d8fb479fd84631138f83b7aa5b6