use Sk2s for EvalQuadTangent and ChopQuadAt

use Sk2s for EvalQuadTangent and ChopQuadAt

abandoned in favor of https://codereview.chromium.org/1024873003/

[reed2, 2015-03-20 03:21:15]

The CQ bit was checked by reed@chromium.org

[reed2, 2015-03-20 03:21:28]

reed@chromium.org changed reviewers:
+ caryclark@google.com, mtklein@google.com

[commit-bot: I haz the power, 2015-03-20 03:21:31]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1026633002/1

[commit-bot: I haz the power, 2015-03-20 03:21:32]

Note for Reviewers:
The CQ is waiting for an approval.
If you believe that the CL is not ready yet, or if you would like to L-G-T-M
with comments then please uncheck the CQ checkbox.

Waiting for LGTM from valid reviewer(s) till 2015-03-20 09:21 UTC

[commit-bot: I haz the power, 2015-03-20 03:25:38]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-03-20 03:25:38]

Try jobs failed on following builders:
  Build-Win-VS2013-x86-Debug-Trybot on client.skia.compile (JOB_FAILED,
http://build.chromium.org/p/client.skia.compile/builders/Build-Win-VS2013-x86...)

[mtklein, 2015-03-20 11:39:48]

https://codereview.chromium.org/1026633002/diff/1/bench/GeometryBench.cpp
File bench/GeometryBench.cpp (right):

https://codereview.chromium.org/1026633002/diff/1/bench/GeometryBench.cpp#new...
bench/GeometryBench.cpp:188: SkEvalQuadAt(fPts, 0.5f, NULL, &result);
Is this intentional, comparing SkEvalQuadTangentAt against SkEvalQuadAt?

https://codereview.chromium.org/1026633002/diff/1/src/core/SkGeometry.cpp
File src/core/SkGeometry.cpp (right):

https://codereview.chromium.org/1026633002/diff/1/src/core/SkGeometry.cpp#new...
src/core/SkGeometry.cpp:185: static Sk2s interp(Sk2s A, Sk2s B, Sk2s T) {
Interesting.

The x86 Windows warnings are telling us it has no way to guarantee these
parameters will be passed with 16-byte alignment.  The x86-64 calling convention
requires these to pass by xmm registers, which is perfect, and other 32-bit
platforms probably 16-byte align the stack, which is fine, but here MSVC is
throwing up its hands and saying it can do neither.

Annoyingly, I think we have to write

static Sk2s interp(const Sk2s& A, const Sk2s& B, const Sk2s& T) {
   return A + (B - A) * T;
}

https://codereview.chromium.org/1026633002/diff/1/src/core/SkGeometry.cpp#new...
src/core/SkGeometry.cpp:192: Sk2s P0 = Sk2f::Load(&src[0].fX);
You might call these A,B,C to match AB and BC.

[mtklein, 2015-03-20 11:42:03]

I'm seeing all them at about an ideal speedup:

maxrss	loops	min	median	mean	max	stddev	samples   	config	bench
  10M	657	11.4ns	11.4ns	11.4ns	11.5ns	0%	█▂▂▃▂▂▂▁▂▁	nonrendering	geo_chopquadat1
  10M	490	19.5ns	19.5ns	19.5ns	19.5ns	0%	▅▅▁▇█▂▃█▇█	nonrendering	geo_chopquadat0
 
10M	1748	6.91ns	7.13ns	8.57ns	12.2ns	29%	▁▁█▁▁█▁█▁▁	nonrendering	geo_evalquadtangentat1
 
10M	807	14.5ns	14.5ns	14.5ns	14.5ns	0%	█▁▁▂▁▄▁▁▁▁	nonrendering	geo_evalquadtangentat0
  10M	2652	7.97ns	8ns	7.99ns	8.01ns	0%	▄█▂▇▂▆▁█▂█	nonrendering	geo_evalquadat1
 
10M	2240	15.8ns	15.9ns	15.9ns	15.9ns	0%	█▄▅▁▁▆▅▅▅▅	nonrendering	geo_evalquadat0

You too?

[reed1, 2015-03-20 12:32:03]

reed@google.com changed reviewers:
+ reed@google.com

[reed1, 2015-03-20 12:32:03]

https://codereview.chromium.org/1026633002/diff/1/bench/GeometryBench.cpp
File bench/GeometryBench.cpp (right):

https://codereview.chromium.org/1026633002/diff/1/bench/GeometryBench.cpp#new...
bench/GeometryBench.cpp:188: SkEvalQuadAt(fPts, 0.5f, NULL, &result);
On 2015/03/20 11:39:48, mtklein wrote:
> Is this intentional, comparing SkEvalQuadTangentAt against SkEvalQuadAt?

The NULL is the quad position parameter, the next param is the tangent (EvalQuad
does double duty).