Add a new bench that tests a simple image filter DAG.

Add a new bench that tests a simple image filter DAG.

This bench exercises a blur filter connected to 5 inputs of the same
merge filter.

R=junov@chromium.org

Committed: https://code.google.com/p/skia/source/detail?r=14135

[Stephen White, 2014-04-10 15:24:36]

reed@: PTAL. Thanks!

[Stephen White, 2014-04-10 15:37:21]

Never mind; I'll bug junov@ instead since this doesn't have any API changes.

junov@: PTAL. :)

[Justin Novosad, 2014-04-10 15:51:28]

lgtm with nits.

I get what is being tested here because we talked about it yesterday, but it
would be a good idea to explain it in a comment.  The concept that there is a
caching opportunity when the output of a node is wired to multiple inputs need
to be stated.

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cpp
File bench/ImageFilterDAGBench.cpp (right):

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cp...
bench/ImageFilterDAGBench.cpp:13: #define NUM_COPIES 5
skia style: enum { kNumCopies = 5 };

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cp...
bench/ImageFilterDAGBench.cpp:29: inputs[i] = blur.get();
num copies is a misleading name.  You are not making a copy of blur. You are
just connecting it to multiple inputs.  num inputs?

[Stephen White, 2014-04-10 16:13:44]

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cpp
File bench/ImageFilterDAGBench.cpp (right):

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cp...
bench/ImageFilterDAGBench.cpp:13: #define NUM_COPIES 5
On 2014/04/10 15:51:28, junov wrote:
> skia style: enum { kNumCopies = 5 };

Done.

https://codereview.chromium.org/233383002/diff/1/bench/ImageFilterDAGBench.cp...
bench/ImageFilterDAGBench.cpp:29: inputs[i] = blur.get();
On 2014/04/10 15:51:28, junov wrote:
> num copies is a misleading name.  You are not making a copy of blur. You are
> just connecting it to multiple inputs.  num inputs?

Done.

[Stephen White, 2014-04-10 16:16:18]

Committed patchset #3 manually as r14135 (presubmit successful).