cc: Add drawing for testing raster perf tests.

cc: Add drawing for testing raster perf tests.

Current code does not add drawing, so it is difficult to assess the
raster speed for the given platform. New tests are added for checking
raster performance when there is drawng. This would help in assessing
actual raster speeds.

BUG=500624
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

[prashant.n, 2015-07-15 15:26:53]

prashant.n@samsung.com changed reviewers:
+ danakj@chromium.org, enne@chromium.org

[prashant.n, 2015-07-15 15:26:54]

PTAL.

[enne (OOO), 2015-07-15 17:36:13]

In a general sense, I am not sure this needs to be committed.  It seems like
something that could be run to generate some numbers and we can move on from
there.

Also, it was my impression that slower tile sizes had to do a lot with Skia
needing to reraster expensive things (filters, save layers, etc) that crossed
clip boundaries.  I can wait for data, but I would suspect that for software
raster this simple test will probably take time proportional to the number of
pixels drawn.

How are you going to look at hardware raster times?

[prashant.n, 2015-07-23 14:06:37]

On 2015/07/15 17:36:13, enne wrote:
> In a general sense, I am not sure this needs to be committed.  It seems like
> something that could be run to generate some numbers and we can move on from
> there.
> 


1. I think we should have these tests as current tests check raster speeds on no

drawing.

2. My new tests which will have different tile sizes and need this basic
modifications. So I pushed this patch.

Kindly do let me know your opinion.


> Also, it was my impression that slower tile sizes had to do a lot with Skia
> needing to reraster expensive things (filters, save layers, etc) that crossed
> clip boundaries.  I can wait for data, but I would suspect that for software
> raster this simple test will probably take time proportional to the number of
> pixels drawn.


Yes. lower tile sizes will get hit from Skia needing to reraster expensive
things. I'll upload the data soon.

 
> How are you going to look at hardware raster times?

Hardware raster times not yet checked.

[enne (OOO), 2015-07-23 17:47:59]

What do the numbers look like?