cc: Add tiling raster tile iterator construction perftest.

cc: Add tiling raster tile iterator construction perftest.

This test measures how expensive it is to simply create a tiling raster
tile iterator.

Initial run from a macbook pro:
[ RUN      ] PictureLayerTilingPerfTest.TilingRasterTileIteratorConstruction
*RESULT tiling_raster_tile_iterator_construction: 0_0_100x100= 5843577 runs/s
*RESULT tiling_raster_tile_iterator_construction: 50_0_100x100= 5940125 runs/s
*RESULT tiling_raster_tile_iterator_construction: 100_0_100x100= 3772066.25 runs/s
*RESULT tiling_raster_tile_iterator_construction: 150_0_100x100= 3712790 runs/s
[       OK ] PictureLayerTilingPerfTest.TilingRasterTileIteratorConstruction (12713 ms)

We can see here that there is some cost (albeit small) associated
with the construction. If the viewport is off the layer bounds,
then we pay more to figure out where the first tile is.

This cost is paid for each tiling iterator that is created (typically two
per layer: high res and low res).

NOTE TO PERF SHERIFFS
--------------------
This patch removes tiling_raster_tile_iterator test and replaces it
it tiling_raster_tile_iterator_construct_and_iterate as well as
tiling_raster_tile_iterator_construct test.

R=reveman

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=283850

[vmpstr, 2014-07-16 17:09:21]

PTAL

[reveman, 2014-07-16 22:09:03]

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:152: it && count;
whatever we decide is the best approach for the eviction perf test, we should do
the same here.

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:159:
perf_test::PrintResult("tiling_raster_tile_iterator",
this should be renamed to make it explicit that it also includes construction
cost now that we have separate test for that

[vmpstr, 2014-07-17 00:17:41]

PTAL

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:152: it && count;
On 2014/07/16 22:09:03, reveman wrote:
> whatever we decide is the best approach for the eviction perf test, we should
do
> the same here.

Done.

https://codereview.chromium.org/398073002/diff/20001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:159:
perf_test::PrintResult("tiling_raster_tile_iterator",
On 2014/07/16 22:09:03, reveman wrote:
> this should be renamed to make it explicit that it also includes construction
> cost now that we have separate test for that

Done.

[reveman, 2014-07-17 16:16:15]

lgtm with nit

https://codereview.chromium.org/398073002/diff/40001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/40001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:153: for (; it && count; ++it) {
nit: maybe a bit easier to read if you just use one variable to determine when
to stop iterating. ie.

while (count--) {
  ASSERT_TRUE(it);
  ++it;
}

also consider adding a ASSERT_TRUE(*it != NULL) to the loop to ensure that the
test stays valid if we ever make bool() operator cheaper than actually getting
the Tile.

[vmpstr, 2014-07-17 16:33:36]

https://codereview.chromium.org/398073002/diff/40001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/40001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:153: for (; it && count; ++it) {
On 2014/07/17 16:16:15, reveman wrote:
> nit: maybe a bit easier to read if you just use one variable to determine when
> to stop iterating. ie.
> 
> while (count--) {
>   ASSERT_TRUE(it);
>   ++it;
> }
> 
> also consider adding a ASSERT_TRUE(*it != NULL) to the loop to ensure that the
> test stays valid if we ever make bool() operator cheaper than actually getting
> the Tile.

Done.

[vmpstr, 2014-07-17 16:33:42]

The CQ bit was checked by vmpstr@chromium.org

[commit-bot: I haz the power, 2014-07-17 16:36:08]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/vmpstr@chromium.org/398073002/60001

[reveman, 2014-07-17 16:39:26]

https://codereview.chromium.org/398073002/diff/60001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/60001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:157: ASSERT_EQ(-1, count);
don't think you need this. it would be pretty hard to get here without this
being the case :)

[vmpstr, 2014-07-17 17:14:52]

The CQ bit was unchecked by vmpstr@chromium.org

[vmpstr, 2014-07-17 17:16:25]

https://codereview.chromium.org/398073002/diff/60001/cc/resources/picture_lay...
File cc/resources/picture_layer_tiling_perftest.cc (right):

https://codereview.chromium.org/398073002/diff/60001/cc/resources/picture_lay...
cc/resources/picture_layer_tiling_perftest.cc:157: ASSERT_EQ(-1, count);
On 2014/07/17 16:39:26, reveman wrote:
> don't think you need this. it would be pretty hard to get here without this
> being the case :)

Hehe, I'm just being overzealous. Removed.

[vmpstr, 2014-07-17 17:16:29]

The CQ bit was checked by vmpstr@chromium.org

[commit-bot: I haz the power, 2014-07-17 17:18:34]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/vmpstr@chromium.org/398073002/80001

[commit-bot: I haz the power, 2014-07-17 20:12:06]

Change committed as 283850