cc: Optimize the tile map key hash for 4-byte size_t.

cc: Optimize the tile map key hash for 4-byte size_t.

This patch changes the hash function for the tile map key to use
16 least significant bits from each index. This makes it faster on
systems where sizeof(size_t) == 4.

The collisions would only start happening after a large enough index.

This makes the TilingSetRasterQueueConstructAndIterate test do
about 10% more iterations per second on N4.

BUG=488636
R=enne, danakj

Committed: https://crrev.com/1dfbd1030bd04e76ba41bff933cd2d8fc6d1e3f0
Cr-Commit-Position: refs/heads/master@{#330448}

[vmpstr, 2015-05-18 20:28:59]

Please take a look.

[enne (OOO), 2015-05-18 21:14:17]

lgtm % danakj hash expert

Also, index_x and index_y should really be index_i and index_j if you want to be
consistent with the rest of the code.

[danakj, 2015-05-18 21:16:54]

I think this is fine, tiles fill some contiguous space within the 32bit integer
space. If there are tiles at positions > 2^16, there are not going to be tiles
with the same lower 16 bits at other positions mod 2^16. And if there was,
because of the contiguous set of tiles, you'd have tiles all in between, you
would have filled the hash key namespace already, and the collisions would be
"well distributed" ie you would only get a 2nd collision once all the keys got
their 1st collision. So this seems pretty well behaved. LGTM

https://codereview.chromium.org/1133243010/diff/1/cc/resources/picture_layer_...
File cc/resources/picture_layer_tiling.h (right):

https://codereview.chromium.org/1133243010/diff/1/cc/resources/picture_layer_...
cc/resources/picture_layer_tiling.h:56: TileMapKey(int i, int j) : index_x(i),
index_y(j) {}
x,y consistently?

[vmpstr, 2015-05-18 21:16:58]

On 2015/05/18 21:14:17, enne wrote:
> lgtm % danakj hash expert
> 
> Also, index_x and index_y should really be index_i and index_j if you want to
be
> consistent with the rest of the code.

I was making it consistent with TilingData::Iterator::index_x() and
...::index_y().
I was actually planning on changing i/j to x/y, if that's cool with everyone...
Wdyt?

[vmpstr, 2015-05-18 21:20:38]

https://codereview.chromium.org/1133243010/diff/1/cc/resources/picture_layer_...
File cc/resources/picture_layer_tiling.h (right):

https://codereview.chromium.org/1133243010/diff/1/cc/resources/picture_layer_...
cc/resources/picture_layer_tiling.h:56: TileMapKey(int i, int j) : index_x(i),
index_y(j) {}
On 2015/05/18 21:16:54, danakj wrote:
> x,y consistently?

Done.

[enne (OOO), 2015-05-18 21:27:53]

On 2015/05/18 at 21:16:58, vmpstr wrote:
> On 2015/05/18 21:14:17, enne wrote:
> > lgtm % danakj hash expert
> > 
> > Also, index_x and index_y should really be index_i and index_j if you want
to be
> > consistent with the rest of the code.
> 
> I was making it consistent with TilingData::Iterator::index_x() and
...::index_y().
> I was actually planning on changing i/j to x/y, if that's cool with
everyone...
> Wdyt?

Sure, whatever.  Consistency is the important part.

[vmpstr, 2015-05-18 21:29:40]

The CQ bit was checked by vmpstr@chromium.org

[vmpstr, 2015-05-18 21:29:41]

The patchset sent to the CQ was uploaded after l-g-t-m from danakj@chromium.org,
enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1133243010/#ps40001
(title: "rebase")

[commit-bot: I haz the power, 2015-05-18 21:30:41]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1133243010/40001

[commit-bot: I haz the power, 2015-05-18 21:45:45]

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

[commit-bot: I haz the power, 2015-05-18 21:45:47]

Try jobs failed on following builders:
  linux_chromium_gn_rel on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[vmpstr, 2015-05-18 22:15:36]

The CQ bit was checked by vmpstr@chromium.org

[vmpstr, 2015-05-18 22:15:37]

The patchset sent to the CQ was uploaded after l-g-t-m from danakj@chromium.org,
enne@chromium.org
Link to the patchset: https://codereview.chromium.org/1133243010/#ps60001
(title: "fixes")

[commit-bot: I haz the power, 2015-05-18 22:16:25]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1133243010/60001

[commit-bot: I haz the power, 2015-05-18 23:08:00]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-05-18 23:09:48]

Patchset 4 (id:??) landed as
https://crrev.com/1dfbd1030bd04e76ba41bff933cd2d8fc6d1e3f0
Cr-Commit-Position: refs/heads/master@{#330448}