It's likely that my change is to blame, but 10% seems a bit much, since it doesn't actually encode any more data.  I can take a look and see if I can figure it out.

Looks like the code is written like malloc(x) is an O(1) operation, which perhaps it should be, but in chrome there is a memset() in free() which means that allocating more memory takes more time.

I also noticed that kLargestQuadSize is used in two different contexts, once to mean "largest possible serialized quad" and once to mean "sizeof(largest quad)". Since those two values are now two different things, I've made two different functions, which should put performance very nearly back to what it was before.

https://codereview.chromium.org/2324723002

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e099ceca4f5145bf650a2293807c585cc717121f

commit e099ceca4f5145bf650a2293807c585cc717121f
Author: hubbe <hubbe@chromium.org>
Date: Fri Sep 09 18:36:33 2016

Fix kLargestDrawQuadSize

This should work, but triggered a bug in the past.
Trying again, if nothing else I'll find out what problem was triggered.

BUG= 644443 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel

Review-Url: https://codereview.chromium.org/2328433005
Cr-Commit-Position: refs/heads/master@{#417646}

[modify] https://crrev.com/e099ceca4f5145bf650a2293807c585cc717121f/cc/quads/draw_quad_unittest.cc
[modify] https://crrev.com/e099ceca4f5145bf650a2293807c585cc717121f/cc/quads/largest_draw_quad.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/208f4013ba1cab8f188beb508915c49abe5ebdf7

commit 208f4013ba1cab8f188beb508915c49abe5ebdf7
Author: dmurph <dmurph@chromium.org>
Date: Fri Sep 09 21:02:01 2016

Revert of Fix kLargestDrawQuadSize (patchset #1 id:1 of https://codereview.chromium.org/2328433005/ )

Reason for revert:
Reverting, I think this caused buffer size issues
https://bugs.chromium.org/p/chromium/issues/detail?id=645634
BUG= 645634 

Original issue's description:
> Fix kLargestDrawQuadSize
>
> This should work, but triggered a bug in the past.
> Trying again, if nothing else I'll find out what problem was triggered.
>
> BUG= 644443 
> CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel
>
> Committed: https://crrev.com/e099ceca4f5145bf650a2293807c585cc717121f
> Cr-Commit-Position: refs/heads/master@{#417646}

TBR=danakj@chromium.org,hubbe@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG= 644443 

Review-Url: https://codereview.chromium.org/2326153002
Cr-Commit-Position: refs/heads/master@{#417708}

[modify] https://crrev.com/208f4013ba1cab8f188beb508915c49abe5ebdf7/cc/quads/draw_quad_unittest.cc
[modify] https://crrev.com/208f4013ba1cab8f188beb508915c49abe5ebdf7/cc/quads/largest_draw_quad.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/cbeebcb824092900816b2b35901fe42b6ff7f7fc

commit cbeebcb824092900816b2b35901fe42b6ff7f7fc
Author: hubbe <hubbe@chromium.org>
Date: Sat Sep 10 02:25:32 2016

Fix kLargestDrawQuadSize

This should be the right fix.
The problem was that two of the classes were not listed in the
static asserts since they were already in the sizeof at the top, which
I missed when I changed it to YUVVideoDrawQuad.

BUG= 644443 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel

Review-Url: https://codereview.chromium.org/2324083003
Cr-Commit-Position: refs/heads/master@{#417802}

[modify] https://crrev.com/cbeebcb824092900816b2b35901fe42b6ff7f7fc/cc/quads/draw_quad_unittest.cc
[modify] https://crrev.com/cbeebcb824092900816b2b35901fe42b6ff7f7fc/cc/quads/largest_draw_quad.cc