[heap] Count bits in markbit cell instead of using a table.

[heap] Count bits in markbit cell instead of using a table.

Iterating live objects using mark bits is memory bound and already pretty heavy
on the cache as we always need to look at the mark bits, the objects (payload),
and their maps.

This CL changes the object start computation from a table lookup, which was yet
another memory location, to a CPU-bound computation.

BUG=chromium:524425
LOG=N

Committed: https://crrev.com/206ba3124573df5a530779af721293853711ad67
Cr-Commit-Position: refs/heads/master@{#32291}

[Michael Lippautz, 2015-11-25 14:30:52]

Patchset #1 (id:1) has been deleted

[Michael Lippautz, 2015-11-25 14:36:54]

mlippautz@chromium.org changed reviewers:
+ hpayer@chromium.org

[Michael Lippautz, 2015-11-25 14:36:54]

Should be supported starting from armv5
  https://en.wikipedia.org/wiki/Find_first_set

Slightly improves the average pause time (~8%) for compacting single pages.
Also, reduces the flakes for those compactions, i.e., we don't see compaction of
pages with <100k live bytes taking up 2ms anymore.

[Michael Lippautz, 2015-11-25 14:44:29]

Description was changed from

==========
[heap] Count bits in markbit cell instead of using a table.

Iterating live objects using mark bits is memory bound and already pretty heavy
on the cache as we always need to look at the mark bits, the objects (payload),
and their maps.

This CL changes the object start computation from a table, which was yet another
memory location, to a CPU-bound computation.

BUG=chromium:524425
LOG=N
==========

to

==========
[heap] Count bits in markbit cell instead of using a table.

Iterating live objects using mark bits is memory bound and already pretty heavy
on the cache as we always need to look at the mark bits, the objects (payload),
and their maps.

This CL changes the object start computation from a table lookup, which was yet
another
memory location, to a CPU-bound computation.

BUG=chromium:524425
LOG=N
==========

[Hannes Payer (out of office), 2015-11-25 16:33:23]

lgtm

[Michael Lippautz, 2015-11-25 16:33:34]

The CQ bit was checked by mlippautz@chromium.org

[commit-bot: I haz the power, 2015-11-25 16:33:51]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1478623003/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1478623003/20001

[commit-bot: I haz the power, 2015-11-25 17:31:31]

Committed patchset #1 (id:20001)

[commit-bot: I haz the power, 2015-11-25 17:32:15]

Patchset 1 (id:??) landed as
https://crrev.com/206ba3124573df5a530779af721293853711ad67
Cr-Commit-Position: refs/heads/master@{#32291}