Allow main thread to contribute to the sweeping phase.

Allow main thread to contribute to the sweeping phase.

BUG=
R=jarin@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=22318

[Jarin, 2014-07-09 11:02:03]

lgtm with a bunch of non-blocking comments.

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc#newc...
src/mark-compact.cc:4069: // Force instantiation of templatized SweepInParallel
method for
Is the templatization of SweepInParallel really necessary? Would not it just
work if you always called with SWEEP_IN_PARALLEL mode?

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc#newc...
src/mark-compact.cc:4110: return max_freed;
For consistency, should not this return the maximum? (Rather than the last value
of returned by the SweepConservatively.)

https://codereview.chromium.org/380653003/diff/40001/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/380653003/diff/40001/src/spaces.cc#newcode2584
src/spaces.cc:2584: HeapObject*
PagedSpace::WaitForSweeperThreadsAndRetryAllocation(
Nit: Now the name is a bit funny - it does not always wait for the sweeper
threads.

[Hannes Payer (out of office), 2014-07-10 12:03:46]

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc#newc...
src/mark-compact.cc:4069: // Force instantiation of templatized SweepInParallel
method for
On 2014/07/09 11:02:02, jarin wrote:
> Is the templatization of SweepInParallel really necessary? Would not it just
> work if you always called with SWEEP_IN_PARALLEL mode?

Done.

It would make sense to remove the template for SweepConservatively. However,
this is a major refactoring of the sweeping code. I will try to do this with a
foll-up cl. I basically have to move all the sweeping modes over to use the lazy
mechanism for bookkeeping instead of eagerly bookkeeping in the non-concurrent
case.

https://codereview.chromium.org/380653003/diff/40001/src/mark-compact.cc#newc...
src/mark-compact.cc:4110: return max_freed;
On 2014/07/09 11:02:02, jarin wrote:
> For consistency, should not this return the maximum? (Rather than the last
value
> of returned by the SweepConservatively.)

Done.

https://codereview.chromium.org/380653003/diff/40001/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/380653003/diff/40001/src/spaces.cc#newcode2584
src/spaces.cc:2584: HeapObject*
PagedSpace::WaitForSweeperThreadsAndRetryAllocation(
On 2014/07/09 11:02:02, jarin wrote:
> Nit: Now the name is a bit funny - it does not always wait for the sweeper
> threads.

Done.

[Hannes Payer (out of office), 2014-07-10 12:07:36]

Committed patchset #4 manually as r22318 (presubmit successful).

[Jarin, 2014-07-10 12:12:20]

Love it, thanks. Small nit (for future work).

https://codereview.chromium.org/380653003/diff/60001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/380653003/diff/60001/src/mark-compact.cc#newc...
src/mark-compact.cc:4077: int max_freed = 0;
Nit: you could remove this line and declare the int type on line 4083.