Don't unlink evacuation candidates before sweeping, move them to the end of their list of pages.

Don't unlink evacuation candidates before sweeping, move them to the end of their list of pages.

BUG=
R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=21042

[Michael Starzinger, 2014-04-29 10:23:57]

LGTM with comments.

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.cc#newc...
src/mark-compact.cc:3711: 
nit: Two lines of white-space ought'a be enough for everybody.

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h
File src/mark-compact.h (right):

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h#newco...
src/mark-compact.h:730: void UnlinkEvacuationCandidates();
nit: The unlink method is gone, let's remove the declaration.

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h#newco...
src/mark-compact.h:731: void ReleaseEvacuationCandidates();
nit: The release method could be moved to near the others for readability.

https://codereview.chromium.org/256743004/diff/100001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/256743004/diff/100001/src/mark-compact.cc#new...
src/mark-compact.cc:4157: space->ClearStats();
As discussed offline: Can we play it safe and set the "end_of_unswept_pages" to
a safe value here? I agree that the first page should always be swept (and not
evacuated), but being defensive is better.

https://codereview.chromium.org/256743004/diff/100001/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/256743004/diff/100001/src/spaces.h#newcode1975
src/spaces.h:1975: Page* end_of_unswept_pages_;
Can we have a comment explaining the semantics of this field?

[Hannes Payer (out of office), 2014-04-29 10:59:26]

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.cc#newc...
src/mark-compact.cc:3711: 
On 2014/04/29 10:23:57, Michael Starzinger wrote:
> nit: Two lines of white-space ought'a be enough for everybody.

Done.

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h
File src/mark-compact.h (right):

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h#newco...
src/mark-compact.h:730: void UnlinkEvacuationCandidates();
On 2014/04/29 10:23:57, Michael Starzinger wrote:
> nit: The unlink method is gone, let's remove the declaration.

Done.

https://codereview.chromium.org/256743004/diff/60001/src/mark-compact.h#newco...
src/mark-compact.h:731: void ReleaseEvacuationCandidates();
On 2014/04/29 10:23:57, Michael Starzinger wrote:
> nit: The release method could be moved to near the others for readability.

Done.

https://codereview.chromium.org/256743004/diff/100001/src/mark-compact.cc
File src/mark-compact.cc (right):

https://codereview.chromium.org/256743004/diff/100001/src/mark-compact.cc#new...
src/mark-compact.cc:4157: space->ClearStats();
On 2014/04/29 10:23:57, Michael Starzinger wrote:
> As discussed offline: Can we play it safe and set the "end_of_unswept_pages"
to
> a safe value here? I agree that the first page should always be swept (and not
> evacuated), but being defensive is better.

Done.

https://codereview.chromium.org/256743004/diff/100001/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/256743004/diff/100001/src/spaces.h#newcode1975
src/spaces.h:1975: Page* end_of_unswept_pages_;
On 2014/04/29 10:23:57, Michael Starzinger wrote:
> Can we have a comment explaining the semantics of this field?

Done.

[Hannes Payer (out of office), 2014-04-29 11:01:56]

Committed patchset #7 manually as r21042 (presubmit successful).