Use smi zero instead of undefine_value to zap dead weak cells.

Use smi zero instead of undefine_value to zap dead weak cells.

It is faster to test for smi zero from generated code.

BUG=
R=erikcorry@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=24758

[ulan, 2014-10-17 11:12:51]

ulan@chromium.org changed reviewers:
+ erikcorry@chromium.org

[ulan, 2014-10-17 11:12:52]

Hi Erik. Could you please take a look?

[Erik Corry Chromium.org, 2014-10-20 08:00:18]

LGTM

https://codereview.chromium.org/637253004/diff/1/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/637253004/diff/1/src/heap/mark-compact.cc#new...
src/heap/mark-compact.cc:2748: if (!MarkCompactCollector::IsMarked(value)) {
IsMarked will assert that the value is a heap object, so you should handle the
non-heap-object case here or make value a HeapObject pointer and use
HeapObject::cast() if you don't think it can happen.

https://codereview.chromium.org/637253004/diff/1/src/heap/objects-visiting-inl.h
File src/heap/objects-visiting-inl.h (right):

https://codereview.chromium.org/637253004/diff/1/src/heap/objects-visiting-in...
src/heap/objects-visiting-inl.h:363: if (weak_cell->next() == undefined &&
!weak_cell->cleared()) {
Perhaps this should just check for Sminess instead of clearedness, since there's
no point in queueing up a cell with a Smi in it.

https://codereview.chromium.org/637253004/diff/1/src/objects-inl.h
File src/objects-inl.h (right):

https://codereview.chromium.org/637253004/diff/1/src/objects-inl.h#newcode1938
src/objects-inl.h:1938: void WeakCell::clear() { WRITE_FIELD(this, kValueOffset,
Smi::FromInt(0)); }
If this should only be called by the GC, can we assert that we are in a GC here?

[ulan, 2014-10-20 12:30:19]

Thanks, landing.

https://codereview.chromium.org/637253004/diff/1/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/637253004/diff/1/src/heap/mark-compact.cc#new...
src/heap/mark-compact.cc:2748: if (!MarkCompactCollector::IsMarked(value)) {
On 2014/10/20 08:00:17, Erik Corry Chromium.org wrote:
> IsMarked will assert that the value is a heap object, so you should handle the
> non-heap-object case here or make value a HeapObject pointer and use
> HeapObject::cast() if you don't think it can happen.

Done.

https://codereview.chromium.org/637253004/diff/1/src/heap/objects-visiting-inl.h
File src/heap/objects-visiting-inl.h (right):

https://codereview.chromium.org/637253004/diff/1/src/heap/objects-visiting-in...
src/heap/objects-visiting-inl.h:363: if (weak_cell->next() == undefined &&
!weak_cell->cleared()) {
On 2014/10/20 08:00:18, Erik Corry Chromium.org wrote:
> Perhaps this should just check for Sminess instead of clearedness, since
there's
> no point in queueing up a cell with a Smi in it.

Since the initialization function of the weak cell accepts only heap objects,
weak_cell->value()->IsSmi() is equivalent to weak_cell->value ==
Smi::FromInt(0).

https://codereview.chromium.org/637253004/diff/1/src/objects-inl.h
File src/objects-inl.h (right):

https://codereview.chromium.org/637253004/diff/1/src/objects-inl.h#newcode1938
src/objects-inl.h:1938: void WeakCell::clear() { WRITE_FIELD(this, kValueOffset,
Smi::FromInt(0)); }
On 2014/10/20 08:00:18, Erik Corry Chromium.org wrote:
> If this should only be called by the GC, can we assert that we are in a GC
here?

Done.

[ulan, 2014-10-21 09:42:28]

Committed patchset #2 (id:20001) manually as 24758 (presubmit successful).