Refactor how embedded pointers are visited.

Refactor how embedded pointers are visited.

This refactoring (almost) gets rid of the requirement to get the target
object address for an object pointer embedded in code objects. This is
not possible on MIPS as pointers are encoded using two instructions. All
usages of RelocInfo::target_object_address() are (almost) obsoleted by
this change. The serializer still uses it, so MIPS will not yet work
with snapshots turned on.

R=danno@chromium.org,vegorov@chromium.org

Committed: http://code.google.com/p/v8/source/detail?r=9597

[danno, 2011-10-12 14:07:39]

this idea looks much better than the #ifdef proposal that Paul sent us, but I
think Slava should take a look for the final OK. Please also verify that this is
performance neutral.

[danno, 2011-10-12 14:08:29]

+vegorov

[Vyacheslav Egorov (Chromium), 2011-10-12 14:21:53]

lgtm

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.cc#newcode3762
src/mark-compact.cc:3762: void
MarkCompactCollector::RecordEmbeddedSlot(RelocInfo* rinfo,
Why not unify with RecordRelocSlot? seems almost identical.

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.h
File src/mark-compact.h (right):

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.h#newcode323
src/mark-compact.h:323: EMBEDDED_OBJECT_SLOT,
micro opt idea: consider placing it first (or in the middle) depending on the
way switch over slot types is compiled in visit function (rationale: there are
currently much more EMBEDDED_OBJECT_SLOTs than anything else)

[Michael Starzinger, 2011-10-12 15:38:36]

Added new patch set.

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.cc#newcode3762
src/mark-compact.cc:3762: void
MarkCompactCollector::RecordEmbeddedSlot(RelocInfo* rinfo,
On 2011/10/12 14:21:53, Vyacheslav Egorov wrote:
> Why not unify with RecordRelocSlot? seems almost identical.

Done.

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.h
File src/mark-compact.h (right):

http://codereview.chromium.org/8245007/diff/1/src/mark-compact.h#newcode323
src/mark-compact.h:323: EMBEDDED_OBJECT_SLOT,
On 2011/10/12 14:21:53, Vyacheslav Egorov wrote:
> micro opt idea: consider placing it first (or in the middle) depending on the
> way switch over slot types is compiled in visit function (rationale: there are
> currently much more EMBEDDED_OBJECT_SLOTs than anything else)

Done. On both ia32 and x64 it is compiled as a lookup table. But I placed it
first to indicate (syntactically) that it is the most frequent type.