Remove unused maps during marking garbage collections.

Remove unused maps during marking garbage collections.


Committed: http://code.google.com/p/v8/source/detail?r=637

[William Hesse, 2008-10-29 08:28:15]

Only change since last time is to add bounds of
instance_type() >= FIRST_JS_OBJECT_TYPE
and
instance_type() <= JS_FUNCTION_TYPE

in three places.

[Kasper Lund, 2008-10-29 09:58:21]

Mostly a rubber stamp LGTM. In the future, it would be great if you could upload
a new patch set (just do gcl upload again) instead of creating a new CL. That
makes it easier to preserve comments across reviews and makes it easier to see
the diff between patch sets. Comments:

http://codereview.chromium.org/8831/diff/209/16
File src/mark-compact.cc (right):

http://codereview.chromium.org/8831/diff/209/16#newcode332
Line 332: // ASSERT(!obj->IsMap());  // Some maps are processed here.
We shouldn't need to carry code in comments. I guess the assertion doesn't hold
anymore?

http://codereview.chromium.org/8831/diff/209/14
File src/objects.h (right):

http://codereview.chromium.org/8831/diff/209/14#newcode1050
Line 1050: static inline Object* &RawField(HeapObject* obj, int offset);
This looks nasty with returning a reference. Can't you rewrite this with a
RawFieldGet and RawFieldSet?

[William Hesse, 2008-10-29 10:15:19]

OK, changes made.  It is a separate patch list because the reversions, merges,
and false starts put me in a different checked-out copy, but I'll try harder to
stay in the same copy in the future.

http://codereview.chromium.org/8831/diff/209/16
File src/mark-compact.cc (right):

http://codereview.chromium.org/8831/diff/209/16#newcode332
Line 332: // ASSERT(!obj->IsMap());  // Some maps are processed here.
On 2008/10/29 09:58:21, Kasper Lund wrote:
> We shouldn't need to carry code in comments. I guess the assertion doesn't
hold
> anymore?

Removed

http://codereview.chromium.org/8831/diff/209/14
File src/objects.h (right):

http://codereview.chromium.org/8831/diff/209/14#newcode1050
Line 1050: static inline Object* &RawField(HeapObject* obj, int offset);
On 2008/10/29 09:58:21, Kasper Lund wrote:
> This looks nasty with returning a reference. Can't you rewrite this with a
> RawFieldGet and RawFieldSet?

It looks beautiful where it is used.  It could be an Object**, but that requires
a dereference where
it is used.  But I'll change it to **.