Merge flush code phase into marking phase.

Merge flush code phase into marking phase.

Committed: http://code.google.com/p/v8/source/detail?r=5284

[Rico, 2010-08-17 10:21:58]

Consider resetting the age counter in the CompileLazy* functions in handles.cc.
Especially CompileLazyInLoop which is the one being called from
Runtime_LazyCompile.
This could potentially help avoid some of the recompilations (I am not sure
about this, put intuitively it would make sense if it did help).

[Mads Ager (chromium), 2010-08-17 10:31:28]

LGTM, but I would like to keep the normal casts instead of using
reinterpret_cast. We should restrict the reinterpret_casts to GC to get the
asserts for normal non-gc code.

http://codereview.chromium.org/3135026/diff/1/6
File src/frames.cc (right):

http://codereview.chromium.org/3135026/diff/1/6#newcode410
src/frames.cc:410: return reinterpret_cast<Code*>(code_slot());
Doing these changes, we loose assertions in the code that is using this when not
during GC. Should we add raw_unchecked accessors for the stuff needed during GC?

http://codereview.chromium.org/3135026/diff/1/8
File src/heap-inl.h (right):

http://codereview.chromium.org/3135026/diff/1/8#newcode422
src/heap-inl.h:422: return Handle<TYPE>(reinterpret_cast<TYPE*>(__object__)),  \
Same here, loosing these assertions seems bad.

http://codereview.chromium.org/3135026/diff/1/15
File src/objects-visiting.h (right):

http://codereview.chromium.org/3135026/diff/1/15#newcode274
src/objects-visiting.h:274: &JSObjectVisitor::template
VisitSpecialized<JSFunction::kSize>);
Long line.

http://codereview.chromium.org/3135026/diff/1/17
File src/objects.h (right):

http://codereview.chromium.org/3135026/diff/1/17#newcode3483
src/objects.h:3483: inline int code_age();
Add comments please.

[Vyacheslav Egorov (Chromium), 2010-08-17 11:09:21]

Returned checked casts.

http://codereview.chromium.org/3135026/diff/1/6
File src/frames.cc (right):

http://codereview.chromium.org/3135026/diff/1/6#newcode410
src/frames.cc:410: return reinterpret_cast<Code*>(code_slot());
On 2010/08/17 10:31:28, Mads Ager wrote:
> Doing these changes, we loose assertions in the code that is using this when
not
> during GC. Should we add raw_unchecked accessors for the stuff needed during
GC?

Done.

http://codereview.chromium.org/3135026/diff/1/8
File src/heap-inl.h (right):

http://codereview.chromium.org/3135026/diff/1/8#newcode422
src/heap-inl.h:422: return Handle<TYPE>(reinterpret_cast<TYPE*>(__object__)),  \
On 2010/08/17 10:31:28, Mads Ager wrote:
> Same here, loosing these assertions seems bad.

Agree. This change is unrelated and will be removed.

http://codereview.chromium.org/3135026/diff/1/15
File src/objects-visiting.h (right):

http://codereview.chromium.org/3135026/diff/1/15#newcode274
src/objects-visiting.h:274: &JSObjectVisitor::template
VisitSpecialized<JSFunction::kSize>);
On 2010/08/17 10:31:28, Mads Ager wrote:
> Long line.

Done.

http://codereview.chromium.org/3135026/diff/1/17
File src/objects.h (right):

http://codereview.chromium.org/3135026/diff/1/17#newcode3483
src/objects.h:3483: inline int code_age();
On 2010/08/17 10:31:28, Mads Ager wrote:
> Add comments please.

Done.

[Mads Ager (chromium), 2010-08-17 11:15:25]

LGTM