Merge (7265, 7271] from bleeding_edge to experimental/gc branch....

Merge (7265, 7271] from bleeding_edge to experimental/gc branch.

This is the second (of three) step towards complete isolates merge into experimental/gc branch.

This merge covers initial isolates landing to bleeding_edge. 

Structures introduced during experimental/gc development were refactored from purely static to instance based.

A number of places with questionable performance (e.g. multiple indirections or TLS lookup) were instroduced. These places were marked with "TODO(gc) isolates merge" marker. All this markers will be removed in a separate change when merge will be completed. 

This CL also includes r7650 from bleeding_edge to fix crashes on testsuite.


Committed: http://code.google.com/p/v8/source/detail?r=7678

[Vyacheslav Egorov (Chromium), 2011-04-18 21:42:58]

This is ready for review. All tests pass. I had to fix another bug in pointer
maps generation which was revealed after merge (r7650 in bleeding_edge).

The CL is enormous but I guess it does not make much sense to review anything
beyond main experimental GC components (Marking, IncrementalMarking,
StoreBuffer) and certain parts of memory management subsystem (Heap,
MarkCompactCollector, spaces, MemoryAllocator).

This CL contains a number of places marked as "TODO(gc) isolates merge". This
markers will be cleaned up in a separate change when we complete isolates merge.
They usually indicate places where heap/isolate is accessed in an inefficient
way.

[Erik Corry, 2011-04-20 20:07:40]

LGTM

http://codereview.chromium.org/6880010/diff/3077/src/heap-inl.h
File src/heap-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/heap-inl.h#newcode478
src/heap-inl.h:478: 
2 blank lines

http://codereview.chromium.org/6880010/diff/3077/src/heap.cc
File src/heap.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/heap.cc#newcode85
src/heap.cc:85: reserved_semispace_size_(8*MB),
We have until now limited max semispace size to 4mb on the gc branch.  this
costs us performance but helps pause time.  if we want to experiment with this
lets do it as a separate change.

http://codereview.chromium.org/6880010/diff/3077/src/heap.h
File src/heap.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/heap.h#newcode272
src/heap.h:272: // All JavaScript contexts by this process share the same object
heap.
This comment is out of date, but thats a bug on bleeding edge

http://codereview.chromium.org/6880010/diff/3077/src/mark-compact-inl.h
File src/mark-compact-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/mark-compact-inl.h#newco...
src/mark-compact-inl.h:56: // TODO(gc) !!!!!!
Marking has a heap_ field so this is easy to fix.

http://codereview.chromium.org/6880010/diff/3077/src/serialize.cc
File src/serialize.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/serialize.cc#newcode516
src/serialize.cc:516: 41,
oooops

http://codereview.chromium.org/6880010/diff/3077/src/spaces-inl.h
File src/spaces-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/spaces-inl.h#newcode280
src/spaces-inl.h:280: return object->map() ==
HEAP->raw_unchecked_free_space_map()
I think this needs a todo

http://codereview.chromium.org/6880010/diff/3077/src/spaces.cc
File src/spaces.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/spaces.cc#newcode1375
src/spaces.cc:1375: set_map(HEAP->raw_unchecked_free_space_map());
Needs a todo

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.cc
File src/store-buffer.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.cc#newcode318
src/store-buffer.cc:318: Address* top =
reinterpret_cast<Address*>(HEAP->store_buffer_top());
Use heap_ here and below

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.h
File src/store-buffer.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.h#newcode43
src/store-buffer.h:43: // TODO(gc) ISOLATESMERGE convert to non static class
Is this not done?

[Vyacheslav Egorov (Chromium), 2011-04-24 11:24:07]

thanks for the review. committing.

http://codereview.chromium.org/6880010/diff/3077/src/heap-inl.h
File src/heap-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/heap-inl.h#newcode478
src/heap-inl.h:478: 
On 2011/04/20 20:07:40, Erik Corry wrote:
> 2 blank lines

Done.

http://codereview.chromium.org/6880010/diff/3077/src/heap.cc
File src/heap.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/heap.cc#newcode85
src/heap.cc:85: reserved_semispace_size_(8*MB),
On 2011/04/20 20:07:40, Erik Corry wrote:
> We have until now limited max semispace size to 4mb on the gc branch.  this
> costs us performance but helps pause time.  if we want to experiment with this
> lets do it as a separate change.

ooops. I thought I paid special attention to merge this correctly.

http://codereview.chromium.org/6880010/diff/3077/src/mark-compact-inl.h
File src/mark-compact-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/mark-compact-inl.h#newco...
src/mark-compact-inl.h:56: // TODO(gc) !!!!!!
On 2011/04/20 20:07:40, Erik Corry wrote:
> Marking has a heap_ field so this is easy to fix.

Done.

http://codereview.chromium.org/6880010/diff/3077/src/serialize.cc
File src/serialize.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/serialize.cc#newcode516
src/serialize.cc:516: 41,
On 2011/04/20 20:07:40, Erik Corry wrote:
> oooops

Done.

http://codereview.chromium.org/6880010/diff/3077/src/spaces-inl.h
File src/spaces-inl.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/spaces-inl.h#newcode280
src/spaces-inl.h:280: return object->map() ==
HEAP->raw_unchecked_free_space_map()
On 2011/04/20 20:07:40, Erik Corry wrote:
> I think this needs a todo

Done.

http://codereview.chromium.org/6880010/diff/3077/src/spaces.cc
File src/spaces.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/spaces.cc#newcode1375
src/spaces.cc:1375: set_map(HEAP->raw_unchecked_free_space_map());
On 2011/04/20 20:07:40, Erik Corry wrote:
> Needs a todo

Done.

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.cc
File src/store-buffer.cc (right):

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.cc#newcode318
src/store-buffer.cc:318: Address* top =
reinterpret_cast<Address*>(HEAP->store_buffer_top());
On 2011/04/20 20:07:40, Erik Corry wrote:
> Use heap_ here and below

here and below done where possible.
GCPrologue/GCEpilogue are static functions. TODO added.

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.h
File src/store-buffer.h (right):

http://codereview.chromium.org/6880010/diff/3077/src/store-buffer.h#newcode43
src/store-buffer.h:43: // TODO(gc) ISOLATESMERGE convert to non static class
On 2011/04/20 20:07:40, Erik Corry wrote:
> Is this not done?

indeed.