Start using store buffers. Handle store buffer overflow situation....

Start using store buffers.  Handle store buffer overflow situation.
Only on Intel architectures at the moment.

Committed: http://code.google.com/p/v8/source/detail?r=6616

[Vyacheslav Egorov (Chromium), 2011-02-02 13:15:47]

LGTM

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h
File src/heap-inl.h (right):

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h#newcode344
src/heap-inl.h:344: Memory::Object_at(dst) = data;
revert please

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h#newcode393
src/heap-inl.h:393: Address cell = reinterpret_cast<Address>(p);
cell is a bit confusing cause we have cell_space and cells.

I usually call them slots.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc
File src/heap.cc (right):

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode642
src/heap.cc:642: PageIterator it(space, PageIterator::PAGES_IN_USE);
indentation

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode3903
src/heap.cc:3903: static void VerifyPointersUnderWatermark(
Why we have function that does not do anything?

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode4113
src/heap.cc:4113: 
If you are not using region marks -> your start and end probably always aligned
to map size so you can get read of all this complicated misalignment handling.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode4133
src/heap.cc:4133: pointer_fields_end,
indent

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc#newcode105
src/mark-compact.cc:105: SweepLargeObjectSpace();
You can now move SweepLargeObjectSpace into SweepSpaces. and probably eliminate
it entirely. Just call FreeUnmarkedObjects() directly.

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc#newcode419
src/mark-compact.cc:419: // substring when page dirty marks do not change. 
TODO(gc): Seems like we
move todo to the new line

http://codereview.chromium.org/6250076/diff/1/src/objects-inl.h
File src/objects-inl.h (left):

http://codereview.chromium.org/6250076/diff/1/src/objects-inl.h#oldcode805
src/objects-inl.h:805: ASSERT(Heap::InNewSpace(object) || \
I there a way to rewrite this assertion to check that store buffer contains this
slot?

http://codereview.chromium.org/6250076/diff/1/src/store-buffer.h
File src/store-buffer.h (right):

http://codereview.chromium.org/6250076/diff/1/src/store-buffer.h#newcode69
src/store-buffer.h:69: kPreserveStoreBufferWhileIterating};
new line + indent closing }

[Erik Corry, 2011-02-03 13:21:17]

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h
File src/heap-inl.h (right):

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h#newcode344
src/heap-inl.h:344: Memory::Object_at(dst) = data;
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> revert please

Done.

http://codereview.chromium.org/6250076/diff/1/src/heap-inl.h#newcode393
src/heap-inl.h:393: Address cell = reinterpret_cast<Address>(p);
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> cell is a bit confusing cause we have cell_space and cells.
> 
> I usually call them slots.

Done.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc
File src/heap.cc (right):

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode642
src/heap.cc:642: PageIterator it(space, PageIterator::PAGES_IN_USE);
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> indentation

Done.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode3903
src/heap.cc:3903: static void VerifyPointersUnderWatermark(
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> Why we have function that does not do anything?

Function restored.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode4113
src/heap.cc:4113: 
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> If you are not using region marks -> your start and end probably always
aligned
> to map size so you can get read of all this complicated misalignment handling.

Done.

http://codereview.chromium.org/6250076/diff/1/src/heap.cc#newcode4133
src/heap.cc:4133: pointer_fields_end,
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> indent

Done.

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc#newcode105
src/mark-compact.cc:105: SweepLargeObjectSpace();
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> You can now move SweepLargeObjectSpace into SweepSpaces. and probably
eliminate
> it entirely. Just call FreeUnmarkedObjects() directly.

Done.

http://codereview.chromium.org/6250076/diff/1/src/mark-compact.cc#newcode419
src/mark-compact.cc:419: // substring when page dirty marks do not change. 
TODO(gc): Seems like we
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> move todo to the new line

Done.

http://codereview.chromium.org/6250076/diff/1/src/objects-inl.h
File src/objects-inl.h (left):

http://codereview.chromium.org/6250076/diff/1/src/objects-inl.h#oldcode805
src/objects-inl.h:805: ASSERT(Heap::InNewSpace(object) || \
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> I there a way to rewrite this assertion to check that store buffer contains
this
> slot?

I brought back the assert but hid it behind a --enable-slow-asserts

http://codereview.chromium.org/6250076/diff/1/src/store-buffer.h
File src/store-buffer.h (right):

http://codereview.chromium.org/6250076/diff/1/src/store-buffer.h#newcode69
src/store-buffer.h:69: kPreserveStoreBufferWhileIterating};
On 2011/02/02 13:15:47, Vyacheslav Egorov wrote:
> new line + indent closing }

Forgot this one :-(