Add a generic remembered set class.

Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO

Committed: https://crrev.com/4fdc19ae33cfb4c9b957585ec35f7dfb6f67177f
Cr-Commit-Position: refs/heads/master@{#34031}

[ulan, 2016-02-11 15:15:07]

Description was changed from

==========
Add remembered set

BUG=
==========

to

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.
==========

[ulan, 2016-02-11 15:20:46]

Description was changed from

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.
==========

to

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO
==========

[ulan, 2016-02-11 15:24:32]

ulan@chromium.org changed reviewers:
+ hpayer@chromium.org, mlippautz@chromium.org

[ulan, 2016-02-11 15:24:33]

PTAL

https://codereview.chromium.org/1683653002/diff/100001/src/heap/store-buffer.cc
File src/heap/store-buffer.cc (left):

https://codereview.chromium.org/1683653002/diff/100001/src/heap/store-buffer....
src/heap/store-buffer.cc:89: heap_->InNewSpace(object);
This verifies only large objects space and looks like no-op. I am removing this
completely.

https://codereview.chromium.org/1683653002/diff/100001/src/heap/store-buffer....
src/heap/store-buffer.cc:131: static SlotSet::CallbackResult
ProcessOldToNewSlot(
This functions are moved to remembered-set.*

[Hannes Payer (out of office), 2016-02-15 16:53:39]

LGTM, just minor things.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc#newco...
src/heap/heap.cc:4492: // the store buffer.  These pages are scanned to find
pointers that point
That comment is outdated.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc#newco...
src/heap/heap.cc:5485: Page* page = Page::FromAddress(slot_addr);
Why don't we move 
Page* page = Page::FromAddress(slot_addr); to Insert and Remove? These methods
would then just take the slot address as parameter.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
File src/heap/remembered-set.cc (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
src/heap/remembered-set.cc:16: template <PointerDirection direction>
Are you using direction just for now for sanity check?

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-set.h
File src/heap/remembered-set.h (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
src/heap/remembered-set.h:23: DCHECK(page->Contains(slot_addr));
You can drop the DCHECK if you move the page lookup to this function.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
src/heap/remembered-set.h:36: DCHECK(page->Contains(slot_addr));
You can drop the DCHECK if you move the page lookup to this function.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/spaces-inl.h
File src/heap/spaces-inl.h (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/spaces-inl.h#...
src/heap/spaces-inl.h:301: if (offset < MemoryChunk::kHeaderSize ||
chunk->owner() == nullptr) {
We already check somewhere else for LO page with owner() == heap->lo_space(). I
think we should add a static check function to large pages that tells us if a
chunk* is a large page.

[ulan, 2016-02-16 10:24:52]

Thanks for the comments.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc#newco...
src/heap/heap.cc:4492: // the store buffer.  These pages are scanned to find
pointers that point
On 2016/02/15 16:53:38, Hannes Payer wrote:
> That comment is outdated.

Thanks. Removed the comment.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc#newco...
src/heap/heap.cc:5485: Page* page = Page::FromAddress(slot_addr);
On 2016/02/15 16:53:38, Hannes Payer wrote:
> Why don't we move 
> Page* page = Page::FromAddress(slot_addr); to Insert and Remove? These methods
> would then just take the slot address as parameter.

I made the page a parameter because of slots buffer (not in this CL, but will be
needed later). When recording slot we cannot infer the page from the slot
address, e.g. slot in a large code object. So there we have to take additional
information - either the object start or the page.

Besides, sometimes the caller knows that it is adding a range of values in the
same page, so it can cache the page.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
File src/heap/remembered-set.cc (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/remembered-se...
src/heap/remembered-set.cc:16: template <PointerDirection direction>
On 2016/02/15 16:53:38, Hannes Payer wrote:
> Are you using direction just for now for sanity check?

Yes.

https://codereview.chromium.org/1683653002/diff/120001/src/heap/spaces-inl.h
File src/heap/spaces-inl.h (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/spaces-inl.h#...
src/heap/spaces-inl.h:301: if (offset < MemoryChunk::kHeaderSize ||
chunk->owner() == nullptr) {
On 2016/02/15 16:53:39, Hannes Payer wrote:
> We already check somewhere else for LO page with owner() == heap->lo_space().
I
> think we should add a static check function to large pages that tells us if a
> chunk* is a large page.

I added HasPageHeader predicate to chunk, which is what we want to test for in
this case.

[Hannes Payer (out of office), 2016-02-16 12:00:53]

lgtm, ship it

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1683653002/diff/120001/src/heap/heap.cc#newco...
src/heap/heap.cc:5485: Page* page = Page::FromAddress(slot_addr);
On 2016/02/16 10:24:52, ulan wrote:
> On 2016/02/15 16:53:38, Hannes Payer wrote:
> > Why don't we move 
> > Page* page = Page::FromAddress(slot_addr); to Insert and Remove? These
methods
> > would then just take the slot address as parameter.
> 
> I made the page a parameter because of slots buffer (not in this CL, but will
be
> needed later). When recording slot we cannot infer the page from the slot
> address, e.g. slot in a large code object. So there we have to take additional
> information - either the object start or the page.
> 
> Besides, sometimes the caller knows that it is adding a range of values in the
> same page, so it can cache the page.

Acknowledged.

[ulan_google, 2016-02-16 12:06:18]

The CQ bit was checked by ulan@google.com

[commit-bot: I haz the power, 2016-02-16 12:06:29]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1683653002/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1683653002/140001

[commit-bot: I haz the power, 2016-02-16 12:34:34]

Description was changed from

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO
==========

to

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO
==========

[commit-bot: I haz the power, 2016-02-16 12:34:35]

Committed patchset #8 (id:140001)

[commit-bot: I haz the power, 2016-02-16 12:34:48]

Description was changed from

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO
==========

to

==========
Add a generic remembered set class.

This new class provides a unified interface for recording and iterating slots in
store and slots buffers:

RememberedSet<OLD_TO_NEW>::Insert(page, slot);
RememberedSet<OLD_TO_OLD>::Insert(page, slot);

RememberedSet<OLD_TO_NEW>::Iterate(heap, callback);
RememberedSet<OLD_TO_OLD>::Iterate(heap, callback);

After this change the store buffer is responsible only for collecting slots from
the generated code.

Subsequent CLs will remove the slots buffer.

BUG=chromium:578883
LOG=NO

Committed: https://crrev.com/4fdc19ae33cfb4c9b957585ec35f7dfb6f67177f
Cr-Commit-Position: refs/heads/master@{#34031}
==========

[commit-bot: I haz the power, 2016-02-16 12:34:49]

Patchset 8 (id:??) landed as
https://crrev.com/4fdc19ae33cfb4c9b957585ec35f7dfb6f67177f
Cr-Commit-Position: refs/heads/master@{#34031}