Faster implementation of Heap::RecordWrites.

Faster implementation of Heap::RecordWrites.

Naive algorithm for to update RSets for a span is rather inefficient
as it performs many unnecessary operations (retrieving a mask, updating
it with the same bit as many pointers go into a single region).

Committed: http://code.google.com/p/v8/source/detail?r=4849

[antonm, 2010-06-11 11:53:29]

Slava,

may you have a look?

I'd esp. appreciate careful check if mask calculating algorithm is correct.

[antonm, 2010-06-11 15:32:33]

Slava, another version with your cool way to calculate the mask.

[Vyacheslav Egorov (Chromium), 2010-06-11 16:06:33]

LGTM with comments addressed.

http://codereview.chromium.org/2727009/diff/5001/6002
File src/spaces-inl.h (right):

http://codereview.chromium.org/2727009/diff/5001/6002#newcode166
src/spaces-inl.h:166: if (!result) result = start_mask | end_mask;
result == 0 instead?

a small comment about why we check result == 0 here might be handy (start can be
greater than end)

http://codereview.chromium.org/2727009/diff/5001/6002#newcode179
src/spaces-inl.h:179: 
One more empty line here.