Support compaction for code space pages.

Support compaction for code space pages.

R=erik.corry@gmail.com
BUG=v8:1550
Committed: http://code.google.com/p/v8/source/detail?r=9155

[Erik Corry, 2011-09-06 14:10:31]

LGTM

You might consider running all the benchmarks with a very small page size to see
if it flushes out any bugs around large pages.

http://codereview.chromium.org/7834018/diff/2002/src/assembler.h
File src/assembler.h (right):

http://codereview.chromium.org/7834018/diff/2002/src/assembler.h#newcode223
src/assembler.h:223: : pc_(pc), rmode_(rmode), data_(data), host_(NULL) {
If you omit this function the compiler will help you find the places it was
called so you can verify that it is correct to pass NULL at this point.

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc#newcode577
src/mark-compact.cc:577: Address slot = candidate->address() +
JSFunction::kCodeEntryOffset;
This is kind of subtle.  Comment?

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc#newcode3480
src/mark-compact.cc:3480: SlotsBuffer::SlotType SlotsBuffer::UpdateSlots(
Somewhere there should be a comment on why this function both takes and returns
a SlotType.

Also perhaps I missed a comment on the way slots and slot types are encoded in
the buffer.

http://codereview.chromium.org/7834018/diff/2002/src/spaces.h
File src/spaces.h (right):

http://codereview.chromium.org/7834018/diff/2002/src/spaces.h#newcode1460
src/spaces.h:1460: bool IsFragmented(Page* p) {
Perhaps this should always return true if the --always-compact flag is used.
Good for testing?

[Vyacheslav Egorov (Chromium), 2011-09-06 15:01:41]

thanks. landing!

http://codereview.chromium.org/7834018/diff/2002/src/assembler.h
File src/assembler.h (right):

http://codereview.chromium.org/7834018/diff/2002/src/assembler.h#newcode223
src/assembler.h:223: : pc_(pc), rmode_(rmode), data_(data), host_(NULL) {
On 2011/09/06 14:10:31, Erik Corry wrote:
> If you omit this function the compiler will help you find the places it was
> called so you can verify that it is correct to pass NULL at this point.

Done.

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc
File src/mark-compact.cc (right):

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc#newcode577
src/mark-compact.cc:577: Address slot = candidate->address() +
JSFunction::kCodeEntryOffset;
On 2011/09/06 14:10:31, Erik Corry wrote:
> This is kind of subtle.  Comment?

Done.

http://codereview.chromium.org/7834018/diff/2002/src/mark-compact.cc#newcode3480
src/mark-compact.cc:3480: SlotsBuffer::SlotType SlotsBuffer::UpdateSlots(
On 2011/09/06 14:10:31, Erik Corry wrote:
> Somewhere there should be a comment on why this function both takes and
returns
> a SlotType.
> 
> Also perhaps I missed a comment on the way slots and slot types are encoded in
> the buffer.

Done.

http://codereview.chromium.org/7834018/diff/2002/src/spaces.h
File src/spaces.h (right):

http://codereview.chromium.org/7834018/diff/2002/src/spaces.h#newcode1460
src/spaces.h:1460: bool IsFragmented(Page* p) {
On 2011/09/06 14:10:31, Erik Corry wrote:
> Perhaps this should always return true if the --always-compact flag is used.
> Good for testing?

Done.