Cache multiple ElementsKind map transition per map.

Cache multiple ElementsKind map transition per map.

R=jkummerow@chromium.org
BUG=none
TEST=none

Committed: http://code.google.com/p/v8/source/detail?r=9417

[Jakob Kummerow, 2011-09-23 12:12:28]

LGTM

http://codereview.chromium.org/8017003/diff/1/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/8017003/diff/1/src/objects.cc#newcode2081
src/objects.cc:2081: new_array->set(i,
FixedArray::cast(descriptor_contents)->get(i));
Simplification: s/FixedArray::cast(descriptor_contents)/array/

http://codereview.chromium.org/8017003/diff/1/src/objects.cc#newcode2163
src/objects.cc:2163: if (!maybe_new_contents->To<Object>(&new_contents)) {
Simplification: s/To<Object>/ToObject/

[danno, 2011-09-23 13:33:45]

Slava, can you please take a look at the GC bug fixes? Thanks.

http://codereview.chromium.org/8017003/diff/1/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/8017003/diff/1/src/objects.cc#newcode2081
src/objects.cc:2081: new_array->set(i,
FixedArray::cast(descriptor_contents)->get(i));
On 2011/09/23 12:12:28, Jakob wrote:
> Simplification: s/FixedArray::cast(descriptor_contents)/array/

Done.

http://codereview.chromium.org/8017003/diff/1/src/objects.cc#newcode2163
src/objects.cc:2163: if (!maybe_new_contents->To<Object>(&new_contents)) {
On 2011/09/23 12:12:28, Jakob wrote:
> Simplification: s/To<Object>/ToObject/

Done.

[Vyacheslav Egorov (Chromium), 2011-09-23 13:49:22]

lgtm

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6509
src/objects.cc:6509: // Verify target.
Can we avoid code duplication by moving these asserts + *RawField() = this to
some auxiliary helper?

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6570
src/objects.cc:6570: for (int i = 0; i < array->length(); ++i) {
Probably you can discard this transition altogether when all maps are discarded.
(it will be picked up by the next GC).

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6579
src/objects.cc:6579: // Getter prototype() is read-only, set_prototype() has
side effects.
long line

[danno, 2011-09-25 09:24:27]

Feedback addressed, landing.

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6509
src/objects.cc:6509: // Verify target.
On 2011/09/23 13:49:23, Vyacheslav Egorov wrote:
> Can we avoid code duplication by moving these asserts + *RawField() = this to
> some auxiliary helper?

Done.

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6570
src/objects.cc:6570: for (int i = 0; i < array->length(); ++i) {
On 2011/09/23 13:49:23, Vyacheslav Egorov wrote:
> Probably you can discard this transition altogether when all maps are
discarded.
> (it will be picked up by the next GC).

Done.

http://codereview.chromium.org/8017003/diff/8001/src/objects.cc#newcode6579
src/objects.cc:6579: // Getter prototype() is read-only, set_prototype() has
side effects.
On 2011/09/23 13:49:23, Vyacheslav Egorov wrote:
> long line

Done.