Implement zone-allocated types

Implement zone-allocated types

More template magic FTW.

R=titzer@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=18543

[titzer, 2013-12-19 13:44:41]

https://codereview.chromium.org/103743004/diff/1/src/types.h
File src/types.h (right):

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode404
src/types.h:404: static Unioned* union_shrink(Unioned* unioned, int size, Zone*
zone) {
So this one mutates, but the heap one creates a new one? Hmm...

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode407
src/types.h:407: for (; size < length; --length) unioned->RemoveLast();
I think you can use Rewind() here.

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode412
src/types.h:412: ASSERT(!is_union(type));
Asserting a union isn't a member of a union when getting...i suppose that's ok,
but how could that happen otherwise? someone monkeying with the fixed array
behind your back?

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode481
src/types.h:481: i::Handle<Unioned> result = union_create(size, isolate);
You actually can chop a FixedArray with set_length() here.

[rossberg, 2013-12-19 17:21:25]

https://codereview.chromium.org/103743004/diff/1/src/types.h
File src/types.h (right):

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode404
src/types.h:404: static Unioned* union_shrink(Unioned* unioned, int size, Zone*
zone) {
On 2013/12/19 13:44:41, titzer wrote:
> So this one mutates, but the heap one creates a new one? Hmm...

Was able to remove that with the Shrink method.

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode407
src/types.h:407: for (; size < length; --length) unioned->RemoveLast();
On 2013/12/19 13:44:41, titzer wrote:
> I think you can use Rewind() here.

Done.

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode412
src/types.h:412: ASSERT(!is_union(type));
On 2013/12/19 13:44:41, titzer wrote:
> Asserting a union isn't a member of a union when getting...i suppose that's
ok,
> but how could that happen otherwise? someone monkeying with the fixed array
> behind your back?

It's a representational invariant, checked here just to assert that the Union
algorithm can't mess things up.

https://codereview.chromium.org/103743004/diff/1/src/types.h#newcode481
src/types.h:481: i::Handle<Unioned> result = union_create(size, isolate);
On 2013/12/19 13:44:41, titzer wrote:
> You actually can chop a FixedArray with set_length() here.

It's a bit more complicated than that. I introduced a suitable Shrink method on
typed arrays.

[titzer, 2013-12-19 18:47:24]

lgtm

[rossberg, 2014-01-10 14:43:59]

Committed patchset #2 manually as r18543 (presubmit successful).