Limit the amount of memory that can be folded together.

Limit the amount of memory that can be folded together.

BUG=

[Michael Starzinger, 2013-07-22 08:46:06]

Looking good, nice catch! Just two comments about additional asserts.

https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:3233: if (new_dominator_size >
Page::kMaxNonCodeHeapObjectSize) {
Can we add an assert here like the following? This makes sure that we are not
statically exceeding the limit for neither new-space nor pretenured allocations.

ASSERT_LE(Page::kMaxNonCodeHeapObjectSize, Page::kMaxObjectSizeInNewSpace);

Also on top of that, could we add asserts to MacroAssembler::Allocate (the one
that takes a static object size) that check we are not exceeding the same limit
through other call-sites? I know it's not part of this change but such an assert
would have caught this violation earlier.

[Michael Starzinger, 2013-07-22 12:28:04]

LGTM.

https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:3233: if (new_dominator_size >
Page::kMaxNonCodeHeapObjectSize) {
On 2013/07/22 08:46:07, Michael Starzinger wrote:
> Can we add an assert here like the following? This makes sure that we are not
> statically exceeding the limit for neither new-space nor pretenured
allocations.
> 
> ASSERT_LE(Page::kMaxNonCodeHeapObjectSize, Page::kMaxObjectSizeInNewSpace);
> 
> Also on top of that, could we add asserts to MacroAssembler::Allocate (the one
> that takes a static object size) that check we are not exceeding the same
limit
> through other call-sites? I know it's not part of this change but such an
assert
> would have caught this violation earlier.

As discussed offline: This is a can of worms and we will clean it up in a
followup CL where we'll try to unify those two constant.

[Hannes Payer (out of office), 2013-07-22 12:37:33]

On 2013/07/22 12:28:04, Michael Starzinger wrote:
> LGTM.
> 
> https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc
> File src/hydrogen-instructions.cc (right):
> 
>
https://codereview.chromium.org/19804002/diff/1/src/hydrogen-instructions.cc#...
> src/hydrogen-instructions.cc:3233: if (new_dominator_size >
> Page::kMaxNonCodeHeapObjectSize) {
> On 2013/07/22 08:46:07, Michael Starzinger wrote:
> > Can we add an assert here like the following? This makes sure that we are
not
> > statically exceeding the limit for neither new-space nor pretenured
> allocations.
> > 
> > ASSERT_LE(Page::kMaxNonCodeHeapObjectSize, Page::kMaxObjectSizeInNewSpace);
> > 
> > Also on top of that, could we add asserts to MacroAssembler::Allocate (the
one
> > that takes a static object size) that check we are not exceeding the same
> limit
> > through other call-sites? I know it's not part of this change but such an
> assert
> > would have caught this violation earlier.
> 
> As discussed offline: This is a can of worms and we will clean it up in a
> followup CL where we'll try to unify those two constant.

committed, Revision: r15804