Pretenure some strings into bump-allocated block

Proof-of-concept pretenuring of some strings:

- Add bump-pointer allocated block in page space.
- Pretenure num.toString whenever >98% of strings are being promoted.
- Fix deadlock in freelist printing.

Committed: https://code.google.com/p/dart/source/detail?r=39688

[koda, 2014-08-28 23:21:52]

koda@google.com changed reviewers:
+ iposva@google.com

[Ivan Posva, 2014-08-29 00:07:12]

LGTMwC -ip

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/freelist.cc
File runtime/vm/freelist.cc (right):

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/freelist.cc#n...
runtime/vm/freelist.cc:239: ASSERT(index >= 0);
Please add
DEBUG_ASSERT(mutex_->Owner() == Isolate::Current());
as the first line.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:287: uword PageSpace::TryAllocateInNewPage(intptr_t size,
TryAllocateInNewPage is a confusing name as it makes you think you are
allocating in the new gen. Please choose a different name.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:649: bump_top_ = bump_end_;
Why not setting these values to 0?

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:762: if (size >= kAllocatablePageSize) {
Please explain in a comment why this is not done as the first thing in this
method.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:768: // side-effect of populating the freelist with a large
block.
...with a large block. Next time we request bump allocation we will have a
chance to consume that large block.

[koda, 2014-08-29 00:52:20]

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/freelist.cc
File runtime/vm/freelist.cc (right):

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/freelist.cc#n...
runtime/vm/freelist.cc:239: ASSERT(index >= 0);
On 2014/08/29 00:07:12, Ivan Posva wrote:
> Please add
> DEBUG_ASSERT(mutex_->Owner() == Isolate::Current());
> as the first line.

Done.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc
File runtime/vm/pages.cc (right):

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:287: uword PageSpace::TryAllocateInNewPage(intptr_t size,
On 2014/08/29 00:07:12, Ivan Posva wrote:
> TryAllocateInNewPage is a confusing name as it makes you think you are
> allocating in the new gen. Please choose a different name.

Done.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:649: bump_top_ = bump_end_;
On 2014/08/29 00:07:12, Ivan Posva wrote:
> Why not setting these values to 0?

Done.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:762: if (size >= kAllocatablePageSize) {
On 2014/08/29 00:07:12, Ivan Posva wrote:
> Please explain in a comment why this is not done as the first thing in this
> method.

Done.

https://codereview.chromium.org/511963007/diff/40001/runtime/vm/pages.cc#newc...
runtime/vm/pages.cc:768: // side-effect of populating the freelist with a large
block.
On 2014/08/29 00:07:12, Ivan Posva wrote:
> ...with a large block. Next time we request bump allocation we will have a
> chance to consume that large block.

Done.

[koda, 2014-08-29 03:54:46]

Committed patchset #5 (id:80001) manually as r39688 (presubmit successful).