Take into account freed global handles and fragmentation when computing heap growing factor.

Take into account freed global handles and fragmentation when computing heap growing factor.

BUG=chromium:493636
LOG=NO

[ulan, 2015-06-01 16:15:45]

ulan@chromium.org changed reviewers:
+ hpayer@chromium.org

[ulan, 2015-06-01 16:15:46]

PTAL

[Hannes Payer (out of office), 2015-06-02 08:23:00]

nice!

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc#newcod...
src/heap/heap.cc:1294: fragmentation_percent >= kLowFragmentationPercent;
Can we move this code into a function with a descriptive name?

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc#newcod...
src/heap/heap.cc:5448: ? 2
Let's make these two value also constants.

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.h
File src/heap/heap.h (right):

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.h#newcode...
src/heap/heap.h:1188: // Allow slack of one page for each space.
I think this function should only return the real percent of fragmentation. I
would prefer passing in the slack as an "additional live memory" parameter.
FLAG_never_compact should be applied outside as well. WDYT?

[ulan, 2015-06-02 12:27:05]

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc#newcod...
src/heap/heap.cc:1294: fragmentation_percent >= kLowFragmentationPercent;
On 2015/06/02 08:23:00, Hannes Payer wrote:
> Can we move this code into a function with a descriptive name?

Done.

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.cc#newcod...
src/heap/heap.cc:5448: ? 2
On 2015/06/02 08:23:00, Hannes Payer wrote:
> Let's make these two value also constants.

Done.

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.h
File src/heap/heap.h (right):

https://codereview.chromium.org/1161603006/diff/20001/src/heap/heap.h#newcode...
src/heap/heap.h:1188: // Allow slack of one page for each space.
On 2015/06/02 08:23:00, Hannes Payer wrote:
> I think this function should only return the real percent of fragmentation. I
> would prefer passing in the slack as an "additional live memory" parameter.
> FLAG_never_compact should be applied outside as well. WDYT?

Done.

[Hannes Payer (out of office), 2015-06-03 06:32:58]

https://codereview.chromium.org/1161603006/diff/80001/src/heap/heap.cc
File src/heap/heap.cc (right):

https://codereview.chromium.org/1161603006/diff/80001/src/heap/heap.cc#newcod...
src/heap/heap.cc:1262:
FragmentationOfCompactedSpacesInPercent(&total_committed);
Can we have a separate function which returns total committed compacted space
memory?

https://codereview.chromium.org/1161603006/diff/80001/src/heap/heap.cc#newcod...
src/heap/heap.cc:5450: 
This function needs more comments about the magic that is going on here :)

https://codereview.chromium.org/1161603006/diff/80001/src/heap/heap.h
File src/heap/heap.h (right):

https://codereview.chromium.org/1161603006/diff/80001/src/heap/heap.h#newcode...
src/heap/heap.h:1194: return 100 - static_cast<int>(100.0 * live / committed);
committed should never be 0 in current v8, but can we still check?