Make GCTracer persistent.

Make GCTracer persistent.

This is essentially https://codereview.chromium.org/390823003/. It turns
out that it's safe to use a persistent GCTracer instead
of creating and destroying GCTracers for each GC, because
Scavenge and Mark-Sweep steps never overlap, and the tracking of
incremental marking also happens separately for the two types of GC.
Tracking of incremental marking will be moved into the GCTracer in a
follow-up patch.

R=hpayer@chromium.org, mstarzinger@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=22465

[ernstm, 2014-07-17 13:37:10]

PTAL

[Hannes Payer (out of office), 2014-07-17 15:26:14]

https://codereview.chromium.org/403543002/diff/1/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/403543002/diff/1/src/heap.cc#newcode6002
src/heap.cc:6002: GCTracer::~GCTracer() {
I guess we can remove the destructor.

https://codereview.chromium.org/403543002/diff/1/src/heap.cc#newcode6031
src/heap.cc:6031: spent_in_mutator_ = 0.0;
That is actually wrong. We would like to record here the time before we invoked
the first gc, i.e. start_time_ - program_start_time. Can we fix that here?

[ernstm, 2014-07-18 08:32:28]

https://codereview.chromium.org/403543002/diff/1/src/heap.cc
File src/heap.cc (right):

https://codereview.chromium.org/403543002/diff/1/src/heap.cc#newcode6002
src/heap.cc:6002: GCTracer::~GCTracer() {
On 2014/07/17 15:26:14, Hannes Payer wrote:
> I guess we can remove the destructor.

Done.

https://codereview.chromium.org/403543002/diff/1/src/heap.cc#newcode6031
src/heap.cc:6031: spent_in_mutator_ = 0.0;
On 2014/07/17 15:26:14, Hannes Payer wrote:
> That is actually wrong. We would like to record here the time before we
invoked
> the first gc, i.e. start_time_ - program_start_time. Can we fix that here?

Done.

[Hannes Payer (out of office), 2014-07-18 08:36:27]

lgtm

[ernstm, 2014-07-18 08:55:47]

Committed patchset #2 manually as r22465 (presubmit successful).