Add counters to track the maximum amount of memory committed by the heap.

Add counters to track the maximum amount of memory committed by the heap.

BUG=None
R=bmeurer@chromium.org, ulan@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=17478

[rmcilroy, 2013-10-18 16:30:10]

I'm hoping after this patch we could enable "--print_max_heap_committed" to the
perf bots and avoid relying on parsing "--trace_gc" for the GC heap measurement
benchmarks, thus enabling us to enable them for all benchmarks and on Android.

Also, by adding samples to the Histogram, we can measure this metric in the
chromium perf tests to track maximum memory usage on real web pages.

PTAL.

[rmcilroy, 2013-10-22 11:39:54]

Didn't realize Ulan was on vacation. Hannes, could you give this a review if you
have a moment please?

[Benedikt Meurer, 2013-10-22 11:44:10]

LGTM with nit

https://codereview.chromium.org/29203003/diff/1/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.h#newcode2384
src/spaces.h:2384: from_space_.MaximumCommittedMemory());
Nit: indentation

[Hannes Payer (out of office), 2013-10-22 14:46:07]

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h#newcod...
src/flag-definitions.h:510: "print statistics of the maxiumum memory committed
by the heap GC "
maximum

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h#newcod...
src/flag-definitions.h:510: "print statistics of the maxiumum memory committed
by the heap GC "
"by the heap GC" => "for the heap"

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1127
src/spaces.cc:1127: heap()->UpdateMaximumCommitted();
Instead of calling heap()->UpdateMaximumCommitted() on every update to a given
space, why don't we call it at the end of a GC? I guess it would make sense to
find a single place in the code.

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1554
src/spaces.cc:1554: }
Instead of duplicating this code, why don't we add as SetCapacity method that
takes care of the the maximum as well.

[rmcilroy, 2013-10-23 14:40:22]

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h#newcod...
src/flag-definitions.h:510: "print statistics of the maxiumum memory committed
by the heap GC "
On 2013/10/22 14:46:07, Hannes Payer wrote:
> "by the heap GC" => "for the heap"

Done.

https://codereview.chromium.org/29203003/diff/1/src/flag-definitions.h#newcod...
src/flag-definitions.h:510: "print statistics of the maxiumum memory committed
by the heap GC "
On 2013/10/22 14:46:07, Hannes Payer wrote:
> maximum

Done.

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1127
src/spaces.cc:1127: heap()->UpdateMaximumCommitted();
On 2013/10/22 14:46:07, Hannes Payer wrote:
> Instead of calling heap()->UpdateMaximumCommitted() on every update to a given
> space, why don't we call it at the end of a GC? I guess it would make sense to
> find a single place in the code.

The heap size can be increased when there is no GC happening.  If we only did
this at GC time we would miss the potential maximum overall heap size wouldn't
we?

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1554
src/spaces.cc:1554: }
On 2013/10/22 14:46:07, Hannes Payer wrote:
> Instead of duplicating this code, why don't we add as SetCapacity method that
> takes care of the the maximum as well.

Done.

https://codereview.chromium.org/29203003/diff/1/src/spaces.h
File src/spaces.h (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.h#newcode2384
src/spaces.h:2384: from_space_.MaximumCommittedMemory());
On 2013/10/22 11:44:10, Benedikt Meurer wrote:
> Nit: indentation

Done.

[Hannes Payer (out of office), 2013-10-24 08:12:35]

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1127
src/spaces.cc:1127: heap()->UpdateMaximumCommitted();
On 2013/10/23 14:40:23, rmcilroy wrote:
> On 2013/10/22 14:46:07, Hannes Payer wrote:
> > Instead of calling heap()->UpdateMaximumCommitted() on every update to a
given
> > space, why don't we call it at the end of a GC? I guess it would make sense
to
> > find a single place in the code.
> 
> The heap size can be increased when there is no GC happening.  If we only did
> this at GC time we would miss the potential maximum overall heap size wouldn't
> we?

Beginning of GC is what I meant, that should be fine, right?

[rmcilroy, 2013-10-24 16:34:30]

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1127
src/spaces.cc:1127: heap()->UpdateMaximumCommitted();
On 2013/10/24 08:12:35, Hannes Payer wrote:
> On 2013/10/23 14:40:23, rmcilroy wrote:
> > On 2013/10/22 14:46:07, Hannes Payer wrote:
> > > Instead of calling heap()->UpdateMaximumCommitted() on every update to a
> given
> > > space, why don't we call it at the end of a GC? I guess it would make
sense
> to
> > > find a single place in the code.
> > 
> > The heap size can be increased when there is no GC happening.  If we only
did
> > this at GC time we would miss the potential maximum overall heap size
wouldn't
> > we?
> 
> Beginning of GC is what I meant, that should be fine, right?

Done.

[rmcilroy, 2013-11-04 11:07:25]

On 2013/10/24 16:34:30, rmcilroy wrote:
> https://codereview.chromium.org/29203003/diff/1/src/spaces.cc
> File src/spaces.cc (right):
> 
> https://codereview.chromium.org/29203003/diff/1/src/spaces.cc#newcode1127
> src/spaces.cc:1127: heap()->UpdateMaximumCommitted();
> On 2013/10/24 08:12:35, Hannes Payer wrote:
> > On 2013/10/23 14:40:23, rmcilroy wrote:
> > > On 2013/10/22 14:46:07, Hannes Payer wrote:
> > > > Instead of calling heap()->UpdateMaximumCommitted() on every update to a
> > given
> > > > space, why don't we call it at the end of a GC? I guess it would make
> sense
> > to
> > > > find a single place in the code.
> > > 
> > > The heap size can be increased when there is no GC happening.  If we only
> did
> > > this at GC time we would miss the potential maximum overall heap size
> wouldn't
> > > we?
> > 
> > Beginning of GC is what I meant, that should be fine, right?
> 
> Done.

Ping?

[ulan, 2013-11-05 09:47:19]

LGTM

[rmcilroy, 2013-11-05 10:34:55]

On 2013/11/05 09:47:19, ulan wrote:
> LGTM

Thanks Ulan.  Could you land for me please.

[ulan, 2013-11-05 10:36:33]

Yep, I'll land when the tree opens.

[ulan, 2013-11-05 11:59:50]

Committed patchset #4 manually as r17478 (presubmit successful).