[heap] Rework and improve object stats tracing

[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

Committed: https://crrev.com/2b2a1694908210ee8083b6112efe82905beae552
Cr-Commit-Position: refs/heads/master@{#37718}

[Michael Lippautz, 2016-07-12 07:51:27]

Patchset #4 (id:60001) has been deleted

[Michael Lippautz, 2016-07-12 07:57:36]

Patchset #4 (id:80001) has been deleted

[Michael Lippautz, 2016-07-12 07:57:57]

Description was changed from

==========
[heap] Better feedback on object stats

BUG=
==========

to

==========
[heap] Improve object stats tracing

BUG=
==========

[Michael Lippautz, 2016-07-12 13:44:11]

Patchset #5 (id:120001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:16]

Patchset #9 (id:220001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:20]

Patchset #8 (id:200001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:25]

Patchset #7 (id:180001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:30]

Patchset #5 (id:140001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:35]

Patchset #6 (id:240001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:40]

Patchset #5 (id:160001) has been deleted

[Michael Lippautz, 2016-07-12 13:44:49]

Patchset #5 (id:260001) has been deleted

[Michael Lippautz, 2016-07-12 14:38:55]

Patchset #6 (id:300001) has been deleted

[Michael Lippautz, 2016-07-12 14:38:59]

Patchset #5 (id:280001) has been deleted

[Michael Lippautz, 2016-07-13 08:42:20]

Patchset #5 (id:320001) has been deleted

[Michael Lippautz, 2016-07-13 08:42:25]

Patchset #5 (id:340001) has been deleted

[Michael Lippautz, 2016-07-13 08:42:30]

Patchset #5 (id:360001) has been deleted

[Michael Lippautz, 2016-07-13 08:42:35]

Patchset #5 (id:380001) has been deleted

[Michael Lippautz, 2016-07-13 10:04:07]

Patchset #8 (id:460001) has been deleted

[Michael Lippautz, 2016-07-13 10:04:12]

Patchset #7 (id:440001) has been deleted

[Michael Lippautz, 2016-07-13 10:06:51]

Description was changed from

==========
[heap] Improve object stats tracing

BUG=
==========

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in 
  a separate phase in MC. (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown   
  bucket to around ~8% (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

BUG=
==========

[Michael Lippautz, 2016-07-13 10:07:07]

Description was changed from

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in 
  a separate phase in MC. (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown   
  bucket to around ~8% (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

BUG=
==========

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

BUG=
==========

[Michael Lippautz, 2016-07-13 10:09:59]

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

[Michael Lippautz, 2016-07-13 10:10:00]

ulan,hannes: ptal
mythri: fyi

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact....
src/heap/mark-compact.cc:2303: SpaceIterator space_it(heap());
Hannes: Your HeapObjectIterator changes have been reverted, so it still causes a
GC and we cannot use it.

https://codereview.chromium.org/2129173002/diff/480001/src/heap/object-stats.cc
File src/heap/object-stats.cc (left):

https://codereview.chromium.org/2129173002/diff/480001/src/heap/object-stats....
src/heap/object-stats.cc:182: CountFixedArray(object->elements(),
DICTIONARY_ELEMENTS_SUB_TYPE,
This looked wrong because we pass DICTIONARY_* as fast_type and FAST_ELEMENTS_*
as dictionary type. I wonder if we see this accounting somewhere (Should be
reflected by syncing instance counters).

[Hannes Payer (out of office), 2016-07-13 11:39:11]

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact....
src/heap/mark-compact.cc:2303: SpaceIterator space_it(heap());
On 2016/07/13 10:10:00, Michael Lippautz wrote:
> Hannes: Your HeapObjectIterator changes have been reverted, so it still causes
a
> GC and we cannot use it.

Yeah, for snapshot reasons. Just use your custom iterator.

https://codereview.chromium.org/2129173002/diff/520001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/520001/src/heap/mark-compact....
src/heap/mark-compact.cc:2291: if (dead_stats_ != nullptr)
Just for fun: add a DCHECK that the object is not grey.

https://codereview.chromium.org/2129173002/diff/540001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/540001/src/heap/mark-compact....
src/heap/mark-compact.cc:2413: if (FLAG_trace_gc_object_stats) {
Do we need both modes here? I guess we can always record live and dead. Why not
unify and get rid of one flag...

[Michael Lippautz, 2016-07-13 12:03:44]

Patchset #6 (id:420001) has been deleted

[Michael Lippautz, 2016-07-13 12:03:48]

Patchset #3 (id:40001) has been deleted

[Michael Lippautz, 2016-07-13 12:20:17]

All addressed. With the current sub typing info for fixed array we lack sub type
accounting for about 9%.

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/480001/src/heap/mark-compact....
src/heap/mark-compact.cc:2303: SpaceIterator space_it(heap());
On 2016/07/13 11:39:11, Hannes Payer wrote:
> On 2016/07/13 10:10:00, Michael Lippautz wrote:
> > Hannes: Your HeapObjectIterator changes have been reverted, so it still
causes
> a
> > GC and we cannot use it.
> 
> Yeah, for snapshot reasons. Just use your custom iterator.

Acknowledged.

https://codereview.chromium.org/2129173002/diff/520001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/520001/src/heap/mark-compact....
src/heap/mark-compact.cc:2291: if (dead_stats_ != nullptr)
On 2016/07/13 11:39:11, Hannes Payer wrote:
> Just for fun: add a DCHECK that the object is not grey.

Done.

https://codereview.chromium.org/2129173002/diff/540001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/2129173002/diff/540001/src/heap/mark-compact....
src/heap/mark-compact.cc:2413: if (FLAG_trace_gc_object_stats) {
On 2016/07/13 11:39:11, Hannes Payer wrote:
> Do we need both modes here? I guess we can always record live and dead. Why
not
> unify and get rid of one flag...

Done: Refactored to also always record the dead set. If we want we can make
allocation of ObjectStats in Heap conditional, based on the tracking flag.

[Hannes Payer (out of office), 2016-07-13 12:56:00]

lgtm

[Michael Lippautz, 2016-07-13 13:09:31]

Patchset #10 (id:570001) has been deleted

[Michael Lippautz, 2016-07-13 13:31:59]

Patchset #10 (id:590001) has been deleted

[Michael Lippautz, 2016-07-13 13:38:49]

Description was changed from

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

BUG=
==========

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

[Michael Lippautz, 2016-07-13 13:39:00]

Description was changed from

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

[ulan, 2016-07-13 13:43:55]

lgtm

[Michael Lippautz, 2016-07-13 13:47:58]

The CQ bit was checked by mlippautz@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-13 13:48:05]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-13 13:50:36]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-13 13:50:37]

Dry run: Try jobs failed on following builders:
  v8_linux64_avx2_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux64_avx2_rel_ng/buil...)
  v8_linux64_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux64_rel_ng/builds/9009)
  v8_linux_nodcheck_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_nodcheck_rel_ng/bu...)

[Michael Lippautz, 2016-07-13 13:51:34]

The CQ bit was checked by mlippautz@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-13 13:51:48]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-13 13:56:37]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-13 13:56:38]

Dry run: Try jobs failed on following builders:
  v8_win_compile_dbg on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_win_compile_dbg/builds/2...)
  v8_win_nosnap_shared_rel_ng on master.tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_win_nosnap_shared_rel_ng...)

[Michael Lippautz, 2016-07-13 14:25:26]

The CQ bit was checked by mlippautz@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-13 14:25:35]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-13 14:48:45]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-13 14:48:46]

Dry run: This issue passed the CQ dry run.

[Michael Lippautz, 2016-07-13 14:50:03]

The CQ bit was checked by mlippautz@chromium.org

[Michael Lippautz, 2016-07-13 14:50:04]

The patchset sent to the CQ was uploaded after l-g-t-m from hpayer@chromium.org,
ulan@chromium.org
Link to the patchset: https://codereview.chromium.org/2129173002/#ps670001
(title: "Fix compilation on Windows")

[commit-bot: I haz the power, 2016-07-13 14:50:17]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-13 14:54:02]

Description was changed from

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

[commit-bot: I haz the power, 2016-07-13 14:54:03]

Committed patchset #13 (id:670001)

[commit-bot: I haz the power, 2016-07-13 14:55:45]

Description was changed from

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

to

==========
[heap] Rework and improve object stats tracing

- Instead of tracing during marking we can now trace in a separate phase in MC.
  (Heap is iterable.)
- Add more subtypes for fixed arrays, reducing the unknown bucket to around ~8%
  (local run).
- Refactor collection calls to have a single bottleneck.
- Provide JSON-based output format that can be "easily" processed in JS.

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

Committed: https://crrev.com/2b2a1694908210ee8083b6112efe82905beae552
Cr-Commit-Position: refs/heads/master@{#37718}
==========

[commit-bot: I haz the power, 2016-07-13 14:55:47]

Patchset 13 (id:??) landed as
https://crrev.com/2b2a1694908210ee8083b6112efe82905beae552
Cr-Commit-Position: refs/heads/master@{#37718}