Clean up mark-compact phases and GC counter names.

Clean up mark-compact phases and GC counter names.

BUG=chromium:568495
LOG=NO

Committed: https://crrev.com/feed943f8176222a8555462766f7dcb42d17bf41
Cr-Commit-Position: refs/heads/master@{#32791}

[ulan, 2015-12-10 15:57:49]

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

[ulan, 2015-12-10 15:57:50]

PTAL, I'll follow up with scavenge clean up and unification with histograms.

[Michael Lippautz, 2015-12-10 16:25:53]

Are you planning on updating
  tools/eval_gc_times.sh
? We should not forget about it ;)

https://codereview.chromium.org/1515503006/diff/60001/src/heap/gc-tracer.cc
File src/heap/gc-tracer.cc (right):

https://codereview.chromium.org/1515503006/diff/60001/src/heap/gc-tracer.cc#n...
src/heap/gc-tracer.cc:514: "clear_code_flush=%.1f "
nit/suggestion: We could use a dot "." to separate the hierarchy levels. This
way clear_code_flush would be clear.code_flush making it clear where the split
is.

(although it's also pretty obvious in the current way)

https://codereview.chromium.org/1515503006/diff/60001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1515503006/diff/60001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:3463: GCTracer::Scope::MC_EVACUATE_CLEAN_UP);
Not all items here are related to evacuation candidate? I am thinking about 
  inner_pointer_to_code_cache()->Flush()
  weak_object_to_code_table()->Rehash(...)
  ClearMarkingStateOfLiveObjects()

[ulan, 2015-12-10 18:51:06]

Thanks!

I updated scripts in tools.

https://codereview.chromium.org/1515503006/diff/60001/src/heap/gc-tracer.cc
File src/heap/gc-tracer.cc (right):

https://codereview.chromium.org/1515503006/diff/60001/src/heap/gc-tracer.cc#n...
src/heap/gc-tracer.cc:514: "clear_code_flush=%.1f "
On 2015/12/10 16:25:53, Michael Lippautz wrote:
> nit/suggestion: We could use a dot "." to separate the hierarchy levels. This
> way clear_code_flush would be clear.code_flush making it clear where the split
> is.
> 
> (although it's also pretty obvious in the current way)

Done.

https://codereview.chromium.org/1515503006/diff/60001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1515503006/diff/60001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:3463: GCTracer::Scope::MC_EVACUATE_CLEAN_UP);
On 2015/12/10 16:25:53, Michael Lippautz wrote:
> Not all items here are related to evacuation candidate? I am thinking about 
>   inner_pointer_to_code_cache()->Flush()
>   weak_object_to_code_table()->Rehash(...)
>   ClearMarkingStateOfLiveObjects()

Done. Moved those to Finish phase.

[ulan, 2015-12-10 18:52:24]

Example of output:

➜  src git:(master) ✗ ./v8/tools/eval_gc_time.sh old-gen-rank max test
pause
  len: 10
  min: 8.5
  max: 23.5
  avg: 13.2
evacuate
  len: 10
  min: 1.4
  max: 15.3
  avg: 7.03
clear
  len: 10
  min: 1.0
  max: 7.0
  avg: 3.1
evacuate.new_space
  len: 10
  min: 0.1
  max: 6.7
  avg: 4.21
mark
  len: 10
  min: 0.2
  max: 6.2
  avg: 1.34
mark.roots
  len: 10
  min: 0.0
  max: 5.1
  avg: 0.59
evacuate.candidates
  len: 10
  min: 0.0
  max: 4.5
  avg: 0.67
evacuate.update_pointers
  len: 10
  min: 0.7
  max: 4.5
  avg: 2.14
external
  len: 10
  min: 0.1
  max: 3.0
  avg: 0.81
clear.store_buffer
  len: 10
  min: 0.0
  max: 2.8
  avg: 0.74
evacuate.update_pointers.weak
  len: 10
  min: 0.1
  max: 2.8
  avg: 1.07
mark.weak_closure
  len: 10
  min: 0.0
  max: 2.4
  avg: 0.55
finish
  len: 10
  min: 0.0
  max: 2.3
  avg: 0.35
clear.slots_buffer
  len: 10
  min: 0.0
  max: 1.9
  avg: 0.26
clear.weak_lists
  len: 10
  min: 0.0
  max: 1.8
  avg: 0.81
clear.code_flush
  len: 10
  min: 0.0
  max: 1.7
  avg: 0.38
evacuate.update_pointers.to_new
  len: 10
  min: 0.1
  max: 1.7
  avg: 0.74
evacuate.update_pointers.between_evacuated
  len: 10
  min: 0.0
  max: 1.3
  avg: 0.22
mark.prepare_code_flush
  len: 10
  min: 0.0
  max: 1.0
  avg: 0.11
clear.weak_cells
  len: 10
  min: 0.0
  max: 0.6
  avg: 0.22
evacuate.update_pointers.to_evacuated
  len: 10
  min: 0.0
  max: 0.6
  avg: 0.12
clear.string_table
  len: 10
  min: 0.1
  max: 0.5
  avg: 0.38
sweep
  len: 10
  min: 0.2
  max: 0.4
  avg: 0.3
sweep.old
  len: 10
  min: 0.1
  max: 0.3
  avg: 0.2
clear.maps
  len: 10
  min: 0.0
  max: 0.2
  avg: 0.14
clear.dependent_code
  len: 10
  min: 0.0
  max: 0.1
  avg: 0.04
mark.finish_incremental
  len: 10
  min: 0.0
  max: 0.1
  avg: 0.01
sweep.code
  len: 10
  min: 0.0
  max: 0.1
  avg: 0.01
clear.global_handles
  len: 10
  min: 0.0
  max: 0.0
  avg: 0.0
clear.weak_collections
  len: 10
  min: 0.0
  max: 0.0
  avg: 0.0
evacuate.clean_up
  len: 10
  min: 0.0
  max: 0.0
  avg: 0.0
sweep.map
  len: 10
  min: 0.0
  max: 0.0
  avg: 0.0
incremental_finalize
  len: 10
  min: 0.0
  max: 0.0
  avg: 0.0

[Michael Lippautz, 2015-12-11 08:25:14]

LGTM from my side

> pause
>   avg: 13.2
> evacuate
>   avg: 7.03
> clear
>   avg: 3.1
> mark
>   avg: 1.34
> external
>   avg: 0.81
> finish
>   avg: 0.35
> sweep
>   avg: 0.3
> incremental_finalize
>   avg: 0.0

Sum: 12.93. Looks like there's no hiding anymore.

[Hannes Payer (out of office), 2015-12-11 09:27:31]

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (left):

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:3832: MoveEvacuationCandidatesToEndOfPagesList();
I would leave this call in here, move them there to avoid sweeping.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:367: #ifdef VERIFY_HEAP
It would be nice to move this one into MarkLiveObjects, to the end. It just
cares about marked objects so that should work.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:381: if (marking_parity_ == EVEN_MARKING_PARITY) {
Move that into finish.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:2085: #ifdef VERIFY_HEAP
We could move the debugging code into ClearInvalidStoreAndSlotsBufferEntries

[ulan, 2015-12-11 09:35:33]

Thanks!

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (left):

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:3832: MoveEvacuationCandidatesToEndOfPagesList();
On 2015/12/11 09:27:31, Hannes Payer wrote:
> I would leave this call in here, move them there to avoid sweeping.

Done.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:367: #ifdef VERIFY_HEAP
On 2015/12/11 09:27:31, Hannes Payer wrote:
> It would be nice to move this one into MarkLiveObjects, to the end. It just
> cares about marked objects so that should work.

It depends on ClearNonLiveReferences which removes (marked -> not marked)
references.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:381: if (marking_parity_ == EVEN_MARKING_PARITY) {
On 2015/12/11 09:27:31, Hannes Payer wrote:
> Move that into finish.

Done.

https://codereview.chromium.org/1515503006/diff/120001/src/heap/mark-compact....
src/heap/mark-compact.cc:2085: #ifdef VERIFY_HEAP
On 2015/12/11 09:27:31, Hannes Payer wrote:
> We could move the debugging code into ClearInvalidStoreAndSlotsBufferEntries

Done.

[Hannes Payer (out of office), 2015-12-11 09:42:52]

lgtm

[ulan, 2015-12-11 10:14:40]

The CQ bit was checked by ulan@chromium.org

[ulan, 2015-12-11 10:14:40]

The patchset sent to the CQ was uploaded after l-g-t-m from
mlippautz@chromium.org
Link to the patchset: https://codereview.chromium.org/1515503006/#ps140001
(title: "Address comments from Hannes")

[commit-bot: I haz the power, 2015-12-11 10:14:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1515503006/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1515503006/140001

[commit-bot: I haz the power, 2015-12-11 11:25:50]

Committed patchset #8 (id:140001)

[commit-bot: I haz the power, 2015-12-11 11:26:28]

Description was changed from

==========
Clean up mark-compact phases and GC counter names.

BUG=chromium:568495
LOG=NO
==========

to

==========
Clean up mark-compact phases and GC counter names.

BUG=chromium:568495
LOG=NO

Committed: https://crrev.com/feed943f8176222a8555462766f7dcb42d17bf41
Cr-Commit-Position: refs/heads/master@{#32791}
==========

[commit-bot: I haz the power, 2015-12-11 11:26:30]

Patchset 8 (id:??) landed as
https://crrev.com/feed943f8176222a8555462766f7dcb42d17bf41
Cr-Commit-Position: refs/heads/master@{#32791}