[heap] Add more tasks for parallel compaction

[heap] Add more tasks for parallel compaction

- We now compute the number of parallel compaction tasks, depending on the
  evacuation candidate list, the number of cores, and some hard limit.
- Free memory is moved over to compaction tasks (up to some limit)
- Moving over memory is done by dividing the free list of a given space up among
  other free lists. Since this is potentially slow we limit the maximum amount
  of moved memory.

BUG=chromium:524425
LOG=N

Committed: https://crrev.com/0e842418835eea85886a06cf37052895bc8a17db
Cr-Commit-Position: refs/heads/master@{#30886}

[Michael Lippautz, 2015-09-23 07:13:08]

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

[Michael Lippautz, 2015-09-23 07:13:09]

WDYT?

[Hannes Payer (out of office), 2015-09-23 07:26:03]

https://codereview.chromium.org/1354383002/diff/80001/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/flag-definitions.h#...
src/flag-definitions.h:663: DEFINE_BOOL(parallel_compaction, true, "use parallel
compaction")
This should be done in a separate cl, makes it easier to revert.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:3399: // Contribute in main thread.
// Perform compaction on the main thread.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h
File src/heap/mark-compact.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h...
src/heap/mark-compact.h:714: int NumberOfParallelCompactionTasks() {
Implement this in the cc file. That makes it faster to recompile when playing
with this metric. The include can then also move to the cc file.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h...
src/heap/mark-compact.h:719: // - a hard limit
Make the magic limits named constants.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h#newco...
src/heap/spaces.h:1634: void Divide(FreeList** other_free_lists, int num,
intptr_t limit);
This code should be mentioned in the CL description.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h#newco...
src/heap/spaces.h:1634: void Divide(FreeList** other_free_lists, int num,
intptr_t limit);
Add a description what divide is doing.

[Michael Lippautz, 2015-09-23 07:40:50]

https://codereview.chromium.org/1354383002/diff/80001/src/flag-definitions.h
File src/flag-definitions.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/flag-definitions.h#...
src/flag-definitions.h:663: DEFINE_BOOL(parallel_compaction, true, "use parallel
compaction")
On 2015/09/23 07:26:03, Hannes Payer wrote:
> This should be done in a separate cl, makes it easier to revert.

Sure, I only enabled it for testing on the perf bots.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:3399: // Contribute in main thread.
On 2015/09/23 07:26:03, Hannes Payer wrote:
> // Perform compaction on the main thread.

Done.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h
File src/heap/mark-compact.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h...
src/heap/mark-compact.h:714: int NumberOfParallelCompactionTasks() {
On 2015/09/23 07:26:03, Hannes Payer wrote:
> Implement this in the cc file. That makes it faster to recompile when playing
> with this metric. The include can then also move to the cc file.

Done.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/mark-compact.h...
src/heap/mark-compact.h:719: // - a hard limit
On 2015/09/23 07:26:03, Hannes Payer wrote:
> Make the magic limits named constants.

Done.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h#newco...
src/heap/spaces.h:1634: void Divide(FreeList** other_free_lists, int num,
intptr_t limit);
On 2015/09/23 07:26:03, Hannes Payer wrote:
> This code should be mentioned in the CL description.

Done.

https://codereview.chromium.org/1354383002/diff/80001/src/heap/spaces.h#newco...
src/heap/spaces.h:1634: void Divide(FreeList** other_free_lists, int num,
intptr_t limit);
On 2015/09/23 07:26:03, Hannes Payer wrote:
> Add a description what divide is doing.

Done.

[Hannes Payer (out of office), 2015-09-23 07:53:12]

lgtm

[Michael Lippautz, 2015-09-23 07:53:33]

The CQ bit was checked by mlippautz@chromium.org

[commit-bot: I haz the power, 2015-09-23 07:53:48]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1354383002/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1354383002/100001

[commit-bot: I haz the power, 2015-09-23 08:06:04]

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

[commit-bot: I haz the power, 2015-09-23 08:06:05]

Try jobs failed on following builders:
  v8_mac_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_mac_rel/builds/9969)

[Michael Lippautz, 2015-09-23 08:38:17]

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

[commit-bot: I haz the power, 2015-09-23 08:38:26]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1354383002/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1354383002/120001

[Hannes Payer (out of office), 2015-09-23 08:47:22]

https://codereview.chromium.org/1354383002/diff/120001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/120001/src/heap/spaces.h#newc...
src/heap/spaces.h:2006: void AddExternalMemory(Address start, int size_in_bytes)
{
Call it AddMemory, to get the naming in sync with RemoveMemory.

https://codereview.chromium.org/1354383002/diff/120001/src/heap/spaces.h#newc...
src/heap/spaces.h:2011: void RemoveMemory(int size) {
accounting_stats_.DecreaseCapacity(size); }
To make this one dual with AddMemory, it should actually also remove memory from
the free list. WDYT?

[commit-bot: I haz the power, 2015-09-23 09:03:07]

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

[commit-bot: I haz the power, 2015-09-23 09:03:08]

Dry run: This issue passed the CQ dry run.

[Michael Lippautz, 2015-09-23 11:49:02]

Please have a look at the restructured methods for adding and removing memory.

https://codereview.chromium.org/1354383002/diff/120001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/120001/src/heap/spaces.h#newc...
src/heap/spaces.h:2011: void RemoveMemory(int size) {
accounting_stats_.DecreaseCapacity(size); }
On 2015/09/23 08:47:22, Hannes Payer wrote:
> To make this one dual with AddMemory, it should actually also remove memory
from
> the free list. WDYT?

Done.

[Hannes Payer (out of office), 2015-09-23 12:00:43]

lgtm

https://codereview.chromium.org/1354383002/diff/180001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/180001/src/heap/spaces.h#newc...
src/heap/spaces.h:1697: enum FreeListCat { kSmall, kMedium, kLarge, kHuge };
FreeListCategoryType?

[Michael Lippautz, 2015-09-23 12:03:59]

https://codereview.chromium.org/1354383002/diff/180001/src/heap/spaces.h
File src/heap/spaces.h (right):

https://codereview.chromium.org/1354383002/diff/180001/src/heap/spaces.h#newc...
src/heap/spaces.h:1697: enum FreeListCat { kSmall, kMedium, kLarge, kHuge };
On 2015/09/23 12:00:43, Hannes Payer wrote:
> FreeListCategoryType?

Done.

[Michael Lippautz, 2015-09-23 12:04:15]

The CQ bit was checked by mlippautz@chromium.org

[Michael Lippautz, 2015-09-23 12:04:15]

The patchset sent to the CQ was uploaded after l-g-t-m from hpayer@chromium.org
Link to the patchset: https://codereview.chromium.org/1354383002/#ps200001
(title: "Addressed comments round 2")

[commit-bot: I haz the power, 2015-09-23 12:04:23]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1354383002/200001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1354383002/200001

[commit-bot: I haz the power, 2015-09-23 12:28:59]

Committed patchset #11 (id:200001)

[commit-bot: I haz the power, 2015-09-23 12:29:21]

Patchset 11 (id:??) landed as
https://crrev.com/0e842418835eea85886a06cf37052895bc8a17db
Cr-Commit-Position: refs/heads/master@{#30886}

[Michael Achenbach, 2015-09-23 13:46:22]

A revert of this CL (patchset #11 id:200001) has been created in
https://codereview.chromium.org/1356363005/ by machenbach@chromium.org.

The reason for reverting is: [Sheriff] May have caused this new flake:
http://build.chromium.org/p/client.v8/builders/V8%20Linux/builds/5412
.