Update the PartitionAlloc documentation.

Update the PartitionAlloc documentation.

It was written assuming its only caller would be Blink, but that's no longer the
case. Separate the Blink-specific stuff into a new Blink documentation file.

BUG=NONE
R=haraken@chromium.org

Review-Url: https://codereview.chromium.org/2870213002 .
Cr-Commit-Position: refs/heads/master@{#471070}
Committed: https://chromium.googlesource.com/chromium/src/+/0d001e69ad6e768eda29b9c3b7551ffaadcf6de6

[Chris Palmer, 2017-05-09 20:40:44]

Description was changed from

==========
Update the PartitionAlloc documentation.

It was written assuming its only caller would be Blink, but that's no longer the
case. Separate the Blink-specific stuff into a new Blink documentation file.

BUG=NONE
==========

to

==========
Update the PartitionAlloc documentation.

It was written assuming its only caller would be Blink, but that's no longer the
case. Separate the Blink-specific stuff into a new Blink documentation file.

BUG=NONE
==========

[Chris Palmer, 2017-05-09 20:40:45]

palmer@google.com changed reviewers:
+ haraken@chromium.org

[Chris Palmer, 2017-05-09 20:41:07]

palmer@google.com changed reviewers:
+ palmer@google.com

[Chris Palmer, 2017-05-09 20:41:07]

PTAL. Thank you!

[haraken, 2017-05-10 01:02:21]

LGTM

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
File base/allocator/partition_allocator/PartitionAlloc.md (right):

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:15: ## Partitions And
Buckets

Add the following points somewhere:

- The virtual memory overhead of the partition is huge because PartitionAlloc
guarantees that address spaces used for one partition are never reused for other
partitions.

- For performance reasons, it is important to group similar objects in the same
partition. It is not a good idea to mix various random objects in the same
partition.

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:36: ## Performance

Also explicitly mention that the current PartitionAlloc is optimized for the
main thread use case. e.g., PartitionAlloc doesn't have threaded caches.

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:53: Callers can get
lock-free performance using a `SizeSpecificPartitionAllocator`

"lock-free" is a bit confusing. It normally means a data structure that achieves
a proper mutual exclusion without acquiring a lock.

"thread-unsafe" would be a more proper wording.

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
File third_party/WebKit/Source/platform/wtf/allocator/Allocator.md (right):

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
third_party/WebKit/Source/platform/wtf/allocator/Allocator.md:1: # Blink
Allocator Design

Design of PartitionAlloc in Blink

(Blink Allocator is a bit confusing with Oilpan.)

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
third_party/WebKit/Source/platform/wtf/allocator/Allocator.md:20: use this
partition, but may in the future.

Remove this. Node partition is gone.

[Chris Palmer, 2017-05-10 23:45:44]

Thanks!

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
File base/allocator/partition_allocator/PartitionAlloc.md (right):

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:15: ## Partitions And
Buckets
On 2017/05/10 01:02:20, haraken wrote:
> 
> Add the following points somewhere:
> 
> - The virtual memory overhead of the partition is huge because PartitionAlloc
> guarantees that address spaces used for one partition are never reused for
other
> partitions.
> 
> - For performance reasons, it is important to group similar objects in the
same
> partition. It is not a good idea to mix various random objects in the same
> partition.
> 

Added these notes under ## Performance.

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:36: ## Performance
On 2017/05/10 01:02:20, haraken wrote:
> 
> Also explicitly mention that the current PartitionAlloc is optimized for the
> main thread use case. e.g., PartitionAlloc doesn't have threaded caches.

Done.

https://codereview.chromium.org/2870213002/diff/1/base/allocator/partition_al...
base/allocator/partition_allocator/PartitionAlloc.md:53: Callers can get
lock-free performance using a `SizeSpecificPartitionAllocator`
On 2017/05/10 01:02:20, haraken wrote:
> 
> "lock-free" is a bit confusing. It normally means a data structure that
achieves
> a proper mutual exclusion without acquiring a lock.
> 
> "thread-unsafe" would be a more proper wording.
> 

Done.

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
File third_party/WebKit/Source/platform/wtf/allocator/Allocator.md (right):

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
third_party/WebKit/Source/platform/wtf/allocator/Allocator.md:1: # Blink
Allocator Design
On 2017/05/10 01:02:20, haraken wrote:
> 
> Design of PartitionAlloc in Blink
> 
> (Blink Allocator is a bit confusing with Oilpan.)

Done.

https://codereview.chromium.org/2870213002/diff/1/third_party/WebKit/Source/p...
third_party/WebKit/Source/platform/wtf/allocator/Allocator.md:20: use this
partition, but may in the future.
On 2017/05/10 01:02:21, haraken wrote:
> 
> Remove this. Node partition is gone.

Done.

[Chris Palmer, 2017-05-10 23:47:54]

The CQ bit was checked by palmer@google.com

[Chris Palmer, 2017-05-10 23:47:54]

The patchset sent to the CQ was uploaded after l-g-t-m from haraken@chromium.org
Link to the patchset: https://codereview.chromium.org/2870213002/#ps20001
(title: "Respond to comments.")

[commit-bot: I haz the power, 2017-05-10 23:48:52]

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, 2017-05-11 03:23:22]

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

[commit-bot: I haz the power, 2017-05-11 03:23:23]

Try jobs failed on following builders:
  linux_chromium_rel_ng on master.tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[Chris Palmer, 2017-05-11 15:53:15]

The CQ bit was checked by palmer@google.com

[commit-bot: I haz the power, 2017-05-11 15:53:54]

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, 2017-05-11 18:09:33]

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

[commit-bot: I haz the power, 2017-05-11 18:09:34]

Try jobs failed on following builders:
  linux_chromium_rel_ng on master.tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[palmer, 2017-05-11 21:06:15]

Description was changed from

==========
Update the PartitionAlloc documentation.

It was written assuming its only caller would be Blink, but that's no longer the
case. Separate the Blink-specific stuff into a new Blink documentation file.

BUG=NONE
==========

to

==========
Update the PartitionAlloc documentation.

It was written assuming its only caller would be Blink, but that's no longer the
case. Separate the Blink-specific stuff into a new Blink documentation file.

BUG=NONE
R=haraken@chromium.org

Review-Url: https://codereview.chromium.org/2870213002 .
Cr-Commit-Position: refs/heads/master@{#471070}
Committed:
https://chromium.googlesource.com/chromium/src/+/0d001e69ad6e768eda29b9c3b755...
==========

[palmer, 2017-05-11 21:06:16]

Committed patchset #2 (id:20001) manually as
0d001e69ad6e768eda29b9c3b7551ffaadcf6de6 (presubmit successful).