Simplify OnReadCompleted() logic that appends bytes to the client buffer

Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer
and copies the data in one step. Unfortunately, there are no performance
benefits in doing so, because the underlying implementation uses
memmove(), which is less efficient than memcpy(); however, the change
simplifies the code.

BUG=724546
Review-Url: https://codereview.chromium.org/2939993002
Cr-Commit-Position: refs/heads/master@{#480455}
Committed: https://chromium.googlesource.com/chromium/src/+/f471163c1112a5ed7685ac2d1f911c922cb08aa0

[kapishnikov, 2017-06-14 22:29:34]

kapishnikov@chromium.org changed reviewers:
+ droger@chromium.org, mef@chromium.org

[kapishnikov, 2017-06-14 22:29:35]

PTAL! This should improve the performance and simplify the code a little.

[kapishnikov, 2017-06-14 23:03:00]

On 2017/06/14 22:29:35, kapishnikov wrote:
> PTAL! This should improve the performance and simplify the code a little.

I actually don't see any performance improvement. The reason is probably because
|appendBytes| uses memmove instead of memcpy, which is more efficient. The
benefit of this CL is only code simplification.

[mef, 2017-06-15 15:29:08]

lgtm, although it is unfortunate that it doesn't save any CPU.

[droger, 2017-06-19 08:45:30]

lgtm

Can you update the CL description if it is incorrect w.r.t. preformance
improvements?

[kapishnikov, 2017-06-19 15:38:29]

Description was changed from

==========
Do not fill with zeros when growing buffer in OnReadCompleted()

Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, we can append the data to the buffer by calling
[NSMutableData appendBytes...]

BUG=724546
==========

to

==========
Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer and
copies
the data in one step. Unfortunately, there are no performance benefits in doing
so, because the underlying implementation uses memmove(), which is less
efficient than memcpy(); however, the change simplifies the code.

BUG=724546
==========

[kapishnikov, 2017-06-19 15:38:56]

Description was changed from

==========
Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer and
copies
the data in one step. Unfortunately, there are no performance benefits in doing
so, because the underlying implementation uses memmove(), which is less
efficient than memcpy(); however, the change simplifies the code.

BUG=724546
==========

to

==========
Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer
and copies the data in one step. Unfortunately, there are no performance
benefits in doing so, because the underlying implementation uses
memmove(), which is less efficient than memcpy(); however, the change
simplifies the code.

BUG=724546
==========

[kapishnikov, 2017-06-19 15:39:24]

On 2017/06/19 08:45:30, droger wrote:
> lgtm
> 
> Can you update the CL description if it is incorrect w.r.t. preformance
> improvements?

Thanks! Done.

[kapishnikov, 2017-06-19 15:39:29]

The CQ bit was checked by kapishnikov@chromium.org

[commit-bot: I haz the power, 2017-06-19 15:39:50]

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, 2017-06-19 15:59:31]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1497886769334610, "parent_rev":
"6d3c6496933a57ef2aeebc9772bb1d3e18fec352", "commit_rev":
"f471163c1112a5ed7685ac2d1f911c922cb08aa0"}

[commit-bot: I haz the power, 2017-06-19 15:59:47]

Description was changed from

==========
Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer
and copies the data in one step. Unfortunately, there are no performance
benefits in doing so, because the underlying implementation uses
memmove(), which is less efficient than memcpy(); however, the change
simplifies the code.

BUG=724546
==========

to

==========
Calling [NSMutableData increaseLengthBy:bytes_read] increases the
NSMutableData length and fills the newly allocated memory with zeros.
According to profiling, filling with zeros may take approximately 1.4%
of the total network thread CPU utilization. Instead of growing
the buffer, call [NSMutableData appendBytes...], which grows the buffer
and copies the data in one step. Unfortunately, there are no performance
benefits in doing so, because the underlying implementation uses
memmove(), which is less efficient than memcpy(); however, the change
simplifies the code.

BUG=724546

Review-Url: https://codereview.chromium.org/2939993002
Cr-Commit-Position: refs/heads/master@{#480455}
Committed:
https://chromium.googlesource.com/chromium/src/+/f471163c1112a5ed7685ac2d1f91...
==========

[commit-bot: I haz the power, 2017-06-19 15:59:49]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/chromium/src/+/f471163c1112a5ed7685ac2d1f91...