CopyOnWriteBuffer::SetSize to smaller size memcpy less.

CopyOnWriteBuffer::SetSize to smaller size memcpy less.

R=kwiberg@webrtc.org

Committed: https://crrev.com/291cd8fac37a2efa026713d198d653bc083881db
Cr-Commit-Position: refs/heads/master@{#14166}

[danilchap, 2016-09-08 16:59:35]

Patchset #1 (id:1) has been deleted

[danilchap, 2016-09-08 17:06:27]

danilchap@webrtc.org changed reviewers:
+ kwiberg@webrtc.org

[danilchap, 2016-09-08 17:06:28]

This optimization checks that when SetSize(12) called on a shared buffer of size
1212, only 12 bytes would be memcopied, where previous version would copy full
1212 and only then discard 1200.
Would help to copy header of an RtpPacket without copying payload.

[kwiberg-webrtc, 2016-09-09 08:28:29]

lgtm

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
File webrtc/base/copyonwritebuffer.h (right):

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
webrtc/base/copyonwritebuffer.h:229: std::max(buffer_->capacity(), size));
You maintain the capacity of the original buffer here. Is that supposed to
happen everywhere? (Because it seems to be done in many but not all places. And
it does cost some, so we shouldn't be doing it unless we intend to.)

[danilchap, 2016-09-09 11:25:33]

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
File webrtc/base/copyonwritebuffer.h (right):

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
webrtc/base/copyonwritebuffer.h:229: std::max(buffer_->capacity(), size));
On 2016/09/09 08:28:29, kwiberg-webrtc wrote:
> You maintain the capacity of the original buffer here. Is that supposed to
> happen everywhere? (Because it seems to be done in many but not all places.
And
> it does cost some, so we shouldn't be doing it unless we intend to.)

Yes. there is no need to reallocate with full capacity after this change.
Code adjusted to use smaller buffer.

[danilchap, 2016-09-09 12:58:44]

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
File webrtc/base/copyonwritebuffer.h (right):

https://codereview.webrtc.org/2317283004/diff/20001/webrtc/base/copyonwritebu...
webrtc/base/copyonwritebuffer.h:229: std::max(buffer_->capacity(), size));
On 2016/09/09 11:25:33, danilchap wrote:
> On 2016/09/09 08:28:29, kwiberg-webrtc wrote:
> > You maintain the capacity of the original buffer here. Is that supposed to
> > happen everywhere? (Because it seems to be done in many but not all places.
> And
> > it does cost some, so we shouldn't be doing it unless we intend to.)
> 
> Yes. there is no need to reallocate with full capacity after this change.
> Code adjusted to use smaller buffer.

Actually, it is debatable if SetSize allowed to decrease capacity (realized I
need a version that keeps capacity).
Since capacity is a topic for neighbor CL, I'll keep old behavior where capacity
is kept and land just less memcpy change here.

[danilchap, 2016-09-09 15:32:46]

Description was changed from

==========
CopyOnWriteBuffer::SetSize to smaller size memcpy less.
==========

to

==========
CopyOnWriteBuffer::SetSize to smaller size memcpy less.

R=kwiberg@webrtc.org

Committed:
https://chromium.googlesource.com/external/webrtc/+/291cd8fac37a2efa026713d19...
==========

[danilchap, 2016-09-09 15:32:50]

Committed patchset #3 (id:60001) manually as
291cd8fac37a2efa026713d198d653bc083881db (presubmit successful).

[commit-bot: I haz the power, 2016-09-09 15:32:53]

Description was changed from

==========
CopyOnWriteBuffer::SetSize to smaller size memcpy less.

R=kwiberg@webrtc.org

Committed:
https://chromium.googlesource.com/external/webrtc/+/291cd8fac37a2efa026713d19...
==========

to

==========
CopyOnWriteBuffer::SetSize to smaller size memcpy less.

R=kwiberg@webrtc.org

Committed: https://crrev.com/291cd8fac37a2efa026713d198d653bc083881db
Cr-Commit-Position: refs/heads/master@{#14166}
==========