Optimize ParseRbsp method in H264 bitstream parser.

Optimize ParseRbsp method in H264 bitstream parser.

After profiling, sakal@ found that this method was taking very long,
and causing the bitstream parsing to take up to 1ms per frame. The
culprit proved to be rtc::Buffer::AppendData, which was called for
every byte and subsequently calls memcpy.

BUG=webrtc:7293
Review-Url: https://codereview.webrtc.org/2728093002
Cr-Commit-Position: refs/heads/master@{#17051}
Committed: https://chromium.googlesource.com/external/webrtc/+/156e3afa7f945069b206c3e252ea3609869bfad5

[kthelgason, 2017-03-03 10:16:56]

The CQ bit was checked by kthelgason@webrtc.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-03 10:17:07]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[kthelgason, 2017-03-03 10:18:46]

kthelgason@webrtc.org changed reviewers:
+ magjed@webrtc.org

[kthelgason, 2017-03-03 10:18:47]

Magnus, please take a look when you can.

The perf improvement here is quite significant. In release configuration, the
bitstream parsing goes from taking roughly 2% of total cpu time to less than
0.2% in my testing.

[commit-bot: I haz the power, 2017-03-03 10:41:02]

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

[commit-bot: I haz the power, 2017-03-03 10:41:03]

Dry run: Try jobs failed on following builders:
  linux_asan on master.tryserver.webrtc (JOB_FAILED,
http://build.chromium.org/p/tryserver.webrtc/builders/linux_asan/builds/22716)

[kthelgason, 2017-03-03 10:46:59]

The CQ bit was checked by kthelgason@webrtc.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-03 10:47:03]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[commit-bot: I haz the power, 2017-03-03 11:32:45]

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

[commit-bot: I haz the power, 2017-03-03 11:32:46]

Dry run: This issue passed the CQ dry run.

[magjed_webrtc, 2017-03-03 13:14:14]

Awesome! Maybe we should file a bug just for tracking purposes so we don't
forget this.

https://codereview.webrtc.org/2728093002/diff/20001/webrtc/common_video/h264/...
File webrtc/common_video/h264/h264_common.cc (right):

https://codereview.webrtc.org/2728093002/diff/20001/webrtc/common_video/h264/...
webrtc/common_video/h264/h264_common.cc:64: std::unique_ptr<uint8_t[]> out(new
uint8_t[length]);
I think it would be better with an std::vector<uint8_t> here, and use push_back
instead of |j|. Also, we can change the return type to std::vector and that
removes an extra copy.

[kthelgason, 2017-03-03 14:50:15]

Description was changed from

==========
Optimize ParseRbsp method in H264 bitstream parser.

After profiling, sakal@ found that this method was taking very long,
and causing the bitstream parsing to take up to 1ms per frame. The
culprit proved to be rtc::Buffer::AppendData, which was called for
every byte and subsequently calls memcpy.

BUG=None
==========

to

==========
Optimize ParseRbsp method in H264 bitstream parser.

After profiling, sakal@ found that this method was taking very long,
and causing the bitstream parsing to take up to 1ms per frame. The
culprit proved to be rtc::Buffer::AppendData, which was called for
every byte and subsequently calls memcpy.

BUG=webrtc:7293
==========

[kthelgason, 2017-03-03 14:50:54]

On 2017/03/03 13:14:14, magjed_webrtc wrote:
> Awesome! Maybe we should file a bug just for tracking purposes so we don't
> forget this.

I filed https://bugs.chromium.org/p/webrtc/issues/detail?id=7293 to track this
work.

> I think it would be better with an std::vector<uint8_t> here, and use
push_back
> instead of |j|. Also, we can change the return type to std::vector and that
> removes an extra copy.

Done.

[kthelgason, 2017-03-03 15:05:59]

The CQ bit was checked by kthelgason@webrtc.org to run a CQ dry run

[commit-bot: I haz the power, 2017-03-03 15:06:02]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[magjed_webrtc, 2017-03-03 15:10:32]

lgtm

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
File webrtc/common_video/h264/h264_bitstream_parser.cc (right):

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
webrtc/common_video/h264/h264_bitstream_parser.cc:16: #include
"webrtc/base/bytebuffer.h"
Can you remove this include?

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
File webrtc/common_video/h264/pps_parser.cc (right):

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
webrtc/common_video/h264/pps_parser.cc:18: #include "webrtc/base/buffer.h"
Can you remove this include?

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
File webrtc/common_video/h264/sps_parser.cc (right):

https://codereview.webrtc.org/2728093002/diff/40001/webrtc/common_video/h264/...
webrtc/common_video/h264/sps_parser.cc:18: #include "webrtc/base/bytebuffer.h"
can you remove this include?

[commit-bot: I haz the power, 2017-03-03 16:15:04]

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

[commit-bot: I haz the power, 2017-03-03 16:15:05]

Dry run: This issue passed the CQ dry run.

[kthelgason, 2017-03-06 07:42:57]

The CQ bit was checked by kthelgason@webrtc.org

[kthelgason, 2017-03-06 07:42:57]

The patchset sent to the CQ was uploaded after l-g-t-m from magjed@webrtc.org
Link to the patchset: https://codereview.webrtc.org/2728093002/#ps60001 (title:
"Remove unused includes")

[commit-bot: I haz the power, 2017-03-06 07:43:00]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[commit-bot: I haz the power, 2017-03-06 08:04:33]

CQ is committing da patch.

Bot data: {"patchset_id": 60001, "attempt_start_ts": 1488786177370160,
"parent_rev": "0a3e69a9a4ab1c4d5fa8fe523a8ce960846605ef", "commit_rev":
"156e3afa7f945069b206c3e252ea3609869bfad5"}

[commit-bot: I haz the power, 2017-03-06 08:04:36]

Description was changed from

==========
Optimize ParseRbsp method in H264 bitstream parser.

After profiling, sakal@ found that this method was taking very long,
and causing the bitstream parsing to take up to 1ms per frame. The
culprit proved to be rtc::Buffer::AppendData, which was called for
every byte and subsequently calls memcpy.

BUG=webrtc:7293
==========

to

==========
Optimize ParseRbsp method in H264 bitstream parser.

After profiling, sakal@ found that this method was taking very long,
and causing the bitstream parsing to take up to 1ms per frame. The
culprit proved to be rtc::Buffer::AppendData, which was called for
every byte and subsequently calls memcpy.

BUG=webrtc:7293

Review-Url: https://codereview.webrtc.org/2728093002
Cr-Commit-Position: refs/heads/master@{#17051}
Committed:
https://chromium.googlesource.com/external/webrtc/+/156e3afa7f945069b206c3e25...
==========

[commit-bot: I haz the power, 2017-03-06 08:04:37]

Committed patchset #4 (id:60001) as
https://chromium.googlesource.com/external/webrtc/+/156e3afa7f945069b206c3e25...