android: Add interactive surface statistics viewer

android: Add interactive surface statistics viewer

Add a simple tool for viewing graphics surface statistics interactively.
Sample output:

avg_surface_fps (fps)   frame_lengths (vsyncs)   jank_count (janks)   max_frame_delay (vsyncs)
---------------------   ----------------------   ------------------   ------------------------
            58.00                    1.158                    1                          4
            55.00                    1.215                    2                          7
            57.00                    1.174                    2                          3
            56.00                    1.192                    2                          6
            56.00                    1.196                    3                          3
            54.00                    1.239                    1                          8

BUG=https://b/8364918


Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=190930

[bulach, 2013-03-27 10:05:03]

lgtm, thanks! useful utility.. just some nits and suggestions below:

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
File build/android/surface_stats.py (right):

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:29: merged_results = {}
import collections, then merged_results = collections.defaultdict(list), then
38-41 becomes:
merged_results[name] += value

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:42: for name in merged_results.keys():
nit:
for name, values in merged_results.iteritems():
  merged_results[name] = sum(values) / float(len(values))

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:58: for name, value in results.iteritems():
nit: for name in results.keys() (and below)

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:69: print '%s  ' % value,
nit:
print value.rjust(len(name)) + ' ',

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:71: 
nit: add another \n between all of the functions above (two \n between
top-levels), and also, could do with a _ prefix for all but the main function

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:92: default=1,
nit: type='float'

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:93: help='Time to sleep between statistics
updates.')
nit: (in seconds)

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:108: time.sleep(float(options.delay))
nit: with type='float' above, no need to convert here.

[Sami, 2013-03-27 11:47:39]

Thanks for the great suggestions Marcus. This is much more pythonic now.

I noticed that on more recent Android versions SurfaceFlinger sometimes gives a
bogus timestamp ((1 << 63) - 1) for the very latest frame if it's still being
presented. I've added a workaround to this version. PTAL.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
File build/android/surface_stats.py (right):

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:29: merged_results = {}
On 2013/03/27 10:05:03, bulach wrote:
> import collections, then merged_results = collections.defaultdict(list), then
> 38-41 becomes:
> merged_results[name] += value

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:42: for name in merged_results.keys():
On 2013/03/27 10:05:03, bulach wrote:
> nit:
> for name, values in merged_results.iteritems():
>   merged_results[name] = sum(values) / float(len(values))

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:58: for name, value in results.iteritems():
On 2013/03/27 10:05:03, bulach wrote:
> nit: for name in results.keys() (and below)

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:69: print '%s  ' % value,
On 2013/03/27 10:05:03, bulach wrote:
> nit:
> print value.rjust(len(name)) + ' ',

Comes with batteries indeed :)

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:71: 
On 2013/03/27 10:05:03, bulach wrote:
> nit: add another \n between all of the functions above (two \n between
> top-levels), and also, could do with a _ prefix for all but the main function

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:92: default=1,
On 2013/03/27 10:05:03, bulach wrote:
> nit: type='float'

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:93: help='Time to sleep between statistics
updates.')
On 2013/03/27 10:05:03, bulach wrote:
> nit: (in seconds)

Done.

https://codereview.chromium.org/13046007/diff/2001/build/android/surface_stat...
build/android/surface_stats.py:108: time.sleep(float(options.delay))
On 2013/03/27 10:05:03, bulach wrote:
> nit: with type='float' above, no need to convert here.

Done.

[bulach, 2013-03-27 11:53:50]

lgtm, nice! tiny nit:

https://codereview.chromium.org/13046007/diff/5001/build/android/surface_stat...
File build/android/surface_stats.py (right):

https://codereview.chromium.org/13046007/diff/5001/build/android/surface_stat...
build/android/surface_stats.py:50: height, width, _, _ = struct.unpack('HHHH',
nit: just ignore width too and return height directly

[commit-bot: I haz the power, 2013-03-27 12:03:01]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/skyostil@chromium.org/13046007/10002

[commit-bot: I haz the power, 2013-03-27 12:08:42]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/skyostil@chromium.org/13046007/5004

[Sami, 2013-03-27 12:09:11]

Looks like we can get more than one bad timestamp at the end of the data if
we're unlucky, so I've removed the assertion.

[commit-bot: I haz the power, 2013-03-27 14:05:51]

Change committed as 190930