Adding a short-running pyauto-based Angry Birds performance test to perf.py.

Adding a short-running pyauto-based Angry Birds performance test to perf.py.

The test loads the Angry Birds webapp, lets it run for a minute, then measures
CPU utilization and v8 heap size.

BUG=chromium-os:23230
TEST=None

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=111258

[Nirnimesh, 2011-11-19 02:31:14]

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf.py
File chrome/test/functional/perf.py (right):

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf....
chrome/test/functional/perf.py:1153:
self.assertTrue(loaded_tab_title.find(url_title_substring) >= 0,
better written as: url_title_substring in loaded_tab_title

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf....
chrome/test/functional/perf.py:1158: # Let the app run for 1 minute.
Don't you want to measure cpu usage over a range of 1 min (like you with FPS for
webgl tests)? You could still emit one value -- the average, which might be more
repeatable.

[dennis_jeffrey, 2011-11-22 23:36:37]

Only addressed the review comments; other changes are due to syncing with
tip-of-tree.

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf.py
File chrome/test/functional/perf.py (right):

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf....
chrome/test/functional/perf.py:1153:
self.assertTrue(loaded_tab_title.find(url_title_substring) >= 0,
On 2011/11/19 02:31:14, Nirnimesh wrote:
> better written as: url_title_substring in loaded_tab_title

Done.

http://codereview.chromium.org/8566053/diff/2001/chrome/test/functional/perf....
chrome/test/functional/perf.py:1158: # Let the app run for 1 minute.
On 2011/11/19 02:31:14, Nirnimesh wrote:
> Don't you want to measure cpu usage over a range of 1 min (like you with FPS
for
> webgl tests)? You could still emit one value -- the average, which might be
more
> repeatable.

Ignoring this as per our offline discussion.  With the current implementation,
we're getting relatively consistent results.  For Chrome desktop and ChromeOS,
running this test 10 times each:

Chrome desktop:
CPU utilization result was 0.03 (8 instances), 0.04 (2 instances)

ChromeOS:
CPU utilization result was 0.50 (4 instances), 0.51 (5 instances), and 0.54 (1
instance)

[Nirnimesh, 2011-11-22 23:59:06]

LGTM