Telemetry: Add a reduced compositor cases page set.

Telemetry: Add a reduced compositor cases page set.

These are setup to have almost no cpu-time noise. There is
always only one frame source, and we either scroll 4 times
or wait for 10 seconds worth of frames.

Since these are jsbin pages, I've added a '?' label to give
the URL more meaning.

BUG=None

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=269674

[epenner, 2014-04-26 03:46:40]

Ptal.

[epenner, 2014-04-28 22:01:26]

On 2014/04/26 03:46:40, epenner wrote:
> Ptal.

Polite ping. I'm going to be off for a while but I'd like this to land as it
represents a much greater variety of hotspots in the compositor, in addition to
having almost no noise such that regressions can be more easily noticed.

[vmiura, 2014-04-29 01:42:01]

On 2014/04/28 22:01:26, epenner wrote:
> On 2014/04/26 03:46:40, epenner wrote:
> > Ptal.
> 
> Polite ping. I'm going to be off for a while but I'd like this to land as it
> represents a much greater variety of hotspots in the compositor, in addition
to
> having almost no noise such that regressions can be more easily noticed.

I'm not really qualified, but question:

 - Is this removing the old fastpath measure?  Are we OK with losing continuity
on that?

[epennerAtGoogle, 2014-04-29 02:11:40]

On 2014/04/29 01:42:01, vmiura wrote:
> On 2014/04/28 22:01:26, epenner wrote:
> > On 2014/04/26 03:46:40, epenner wrote:
> > > Ptal.
> > 
> > Polite ping. I'm going to be off for a while but I'd like this to land as it
> > represents a much greater variety of hotspots in the compositor, in addition
> to
> > having almost no noise such that regressions can be more easily noticed.
> 
> I'm not really qualified, but question:
> 
>  - Is this removing the old fastpath measure?  Are we OK with losing
continuity
> on that?

Yes it will clobber the old one. I figured there is going to be a discontinuity
anyway when I add new pages, so may we well reset it... However, another
possibility would be to just enable it for _all_ pages and just deal with
sifting through more data for now.

It would be great if we could just collect this on all pages together with
smoothness, and filter after the fact.

[epenner, 2014-05-01 18:22:18]

Rebased. Ptal.

[epenner, 2014-05-01 18:38:43]

On 2014/05/01 18:22:18, epenner wrote:
> Rebased. Ptal.

Oops, I didn't upload the rebase. Not much changed though anyway. Ptal.

[tonyg, 2014-05-02 04:01:04]

I thought that key_mobile_sites was supposed to represent important (or key)
mobile sites that we want to optimize for. The reduced test cases sound to me
like they belong in a new page set, maybe named tough_compositor_cases.py? WDYT?

[epennerAtGoogle, 2014-05-02 18:44:53]

On 2014/05/02 04:01:04, tonyg wrote:
> I thought that key_mobile_sites was supposed to represent important (or key)
> mobile sites that we want to optimize for. The reduced test cases sound to me
> like they belong in a new page set, maybe named tough_compositor_cases.py?
WDYT?

That works for me too. I'll do it that way for these pages. Perhaps then I can
just remove the filter for key-mobile-sites thread-times and track everything.
The filter was there to track a more relevant set initially, which will be
replaced by the reduced cases.

On the flip side, one thing I'm finding is that there is a lot of different
subsets people want to test, or track over time, and inevitably it will create
duplicate work if each one requires a different page set. However, given how
results are presented currently this still sounds like the most reasonable thing
to do.

[epenner, 2014-05-02 20:49:12]

Ptal. I've added a new page set.

I'll add the recordings before landing, and I'll remove the filter on
thread_times.key_mobile_sites in another CL.

[epenner, 2014-05-02 21:56:33]

Added the recordings etc.

[epenner, 2014-05-08 22:19:13]

On 2014/05/02 21:56:33, epenner wrote:
> Added the recordings etc.

Ptal (pretty please with sugar on top)

[vmiura, 2014-05-08 23:19:24]

https://codereview.chromium.org/250813002/diff/220001/tools/perf/page_sets/to...
File tools/perf/page_sets/tough_compositor_cases.py (right):

https://codereview.chromium.org/250813002/diff/220001/tools/perf/page_sets/to...
tools/perf/page_sets/tough_compositor_cases.py:10: class
ToughCompositorPage(page_module.PageWithDefaultRunNavigate):
PageWithDefaultRunNavigate seems to have gone away.

Use page_module.Page?

[dtu, 2014-05-09 00:05:39]

lgtm with nit

https://codereview.chromium.org/250813002/diff/220001/tools/perf/benchmarks/t...
File tools/perf/benchmarks/thread_times.py (right):

https://codereview.chromium.org/250813002/diff/220001/tools/perf/benchmarks/t...
tools/perf/benchmarks/thread_times.py:43: class
ThreadTimesCompositorCases(test.Test):
nit: two blank lines between top-level blocks

https://codereview.chromium.org/250813002/diff/220001/tools/perf/page_sets/to...
File tools/perf/page_sets/tough_compositor_cases.py (right):

https://codereview.chromium.org/250813002/diff/220001/tools/perf/page_sets/to...
tools/perf/page_sets/tough_compositor_cases.py:10: class
ToughCompositorPage(page_module.PageWithDefaultRunNavigate):
On 2014/05/08 23:19:24, vmiura wrote:
> PageWithDefaultRunNavigate seems to have gone away.
> 
> Use page_module.Page?

Yup, that's the right thing.

[epenner, 2014-05-10 06:38:27]

Thanks!

> ThreadTimesCompositorCases(test.Test):
> nit: two blank lines between top-level blocks
> 

Done.

> > Use page_module.Page?
> 
> Yup, that's the right thing.

Done.

[epenner, 2014-05-10 06:38:33]

The CQ bit was checked by epenner@chromium.org

[commit-bot: I haz the power, 2014-05-10 06:43:23]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/epenner@chromium.org/250813002/260001

[commit-bot: I haz the power, 2014-05-11 01:18:34]

Change committed as 269674