Refine bench_record and bench_playback:

Refine bench_record and bench_playback:

  - use high-precision wall timer only
  - warm caches once before measuring
  - measure independent samples, calculating statistics
  - add --verbose to control how much data we output

Also removed some unloved features from bench_record.

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/90c471e73fb0bd09aef8ad1e3f0842e5d46fb342

[mtklein, 2014-06-16 19:51:57]

These tools are similar to some modes in bench_pictures, but stand alone mostly
because I want to be able to fiddle with them without breaking the bots.  I've
had some good success here getting better measurements out of these tools, and I
figure you might be a good person to take a look at this in case anything seems
applicable for bench/bench_pictures.

[jcgregorio, 2014-06-16 20:12:48]

https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp
File tools/bench_playback.cpp (right):

https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp#new...
tools/bench_playback.cpp:78: draw(*record, *picture, canvas.get());
I know there was talk of throwing away the first draw in bench pictures, but it
was never actually done?

[mtklein, 2014-06-16 20:21:28]

https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp
File tools/bench_playback.cpp (right):

https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp#new...
tools/bench_playback.cpp:78: draw(*record, *picture, canvas.get());
On 2014/06/16 20:12:48, jcgregorio wrote:
> I know there was talk of throwing away the first draw in bench pictures, but
it
> was never actually done?

I would strongly recommend it if any of the data we're recording are affected by
slow outliers (mean, max?).  I saw the first measurement generally as a huge
outlier.  My only hypothesis right now is that this is when the SKP is actually
paged in from disk.

[mtklein, 2014-06-16 20:29:45]

On 2014/06/16 20:21:28, mtklein wrote:
> https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp
> File tools/bench_playback.cpp (right):
> 
>
https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp#new...
> tools/bench_playback.cpp:78: draw(*record, *picture, canvas.get());
> On 2014/06/16 20:12:48, jcgregorio wrote:
> > I know there was talk of throwing away the first draw in bench pictures, but
> it
> > was never actually done?
> 
> I would strongly recommend it if any of the data we're recording are affected
by
> slow outliers (mean, max?).  I saw the first measurement generally as a huge
> outlier.  My only hypothesis right now is that this is when the SKP is
actually
> paged in from disk.

I'm not entirely sure (one reason I wrote bench_record and bench_playback is
that I get lost in bench_pictures code) but it looks like we do already run a
throwaway iteration at tools/PictureBenchmark.cpp, lines 73-82.

[jcgregorio, 2014-06-16 20:42:37]

On 2014/06/16 20:29:45, mtklein wrote:
> On 2014/06/16 20:21:28, mtklein wrote:
> > https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp
> > File tools/bench_playback.cpp (right):
> > 
> >
>
https://codereview.chromium.org/338203002/diff/1/tools/bench_playback.cpp#new...
> > tools/bench_playback.cpp:78: draw(*record, *picture, canvas.get());
> > On 2014/06/16 20:12:48, jcgregorio wrote:
> > > I know there was talk of throwing away the first draw in bench pictures,
but
> > it
> > > was never actually done?
> > 
> > I would strongly recommend it if any of the data we're recording are
affected
> by
> > slow outliers (mean, max?).  I saw the first measurement generally as a huge
> > outlier.  My only hypothesis right now is that this is when the SKP is
> actually
> > paged in from disk.
> 
> I'm not entirely sure (one reason I wrote bench_record and bench_playback is
> that I get lost in bench_pictures code) but it looks like we do already run a
> throwaway iteration at tools/PictureBenchmark.cpp, lines 73-82.

Cool, lgtm.

[mtklein, 2014-06-16 20:52:26]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2014-06-16 20:53:14]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/mtklein@chromium.org/338203002/1

[commit-bot: I haz the power, 2014-06-16 21:04:48]

Change committed as 90c471e73fb0bd09aef8ad1e3f0842e5d46fb342