nanobench: add median and --cpu/--gpu

nanobench: add median and --cpu/--gpu

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/40b32be3718f0f2e01c4a21bb0004b7f93670c42

[mtklein, 2014-07-09 14:46:05]

As the only active user of nanobench I know today, if you've got opinions on
changes I'd love to know about them.

[krajcevski, 2014-07-09 15:12:36]

On 2014/07/09 14:46:05, mtklein wrote:
> As the only active user of nanobench I know today, if you've got opinions on
> changes I'd love to know about them.

LGTM

More statistics are always nice. One additional thing that could be nice is an
--outliers flag that reports:

Outlier: Sample 2/15 (Median: 20s, Sample: 200s)

And then throwing away the outliers determines what the mean is.

[mtklein, 2014-07-09 15:23:10]

On 2014/07/09 15:12:36, krajcevski wrote:
> On 2014/07/09 14:46:05, mtklein wrote:
> > As the only active user of nanobench I know today, if you've got opinions on
> > changes I'd love to know about them.
> 
> LGTM
> 
> More statistics are always nice. One additional thing that could be nice is an
> --outliers flag that reports:
> 
> Outlier: Sample 2/15 (Median: 20s, Sample: 200s)
> 
> And then throwing away the outliers determines what the mean is.

Agreed.  Adding the median here is expressly to have an outlier-resistant
statistic to look at (though min serves this pretty well too).

I'm also interested in adding bootstrapping to get an estimate of how much of
the variance comes from outliers.  My stats knowledge, though, means this will
be at least another few quiet weekends away from understood enough to get it
working reliably.

[mtklein, 2014-07-09 15:25:41]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2014-07-09 15:26:25]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/mtklein@chromium.org/377283002/1

[commit-bot: I haz the power, 2014-07-09 15:46:53]

Change committed as 40b32be3718f0f2e01c4a21bb0004b7f93670c42