[PATCH 2 of 7] run-tests: add more scheduling weight hints

[Matt Mackall]

On Thu, 2015-12-17 at 05:40 +0000, Martin von Zweigbergk wrote:
> On Wed, Dec 16, 2015 at 7:02 PM Matt Mackall <mpm at selenic.com> wrote:
> 
> > # HG changeset patch
> > # User Matt Mackall <mpm at selenic.com>
> > # Date 1449270320 21600
> > #      Fri Dec 04 17:05:20 2015 -0600
> > # Node ID 38d094b896775480d5097c38c3383e8dcab5825b
> > # Parent  dd64b1794321538bc5c1d20045376ff75c6d76d4
> > run-tests: add more scheduling weight hints
> > 
> > The scheduler would like to order test execution by expected run-time,
> > but doesn't know much about how long a test will run. It thus uses
> > test size as a proxy for run-time. By tweaking these weights we can
> > keep CPUs more evenly busy and thus finish sooner.
> > 
> 
> Do you think it would make sense to save the run-time of each test to a
> file and use the previous run-time as an estimate for the current run?

Ooh, yes. I hadn't considered doing it dynamically.

I had been considering a linear regression of length vs runtime to
discover outlier weights and statically update them somehow, but simply
storing the last runtime or a history of runtimes is much more
straightforward.

-- 
Mathematics is the supreme nostalgia of our time.