dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Nov 11, 2010 at 11:02:21AM +0000, Garth N. Wells wrote:
>
>
> On 11/11/10 10:39, Anders Logg wrote:
> >On Wed, Nov 10, 2010 at 01:27:02PM -0800, Johan Hake wrote:
> >>On Wednesday November 10 2010 13:14:26 noreply@xxxxxxxxxxxxx wrote:
> >>>------------------------------------------------------------
> >>>revno: 5256
> >>>committer: Anders Logg<logg@xxxxxxxxx>
> >>>branch nick: dolfin
> >>>timestamp: Wed 2010-11-10 22:11:59 +0100
> >>>message:
> >>>   Use boost::timer in place of std::clock. Simplifies implementation,
> >>>   but same (slow) speed as std::clock and equally bad precision.
> >>>   boost:timer most likely uses std::clock but at least the implementation
> >>>   is simplified.
> >>
> >>The boost::timer _is_ using std::clock and the implementation is more or less
> >>exactly the same as ours (where) :). I looked into this when I tried to fix
> >>the strange timer bug in PyDOLFIN.
> >
> >Yes, but the implementation is "much" simpler using boost.
> >
> >It's now been replaced by a system-specific call to getttimeofday
> >which has much better speed and accuracy.
> >
>
> Do we have a benchmark to measure the cost of the timing?

No but we should have. I'll see if I can add one.

We do have one for the progress bar, which is affected by the speed of
timing.

--
Anders