dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Nov 30, 2010 at 11:00:20AM +0100, Johannes Ring wrote:
> On Tue, Nov 30, 2010 at 10:49 AM, Garth N. Wells <gnw20@xxxxxxxxx> wrote:
> >
> >
> > On 30/11/10 09:40, Johannes Ring wrote:
> >>
> >> On Mon, Nov 29, 2010 at 10:12 PM, Anders Logg<logg@xxxxxxxxx>  wrote:
> >>>
> >>> On Mon, Nov 29, 2010 at 04:47:47PM +0000, Garth N. Wells wrote:
> >>>>
> >>>>
> >>>> On 29/11/10 16:38, Anders Logg wrote:
> >>>>>
> >>>>> Why were those not detected by my compiler? Isn't DDEBUG set by
> >>>>> cmake.local?
> >>>>>
> >>>>
> >>>> It should be. cmake.local has:
> >>>>
> >>>>     -DCMAKE_BUILD_TYPE=Developer
> >>>>
> >>>> Garth
> >>>
> >>> It does set DEBUG, but it also sets NDEBUG, which is what
> >>> /usr/include/assert.h looks for.
> >>>
> >>> Why do we set NDEBUG? I can't see that we set it anywhere. Johannes?
> >>
> >> The problem was that in bench/fem/multicore/cpp/CMakeLists.txt the
> >> CMAKE_BUILD_TYPE variable was forced to 'Release' and thereby
> >> overwriting the 'Developer' build type that is set in cmake.local.
> >> This should be fixed now.
> >>
> >
> > It should be set to 'Release' since it's a performance benchmark. I build
> > DOLFIN out-of-source (which is the right way to do things), so whatever is
> > in bench/fem/multicore/cpp/CMakeLists.txt doesn't affect the building of the
> > DOLFIN libraries.
>
> Yes it does when benchmarks are enabled like in cmake.local.

This makes some sense. If we want to run benchmarks, then the whole of
DOLFIN should be compiled in 'Release' mode or otherwise it's not much
use.

Which implies cmake.local should remain in 'Developer' mode and those
the flags should not be touched in the individual benchmarks (so that they
are the same as for the rest of DOLFIN).

I assume the benchbot is building in Release mode. Is that correct?

--
Anders