On Fri, Apr 05, 2013 at 01:29:36AM +0200, Roland Kaufmann wrote:
> I checked with the master branch of opm-core, as well as the DUNE
> modules. It seems that they all build release builds as default, so I
> guess we should do it here too. The pull-request OPM/opm-core#226
> implement this change. (And documents it in the README).
> 

But this due to auto-tools having this as the default. As far as I
understand the CMake documentation, cmakes default is to not choose a
buildtype and in this case honor user supplied flags.

I think it is a matter of taste which default to use (auto-tools or
CMake's). In the light of OPM using CMake adhering to the CMake policy
seems to be more intriguing. No surprise for CMake users and a little
surprise for DUNE users. (I usually set my flags manually when
debugging and for productive runs.)

Anyway, if it is prominently documented and I can overwrite the
default values, then I am OK with it.


> As of pull request OPM/opm-core#225 you should now be able to specify
> optimization flags through the environment variable CXXFLAGS, or the
> CMake "cache" variable CMAKE_CXX_FLAGS_RELEASE (or _DEBUG).
> 

Nice.

> >>>Suspects are these options: -mtune and -flto .
> 
> >>I rather suspect that it is -flto.
> >> If you have a version of GCC < 4.7
> 

4.7.2

> I am currently on Ubuntu Precise, which is (sort of) a derivative of
> Wheezy, but I cannot reproduce this problem in neither GCC 4.6, 4.7
> nor 4.8, all of them with -flto. binutils is 2.22 and Boost is 1.53.
> 
> However, you can now disable lto altogether with --disable-lto, or
> -DWHOLE_PROG_OPTIM=OFF.
> 

Might be a binutils bug
http://sourceware.org/bugzilla/show_bug.cgi?id=14272 

If I find the time, will test it on Debian stable, too.

Markus

-- 
Dr. Markus Blatt - HPC-Simulation-Software & Services http://www.dr-blatt.de
Hans-Bunte-Str. 8-10, 69123 Heidelberg, Germany,  USt-Id: DE279960836
Tel.: +49 (0) 160 97590858  Fax: +49 (0)322 1108991658

Please notice the new address!

_______________________________________________
Opm mailing list
[email protected]
http://www.opm-project.org/mailman/listinfo/opm

Reply via email to