[Pw_forum] Fortran compiler

Henning Glawe glaweh at physik.fu-berlin.de
Sun Mar 28 21:05:35 CEST 2010


On Wed, Mar 24, 2010 at 12:25:01PM +0100, Miguel Martinez Canales wrote:
> Dear all,
> 
> Following Paolo's warning, I've performed some tests with a not so  
> surprising result: Paolo was right. For some reason, the ubuntu 9.10  
> builds of gfortran 4.3 and 4.4.1 give astonishingly different results  
> for example07. Well, the pw.x output is the same as in the reference,  
> but the lambda values... at least they are of the same order of  
> magnitude. Mind you, for once this is not an ubuntu bug, as it also  
> happens in self-compiled gfortran 4.4.3 (Red Hat 5.4 in this case).  
> However, the differences in lambda don't seem important in the case of  
> gfortran 4.1.2 (the default in Red Hat) or 4.2.4 (the oldest version  
> available in ubuntu 9.10).
> 
> I'm going to check other examples to see whether there are very  
> different phonon frequencies elsewhere, but I don't expect anything  
> good at all.

I am in contact with one of the gfortran developers, and reported the problem
with ph.x to him; he now spent quite some time tracing the problem down, and
apparently there is a patch on-queue:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43551#c10

-- 
c u
henning


More information about the Pw_forum mailing list