[Wien] Segmentation fault in xcpot3, running lapw0 (ifc8.1, AMDopteron,

Alessandro Mirone mirone at esrf.fr
Wed Jan 25 18:55:13 CET 2006


Dear Wien community,

   first of all thanks for this great mailing list from which I learned 
a lot.

I would like to ask your help about this problem  that, after searching 
the archives ,  it seems new to me    :

The problem happens
 for a beryllium supercell, containing 28 equivalents atom ( more 
precisely an hexagonal supercell with 96 atoms
 and one atom slightly shifted)

For smaller supercells the program runs smoothly.

The crash ( segmentation fault )  happens in lapw0, in routine xcpot3.

I suspect that it happens during the creation of temporary arrays 
because  ( putting print' s inside )
no single line of xcpot3 is executed. ( and the temp. arrays have size 
around 10Mb )
 
My configuration is (ifc8.1, AMDopteron, suse9.2, libgoto)

I have read a last year post from Efimov reporting a problem with 
reallocate and ifc8.0/1.
Reallocation happens at the beginning of lapw0, reading the structure : 
the array pos
is reallocated to pos(3,index).

As segmentation fault does not occur inside reallocation routines I am 
wondering
if it is worth to try  to patch the reallocation routines (but it could 
be because of the strange behaviour
of  segmentation problems which may  manifest themselves later)

What do you think about this problem and about the way I could fix it or 
try to fix it?

Best regards
    Alessandro Mirone




More information about the Wien mailing list