[Wien] SegFault in lapw2 when using the -fc switch

[Stefaan Cottenier]

This probleem seems to appear on some architectures in combination with 
particular mkl/ifort versions. It shows up indeed when forces are 
calculated, and (probably) only in cases without inversion symmetry (-c).

I know no systematic cure (anyone?). Lowering the optimization level of 
lapw2 during compilation, and/or trying several mkl/ifort versions does 
help. But which combination works and which one does not, seems to 
depend on the computer system.

(anecdote: one of our otherwise very well-informed sysadmins blamed a 
programming bug for this, and refused to believe that this code was able 
to run on any architecture -- until being demonstrated a counter-example).

Stefaan


On 16/05/2012 9:22, vivien badaut wrote:
> Dear WIEN2k users
>
> I am using WIEN2k since a few month, and I am studying the compound
> IrTe2 on a local cluster (Centos, with compilator with ifort 11.1) since
> a few weeks without any significant problem.
> I converged successfully the system to -ec 0.00000001 and -cc 0.0001
> without any problem.
> However, I recently tried to minimize the positions of the atoms in the
> cell.
> During the procedure, after a few iterations, I get a SegFault in lapw2.
> I was able to trace it back to the use of the -fc switch un run_lapw,
> which causes this error - independently of the value set for the forces
> convergence criteria.
> Additionally, when I get this SegFault, the files are then corrupted and
> I am unable to relaunch the calculation even after suppressing:
> *broyd* and *vector. I am now very stuck and I might say a bit
> frustrated, as I have no idea of where to look to solve it.
>
> Any ideas of parameters/values to tests or modules to check are welcome.
>
> I attached the *struct, and you can use the default setting for
> generating in1 and in2.`
>
> Regards,
>
> Vivien Badaut
> Computational Condensed Matter Physics Laboratory,
> RIKEN, ASI, Japan
>
>
>
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien