[Pw_forum] Pw_forum Digest, Vol 24, Issue 58

[Axel Kohlmeyer]

On Mon, 2009-06-29 at 19:04 +0200, Lorenzo Paulatto wrote:
> In data 29 giugno 2009 alle ore 18:47:11, leila salimi  
> <l.salimi at ph.iut.ac.ir> ha scritto:
> > from ewald : error #         1
> >      optimal alpha not found
> >
> > what should I do now?
> 
>  From what I can see, it looks like the xlf version of function erfc is not  
> working as it should, or at least as we expect it to. It may take a while  
> before it is possible to fix or workaround the problem in an "official"  
> way.

you may want to test if the 

--allow-multiple-definition

flag works. this is usually required if you compile
with xlf on a Linux OS with power cpu and you want
to use the optimized math functions from -lmass.

the linux linker by default does not allow multiple
definitions of the same "symbol" that are different
(checked by comparing the sizes). other linkers are
more forgiving and use the first occurance of a
symbol after it has been flagged as undefined.

the flag from above is needed on BG/L for example
when using the IBM optimized library versions.

cheers,
   axel.

> 
> For now, I can propose this other temporary workaround:
> 1. put the line you've removed from Makefile back in there, so that the  
> QE-internal version of erfc is compiled,
> 2. open include/f_defs.h and add the following line (it does not matter  
> where):
> #define erfc erfc_qe
> 3. open PW/ewald.f90 and add the following line at the very beginning:
> #include "f_defs.h"
> 4. add the same line also at the beginning of flib/erf.f90
> 4. recompile
> 
> This shall be enough for now (not for car-parrinello, though).
> 
> I'm not sure why you're having this weird problem, I've personally  
> compiled QE on SP5 and SP6 machines, using the xlf compiler without any  
> such problem (although editing make.sys by hand). Can you please attach  
> you copy of make.sys (from the QE source directory)?
> 
> best regards
> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.