[Pw_forum] Re: Solved! Re: Problems running PWscf in parallel

stewart at cnf.cornell.edu stewart at cnf.cornell.edu
Tue Nov 1 23:03:20 CET 2005


Hi Paolo, 

  I took a look at my make.sys file and you are correct.  The LAM variable 
was not defined by the configuration utility in my case and the lack of it 
was the source of my problem.  The fix below did take care of the problem in 
my case, however I agree that use of it in other scenarios could cause 
problems.  I was glad to see the code running in parallel and sent the email 
off before I followed through the implications.  Does the configure utility 
check for the presence of lam-mpi or does this have to be added in manually? 

Thanks, 

Derek 

 

Paolo Giannozzi writes: 

> On Tuesday 01 November 2005 22:06, stewart at cnf.cornell.edu wrote: 
> 
>> Marco Fornari suggested changing the following three lines of f_def.h
>> in the include directory from 
>>
>> #if defined(__LAM) && ( defined (__LINUX) || defined (__LINUX64) )
>> #define MPI_REAL8 MPI_DOUBLE_PRECISION
>> #endif 
>>
>> to 
>>
>> #if defined(__LAM) || ( defined (__LINUX) || defined (__LINUX64) )
>> #define MPI_REAL8 MPI_DOUBLE_PRECISION
>> #endif
> 
> I don't see the difference: if you have a Linux PC, either a 32- or a 64-bit
> machine, and if you #define __LAM, in both cases you have the same result
> (i.e. redefine  MPI_REAL8, which some implementation of LAM don't like,
> to MPI_DOUBLE_PRECISION). If you don't define __LAM you may have a 
> different result, but this is exactly why __LAM is there and why it should
> be #define'd if needed 
> 
>> This change might be a good thing to implement in the next version
>> of the espresso package. 
> 
> it would break other cases, for sure 
> 
> Paolo 
> 
> -- 
> Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
> Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
> Piazza dei Cavalieri 7      I-56126 Pisa, Italy
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
 




More information about the Pw_forum mailing list