> > There is protection in the openmp code to force runtime
> OMP_NUM_THREADS to 1
> > (using an openMP call) for all the other MKL routines, like
> vdinvsqrt.
>
> Does it make sense to use the openMP call to set OMP_NUM_THREADS to 1
> if mpi
> is also running? Could that eliminate the "disaster" in the first
> paragraph?
> And/or allow us always to link to the openmp version of the library
> (assuming
> it is available)?
Yes except that it only works if one compiles with openmp and links with an
openmp library. So we could perhaps add it for all compilers that support
openmp. Say ifort for example. With the current flag -DOPENMP. Without
compiling with openmp the call is not available.
Line 1017 of sander.f:
call omp_set_num_threads(1)
All the best
Ross
/\
\/
|\oss Walker
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk | PGP Key available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Aug 19 2009 - 22:17:51 PDT
