In LAM, yes, they are different. LAM's datatype support is "good but
not great." Their difference is really an artifact of that.
In Open MPI, when REAL*8 and DOUBLE PRECISION are actually the same
back-end representation, then the MPI datatypes are equivalent as well
(i.e., they generate the same MPI type map). They might not be the
same on 64 bit architectures, for example, where a DOUBLE PRECISION may
be the same thing as a REAL*16.
However, it is always safer to program using an MPI datatype that
matches your real datatype -- so if you're using REAL*8s, you should
use MPI_REAL8. If you're using DOUBLE PRECISION's, you should use
MPI_DOUBLE_PRECISION.
On May 13, 2005, at 11:06 AM, Christophe Peyret wrote:
> does anyone know what is the differnce between MPI_REAL8 and
> MPI_DOUBLE_PRECISION. Indeed, on mpich and other MPI distribution,
> MPI_REAL8 is equivalent to MPI_DOUBLE_PRECISION but not on LAM-MPI ?
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
