* Richard Treumann <treumann_at_[hidden]> [2008-03-03 12:01:12]:
> The real concern here is:
> Does the MPI implementation need to provide data conversion services
> between any pair of tasks in the communiator? If I have some tasks on a
> slow node and other tasks on a fast one we could debate whether that is
> heterogeneous.
> I think the proposal has merit but we need to be specific that only data
> representation conversion for data transfer routines is involved in the
> meaning of "heterogeneous".
This is what I meant:
that's why I talked about not specifying the correct datatype, but instead
using MPI_BYTE and relying on the fact that the in-memory representation
on both ranks is the same.
So, my definition of 'heterogenous':
not having the same in-memory data representation.
Greetings,
Dries
> > > Some applications are not written with heterogenous environments in
> > > mind;
> > > They do not (always) use correct datatype descriptions when sending or
> > > receiving data, but instead treat the data as an array of bytes,
> > > relying on all datatypes having the same memory representation on both
> > > sender and receiver. Most often, this is done to avoid the added
> > > complexity
> > > of creating the correct datatypes.
- application/pgp-signature attachment: stored
|
![[Flag]](../../../../../images/flag-us.gif){kind=small}