> >Well, calling an MPI function by remote process to assure progress is a
> >SPECIFIC ACTION. This requirement imposed on the MPI-2 applications is again
> >the principles of 1-sided communication.
> Well, waiting for my next breath is a specific action, but one that will
> satisfy most applications. Exactly where is your application going to
> break on a polled implementation? We're not talking about realtime MPI, yet.
The point is that some applications (including computational chemistry codes
developed at PNL) need truly 1-sided communication. These applications are not
looking for an MPI API to functionality that appears like 1-sided communication
but has unpredictable performance characteristics of implicit 2-sided
communication. The polling in the communication library might yield extremely
high latencies which are very hard to incorporate into the performance model and
make the programming model very complex.