[MPICH] About syncronism.

[Bruno Simioni]

Hi ererybody!

How can I solve this problem

At id = 0

double x1,fr;

Do
{
   MPI_Recv(&x2, 1, MPI_DOUBLE, 1, 1, MPI_COMM_WORLD, &status);
   fr=x1+x2;
   MPI_Send(&fr, 1, MPI_DOUBLE, 1, 1, MPI_COMM_WORLD,&status);
   .
   .
}

At id = 1

double x2,calc;

Do
{
   MPI_Send(&x2, 1, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD,&status); //
MPI_Send(x2 em 1 -> 0.).
   MPI_Recv(&fr, 1, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD, &status);
   calc=fr*2;
   .
   .
}

Well, that is a slow solution. And that is my present problem. How can I
optimize the comunication using non-blocking comunication?
I've tried MPI_Isend and MPI_Irecv, but i thing that I should use MPI_Wait
somewhere (i don't know yet), to provide some syncronism. So, how can I
solve the problem of time and became that syncronous?

Thanks!

-- 
Bruno.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20061014/f074fe2a/attachment.htm>