MPI package questions / discussion

[Ouzounis_Georgios]

Hello,

I have been reading the code from the existing mpi package on sourceforge
and I would like to ask a few questions.

To my understanding, the goal is to give the octave user the ability to use
the mpi functions as he would do if he was writting code on C/C++. So the
octave user should be able to use functions like MPI_Init() and
MPI_Finalize().

I don't quite get the way MPI_Init() is implemented in the MPI_Init.cc file.
The DLD function that is written in this file simply calls the MPI_Init()
and then it returns. Is this functional? Shouldn't there be an
MPI_Finalize() call in the same function that MPI_Init() is called? 
I can't see the way an MPI_Init() call can be wrapped in a DLD function
without the concurrent existence of an MPI_Finalize() call. After all, a DLD
function is still a C++ function right? Calling MPI_Init() in a C++ function
means you have to call MPI_Finalize() too.

And also, in the future, when a user will want to start a number of parallel
jobs on octave using mpi, how will he be able to do it? What commands need
to be implemented? Will he have to write something like "mpi open 4" (4
jobs) or something else?





--
View this message in context: 
http://octave.1599824.n4.nabble.com/MPI-package-questions-discussion-tp4663098.html
Sent from the Octave - Maintainers mailing list archive at Nabble.com.