Hello, I'm not sure you'll like all the things I've done with your pypar, so any remark you've to to, it's welcome. My problem: I'm trying to do a little master-slave parallel program with pypar, something like the standard LAM one-step tutorial. To do so, I've found that the actual pypar implementation lacks of two things (please, please correct me if I'm wrong): 1. MPI constants in the pypar module (like MPI_ANY_SOURCE, MPI_ANY_TAG) 2. The capability to read the MPI_Status structure after an MPI_Recv. for 1. one can always stick the constant -1 inside the code but for 2. I've found that the actual implementation doesn't export MPI_Status to python. So I've just modified mpi.c (and pypar.py, testpypar.py) to 1. export some MPI constants (trivial) 2. export MPI_Status structure members along with the error code, in a "i(iiii)" (touple of an integer and a four-integer other touple) (ugly) 3. modify raw_receive methods accordingly (trivial again, but pypar API slighty changes). Even worse, I've modified an old mpi.c version (before bcat), so please take into account this. I've attached all the stuff (along with the ms.py demo program). If you want to spend some time reviewing it, I'll be thankful. Best Regards /gp -- "customers, snakes." g.p. "Non è il caldo ma sei tu che alzi la temperatura, non i fiori ma sei tu che profumi l'atmosfera, alzati, girati, muoviti, risiediti, non è il caldo ma sei tu che sei bella vera." max "anyone can create a cryptography product that he himself cannot break." Bruce Schneier "If you don't know the things, study them." g.p. "But writing non-free software is another matter. Non-free software is a plan for subjugating and dividing the people who use it--an antisocial practice. Our goal is to bring that practice to a halt." Richard Stallman Gian Paolo Ciceri via B.Diotti 45 - 20153 Milano MI ITALY CTO @ Louise mobile : ++39 348 3658272 eMail : gp.ciceri@acm.org, gp.ciceri@louise.it webSite: http://www.louise.it ICQ : 94620118