Newsgroups: comp.parallel.mpi
From: Jennifer Phillips <j.k.phillips@larc.nasa.gov>
Subject: implenmenation question: myserious NaNQ's
Organization: NASA Langley Research Center, Hampton, VA  USA
Date: 22 Aug 1995 21:25:57 GMT
Message-ID: <41di15$999@reznor.larc.nasa.gov>

hi all

as of yet i don't know a lot about MPI
but i'm learning
i am adapting a matrix solver written with MPL to use MPI
to date, it seems to work fine for small matrices 
(i.e. up to 1802 equations)
however, when i get up to larger matrices (16,152 eqs)
i get a NaNQ with both MPICH and LAM
yet i get an answer with the original MPL code
there are no errors and the program runs to completion

i'm basically stumped
right now i don't see how i'm running to problems with
the matrix solver itself since i get correct answers 
for small mattrices (also, the array sizes are identical 
for both the MPL and MPI version of the solver)
so i guess my question is: is there any buffer size limits 
in any of the implementations i may be running into 
that i haven't seen?

thanks for your patience

jennifer


