Dear All,

After an initial misunderstanding on my part, we did establish that
our FiPy/Trilinos configuration allowed the intended distribution of
calculations across multiple processors (at least for the
../examples/diffusion/mesh1D example), after much patient help from
Drs. Guyer and Wheeler and Igor.

Moving on to our own code, which at present subclasses Grid1D, we are
unable to achieve any decrease in execution time.  In fact the
opposite:  we see a slight increase in all our trials, across an order
of magnitude in the number of grid points and 2 orders in time steps.
(Sigh...)  To help understand what's going on, we have 3 general
questions:

1) Is there an example in the FiPy 2.1 release that demontrates a
performance improvement from using the trilinos solvers in parallel?
(That is, is there an example that runs in a certain amount of time
with the pysparse solvers and then less time using mpirun and the
trilinos solvers?)  I'm hoping to compare the set-up for something
that "works" with what we are doing in order to look for clues as to
the difference.

2) Just in a very general way, what features would you expect a FiPy
problem to have that would lend themselves to improved performance
under the parallelization scheme FiPy 2.1 implements?

Thanks and regards...
+jtg+

Reply via email to