[GRASS-dev] Parallel computing for r.sun

[Hamish]

oh yeah, and I should probably say a little about MPI as that's
most of what we're running on our cluster here.

well, it's dominated by the I/O problem: let's get multithreading
working first. :) MPI is good for the massively-CPU bound problems,
which again I think are mostly the v.surf.* interpolation ones.
It can be more or less efficent than OpenMP depending on the
class of problem, but MPI requires much deeper changes to the
modules and a lot more work setting up the ring of systems.


Hamish