[GRASS5] milestones

Helena Mitasova - staff helena at gis.uiuc.edu
Sat Jan 27 17:09:30 EST 2001

>Parallelize math-intense GRASS modules (interpolation etc)

This is just my opinion/comment from previous experiences:

I would not put this into milestones - there are more 
important things to do (e.g. fix and substantially improve
NVIZ and map making capabilities, support for dynamic
simulations and fix and develop the volume capabilities
to a usable stage, etc...)
There have been people parallelizing GRASS functions including 
interpolation for years with various success, however the impact is
minimal because computing on parallel machines is still
such hassle that it is easier to run it 2 days on your workstation.
Untill many people have parallel machines on their desks (with more than
2 processors) or distributed computing through internet becomes as
accessible as www such a work will have few users.
Also scientifically, parallelizing interpolation is trivial, 
so I would not even do it as a research project. 
Lubos has been working with parallel
computers (from Cray to PC clusters) for more than 10 years
for his physics research so we occasionally run some geoscience
stuff too, however, so far, it hasn't grown into anything that
I would want to do all time. Some people may have different
applications for which it may be more useful.

There are many ways how to make
e.g. s.surf.rst run faster on one processor. In fact,
what would be really nice for interpolation is to take all
what we have - 2D, 3D, 4D, + crossvalidation for automatic
parameters optimization, etc.  and re-implement it as a library
for building multivariate interpolation commands (as it was 
originally intended) - with
2D, 3D, 4D RST as examples of such commands. It would be
a nice PhD work, but my experience is that because you
cannot really show too much new results with it - not too many people 
notice/care that it runs faster or that the implementation is cleaner
and there has been little stimulation to do such a project -
there are probably more important things to do, such as the vector stuff
and Lubos and I need to do the path sampling modules - see my next


If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo at geog.uni-hannover.de with
subject 'unsubscribe grass5'

More information about the grass-dev mailing list