[GRASS-dev] Multi-threading and use of multiple processors

[Hamish]

Colin wrote:
> >> A colleague of mine would like to run a cost analysis
> >> (with either r.walk or r.cost) on a very very large
> >> raster, ~ 1500 km x 800 km with about 30 m resolution
> >> equals about 40 billion cells.
> >
> > [ see my other today's email about r.terracost now in
> > Addons ]
> >
> >> Is there any plan to add multi-threading to these or other
> >> modules (I think I remember something about r.mapcalc and
> >> another a while back)?
Markus:
> > Yann Chemin parallelized i.atcorr for me with openMP which
> > required the addition of a few lines and a gcc >= 4.2
> > compiler. For hints, see
> > http://grass.osgeo.org/wiki/OpenMP
> >
> > Perhaps this can be introduced at library level at some
> > point. The behavior or a pixel oriented module such as
> > i.atcorr is of course different from that of r.cost which
> > has to consider more than a pixel in a cycle.

Colin wrote:
> In my experience, the main drain in computation time is the
> IO not the actual calculation of the btree. Is there any way
> to parallelize the IO by passing rows to different cores or
> something like that ie. "for (row = 0; row < nrows; row++) {"?
> 
> Or a more efficient segment library sounds good as well. Is
> there a way to estimate how much more efficient would it be and
> would it make use of multiple cores?


FYI there is a Google Summer of Code student proposal to add
OpenMP support to a number of GRASS modules and/or libraries.

I would encourage anyone interested or knowledgeable in such things to get in touch with Wolf to sign up to be a mentor. Then you can review and comment on the OSGeo applications etc.

The better the plan & the better the mentoring, the more successful the student's application and (if accepted) project
will be.


regards,
Hamish