[GRASS5] SOurce code and clustering

[Andreas Lange]

Rich Shepard wrote:
> 
> On Mon, 17 Apr 2000, Bernhard Reiter wrote:
> 
> > It would make sense for some modules to utilise more than one process, but
> > this is not completly trivial to do.
> 
> Bernhard,
> 
>   Local resources include the Oregon Graduate Institute (a private graduate
> school in the middle of the Silicon Forest here) and The Portland Group (a
> high-performace compiler developer), as well as interested individuals in
> the local linux community. This is something I want to investigate because I
> believe that it could be very useful in many simulation or modeling
> situations.
> 
> Rich

Hi,

as far as i remember a group from the university of heidelberg
investigated/evaluated this for grass4.x. But the link to the homepage
no longer exists (or i could'nt find it again). See the documentation in
grass4.3/src.alpha/doc_modules_heidelberg (in german!). 

If i remember correct they found that it is not possible to run grass on
a multiprocessor system or cluster with reasonable effort. I have no
experience with that, but i think that it is very complicated as the
grass libraries have not been developed with this in mind. Think of map
calculations for raster maps: for some calculations you need the value
for the nth row to calculate the nth+1 row, which will produce
problems/overhead in a distributed enviroment. 
I fear that even a multi-processor machine will spent most time waiting
for input/output or memory operations to complete, so that a faster bus
system (e. g. crossbar-architecture in SGIs big irons) will give better
performance than more CPUs. 

Only some thoughts, 

Andreas
-- 
Andreas Lange, 65187 Wiesbaden, Germany, Tel. +49 611 807850
Andreas.Lange at Rhein-Main.de, A.C.Lange at GMX.net

---------------------------------------- 
If you want to unsubscribe from GRASS Development
Team internal mailing list write to:
minordomo at geog.uni-hannover.de with
subject 'unsubscribe grass5'
length: 3109
max: 0