[GRASS-dev] s.surf.idw.mpi equivalent for grass6.x

[Paul Kelly]

Hello Yann
The algorithm in s.surf.idw was changed in 5.3.x to be much more 
efficient. IMHO this made the s.surf.idw.mpi not so much use - it seems
to me a much more logical improvement to make the point searching and 
indexing algorithm more efficient in running on all systems than to keep 
the inefficient algorithm and have it run faster on parallel systems! The 
improved algorithm is used in v.surf.idw in 6.0+

Have you any quantitative data, that shows the improved s.surf.idw in 5.3+ 
is still slower than s.surf.idw.mpi. I would be very interested to see 
that and see how it can be improved further. The improved version only 
uses points within the region, and indexes them according to raster cell 
to make them quick to find. The old version searched through *every* site 
in the whole map for each raster cell, in order to find the 12 closest. It 
was very slow for large site maps. I wrote a little bit about the improved 
algorithm in my PhD thesis - see the relevant section here (gzipped 
PostScript), section 4.4.2: 
ftp://ftp.qub.ac.uk/pub/users/paulk/grass/idw.ps.gz

Let me know your thoughts

Paul

On Wed, 10 Jan 2007, Yann Chemin wrote:

> Hi list,
>
> is there any "port" or new version of s.surf.idw.mpi to grass6.x.
> If not, what would be the (rough) direction to take to make it a 6.x program?
>
> thanks
> Yann
>
> _______________________________________________
> grass-dev mailing list
> grass-dev at grass.itc.it
> http://grass.itc.it/mailman/listinfo/grass-dev
>