[GRASSLIST:8830] Re: GRASS and xgrid

[Paul Kelly]

Hello James

On Tue, 1 Nov 2005, James Conolly wrote:

> xgrid is a method for parallel processing, not a data format. The issue is 
> that I'd like to run many (many) r.los (or r.cva on a large sample) - I've 
> estimated what I need is in the region of a few months+ processing time. I've 
> got xgrid running in my lab, but there are issues (a) with the grass shell 
> and (b) the fact that the code for r.cva or r.los isn't parallel. I could 
> break the job down into chunks and send it to xgrid to have it farmed out, 
> but I'd like to know if anybody has already tried this.

I feel a better long-term solution would be to re-write r.los/r.cva to use 
a more efficient algorithm. For a long time I've wanted to try and 
implement the algorithm described in

"A fast algorithm for approximate viewshed computation" by David 
Izraelevitz (U.S. Army Corps of Engineers) in Photogrammetric Engineering 
and Remote Sensing vol. 69 no. 7, July 2003 pp 767-774

but never had time and I don't really do any GRASS development work now so 
it won't happy any time soon. I think I've mentioned it before on the list 
but just thought it would be worthwhile mentioning it again.

Paul