the -n option of r.mapcalc

[Lars Schylberg]

I have lately been doing a lot of processing on a very large 
data set ( raster images 10000 x 20000 pixels ).  I have started
to think how I could select the most effective value of the
-n option of r.mapcalc.  I am doing neighbourhood operations
that in some instances uses up 6 row above and 6 rows below current 
line.  That would imply that I should use at least use "-n 13" I think.

But how much more could I raise this value before the program
starts to swap to disk ?  I am using a Sun sparc 1+ with 16 Mb of
memory.  How much of the resources can the r.mapcalc
program can use of that memory.

Could some rule of thumb be specified how to figure the most
effective -n value given a certain image width.

Lars Schylberg

National Land Survey of Sweden           Dept. of Photogrammetry
R&D Department                           Royal Inst. of Technology
S-80112 Gavle, Sweden                    10044 Stockholm, Sweden

larss at lmv.lm.se                          larss at fmi.kth.se