[GRASS-user] Hardware config/limits

[Chemin, Yann (IWMI)]

Maybe try to sample points from your contours lines and interpolate that sample of points with standard interpolation modules.

-----Original Message-----
From: grass-user-bounces at lists.osgeo.org [mailto:grass-user-bounces at lists.osgeo.org] On Behalf Of Markus Metz
Sent: Wednesday, June 01, 2011 12:59 PM
To: Glynn Clements
Cc: grass-user at lists.osgeo.org; nunosousa84
Subject: Re: [GRASS-user] Hardware config/limits

On Tue, May 31, 2011 at 11:15 PM, Glynn Clements <glynn at gclements.plus.com> wrote:
>
> nunosousa84 wrote:
>
>> I've noticed that i have serious problems dealing with large raster 
>> maps and some GRASS commands in Grass plugin, like r.surf.contour, 
>> shaded relief, some raster calcs, etc... Im using 1 by 1 resolution, 
>> maybe it is a bit exagerated, but even at 5 by 5 the calculation 
>> process is very slow. Im using a Intel Core 2 Duo E8400 and 4Gb 
>> 8500mhz DDR2 Memory. Im about to do an upgrade, should i opt for 8Gb 
>> memory, or i better opt for a good CPU (maybe quad core)? Whats your experience dictates to improve performance?
>
> Multiple cores won't help; GRASS modules are single-threaded.
>
>> The strange thing is that Ubuntu System Monitor says only one core is 
>> at 100% full capacity but the other is only 20%.
>
> That's to be expected.
>
>> Ans memory usage is only 700Mb, so whats causing all this lag and 
>> slow calc, the bar almost dont moves?
>
>> I just run the r.surf.contour command and it takes infinite time
>
> r.surf.contour is the problem. The documentation says:
>
>        The running of r.surf.contour is very sensitive to the
>        resolution of rasterized vector map. If multiple contour lines
>        go through the same raster, slight anomalies may occur. The
>        speed of r.surf.contour is dependent on how far "apart" the
>        contour lines are from each other (as measured in raster
>        cells). Since a flood fill algorithm is used, the program's
>        running time will grow exponentially with the distance between
>        contour lines.
>
> Maybe there's a faster approach along the lines of r.grow.distance?

Interpolating contour lines to get a surface should be a means of last resort if nothing else is available. For elevation, there are various datasets available, most importantly SRTM which easily beats contour lines in terms of accuracy.

Markus M
_______________________________________________
grass-user mailing list
grass-user at lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user