[GRASS-user] Re: [Qgis-user] error with r.cost

[Markus Metz]

On Mon, Apr 11, 2011 at 3:19 PM, Hamish <hamish_b at yahoo.com> wrote:
> Markus wrote:
>> > you need to check the region extent + resolution, probably
>> > you are exceeding the available RAM if you are having too
>> > many rows and columns.
>> >
>> > Get the information with:
>> > g.region -p
>> >
>> > or the respective menu entry.
>
> ambijat wrote:
>> I am using a mask so as to limit my operation.
> ...
>> But, now I am stuck with a new thing and this is with
>> r.walk. I am still performing this operation on F14. But,
>> there is a crash everytime.
>
> is it the the same G_alloc() error message? if not, what does
> the crash report say? if so it means you do not have enough
> memory, which usually happens when the raster region/resolution
> is set to crazy values.
>
> As Markus asked, what does the region information from
> "g.region -p" say? Is it more than ~ 40000x40000 rows and
> columns?
>
Just for reference: r.walk needs about 30MB per 1 million cells, plus
a little overhead. The memory requirements of r.cost and r.walk are
independent of any mask or of the range of input values, and solely
dependent on the number of cells in the current region. The
percent_memory option gives some control over the amount of memory
used by these two modules.

Also, the friction cost input for r.walk should indicate the friction
costs per cell. The output of r.cost is thus not suitable as input for
r.walk because the output of r.cost is cumulative: cumulative costs to
reach a given cell from the nearest start point. Rather calculate
friction costs for each cell from r.slope.aspect, as suggested by
Markus.

Markus M