[GRASS-user] issues with r.horizon incombination with multiprocessing

[Anna Petrášová]

On Tue, May 2, 2017 at 10:09 AM, Tinsley Wolfs <twolfs15 at student.aau.dk> wrote:
> Hi list,
>
> Based on a discussion in the list archive, I copied and edited a script that
> parallelizes the r.horizon module across multiple cores (link). I have
> updated this script to work with the current module parameter naming schemes
> where required. Additionally I added a for loop to the main() function, as I
> want it to do the r.horizon module for several areas sequentially (setting
> g.region, etc...).
>
> I have several issues:
>
> 1. Each time it creates the workers for the last r.horizon angles for a list
> item, it also creates the workers for the first angles of the next item in
> the list. But it should finish the first workers, copy the raster from the
> temporary mapsets to the current mapset and delete the temporary mapsets
> before creating new workers and starting the process again.
> 2. It doesn't calculate the last angles of the last area part of the list
> that it loops over. Though I think this is because of the first issue which
> is that the order of execution is off.
>
> Why doesn't it follow the correct order as per the script but move on to the
> next step before finishing the last? I run the script from the simple python
> editor in Grass 7.2.0.


Not sure if that's the problem but it seems this condition on line 232
(unless it messed up the line numbers):

 if(c < cores):

won't ensure all processes will be joined, for example with 4 cores
and 22 processes, you will process and join 5 chunks of 4 processes
and the 2 last ones won't be joined. That's at least what I guessed
from your code, might be something different too.

Anna


>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-user