r.watershed.fast new version

[Markus Metz]

Hi all,

some time ago I posted a new version of r.watershed with substantial
speed increase, but still with slight differences in the outputs
compared to the original version. This problem is solved, whether or not
these differences in flow direction and flow accumulation were critical.
Now the results in flow direction and flow accumulation are 100%
identical to the original version. All other outputs are based on flow
direction and flow accumulation and therefore also identical.

The speed increase for the ram version is the same as with my first
version: e.g. a region with 22,500,000 cells is processed in 1minute 6
seconds, the original version needs 10 hours. The modified seg version
(-m flag set) needs 16 minutes (North Carolina sample data, elevation
resampled to 3 m). No idea how long the original seg version needs for
such a region, more than a day I guess.

The source code of r.watershed.fast available here
http://markus.metz.giswork.googlepages.com/r.watershed_fast_version.tar.gz

If you give this modified version a try, please let me know if it works
or not, I tested it only on Linux 32bit and Linux 64bit.

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

[Dylan Beaudette-2]

On Thursday 09 October 2008, Markus Metz wrote:

> Hi all,
>
> some time ago I posted a new version of r.watershed with substantial
> speed increase, but still with slight differences in the outputs
> compared to the original version. This problem is solved, whether or not
> these differences in flow direction and flow accumulation were critical.
> Now the results in flow direction and flow accumulation are 100%
> identical to the original version. All other outputs are based on flow
> direction and flow accumulation and therefore also identical.
>
> The speed increase for the ram version is the same as with my first
> version: e.g. a region with 22,500,000 cells is processed in 1minute 6
> seconds, the original version needs 10 hours. The modified seg version
> (-m flag set) needs 16 minutes (North Carolina sample data, elevation
> resampled to 3 m). No idea how long the original seg version needs for
> such a region, more than a day I guess.
>
> The source code of r.watershed.fast available here
> http://markus.metz.giswork.googlepages.com/r.watershed_fast_version.tar.gz
>
> If you give this modified version a try, please let me know if it works
> or not, I tested it only on Linux 32bit and Linux 64bit.
>
> Markus

Fantastic news. This should be integrated back into GRASS given that the
output now exactly matches the original. Perhap a temporary move to the GRASS
addons page would be a good start.

Thanks for making this possible.

Cheers,
--
Dylan Beaudette
Soil Resource Laboratory
http://casoilresource.lawr.ucdavis.edu/
University of California at Davis
530.754.7341
_______________________________________________
grass-user mailing list
grass-user@...
http://lists.osgeo.org/mailman/listinfo/grass-user

[Giovanni A.]

I'm already using your fast version for a big work on half the Italy
territory. Only your improvement let me make it with grass, so THANKS!

Giovanni

2008/10/10 Dylan Beaudette <dylan.beaudette@...>:

> On Thursday 09 October 2008, Markus Metz wrote:
>> Hi all,
>>
>> some time ago I posted a new version of r.watershed with substantial
>> speed increase, but still with slight differences in the outputs
>> compared to the original version. This problem is solved, whether or not
>> these differences in flow direction and flow accumulation were critical.
>> Now the results in flow direction and flow accumulation are 100%
>> identical to the original version. All other outputs are based on flow
>> direction and flow accumulation and therefore also identical.
>>
>> The speed increase for the ram version is the same as with my first
>> version: e.g. a region with 22,500,000 cells is processed in 1minute 6
>> seconds, the original version needs 10 hours. The modified seg version
>> (-m flag set) needs 16 minutes (North Carolina sample data, elevation
>> resampled to 3 m). No idea how long the original seg version needs for
>> such a region, more than a day I guess.
>>
>> The source code of r.watershed.fast available here
>> http://markus.metz.giswork.googlepages.com/r.watershed_fast_version.tar.gz
>>
>> If you give this modified version a try, please let me know if it works
>> or not, I tested it only on Linux 32bit and Linux 64bit.
>>
>> Markus
>
> Fantastic news. This should be integrated back into GRASS given that the
> output now exactly matches the original. Perhap a temporary move to the GRASS
> addons page would be a good start.
>
> Thanks for making this possible.
>
> Cheers,
> --
> Dylan Beaudette
> Soil Resource Laboratory
> http://casoilresource.lawr.ucdavis.edu/
> University of California at Davis
> 530.754.7341
> _______________________________________________
> grass-user mailing list
> grass-user@...
> http://lists.osgeo.org/mailman/listinfo/grass-user
>

_______________________________________________
grass-user mailing list
grass-user@...
http://lists.osgeo.org/mailman/listinfo/grass-user

[hamish_b]

re. putting it into grass-addons for testing (go for it!), it would be nice
to have some schedule for moving things into main SVN.

i.e. put it on the official TODO list for 6.4.0 release, which means merging sooner rather than later so it isn't released lightly tested.


the same is true for the Google SoC module replacements and r.viewshed.
those seem to linger in -addons without much feedback.


Hamish



     

_______________________________________________
grass-user mailing list
grass-user@...
http://lists.osgeo.org/mailman/listinfo/grass-user

[Markus Neteler OSGeo]

On Fri, Oct 10, 2008 at 7:14 AM, Hamish <hamish_b@...> wrote:
> re. putting it into grass-addons for testing (go for it!), it would be nice
> to have some schedule for moving things into main SVN.
>
> i.e. put it on the official TODO list for 6.4.0 release, which means merging sooner rather than later so it isn't released lightly tested.
>
>
> the same is true for the Google SoC module replacements and r.viewshed.
> those seem to linger in -addons without much feedback.

It would be just great to have a script to fetch addons into an existing GRASS
installations (like "install.packages()" in R). A first approach could be just
simple, less sophisticated/complex than GEM.

Well, not a really new idea,
Markus
_______________________________________________
grass-user mailing list
grass-user@...
http://lists.osgeo.org/mailman/listinfo/grass-user