[GRASS-dev] Candidate modules for OpenMP parallelization

[Anna Petrášová]

Great you can be a mentor! Have you registered as a mentor yet? If not
please ask OSGeo GSoC admins to help you with that.

Regarding the modules, I consider r.watershed as too complex for this. The
student doesn't have extensive experience with OpenMP, neither have I, so I
would prefer to choose something less complex. If you feel confident it's
doable with your help, I am open to that. r.stats looks easier. What I so
far suggested:
* r.proj - Markus Metz tried it 10 years ago, it was not quite working, but
perhaps we could try again with different approach, it would be an
impactful module to speed up
* r.horizon - seems doable, computes maps for different angles, so it looks
like a good fit, likely some rewrite needed
* r.texture - not as impactful, but perhaps easy when we use a similar
approach Aaron used
* v/r.surf.idw - IDW is slow, this could help a lot and it could be
impactful as well.
* r.fill.stats - it's already pretty fast, but it could be relatively easy

The student is interested in the shorter project, so realistically, I don't
think we can do more than 3 modules.
As I said, I am definitely not an expert on this, so I would appreciate
some feedback for others.

Anna

On Fri, Mar 31, 2023 at 7:24 AM Huidae Cho <grass4u at gmail.com> wrote:

> Hi Anna,
>
> I'm not sure if these modules are "low hanging fruit" and even fairly
> parallelizable, but r.stats and r.watershed might be good candidates
> because r.stats is generic with no output rasters, and may be used a lot in
> different fields, and r.watershed calculates several important hydrologic
> rasters that are often used as input to other modules (i.e., as a first
> step for most hydrologic analyses based on my experience). Maybe,
> r.statistics as well?
>
> Since I now have access to HPC, I can help mentor him with some testing.
>
> Best,
> Huidae
>
> On Thu, Mar 30, 2023 at 9:42 AM Anna Petrášová <kratochanna at gmail.com>
> wrote:
>
>> Hi devs,
>>
>> We have a GSoC candidate who would like to do OpenMP parallelization. I
>> could mentor, anybody else is interested?
>>
>> At this point, we need to figure out (at least tentatively) which modules
>> he could work on. After Aaron's work in GSoC 2021, there are not that many
>> modules left that could be described as "low hanging fruit". Do you have
>> any suggestions?
>>
>> The proposal is due very soon.
>>
>> Thanks,
>> Anna
>>
>
>
> --
> Huidae Cho, Ph.D., GISP, /hidɛ <http://ipa-reader.xyz/?text=hid%C9%9B>
> t͡ɕo/, 조희대, 曺喜大
> GRASS GIS Developer
> https://idea.isnew.info/
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20230331/d1be5347/attachment.htm>