[Dune-devel] [SOCIS 2015]: Application

Aleksejs Fomins aleksejs.fomins at lspr.ch
Wed Apr 29 09:40:02 CEST 2015


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear Philipp,

I proposed the project 2.

CurvilinearGrid is a new grid manager we have been working on now for almost a year. It has now passed most consistency checks, and we are waiting to run a FEM code on it soon to make sure it can deal with real world problems, before releasing it to the community.

As can be derived from the name, this grid manager can handle curvilinear elements, that is, reading them from a GMSH .msh file, creating a mesh, and performing all sorts of necessary operations like integrating over such elements, or transforming between local and global coordinates.

As part of this grid, a powerful CurvilinearGMSHReader was developed, that can
* Read GMSH .msh files in parallel on multiple processes
* Read Curvilinear meshes
* Pre-partition the mesh using ParMetis

Initial idea of the summer project was to be centered about this reader, namely
* Adding optional functionality to read internal boundaries and insert them into a grid. That is, additional (physical) surfaces, defined by having same entity tag.
* Adding functionality to read periodic boundaries from GMSH. To the best of my knowledge, currently periodic boundaries in Dune are available using DGF format, so you can re-use many ideas to implement it for the GMSH reader. Further, this will include extending the functionality of the CurvilinearGrid to cope with such boundaries (to a large extent writing additional the parallel communication)

However, depending on your interests, quite a few other topics are of large interest to us, namely
* Helping integrate CurvilinearGrid into community. When the grid will be released, adjusting the interface to make it comfortable for users will be very beneficial.
* Recursive integration. Currently, Dune provides quadrature rules, which allow integrating over an entity effectively by approximating the function with a n-th order polynomial and obtaining an exact result. However, in CurvilinearGrid, non-polynomial integrands naturally arise (e.g. surface areas). Therefore, the optimal quadrature order is not known a priori. Therefore, it is very useful to have a tool that can integrate over non-polynomial smooth functions efficiently.
* Point location. In order to locate the element in which the point resides, one usually iterates over all elements and checks them one-by-one. But for the curvilinear grid a single check is very expensive, since Newton's method has to be employed to locate the local coordinate of a point inside the element. To reduce the computational cost, an OCTree must be implemented, which subdivides all elements into a tree by their geometric location, and thus cuts the number of elements to be checked dramatically.


Since it is unlikely that you will have time for all of the above, it is good if you choose according to your own interests when writing a proposal. You are of course encouraged to make your own suggestions for the aims as well.



Do not hesitate to contact me or other members of the community if you have further questions.
I appreciate your interest

Kind regards,
Aleksejs




On 29/04/15 08:16, Christoph GrĂ¼ninger wrote:
> Hi Philipp,
> I am glad to hear that you apply for Dune's SOCIS 2015 slot.
> There are no more information publicly available then at
> http://users.dune-project.org/projects/esa-socis-2015/wiki/Project_ideas
> The idea is, that you contact your mentors, present them a proposal
> which includes your ideas. This will be difficult as the deadline is
> going to end tomorrow.
> You will be enrolled as a student during the summer? That's a
> requirement of SOCIS.
> 
> What you should do:
> 1. Contact the mentor listed on the project page. They might provide you
> addition information.
> 2. Prepare a proposal, you can share your proposal either with the
> mentors or here on the mailing list. We might give you hints how to
> improve your proposal.
> 3. We ask students for a proof that they are capable of fulfilling their
> task. That can be a junior job
> http://users.dune-project.org/projects/esa-socis-2015/wiki/Junior_jobs
> You can also send me your master's thesis in a private mail. I'll treat
> it confidential and will only share it with potential mentors.
> 4. You have to fill out the application form on the SOCIS homepage
> http://sophia.estec.esa.int/socis/?q=node/21
> 
> Some help on how to write a good application:
> http://teom.org/blog/kde/how-to-write-a-kick-ass-proposal-for-google-summer-of-code/
> 
> Please always ask on the mailing list, if you are in doubt. We will help
> with your application.
> 
> Bye
> Christoph
> 
> Am 28.04.2015 um 23:37 schrieb Philipp Stekl:
>> Hello everybody,
>>
>>
>> my name is Philipp and I just finished my master thesis 
>> "Solution of the time harmonic, acoustic wave equation with the Finite Element Method" 
>> in the group of Prof Peter Bastian at Heidelberg University.
>>
>>
>> I am interested to participate in SOCIS 2015, in particular 
>> Project 2: CurvilinearGrid and CurvilinearGMSHReader: Internal Boundaries and Periodic Boundaries.
>>
>>
>> I would like to get all the important information of this project.
>>
>>
>> Which steps are necessary for the application?
>>
>> Best,
>>
>> Philipp
> 
> 
> 
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
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=creg
-----END PGP SIGNATURE-----




More information about the Dune-devel mailing list