Hey Stefan, all, I'm also working on the same problem (SPH and Cactus). I have the critical pieces in place that form the basis for working on an unstructured mesh in an HPC context (MPI, ghost handling, IO, etc), the kind of things carpet provides for AMR grids. Actually, I'm currently in the process of writing my thesis on precisely this. I would be happy to get in on this discussion too.
NB: It's currently not written as 'native cactus,' but instead as a shared object that gets linked in and provides cactus with r/w access to its memory space. This is the result of it starting as a generic unstructured mesh driver for use with various other codes as well as 'accelerators' (e.g. MICs). I've since lost sight of that goal (temporarily) as the devel version is currently being littered with hackish crud for my thesis. It then uses a cactus thorn just to handle some of the interfacing (calling the interpolator, calculating Tmunu). It shouldn't take much effort to move everything over to the thorn itself. Ill be on the call this morning if anyone would like to discuss it, Matt Kinsey ________________________________ From: [email protected] <[email protected]> on behalf of Erik Schnetter <[email protected]> Sent: Monday, August 31, 2015 7:54 AM To: Stefan Ruehe Cc: Users Subject: Re: [Users] SPH in EinsteinToolkit Stefan Please wait with creating a new group. Once the group is created, only those that are subscribed will participate in discussion, essentially limiting this discussion to very few people. Please use this list here instead -- it is an open list, and you should not feel bad for using it for technical discussions and questions. -erik On Mon, Aug 31, 2015 at 4:30 AM, Stefan Ruehe <[email protected]<mailto:[email protected]>> wrote: Hallo, first thank you for your advice. I will make a google group until tomorrow, where we could discuss how to integrate SPH in the ET. Roland, to your question, I am not sure which major problem you mean regarding the MPI process, but I have made test how to couple SPH with a grid based background and how to take advantages from the grid. I am afraid I could not call in into today´s ET phone call. Regards, Stefan Am 2015-08-30 15:21, schrieb Roland Haas: Hello Stefan, long time ago i asked about a SPH modul in EinsteinToolkit for my PhD thesis. It took more time i thougt, but i am still working. From what I can tell from the email thread from January 2014 the discussion there was on how to couple and SPH hydro code to the spacetime solver in Cactus and how to handle the particles moving between MPI processes. Judging from your current email you made some progress in the direction? First i have some questions about citing, because i will present a poster at a meeting and i will publish some prework. I want to mention the EinsteinToolkit, but there are no results using it yet. Should i taking something into account? For a post I think it is fine if you just add the Einstein Toolkit logo (the little Einstein https://einsteintoolkit.org/global/images/einstein.png). If you want to add a reference to the poster, I would use the ET website URL. We have a section on citing policies on the website: https://einsteintoolkit.org/documentation/licenses/ where the main publication is Loffler:2011ay in https://einsteintoolkit.org/manifest/einsteintoolkit.bib Second i made necessery preparatory work and will now try to integrate the SPH algorithm. I am sure it could be helpful not working alone on it, so i could open a google group for communication, if somebody want. It will probably be helpful to get some support from at least one of the ET people (incl. me). I am fine trying to help out with anything that has to do with how to integrate an SPH code into the existing framework and the technical details on eg how the scheduling and the interpolator works, and given that I am (usually) the same timezone as you are you may expect me to reply to a lot of questions first. Matt Kinsey (GeorgiaTech) who also worked with many particles has experience in how to make this work fast. A Google group would be fine to unclutter the mailing list with the technical discussions. Would you be able to call in into Mondays ET phone call once more? Yours, Roland _______________________________________________ Users mailing list [email protected]<mailto:[email protected]> http://lists.einsteintoolkit.org/mailman/listinfo/users -- Erik Schnetter <[email protected]<mailto:[email protected]>> http://www.perimeterinstitute.ca/personal/eschnetter/
_______________________________________________ Users mailing list [email protected] http://lists.einsteintoolkit.org/mailman/listinfo/users
