65,000 floating bodies

Dear DSPH team,

Thanks very much for releasing v4.2. I tried to simulate many floating objects (~9,000) and encountered the error message "number of floating particle blocks exceeds the maximum".

To make sure that I am not doing anything wrong, I ran the example file (CaseManyFloatings_Def) by increasing the number of floating objects to 5,000 and got the same error message. It looks like the limit is ~2,000 floating objects, not 65,000. Does it need any special treatment to go beyond 2,000?

Btw, the example .def and .bat files are confusing. Floating property was assigned to 42,000 objects, where only 2,000 spheres were actually created.

Thanks.

Comments

  • Excellent. Thanks.
  • If any problem, do not hesitate to contact us again
    via forum is the doubt can be of interest to be solved to many others or email us to dualsphysics@gmail.com

    Regards
  • samsam
    edited May 2018
    I am afraid DualSPHysics4Re_vs2015.sln does not contain any project. I also tried downloading the same file from the Github repository and opened in VS2015 only to get "the file is not a valid solution file" message.
  • Have you installed Visual Studio Community 2015? It is free
  • Yes. As I mentioned before, I opened it in VS2015 and I meant Visual Studio Community 2015. Sorry for using an abbreviation.
  • It seems that only you have that problem.
    We have checked again the package and everything is ok.

    Let us see if someone else report the same problem since it seems that it can be an external problem

    Regards
  • The file size of DualSPHysics4Re_vs2015.sln is only 2 kb that tells me probably it has nothing in it. I was able to modify and compile v4.0
  • Yes 2KB is the size of the .sln file
  • I see. It's little confusing because the same file on Github is 48kb, and the older version of it is 39kb. I will update this thread if I find a solution.
  • samsam
    edited May 2018
    Problem solved! Just for the benefit of other users, I am sharing my solution.

    First, CUDA toolkit 9.1 is actually replaced by CUDA 9.2 by NVIDIA. If you try to download v9.1 from their archive, it actually downloads v9.2, and not 9.1. I did not notice it before which created all these confusions.

    The rest is pretty simple. You have to update the project's import directory by simply replacing 9.1 with 9.2.

    Thanks.
  • Thanks Sam
  • Why the number of the floating object is limited? And will it open?
    Thanks!
  • DualSPHysics was not conceived to simulate thousands of floating bodies.

    The limit in the number of solid objects is related with they type of label that is used along the code to identify them. The increase of this number implies a general loss of performance and higher memory consumption.

    That is why by default we comipled a version with limited number of floating and solid objects.

    Regards
  • If the number is unlimited will be more useful. We are working on the coupling between DualSPHysics and other DEM code, however the number of the DEM particles is limited.
    Thanks!
  • The number is now limited for performance reasons.
    You can modify the source code of DualSPHysics to create the number of objects you need.

    Please, also note that number of particles is not limited, but the number of "mk" labels (the number of objects, no number of particles)
  • Hi Alex,
    DualSPHysics team thanks very much for releasing v4.2. I also have a similar requirement, needed to run more then 2000 particles coupled with fluid. The info in the faq section makes seance. But I was wondering, is it possible to gain access to compiled version of the excitable in Linux that can run more then 2000 particles. Greatly appreciated if you can!!
    Thanks
    Nirmal
  • edited May 2018
    access to compiled version of the excutable in Linux?
    not sure what you mean...

    Both versions of GenCase are compiled in bin/linux
    DualSPHysics is also compiled in bin/linux

    If you want to modify DualSPHysics, you only need to
    activate or comment the code line #define CODE_SIZE4 in Types.h
    and compile using makefiles...

    We are not going to add extra binaries because the size of ZIP will drastically increase since we will need to add 4 new executables: CPU and GPU / linux and windows

    Regards
Sign In or Register to comment.