[ESPResSo-devel] Re: LB parallel error

[Ulf Schiller]

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

Hi,

Kai Grass wrote:
> Hallo Ulf,
> 
> die aktuelle Version von D3Q18 läuft leider nicht parallel ...
>
> (für LAMMPI)
> MPI_Type_free: invalid datatype argument: Invalid argument (rank 0,
> MPI_COMM_WORLD)
> Rank (0, MPI_COMM_WORLD): Call stack within LAM:
> Rank (0, MPI_COMM_WORLD):  - MPI_Type_free()
> Rank (0, MPI_COMM_WORLD):  - main()
> 
> (für MPICH)
> 0 - MPI_TYPE_FREE : Datatype is MPI_TYPE_NULL
> [0]  Aborting program !
> [0] Aborting program!
> 
> (für OPENMPI)
> [Simbio-MacBookPro.local:03565] *** An error occurred in MPI_Type_free
> [Simbio-MacBookPro.local:03565] *** on communicator MPI_COMM_WORLD
> [Simbio-MacBookPro.local:03565] *** MPI_ERR_TYPE: invalid datatype
> [Simbio-MacBookPro.local:03565] *** MPI_ERRORS_ARE_FATAL (goodbye)

The D3Q18 model is not supported any more in the latest repository
version. However, looking at the error message I assume that the error
also occurs for D3Q19. It's due to the fact that the initialization of
LB is still somewhat preliminary. There should be a clean
pre-initialization of a zero-size LB lattice (similar to the cell
system), which I didn't have the time to implement yet.
As a workaround, try to comment out the first MPI_Type_free in
lb_create_fluid(); this probably gives rise to a potential memory hole
when recreating the LB structures (as the DEBUG output already states).
Anyway, at the moment there's no 100% guarantee that the latest CVS
version of lattice Boltzmann runs in parallel at all.
I can have a closer look after Statphys.

Cheers,
Ulf

- --
Ulf D. Schiller   *   Room 1.404   *   Phone +49 6131 379-481
Max Planck Institute for Polymer Research
Theory Group
D-55128 Mainz, Germany                   50° 0' N, 008° 16' E
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFGieIftqL0QpvXQjERAqiBAJ9svK59KbwAGZLD/JrFKvOkaW4FGgCg8DQW
ZN4G9Fz/0kzGVuewdpY++Ts=
=I+GA
-----END PGP SIGNATURE-----