MFIX-25.3 fails at writing a vtu 2D slice in DMP

MFiX
1

Hi,

I am passing a 3D simulation from v23.4.1 to v25.3, and realized that MFIX in DMP crashes when trying to write a slice of the following 2D geometry. I had the issue without any user subroutines even.

When deleting the slice outputs, my runs works. This 2D slice approach used to work in v234.1., is that normal that it does not? Do we now need to make it 1 cell wide (a volume for DMP)?

image
image1128×263 5.81 KB

atal error in MPIDI_Cray_shared_mem_coll_bcast: Other MPI error, error stack:
MPIDI_Cray_shared_mem_coll_bcast(453): collective tags 2 and 1 do not match across processes. This can happen if processes are performing different collective operations on the same communicator at the same time.
MPICH ERROR [Rank 21] [job id 11841283.0] [Wed Dec 10 12:09:15 2025] [nid004332] - Abort(26383) (rank 21 in comm 0): Fatal error in MPIDI_Cray_shared_mem_coll_bcast: Other MPI error, error stack:
MPIDI_Cray_shared_mem_coll_bcast(507): collective tags 2 and 1 do not match across processes. This can happen if processes are performing different collective operations on the same communicator at the same time.

fyi: the crash happened at the same time on two different HPC with the same inputs.

2

Hi Eric, can you please attach your setup so we can take a look? Thanks.

3

mfix.dat (15.1 KB) I made the 2D slice a 3D one and it fails with the same issue (see case attached). If i delete the first vtk write out, it write the IC* and proceed as expected. #StrangerThings :slight_smile:

4

@jeff.dietiker I tried in v25.1.1 and the 2D slices I had originally can be written, does not work in v25.3, just tried again. :face_with_spiral_eyes:

6

I think I introduced a bug when adding the STL force data to the VTU files. Please try the attached.

vtk_out.f (175.9 KB)

7

FWIW, the fix works for me on a 2x4x2 test decomposition. (I was getting the same error as Eric previously).

8

It works for me too! Could you add again the T_START and T_END part though? Thanks for the update :slight_smile:

9

You had a fix for T_START and T_END in another post. Can you merge it with the file above? We will fix both issues in the next release.