Bug GSP with DMP? Initial GSP orientations from particle_input.csv do not seem to be broadcast to other processes

MFiX
1

Dear MFiX community,

I think I have come across a bug where the orientations stored in the particle_input.csv are not broadcast to other processes than the “first” when running the DMP solver. I have attached the .zip file with an example of this.

If I run this sim using DMP then all but the particles in the domain of the first processor have the “default” orientation instead of the orientation specified in the particle_input.csv file.

If I run the sim using single threaded or the SMP solver then they have all the correct orientations, indicating the problem is not with the particle_input.csv file.

I couldn’t find anything on the forum regarding this problem and thought that was odd but I have been unable to fix it. Has anyone else experienced this problem?

932_2026-03-31T094420.157658.zip (25.9 MB)

Note I am using MFiX 25.3 as the new 26.1 does not work on our cluster due to a bug related to “OSError: /lib64/libmvec.so.1: version `GLIBC_2.35’ not found” but I will not consider this a bug until the system administrator has confirmed its not a local problem we have.

Best Regards,

Simon

2

Hi Simon,

Thanks for testing it, I confirmed that it was a bug within the GSP implicit representation. There will be a fix if you can add the following line:

CALL bcast(found_quat, pe_io)

into line 2922 in mfix/model/des/read_particle_input.f, it will look like this at the end:

         ELSEIF(GSP_IMPLICIT) THEN

            CALL bcast(quat_start, PE_IO)

            CALL bcast(found_quat, pe_io)

         ENDIF
3

Thanks for the extremely fast response.

I checked and it works.

Best Regards,

Simon

1 Like