Peculiar parallel behavior

When running Palabos programs in parallel, there seems to be little or no mass transfer from one periodic domain to the next. For the shanchenMultiPhase example, segregation2D.cpp, all four quadrants of a four processor simulation have identical density distributions at every timestep. I end up with one droplet in each quadrant. For a droplet spreading simulation, the droplet will not spread into the next processor domain. Do you have any ideas as to what is causing this?

I get identical results using either mpich2 or Intel’s impi library. I’m compiling with MPIparallel = true and I do see a substantial speed up, so the mpi implementation seems to be working. I also compared v1.0 with v1.1, turned off optimization, and toggling SMPparallel with no change in results.

Dear Ludwig,

Is is possible that you are using the code segregation2D.cpp from an older Palabos version?



No, version 1.0.1 was the only version I downloaded at the time. I recently tried version 1.1.0 and found the same results.


As a follow up, I only see this strange behavior when using the Shan Chen processor in parallel. The colliding bubbles example using the He Lee collision works fine in parallel.