Dear Colleague,

Two bugs in the parallel version of the Lattice Boltzmann code in DL_MESO 2.3 
have been discovered:

1. The boundary layers of grid points for sub-domains of each process contain 
   particle distribution function values which are not subjected to collisions 
   and yet are propagated into the main fluid points.

2. Dissimilar values for fluid properties in the boundary layers for each sub-
   domain are produced in output files, showing discontinuities when systems are
   visualized.

Please note that these bugs do not apply to the serial version of DL_MESO_LBE: 
this version of the code is currently believed to work correctly.

Both of these bugs may be fixed by simple changes to the plbe.cpp file. The time 
step loop (starting at line 25) should be modified to the following:

  for(int i=0; i<= lbtotstep; i++) { 
    fNonBlockCommunication();
    if(i%lbsave  == 0) {
      fOutputCA3D();
      if(lbdm.rank == 0) {
        printf("%d \t ", i);
        fPrintDomainMass();
        fPrintDomainMomentum();
      }
    }
    fCollision();
    fNonBlockCommunication();
    fPropagation();
    fColliBoundary();
    fAllReady();
  }

The first bug is fixed by an additional call of fNonBlockCommunication between 
the collision and propagation routines, while the solution for the second bug is 
to move the pre-existing call of the communication routine to before the if-
statement for output file generation.

These bug fixes have been applied to the current DL_MESO release and a corrected 
version may be downloaded by any registered user without re-registering 
(decryption uses the same password). Please note that the serial version of the 
code - including the slbe.cpp file - is believed to work correctly and the bugs 
identified above do not apply to this version.

Michael Seaton
2010-05-24