[FLASH-USERS] Change boundary condition for restarted simulation

[Klaus Weide]

On Tue, 7 May 2019, Magyar Norbert wrote:

> What if I initialize a dummy simulation, with the same amount of
> blocks/refinement, but initially periodic. Then, I just replace the
> variables in this periodic checkpoint file with my variables from the
> checkpoint file at time t. Would that work? Is the topology consistent
> between identical "looking" checkpoint files (same number,size,refinement
> of blocks) coming from two separate setups?

IF you manage to create a periodic grid that is 'identical "looking"' in 
your sense, THEN that could work.  I assume that what you have in mind is 
some sort of stitching together of two HDF5 files, using utilities like 
h5copy.

One issue you may run into is the following: 
A given arrangement of blocks may be valid as a nonperiodic grid,
but not as a PERIODIC grid, because it wouldv violate a requirement for
a well-formed PARAMESH grid.

For example, let the following represent a domain with 3 levels of
LEAF blocks, levels 2-4:


       +--+--+-----+-----------+
       |F1|F3|     |           |
       +--+--+     |           |
       |F2|  |     |           |
       +--+--+-----+    C1     |
       |     |     |           |
       |     |     |           |
       |     |     |           |
       +--+--+-----+-----------+
       |           |           |
       |           |           |
       |           |           |
       |    C2     |    C3     |
       |           |           |
       |           |           |
       |           |           |
       +--+--+-----+-----------+

This is a valid grid arrangement if all boundaries are nonperiodic.
It is not valid if there are PERIODIC boundaries: for example, the
blocks labeled F1 and C1 are direct neighbors if xl_boundary_type and
xr_boundary_type are PERIODIC; but it is not allowed to have a refinement 
change by more than one level between neighbors.

Klaus