We use the field 'neigh' instead of the field 'pneigh'. The field
'neigh' is of the type of 'blockref' which simplifies the reference to
the neighbors blocks when using MPI. Each reference contains two fields:
cpu and id. In MPI version we cannot use pointers stored in pneigh, but
using blockref structures we can refer to the blocks belonging to other
processors.
The new type 'blockref' is a reference to a block by two integer
numbers, the first points to the cpu that block is running, and the
second one is the block identificator.
Now we can control the levels of refinement/derefinement criterion from
the config file. The bounds of the domains are stored in the HDF5 files
from now.
The generation of dx, dy, dz for each level didn't include the base
grid, i.e. they were 2 times larger. Now it is fixed. Also plenty of
minor corrections in different places.
It looks like the refining and derefining work more or less, at least
without interrupting the execution. Nevertheless, there are still some
artifacts, like the lack of symmetry after some time or not efficient
derefining of the mesh. This could be cause by the solver, however.
The refinement criterion is computed using pressure now.
In addition, I've done some fixes to the problem initialization, and I
defined new variables igrids, jgrids, kgrids, which specify the
dimensions of the block.
Now the source file can use some of the values defined in make files,
like NDIMS. The make files, i.e. makefile, make.default, and host files
are simpler and should be easier to manage.
One array U, which is a field in the BLOCK structure, stores all
variables. The number of variables is determined by the nvars parameter.
To access each variable we use variable index now, like idn, imx, imy,
mz, ien, etc.
I replaced the pointer 'pfirst' pointing to the first element of the
list of blocks with 'plist'. I improved formatting of the informations
printed during the initial mesh generation.
I've added 'Z' configuration of the refined block, completed 'N', 'C',
'D', and 'U' configurations. Implemented pointers to the neighbors
correctly. Finally, I've implemented correct initial geometry
generation, by selection of the blocks to refine, their neightbors if of
the lower level, and finally performing the actual refinement.
Usage of IDs to find children requires additional work, i.e. looking
through all blocks and comparing their IDs. Instead of IDs we use an
array of pointer for instantenous selection of children.
Subroutine to refine a selected block has been implemented. This
subroutine creates 4 new blocks, and inserts them before the parent
block. It fills all block arguments like parent, neighbors, leaf flags,
bounds, next and previous pointers, etc.
Implemented the generation of initial blocks in N-configuration with
proper ataching to the list, pointer, neighbors and bounds
initialization, as well as the initial problem setup. Two new files has
been added: mesh.F90 - hadling the adaptive mesh structure, and
problem.F90 - handling the problem initialization.
The new subroutines take care of allocating and deallocating the space
for one block. Allocation uses dimensions provided by configuration
file. The block dimensions and number of ghost cells have been added to
config.F90 as well.
Config module takes care of the configuration file 'config.in'. It
reads the file line by line, parsing each line and extracting the name
and value of parameter. Then it substitutes the parameter with a new
value.
Error module handles error, warning and informative messages. So far it
only prints information on the screen.
There is some improvments for comments formating as well.
Block structure has been implemented. Blocks contain pointers to the next
and previous blocks. This easily allows to create block lists. Two
functions are implemented, one for creating initial structure of blocks
with lists, and the second for destroying all blocks from the list.
File 'blocks.F90' has been added to makefile too.