Output data formats

In Specifying classic run-time parameters in setrun.py, the format for the output data (solutions) can be specified by setting the parameter output_style.

To read the solution stored in these files into Python for plotting or other postprocessing purposes, utilities are provided that are described in python_io.

Setting output_style = ‘ascii’ gives ASCII text output. The data files can then be viewed with any standard text editor, which is particularly useful for debugging. However, ASCII files are generally much larger than is necessary to store the original data in binary form, and so when grid have many grid cells or when many output frames are saved it is often better to use some form of binary output, e.g. Raw binary output data format or NetCDF output data format.

In AMRClaw, ASCII and binary output are both written by the library routine valout.f. The aux arrays are also dumped if requested, see Output of aux arrays.

ASCII output data format

Two output files are created at each output time (each frame). The frames are generally numbered 0, 1, 2, etc. The two files, at frame 2, for example, are called fort.t0002 and fort.q0002.

fort.t0002

This file has the typical form:

 0.40000000E+00    time
 1                 meqn
36                 ngrids
 0                 naux
 2                 ndim
 2                 nghost

This file contains only 6 lines with information about the current time the number of AMR patches at this time.

In the above example, Frame 2 contains 36 patches. If you are using the classic code or PyClaw with only a single patch, then ngrids would be 1.

The data for all 36 patches is contained in fort.q0002. The data from each patch is preceeded by a header that tells where the patch is located in the domain, how many grid cells it contains, and what the cell size is, e.g.

fort.q0002

This header has the typical form:

1                 grid_number
1                 AMR_level
40                mx
40                my
0.00000000E+00    xlow
0.00000000E+00    ylow
0.25000000E-01    dx
0.25000000E-01    dy

This would be followed by 40*40 = 1600 lines with the data from cells (i,j). The order they are written is (in Fortran style):

do j = 1,my
    do i = 1,mx
        write (q(i,j,m), m=1,meqn)

Each line has meqn (change to num_eqn?) values, for the components of the system in this grid cell.

After the data for this patch, there would be another header for the next patch, followed by its data, etc.

In the header, xlow and ylow are the coordinates of the lower left corner of the patch, dx and dy are the cell width in x and y, and AMR_level is the level of refinement, where 1 is the coarsest level. Each patch has a unique grid_number that usually isn’t needed for visualization purposes.

Raw binary output data format

The files for each frame are numbered as for the ASCII file and the fort.t0002 file, for example, is still an ASCII file with 6 lines of metadata. There are also ASCII files such as fort.q0002, but these now contain only the headers for each grid patch and not the solution on each patch. In addition there are files such as fort.b0002 that contain a raw binary dump of the data from all of the grid patches at this time, one after another. In order to decompose this data into patches for plotting, the fort.q0002 file must be used.

Unlike the ASCII data files, the binary output files contain ghost cells as well as the interior cells (since a contiguous block of memory is dumped for each patch with a single write statement).

NetCDF output data format

See pyclaw.io.netcdf.

Output of aux arrays

Describe…