Converting from Clawpack 4.6 to 5.0¶
(If you are converting code from Clawpack 4.3, see Converting from Clawpack 4.3 to 4.6.)
Many input parameters have been renamed and some new options have been added in Clawpack 5.0. See Changes to input parameters in setrun.py from 4.x to 5.0.
Python conversion tool¶
A first pass at the conversion of setrun.py, setplot.py and the Makefile can often be achieved by typing:
$ python $CLAW/clawutil/src/python/clawutil/conversion/convert46to50.py
in your application directory. You should then inspect the files generated and fix any broken links, etc.
The original versions of various files will have _4.x appended to the file name for reference. When conversion is complete, these files can be deleted.
This does not yet work for all variants of the code.
Old AMRClaw codes are often in a subdirectory amr of an application directory, and the directory above may contain Fortran files or other files used by the AMR code. Typically you will want to combine these in one directory.
The Makefile is currently not converted properly – a generic Makefile is added to the directory but must be customized to point to any local Fortran codes, for example. In particular, make sure the Makefile points to the correct Riemann solver(s), either a local file or a library routine from the $CLAW/riemann/src directory.
The indices in q and aux arrays were permuted in Clawpack 5.0 relative to early versions, e.g. the m`th component of `q in grid cell (i,j) is now q(m,i,j) rather than q(i,j,m). The conversion script attempts to do this reordering if it sees a pattern of a suitable form, but you should carefully check your own local routines such as qinit or setaux to make sure this has been done properly.
Calling sequences of several routines have also been changed and will need to be adjusted by hand for any Fortran routines in your application directory. See User files required for the Fortran code for calling sequences of the routines that most frequently are provided by users. If you specify your own Riemann solver, see also Riemann solvers, and if you have custom boundary conditions, see Boundary conditions.
Note in particular that parameter maxmx (and maxmy, maxmz in more dimensions) is no longer in the calling sequence. In earlier versions of Clawpack this indicated the declared dimension of the q and aux arrays. It is now assumed the arrays are dimensioned properly since dynamic memory allocation is generally used at run time based on mx (resp. my, mz). You should remove these from the calling sequence and also modify the declaration of input parameters to use mx in place of maxmx, etc.
The classic code no longer requires a driver routine driver.f. In earlier versions of Clawpack this was used to declare the q and aux arrays to be sufficiently large for the size of the problem to be solved. (And specifying a larger value for mx led to a run-time error.) In Clawpack-5, a library routine driver.f90 is provided that calls the Clawpack routines, which now use dynamic memory allocation to allocate the required arrays at run time.