Changes in Clawpack 5.0¶
Clawpack 5.0 is a major reorganization of the Clawpack code base that has been going on for several years.
PyClaw in 5.0¶
The extensive PyClaw code base is now incorporated into Clawpack. See Which Clawpack solver should I use? for more about how PyClaw relates to the other Clawpack components. For recent changes in PyClaw, see the PyClaw changelog.
Fortran package changes¶
The rest of this page concerns the Fortran components of Clawpack.
There is no complete list of changes since it has evolved to be very different from the 4.x version of Clawpack in organization, but some of major changes that affect users are listed below.
Some tools are available to assist users in converting code from earlier versions. To go from Clawpack 4.6 to 5.0, see Converting from Clawpack 4.6 to 5.0. Some older Clawpack 4.3 code can be first converted to 4.6 form using Converting from Clawpack 4.3 to 4.6.
If you wish to view recent changes on GitHub, note that Clawpack is an organization, meaning that it is comprised of several repositories. Go to the Clawpack GitHub webpage to view all the repositories. Major changes that are specific to PyClaw since its initial release in 2012 are listed in the PyClaw changelog.
You might also view the issues associated with each Clawpack repository on GitHub to see what bugs and features we are working on.
Some of the major changes:
In all of the Clawpack codes, indices of the primary arrays q (for the solution) and aux (for auxiliary variables) have been reordered for better cache performance and to interface better with the PETSc code (used in as an option in PyClaw). For example, in the two-dimensional Clawpack 4.x code, q(i,j,m) denoted the m’th component of the solution vector in the (i,j) grid cell. With this convention the various components of the solution in a single grid cell are scattered in memory with a stride of mx*my. In Clawpack 5.0, the indexing for the same value is q(m,i,j) so that the components of q in a single grid cell are continguous in memory.
Note that this means user routines such as qinit.f, setaux.f, and Riemann solvers must be modified.
The calling sequence of Riemann solvers has been modified by adding maux (the number of auxiliary variables) as another parameter. This is required because of the reordering of indices so that aux(ma,i,j) is now the ma element of the aux vector in the (i,j) cell. The leading dimension of aux is assumed to be maux and is required in declaring this variable.
Calling sequences of many subroutines have changed, so users converting code from Clawpack 4.x should carefully compare the calling sequences in any custom Fortran code to those in relevant examples, or to the default versions in the libraries.
Many Riemann solvers for different applications are now found in the riemann repository. In the future major changes may be made to the form of the Riemann solvers to allow more flexibility.
The names of many input variables in setrun.py have been changed to clean things up and be more systematic. Several options that used to be specified by obscure means have been clarified, and some new options have been added. For details and documentation on the new parameters, see:
The directory structure has been reorganized. See Clawpack components.
Some regression tests have been added to the classic, amrclaw, and geoclaw directories in subdirectories named tests. Travis is now used for continuous integration testing during development.
The 3d single-grid and AMRClaw code, missing since Version 4.3, has been updated to conform with 1d and 2d style. In particular, the inputs can now be specified using setrun.py.
Three-dimensional plotting routines in Python are still under construction, so currently there is no capability to use setplot.py to specify the desired plots or make plots to produce them. However, the Matlab plotting routines have been updated and are now found in Visclaw. See Plotting using Matlab. It is also possible to render 3d plots using the VisIt GUI, see Plotting with VisIt.
The classic single-grid Clawpack code (without AMR) is now in the classic directory and the classic repository on GitHub. Some new capabilities have been added, e.g.:
- OpenMP parallelization has been added to the 3d codes. See Using OpenMP.
The AMRClaw code is now in the amrclaw repository. Some new capabilities have been added, e.g.:
- It is now possible to specify refinement regions, previously only supported in GeoClaw. For a description, see Specifying AMR regions.
The GeoClaw code is now in the geoclaw repository. Some new capabilities have been added, e.g.:
- There is an improved set of tools for monitoring the maximum depth or surface elevation seen over a fixed grid, and the first arrival times. See fgmax.
Changes to input parameters in setrun.py from 4.x to 5.0¶
Changes to general parameters¶
Many names have been changed, e.g.
- ndim to num_dim
- xlower, ylower, zlower are now lower, lower, lower.
- xupper, yupper, zupper are now upper, upper, upper.
- mx, my, mz are now num_cells[0:3].
There are many other such changes. It is best to take a look at the setrun.py for an example in $CLAW/classic/examples.
Changes to AMR parameters¶
- The rundata object generally defined in setrun.py now has an attribute rundata.amrdata and AMR parameters are attributes of this object. Most names of attributes have changed from those used in 4.x.
- Setting mxnest negative to indicate that anisotropic refinement in different directions might be used has been eliminated. Now this is always assumed and one must always specify refinement ratios in each direction and in time.
- New attributes have been added to indicate whether Richardson extrapolation and/or the routine ins flag2refine should be used to flag cells for refinement. See AMR refinement criteria.
- The capability of using “regions” to specify areas where refinement is forced or prohibited has been extended from GeoClaw to AMRClaw. See Specifying AMR regions.
Changes to plotting routines¶
The plotting routines are now in Visclaw, see Visclaw Plotting options.
The Matlab tools from version 4.x have been updated a bit and many examples once again include .m files for users who wish to plot using Matlab.
The Python routines have also been updated. For the most part older versions of setplot.py should still work, with a few exceptions:
In AMR the individual grids are now called “patches” rather than “grids”. This caused a few changes in attribute names in ClawPlotItem:
- The old plot_type named 2d_grid has been renamed 2d_patch
- The old attirbute gridlines_show has been renamed celledges_show
- The old attirbute grideges_show has been renamed patchedges_show
To use the interactive plotting tool Iplotclaw, you now need to import this via:
from clawpack.visclaw.Iplotclaw import Iplotclaw
See Plotting options in Python for more information.