Specifying AMRClaw run-time parameters in setrun.py¶
It may be useful to look at a specific example, e.g. Sample setrun.py module for AMRClaw.
Note: Many parameters have changed name since Version 4.X and some new ones have been added. See Changes to input parameters in setrun.py from 4.x to 5.0 for a summary.
To convert a Version 4.x setrun.py file to Version 5.0, see Converting from Clawpack 4.6 to 5.0.
Input¶
setrun takes a single argument claw_pkg that should be set to amrclaw.
Output¶
rundata, an object of class ClawRunData, created in the setrun file with the commands:
from clawpack.clawutil import clawdata
rundata = clawdata.ClawRunData(claw_pkg, num_dim)
The rundata object has an attribute rundata.clawdata whose attributes are described in Specifying classic run-time parameters in setrun.py.
In addition, for AMRClaw rundata has an attribute rundata.amrdata whose attributes are described below.
Run-time parameters¶
The parameters needed in 2 space dimensions (ndim=2) are described. In 3d there are analogous parameters in z required, as mentioned below.
In addition to the parameters in rundata.clawdata (see Specifying classic run-time parameters in setrun.py), the AMR parameters that can be set are the following attributes of rundata.amrdata:
Special AMR parameters¶
- amr_levels_max : int
Maximum levels of refinement to use.
- refinement_ratios_x : list of int
Refinement ratios to use in the x direction.
Example: If num_cells[0] = 10 and refinement_ratios_x = [2,4] then the Level 1 grid will have 10 cells in the x-direction, Level 2 patches will be refined by a factor of 2, and Level 3 will be refined by 4 relative to Level 2 (by 8 relative to Level 1).
- refinement_ratios_y : list of int
Refinement ratios to use in the y direction.
- refinement_ratios_t : list of int
Refinement ratios to use in time. For an explicit method, maintaining the Courant number usually requires refining in time by the same factor as in space (or the maximum of the refinement ratio in the different space directions).
Note: Rather than specifying this list, in GeoClaw it is possible to set to set variable_dt_refinement_ratios = True so refinement ratios in time are chosen automatically. This might be ported to AMRClaw?
- aux_type : list of str, of length num_aux
Specifies the type of variable stored in each aux variable. These are used when coarsening aux arrays. Each element can be one of the following (but at most one can be ‘capacity’):
‘center’ for cell-centered values (e.g. density)
‘capacity’ for a cell-centered capacity function (e.g. cell volume)
‘xleft’ for a value centered on the left edge in x (e.g. normal velocity u)
‘yleft’ for a value centered on the left edge in y (e.g. normal velocity v)
- flag_richardson : boolean
Determines whether Richardson extrapolation will be used as an error estimator. If True, patches will be coarsened by a factor of 2 each time regridding is done and the result from a single step on the coarsened patch with double the time step will be compared to the solution after 2 steps on the original patch in order to estimate the error.
- flag_richardson_tol : float
When flag_richardson == True, cells will be flagged for refinement if the absolute value of the estimated error exceeds this value.
When flag_richardson == False, this value is not used.
- flag2refine : boolean
Determines whether the subroutine flag2refine is used to flag cells for refinement.
- flag2refine_tol : float
When flag2refine == True, the default library version flag2refine.f checks the maximum absolute value of the difference between any component of q in this cell with the corresponding component in any of the neighboring cells. The cell is flagged for refinement if the maximum value is greater than this tolerance.
- regrid_interval : int
The number of time steps to take on each level between regridding to the next finer level.
- regrid_buffer_width : int
The number of points to flag for refining around any point flagged by error estimation or flag2refine. This buffer zone is to insure that waves do not leave the refined region before the next regridding and so is generally chosen based on the value of regrid_interval, typically to be the same value since waves can travel at most one grid cell per time step.
- clustering_cutoff : float between 0 and 1
Cut-off used in clustering flagged points into rectangular patches for refinement. Clusters are chosen to minimize the number of patches subject to the constraint:
(# flagged pts) / (total # of cells refined) < clustering_cutoff
If clustering_cutoff is close to 1, only flagged cells will be refined, which could lead to many 1 x 1 patches.
The default value 0.7 usually works well.
- verbosity_regrid : int
Additional information is printed to the terminal each time regridding is done at this level or coarser. Set to 0 to suppress regridding output.
- regions : list
List of lists of the form [minlevel,maxlevel,t1,t2,x1,x2,y1,y2]. See Specifying AMR regions
Debugging flags for additional printing¶
Setting one or more of these to True will cause additional information to be written to the file fort.amr in the output directory.
- dprint : boolean
Print domain flags
- eprint : boolean
Print error estimation flags
- edebug : boolean
Print even more error estimation flags
- gprint : boolean
Print grid bisection and clustering information
- nprint : boolean
Print proper nesting output
- pprint : boolean
Print projection of tagged points
- rprint : boolean
Print regridding summary
- sprint : boolean
Print space/memory output
- tprint : boolean
Print time step info on each level
- uprint : boolean
Print update/upbnd information
