The restart pointer file is an ASCII text file containing the Mass Store pathname of the most recently written master restart dataset. Thus, for a "restart run" (where the namelist variable NSREST equals 1), the model will read the restart pointer file, acquire the appropriate restart files, and commence running the model from the end point of the previous run. The restart pointer file is only written to disk. Its full pathname can be specified by the namelist variable REST_PFILE. If this namelist variable is not specified, restart pointer files will be written to /HOMEDIR/ccm3.CASEID.rpointer, where HOMEDIR is the users home directory (environment variable $HOME) and CASEID is the case name for the run as specified by the namelist variable, CASEID. The restart pointer file is written from routine wrapup.
The master restart file contains information necessary to restart the model. Variables saved in this file include the current timestep of a restart run (nstepr), the history buffer length, model date and time, history file specifications, and primary and secondary restart file specifications. The master restart file is written from routine wsds, which is in turn called by wrapup.
The primary restart file contains data from the main model buffers and 3-dimensional arrays. This allows the model to set all variables in a continuation run as they were at the time the restart datasets were written. Data from the following common blocks are saved on the primary restart file: /coreioab/, /com3d/, /comslt/, and /comsrf/ (see the header files of the same name for details of these common blocks). The primary restart file is written from routine wrtrs1, which is called by scan1bc.
A secondary restart file is written if the absorptivities/emissivities will not be calculated on the first timestep after restart, and therefore must be saved on a restart dataset. For a stand-alone run this occurs if the primary history file write frequency, NHTFRQ(1), is not a multiple of the absorptivity/emissivity calculation frequency, IRADAE (note that for a CSM flux coupled run, only the flux coupler determines when the restart files are written). It is advisable to avoid this situation if possible, since this dataset is relatively large even for the standard T42 model. On high-resolution runs, the model will split the secondary restart file over latitudes if necessary. This file is written from routine wrtrs1.
A history buffer restart file is written in order to retain the accumulated values in the history buffer if restart files are to be written on a timestep when one or more history file time samples are not written. A separate restart dataset is written for each such history file (primary and/or auxiliary). Each history buffer restart file contains the portion of the history buffer pertaining to that history file. If necessary, the model will split the history buffer restart file over latitudes. History buffer restart files are also written from routine wrtrs1.
Writing of the restart files occurs
from within the time integration loop of routine stepon .
Restart files are not written every timestep. The logical variable rstwr
is used to determine if it is time to write the restart files. No
restart data is written on timestep zero. Routine wrapup completes
the task of writing the restart files begun in scan1bc . Common
blocks associated with restart files are defined
in the header files comlun.h and comloc.h .
Questions on these pages can be sent to... erik@ucar.edu .
Search page