| Home > Research > Ifsdocs > TECHNICAL > |
IFS documentation front page
Chapter 1. Technical overview
Chapter 2. FULL-POS post-processing and interpolation
Chapter 3. Parallel implementation
REFERENCES
Previous Section
1.8 Restart capability
The restart option is designed so that long-running IFS executions (typically configuration 1 with long integrations) can be made to write the model state to files. This data is such that it is possible subsequently to restart from these files and exactly reproduce the same model computations. An additional flexibility is retained by allowing the restarted run to use a different number of processors to the original run. To achieve this, all data is communicated to the master processor who is then responsible for writing it to files. The existence of these restart files is recorded by writing a `restart control file' which contains additional control information to enable a clean restart to be made. Control over when restart files are written is achieved in two ways:
| 1) by predefining the timesteps at which the files should be written. This is achieved using the NAMRES namelist array NRESTS. |
| 2) by sending a signal to the running job. |
| Signal 1 means "write restart files and continue". |
| Signal 2 means "write restart files and stop" |
The files written and their contents are described in the following tables. All file names may be modifed using NAMIOS namelist variables. The file names are all appended with a timestamp in the form of `ddddhhmm' to reflect the model timestep.
Next Section
Previous Section
 |
23.04.2002 |
 |
© ECMWF |
