############################################################################## # Flags: nomenu, noprompt, nomessage, # ############################################################################## # section: -general # ############################################################################## # outfile = /tmp/--username--/--SomeDirectory--/Results # # scratchbase= /tmp/--username--/--SomeDirectory--/scratch # # restartfiles= -none- # # t1restartfiles= 1440 -- Minutes. When to write the first set. # # dtrestartfiles= 1440 -- Minutes. Distance between writing. # # stopafterrestartfiles=1000000000 -- Stop after writing so often. # # text( 1)= ' ' # # text( 2)= ' ' # # text( 3)= ' ' # # text( 4)= ' ' # # text( 5)= ' ' # # text( 6)= ' ' # # text( 7)= ' ' # # text( 8)= ' ' # # text( 9)= ' ' # # text(10)= ' ' # # text(11)= ' ' # # text(12)= ' ' # # text(13)= ' ' # # text(14)= ' ' # # text(15)= ' ' # # text(16)= ' ' # # text(17)= ' ' # # text(18)= ' ' # # text(19)= ' ' # # text(20)= ' ' # # dice = yes -- PVM/MPI: use many subvolumes # # ndpw = auto -- PVM/MPI: number of dice per worker # # iodice= no -- PVM : each worker writes its own results # ############################################################################## # ?, return, help # ##############################################################################
TMPDIR
set,
gd1 will as default set scratchbase
to the string
$TMPDIR/gdfidl-scratch-pid=XXXXX-Here
$TMPDIR
is the value of the environment variable TMPDIR
,
and XXXXX
is the numerical value of the process-id of gd1.
restartfiles
must be a valid filename, in a
filsystem which survives a system crash, ie. a filesystem which is not
cleared at boot time.
If a value for restartfiles
is given
the solver will write a set of at selected times.
There will be up to two sets of restartfiles.
There are two sets because according to murphy's law it will happen that a
system crash just occurs while one set is being written. In such a case,
the older set can still be used. When the restartfiles are written,
a comment is given how to use the restartfiles.
Essentially it is: you have to restart the computation
on the same computer system (If using a parallel system: with the same
number of tasks, using the same nodes, using the same executable), with
an additional parameter on the command line of the master task.
The additional command line parameter is
-restartfiles=NAME-OF-SET
.
text()= ANY STRING, NOT NECESSARILY QUOTED
text(NUMBER)= ANY STRING, NOT NECESSARILY QUOTED
ANY STRING, NOT NECESSARILY QUOTED
:
NUMBER
:
NUMBER
, the string following
text()=
is placed in the next free line.
In the case with NUMBER
, it is guaranteed, that the string
is placed in the NUMBER
.st line.
You can specify up to 20 annotation strings,
the maximum length of each annotation string is 80 characters.
[yes|no]
When computing in parallel, and dice= yes
,
the computational volume is subdivided
in many more subvolumes than available processors.
This leads to a better loadbalancing.
[auto|NUMBER]
When computing in parallel, and dice= yes
,
the computational volume is subdivided in NUMBER subvolumes per processor.
If ndpw=auto
, the number of subvolumes (dice) is chosen such
that each subvolumes contains about 1 million gridcells.
[yes|no]
When computing in parallel, and dice= yes
,
the results of each subvolume is stored on the local disk of each
processor. This leads to faster writing of the results, but longer time
for postprocessing of the results.
This might be needed for very large parallel computations where the
computed electromagnetic fields must be loaded into the postprocessor
for computing eg. the Q-value.
For very large parallel computations, it can happen that the computed
datasets are so large that the postprocessor cannot handle the amount of
data in its limited adress space (32 bit processors).
When dice= yes, iodice= yes
, the results are written to the local
disks of the compute nodes, and when the results shall be analysed by
gd1.pp, on each compute node an instance of gd1.pp will be running,
handling the local amount of data.
Only the PVM version of gd1.pp can be run in parallel. There is no MPI version of gd1.pp, as MPI lacks some needed functionality.
/tmp/UserName/restartfile
.
Together with the plots that gd1.pp will produce,
the strings 'Parameter is 45' and '2*Parameter is 90' shall appear.
-general outfile= /tmp/UserName/resultdirectory scratch= /tmp/UserName/delete-me- restartfiles= /tmp/UserName/restartfile define(PARAMETER, 45) text(1)= Parameter is PARAMETER text(2)= 2*Parameter is eval(2*PARAMETER)