To run MG convergence tests:
./runAndTestMG.pl [xmlfile] [ OPTIONS ]
where OPTIONS:
-
-myOwnOptFiles [specifiedmgoptfile]
- where [specifiedmgoptfile]
is a .dat file with the list of the options files you want to use and
this will call a separate subroutine to include your options files in
an array by reading the .dat file and storing the lines with .opt in an
array. See options/exampleOptionFile.opt for a list of all the possible
options you can set and exampleOptFilesMG.dat for an example of how to
list the option files you want to use.
-paramFile [filename]
- where [filename]
is the param file where both resolution and mgLevels are set. By
default serial runs of resolution 64sq, 128sq with mgLevels of 4 and 5
are done- REMEMBER TO SET the mgLevels parameter in the command line.
See OFileMG2D.dat. A recommended method for setting mgLevels is to find
the number of levels where the coarsest grid is about 8x8 and the
number of levels should be calculated as if you are running the job on
1 proc. For example for a 64sq job, the mgLevels is set to 4 regardless
of how many procs the job is on (the MG solver takes care to ensure
that the coarsest grid on each proc is as close to 8x8 as possible).
-nonAnalytic [reffilepath]
- where [reffilepath]
is the high resolution reference solution file path where the high
resolution solution field variable is compared against the run time
solutions instead of the analytic solution. IF THIS IS SET, YOU MUST
SET THE TEST TIMESTEP variable. See pluginForMGTest.xml and the
pluginData struct for how to set out the numeric and reference fields
for comparison.
-testTimestep [testtimestep]
- The timestep at which you want your high resolution reference solution and your numeric solution compared at.
-againstExpected [expfilename]
- will check the generated numbers against a file you specify, assuming the file is in the 'expected/MG/' dir. It can be a frequent output file or a convergence file. By default a check will not occur.
-againstEachOther
- will
compare the generated convergence results and frequent output results
for each option file (assumes these results are generated in the log
dir). Finds the quickest inner, outer and if non-linear, non-linear
iterations, the least number of inner, outer and non-linear iterations
and the best convergence rates for each field. The log dir files will
be saved with the name of the particular option file concatinated onto
the end of the file name, for user reference. For example:
log/Underworld_ExtensionFMM.xml-analysis.frequentout-MG-opt_1.opt. The
best option files for each field variable (convergence or speed) will
be printed to screen.
EXAMPLES: 1 analytic, 2 nonanalytic
-
./runAndTestMG.pl testVelicSolCx.xml -myOwnOptFiles exampleOptFilesMG.dat -againstEachOther
Runs
with user specified option files, NO param file OFile.dat, NO check
against a specified expected file although the generated convergence
and timing variables for each option file specified, are compared
against each other.
-
UNCOMMENT OUT PLUGIN DATA STRUCT IN pluginForMGTest.xml
./runAndTestMG.pl ExtensionFMM.xml -paramFile OFileMG2D.dat -nonAnalytic ./output/HRRefSoln_opt_2 -testTimestep 1
Runs
with default option files, a user specified resolution and mgLevel via
a param file OFileMG2D.dat, NO check against a specified expected file
and the timestep the high resolution and low resolution field variables
are compared at is 1.
These scripts measure the convergence rates in numerical fields
generated in Underworld either against analytic solutions supplied by
Mirko Velic OR a fine resolution grid that was run earlier.
The error of FEM fields should decrease with higher resolution -
depending on the choice of element type that represents a given field.
These scripts run error measures on given field at several resolutions and then process how the error converges.
THESE SCRIPTS REQUIRE EXPERIMENTAL TO BE BUILT