2) Generating EOS table from EOS Fortran code
Version 1 Created by Jonathan
M. on 03 Jul 2009.
Overall, this document explains how to generate EOS tables that are ultimately used in HARM.
*Create Shen EOS table that is interpolated onto uniform grid with desired range and resolution*
1) Using relativity
for Matlab because 32-bit and problem with 64-bit when
using "mex" function calls. Matlab core
dumps.
below uses matlabscripts/shen_interp.m that should be copied using gethelmmatlabs.sh on the system using Matlab so it's located in matlab@<machine>:matlab/
2) Run Matlab interpolation script:
cd helmeoscode
sh sheneosinterp.sh ""
*Below not required anymore since NaN or Inf's are taken out in Matlab:*
3) sed 's/NaN/0.0/g' sheneos.dat > sheneosnonnan.dat ; cp sheneosnonnan.dat sheneos.dat
*Review Shen EOS table in SM*
1) go to shen_helm_eostable directory
2) rdshentable0
3) plc 0 xalfa
To see values on contour plot do: image cursor
Then click on a point and get the reported value
4) Set WHICHLEV and PLANE to see other slices of 3D data
*And also look at Matlab interpolated/processed Shen Table
1) go to shen_helm_eostable directory
2) rdshenmatlab
3) plc 0 aheav
To see values on contour plot do: image cursor
Then click on a point and get the reported value
4) Set WHICHLEV and PLANE to see other slices of 3D data
*Create full EOS table*
1) Go to helmeoscode directory
2) Can change things in eosparms.f, kazeos.parms.dek and kazeos.loopparms.dek
3) In kazeos.parms.dek:
NORMAL FULL Ye,Ynu dependent EOS has:
parameter (whichrnpmethod = 1)
parameter (whichynumethod = 1)
parameter (whichhcmmethod = 0)
SIMPLE Ynu quasi-thermalization table has [also used for stellar model compilation]
parameter (whichrnpmethod = 1)
parameter (whichynumethod = 0)
parameter (whichhcmmethod = 1)
4) kazeos.loopparms.dek: Sets ranges of table in rho,T,(
H if whichhcmmethod==1), Y_e, Y_nu space
For whichynumethod==0, must set ntydynorynu=1 and probably set value range from 1 to 1
5)
cd ../joneoscode/jonmod/ ; make clean ; make ;
cd ../../helmeoscode/ ; make clean ; make
6) sh copyjonhelm.sh <
DATADIR> (e.g. sh copyjonhelm.sh ../fulltabletest1)
8) rm -
rf eos.head eosdetails.dat eosother.dat eos.dat eoscoulomb.dat eosazbar.dat
9) nohup ./helmeos.exe &> output.txt &
9.5) If want to generate large table that takes a lot of time, can run on multiple CPUs on cluster and just collate files in the end. For example, to chunk with 200 CPUs, do:
echo "1 200" > eoschunk.dat
nohup ./helmeos.exe &> output.txt &
This will produce chunk #1 of 200 chunks. Iterate 1 through 200 for each of the 200 CPUs. One could script the creation of the eoschunk.dat file. Ensure to collate in order of the chunk number or have to sort after and that will be slow and require use of bash's "sort" command.
9.7) If want to submit a batch job with either many jobs with some number of CPUs per job or as a single job
using a
C MPI wrapper, see helmeoscode/scripts/chunkbunch.sh . For example, following install0.sh on lonestar.tacc.utexas.edu, one does in the data directory:
sh chunkbunch.sh 5 . 256
This will generate a bsub job with 256 cores using MPI but running a single chunk (out of total 256 chunks) on each core.
This is the most efficient way. See scripts.
*Review FULL EOS table in SM*
1) load SM
2) load macros:
gogrmhd
jre kaz.m
jre grbmodel.m
3) Run macros:
rdmykazeos eos.dat
rdmykazeosother eosother.dat
rdhelmcou eoscoulomb.dat
rdhelmextra eosazbar.dat
4) Plot few things setup to setup in 3D space from 4D data (e.g.):
plc 0 (LG(mydptot))
plc 0 (LG(mydutot))
plc 0 (LG(mydstot))
5) Some other things not yet with 4D->3D copy, but still can view:
showpressure 1 1
plcs2helm
agzplc 'dump' cs2helm
6) Show where nuclear EOS converged:
pleos2dconv
*Process EOS with Matlab to produce "new" table used by HARM*
*Assume have access via ssh; probably have matlab's ~/.ssh/authorized_keys with that users public key inside added to the file*
1) *Create eosparms.head*
eosparms.head filled with (e.g.):
1E-8 1E-8 1E-8 1E-8 1E-14 1E-14 1E-14 1E-14
2) *If desired, copy data-directory to another computer (say from ki-rh39 to ki-rh42) before processing on the new computer*
e.g. scp -rp <dir> jon@ki-rh42:<otherdir>
3) Ensure latest Matlab scripts.
a) If one believes Matlab scripts may have changed, then do:
ssh matlab@ki-rh42
cd matlab
export HOST=<your host>
export USER=<your username>
export REMOTEDIR=<your directory with matlabscripts and helmeoscode directory on your machine>
scp $USER@$HOST:"$REMOTEDIR/matlabscripts/eos_extract.m $REMOTEDIR/matlabscripts/monotonize.m $REMOTEDIR/matlabscripts/shen_interp.m" .
scp $USER@$HOST:"$REMOTEDIR/helmeoscode/enforce_x_consistency.mex.f $REMOTEDIR/helmeoscode/enforce_ye_consistency.mex.f $REMOTEDIR/helmeoscode/compute_nuclear_azbar.mex.f $REMOTEDIR/helmeoscode/general_compute_functions.f $REMOTEDIR/helmeoscode/eosparms.f $REMOTEDIR/helmeoscode/const.dek $REMOTEDIR/helmeoscode/gethelmmatlabs.sh $REMOTEDIR/helmeoscode/sheneosinterprunmatlab.sh" .
exit
b) If using own version of Matlab
Can get matlab scripts if using local Matab:
svn checkout https://harm.unfuddle.com/svn/harm_matlabscripts/ <some directory>
But note that original source for eos_extract and mex files is actually in joneoscodes/helmeoscode/ and joneoscodes/joneoscode/jonmod/
5) *Convert to HARM form*
*sh eosextract???.sh : Choose the version of script to run. Normal is (with mod of DIR=) (takes < 1 hour for 50x50x50x1x1): e.g.:*
sh eosextract2c.sh remotedirtest1 test1
6) If Matlab script doesn't run, then ensure matlab server running:
ssh matlab@ki-rh42
/data/opt/matlab/etc/lmstart
exit
*Check that files exist and make sense: eosmonodegen*.dat , eosnew*.head , eosnew*.dat , eosdegennew*.dat
7) Link
up to standard names (e.g.):
ln -s eosnewtest1.head eosnew.head
ln -s eosnewtest1.dat eosnew.dat
ln -s eosmonodegentest1.dat eosmonodegen.dat
ln -s eosdegennewtest1.dat eosdegennew.dat
*Review Matlab-generated EOS files in SM:*
1) Be in directory with eosnew* files generated above
2) load sm
3) load macros:
gogrmhd
jre kaz.m
jre kazpostmatlab.m
4) Run macros:
rdjoneos 'test1'
rdjondegeneos 'test1'
5) Check some things about EOS:
jre kaz.m
jre grbmodel.m
jre kazpostmatlab.m
checkeossimplenew
plc 0 (LG(rhob))
6) Check monotonicity of pre-Matlab EOS
jre kaz.m
jre grbmodel.m
jre kazpostmatlab.m
*Pre-Matlab plot of EOS near degeneracy*
checkhelmmonoeos 1 0
*Pre-Matlab with degendiff shown: Bad points are red -- not working as desired yet*
redohelmmono 1 0
*See whether monotonic: Any visible lines are not monotonic somewhere*
nonmonosequence2
------------------------------------------------------------------------------------
Example EOS table from HELM and from Matlab script to be used by HARM:
http:
1.8.3.1
