Skip to content
Snippets Groups Projects

The ASTRA code

Last edited by Giovanni Tardini

Download & install

If you wish an ssh key, follow the instruction

cd <some_path>
git clone https://gitlab.mpcdf.mpg.de/git/astra.git a8
# git clone git@gitlab.mpcdf.mpg.de:git/astra.git a8 # if you have a ssh key
cd a8
chmod u+x install.sh
./install.sh

The install.sh command installs the relevant directories, compiles the fixed and variable parts of the code, and runs the default test case.

Pull updates

In the following, $AWD represents the ASTRA working directory - in the example above, it would be <some_path>/a8.

cd $AWD
git pull
./install.sh

If git pull fails due to conflicts, you can do either:

  1. a new installation from scratch (recommended)
  2. git checkout <filename causing conflict> before git pull

Browse commits history

You can browse the previous commits with the command

cd $AWD
module load git # necessary only in toki
gitk

You can see the full history of the past commits, with a list of the summary labels in the user-friendly gitk GUI, so that you can track all code changes since the repo was created. But you can also view (and copy) any file of any previous version.

Execute

cd $AWD
exe/as_exe -m <equ> -v <exp> -s <tbeg> -e <tend> [-tpause <tpause>] [-debug] [-resize <factor>]

Note that exe/as_exe -h prints help for the user in the shell.

Support

Please use the support email
astra.support@ipp.mpg.de

Boundary + initial conditions

An overview of the options hierarchy for equation-statements in the equ file.

New with respect to ASTRA7

User's end

  • The syntax TE:EQ in the equ file is not valid anymore. Please always write the full TE:EQ[2,1.] instead, the parser needs that

  • Need to replace include status.inc and similar statements with use status_inc, before the implicit none statement in sbr, xpr

  • In fml, make code f90-compatible:

    • C is no comment tag, use ! instead
    • new-line is done with & at the (upper) line end, not at 6th place of the new line

  • In the equ files:

    • Remove NETCDF, A2CDF, A2JSON

    • In the TGLF calls, replace work(j,21)->mem_tglf(j,1), work(j,22)->mem_tglf(j,2)...

    • In the QuaLiKiZ calls, replace work(j,21)->mem_qlkz(j,1), work(j,22)->mem_qlkz(j,2)...

    • In the NEO calls, replace work(j,51)->mem_neo(j,1), work(j,52)->mem_neo(j,2)...

  • ./clean.sh for proper full recompilation (avoiding left-overs, change of platform in same $AWD)

  • Removed the ObjectCode variable: use different installations for different platforms

  • tmp/<exp><equ>.nml instead of tmp/astra.log; tmp/*.f90 instead of tmp/*.inc, so the generated code is all enclosed in fortran subroutines, no code blocks

  • exe/as_exe -debug option: localise crash and track sequence of calls

  • exe/as_exe -resize 1.5 option: larger graphic window!

  • Removed work, work1 arrays from the modules neocl4, a2strahl, neut as well as from tglf, qlkz, neo

  • The ASTRA driver (astra_main.f90, stepup.f90) as well as the variables tables (*.txt) are in $AWD/main

  • The geom2d array does not exist anymore: everywhere you can get equilibrium quantities with
    use parameters_a2equil, only : equil_now.
    equil_now is an IMAS-like structure with a full set of equilibrium parameters, 1D profiles and R-z based 2D matrices (Psi, grids).

Default output format

ASTRA writes json files including all plasma variables automatically, every dPout seconds (you set dPout in the $AWD/equ/log/* files). At the end of the ASTRA run, all these snapshots are concatenated along the time axis into a NetCDF file, located in $AWD/ncdf_out.

Behind the curtains

  1. Using modules instead of commons
  2. Less memory constraints, cleaner programming
  3. All implicit none, even in astra_main.f90, stepup.f90
  4. python parser replacing the historic Fortran parser, for generating the variable fortran code (tmp/)
  5. Removed intermediate layers in metric.f90, including COUPLING_SCHEME

Equilibrium solver switch IPEQL

The choice of how to solve the equilibrium is not done by special values of NEQUIL, MEQUIL anymore, rather by the IPEQL variable.

Value Solver
-2 Cylindrical (no solver)
-1 Input metrics from exp/udb
0 EQGUESS (no solver)
1 EMEQ (3-moments)
3 (Don't use: iterations only)
4 SPIDER
5 FEQIS

New directory trees

User's part

The usual $AWD subdirs equ, exp, sbr, fml, fnc, xpr contain the user part. Scripts and Makefiles are in $AWD/exe (former $AWD/.exe).
Note that these dirs are "duplicated" in $AWD/repo_user_area, which is tracked by the git repo. You do not need to edit this part, unless you really want to contribute/develope.
The level $AWD/equ (exp, ...) is the working area, where users can add and modify things without conflicting with the repo, in order to be able to git pull at any time without losing the work done.
The execution command is $AWD/exe/as_exe

Core code

For normal usage, you can sadfely ignore this core part of the code, which compiles into $AWD/lib the libraries libfor.a, libfeqis.a, libnbi.a, libspider.a .
The source code is in $AWD/src, subdirs feqis, for, nbi, spider. This part is relevant for developers or for debugging, and it is tracked by the git repo.
The python parser, generating Fortran code from the equ file, is in $AWD/pyparse. Also this part is directly linked to the git repo.

Other useful dirs

  • $AWD/tmp contains the temporary parser-generated code
  • $AWD/bin (former $AWD/.tsk) tgit@ghe actual executables

FAQ & workarounds

References

Manuals

Online manual

Astra6.0 astra.pdf

Addendum Astra7 addendum_EF.pdf

Comments

    Please register or sign in to add a comment.