ASTE is a lightweight wrapper around the preCICE API, which allows emulating participants to investigate simulation setups.


ASTE is a collection of tools that can be used to reproduce and evaluate particular setups without using actual solver codes. There are two common use-cases, where this is particularly useful:

  1. Reproducing a specific mapping setup of a coupled case, e.g., a case crashes since the mapping fails or the mapping seems to behave unexpected. ASTE allows to rerun such a case (in parallel if needed) and investigate the mapping in terms of accuracy as well as runtime.

  2. Replay mode, where we replace a participant in a coupled setup with ASTE resulting in a uni-directional coupling. This is useful for debugging, for developing new adapters, but also for efficiency reasons (explicit instead of implicit coupling, no computationally demanding solver needs to be executed).


The core module, which interfaces with preCICE, is called precice-aste-run and written in C++. In addition, ASTE offers several tools for pre- and post-processing purposes written in python.


The C++ core module of ASTE uses similar dependencies as preCICE itself. In particular, ASTE requires a C++ compiler, CMake, MPI and Boost. Have a look at the corresponding preCICE documentation for required versions and on how to install these dependencies if needed. In addition, ASTE relies on preCICE (version >= 2.0) and the VTK library (version >= 7) in order to handle mesh files. The VTK library can be installed using the package manager (libvtk<VERSION>-dev), e.g., on Ubuntu

sudo apt install libvtk9-dev

As an optional dependency for pre-processing, METIS can be installed. METIS is a graph partitioning library used for topological partitioning in the mesh partitioner and can be installed similarly via apt

sudo apt install libmetis-dev

The python tools require

  • NumPy
  • sympy (optional)

which can be installed directly using pip and the requirements.txt file in the repository

pip3 install -r requirements.txt

Building and installation

In order to build ASTE, clone the project repository and use the usual CMake steps to steer the installation procedure:

git clone
mkdir build && cd build
cmake .. && make

In order to install ASTE and the associated tools system-wide, execute

make install

which might require root permission.

Command line interface

After the installation procedure, the following executables are available

  • precice-aste-run: core module interfacing with preCICE
  • precice-aste-evaluate: python tool to compute and store data on mesh files
  • precice-aste-partition: python tool to partition a single mesh file into several ones for parallel runs
  • precice-aste-join: python tool to join several mesh files into a single mesh file for parallel runs.

All ASTE tools are executed from the command line and running a particular executable with --help prints a complete list of available command line arguments and their meaning. There is also an ASTE tutorial in the preCICE tutorials.

The following subsections explain each part of ASTE in more detail. All ASTE modules have the following three command line arguments in common

Flag Explanation
--mesh The mesh filename/prefix used as input
--data Name of data array (input or output depending on the module)
--output The mesh filename used to store the output mesh


precice-aste-run calls the preCICE API and can be executed in serial as well as in parallel (using MPI). As stated in the introduction, there are two different use-cases, one for investigating mappings and one for replacing participants in a coupled scenario (replay mode). Configuring the replay mode in ASTE relies on a json configuration file (see further below). Therefore, the replay mode takes usually only the --aste-config <FILE.json> option as a command line argument. All other command line arguments are mostly used for reproducing mappings.

Flag Explanation
--aste-config ASTE configuration file (only used for replay mode)
-v Enables verbose logging output from preCICE
-c To specify preCICE configuration file (default="precice-config.xml")
-p Participant name, which can take the arguments A or B
--vector A bool switch to specify vector data (default=False)

For example, mapping the data “dummyData” from a mesh named fine_mesh.vtk to an output mesh coarse_mesh.vtk and saving the resulting mesh into the variable “mappedData” on the mesh mappedMesh would read as follows:

precice-aste-run -p A --mesh fine_mesh --data "dummyData"
precice-aste-run -p B --mesh coarse_mesh --data "mappedData" --output mappedMesh

While the example above exexutes the mapping in serial, precice-aste-run can be executed in parallel (using MPI). However, this requires a partitioned mesh (one per parallel rank). In order to decompose a single mesh appropriately, the tools precice-aste-partition and precice-aste-join can be used.


Reads a single mesh file (either .vtk or .vtu extension) and partitions it into several mesh files. The resulting mesh files are are stored as output_1.vtu, output_2.vtu, .... There are there algorithms available in order to execute the partitioning. The meshfree and uniform algorithm are rather simple algorithms, which don’t require any mesh topology information. The topological algorithm relies on the optional dependency METIS and is more powerful, but needs topology information.

Flag Explanation
--directory Output directory (optional)
--numparts The number of parts to split the mesh into
--algorithm Algorithm used for determining the partitioning (options=”meshfree”, “topology”, “uniform”)

Example: to divide a mesh into two parts using the topological partitioning and store it in a directory:

precice-aste-partition --mesh MeshA.vtk --algorithm topology --numparts 2 --output fine_mesh --directory partitioned_mesh


Reads a partitioned mesh from a given prefix (looking for <prefix>_<#filerank>.vtu)) and saves it to a single .vtk or .vtu file. The -r flag also recovers the connectivity information across several ranks from a mesh.

Flag Explanation
--recovery The path to the recovery file to fully recover connectivity information across ranks.
--numparts The number of parts to read from the input mesh. By default, the entire mesh is read.
--log Logging level (default=”INFO”)

For example, to join a partitioned mesh using a recovery file:

precice-aste-join --mesh partitoned_mesh_directory/partitioned_mesh --recovery partitioned_directory --output rejoined_mesh.vtk


While the previous two tools of ASTE handled the meshes for parallel runs, precice-aste-evaluate takes care of pre- and postprocessing the actual data on the meshes. precice-aste-evaluate reads a mesh as either .vtk or .vtu, evaluates a function on the mesh given by --function on it and stores the resulting data on this particular mesh. When using the --diff flag, the tool can also compute the difference between the data values already stored on the mesh and the function values (usually applied after a mapping). The diff flag also reports common error metrics such as the l2-norm and minimum or maximum errors on the mesh

Flag Explanation
--function The function which should be evaluated on the mesh (see below for examples).
--list-functions Prints a list of predefined functions.
--diff Calculates the difference between --diffdata and the given function.
--diffdata The name of the data to compute the difference used in diff mode. If not given, –data is used.
--log Logging level (default=”INFO”)
--dir Output directory (optional)
--stat Store statistics of the difference calculation in a separate file called mesh.stats.json
--gradient Calculate and store gradient data in addition to the given input function on the mesh.

The predefined functions are a collection of common interpolation functions, which are usuually too cumbersome for the command line:

Function Explanation
franke Franke’s function has two Gaussian peaks of different heights, and a smaller dip.
eggholder A function with many local extrema.
rosenbrock A function having a global minimum in a narrow, parabolic valley.

All function provided have 3D and 2D variants (which should be applied depending on your mesh topology). Example: calculate and store the Eggholder function on given mesh

precice-aste-evaluate --mesh 3DMesh.vtk --function "eggholder3d" --data "EggHolder"
precice-aste-evaluate --mesh 2DMeshonXY.vtk --function "eggholder2d(xy)" --data "EggHolder"
precice-aste-evaluate --mesh 2DMeshonXZ.vtk --function "eggholder2d(xz)" --data "EggHolder"
precice-aste-evaluate --mesh 2DMeshonYZ.vtk --function "eggholder2d(yz)" --data "EggHolder"

Example: calculating the function “sin(x)+exp(y)” on mesh MeshA and store the result in “MyFunc”

precice-aste-evaluate --mesh MeshA.vtk --function "sin(x)+exp(y)" --data "MyFunc"

Example: calculating the difference between MappedData and the analytic function “sin(x)” and storing the resulting difference data in the variable “Error”:

precice-aste-evaluate --mesh Mapped.vtk --function "sin(x)" --diff --diffdata "MappedData" --data "Error"

Replay mode

The replay mode is a bit different from the scenarios we have seen so far. Here, we emulate the behavior of individual participants in a coupled simulation. In order to configure such a scenario, each participant you want to replace needs a configuration file in JSON format with the following attributes:

  "participant": "Participant-Name",
  "startdt": "PreCICE mesh dt number (>= 1)",
  "meshes": [
      "mesh": "Mesh name in preCICE config file",
      "meshfileprefix": "/path/to/mesh/file/with/prefix/Mesh-Participant-A",
      "read-data": {
        "vector": ["Vector dataname in preCICE config which has a read type"],
        "scalar": ["Scalar dataname in preCICE config which has a read type"]
      "write-data": {
        "vector": ["Vector dataname in preCICE config which has a write type"],
        "scalar": ["Scalar dataname in preCICE config which has a write type"]
  "precice-config": "/path/to/precice/config/file/precice-config.xml"

The JSON configuration file is similar to an adapter configuration file. The above configuration file is an example of a participant with one mesh. The user can add as many meshes as required.

Step-by-step guide for replay mode

Step 1: Setup export of your original coupling

In a first step we have to generate the required mesh and data files of the participant we want to emulate with ASTE. Therefore, we use the export functionality of preCICE. After adding the export tag in the precice configuration file, start the coupled simulation and run as many time-steps as you need.

Step 2: Prepare ASTE and preCICE configuration files

Prepare an ASTE configuration for the solver you want to replace. See above for the corresponding ASTE configuration format. If your previous simulation used an implicit coupling, make sure to change the configuration to an explicit coupling.

Step 3: Run your solver and ASTE

Run your solver and ASTE as usual, e.g., execute myFluidSolver in one shell and precice-aste-run in another shell:

./myFluidSolver &
precice-aste-run --aste-config solid-config.json

ASTE picks up the correct mesh files, extracts the data and passes the data to preCICE.