|
SpECTRE
v2024.04.12
|
SpECTRE source files for evolution executables are located in src/Evolution/Executables. Executables can be compiled by running the command make EXECUTABLE_NAME where EXECUTABLE_NAME is the name of the executable as defined in the CMakeLists.txt file located in the same directory as the source file. For example, to compile the executable that evolves a scalar wave using a three-dimensional domain, navigate to $SPECTRE_HOME/build/, then one runs the command: make EvolveScalarWave3D, which then results in an executable of the same name in the bin directory of the user's build directory.
Each SpECTRE executable reads a user-provided YAML file that specifies the runtime settings of the evolution. This is where options such as the Domain, AnalyticSolution, TimeStepper, etc are supplied. Example input files are kept in tests/InputFiles, and are written to provide a functional base input file which the user can then modify as desired. Copy the executable and YAML file to a directory of your choice. The YAML file is then passed as an argument to the executable using the flag --input-file. For example, for a scalar wave evolution, run the command:
You can also use the command-line interface (CLI) to run executables:
By default, the example input files do not produce any output. This can be changed by modifying the options passed to EventsAndTriggers or EventsAndDenseTriggers:
This will observe the norms of the errors in the system at times \(0.0\) and \(1.0\) and the field data at the start of every 50th slab. A successful observation will result in the creation of H5 files whose names can be specified in the YAML file under the options VolumeFileName and ReductionFileName. One volume data file will be produced from each Charm++ node that is used to run the executable. Each volume data file will have its corresponding node number appended to its file name. Visualization of the volume data will be described in the next section.
A SpECTRE executable with observers produces volume and/or reduced data h5 files. An XDMF file must be created from the volume data in order to do visualization using ParaView. To this end we provide the tool generate-xdmf in the Python command-line interface. Run it in your build directory as spectre generate-xdmf. It takes two required arguments which are passed to --subfile-name, and --output. The argument passed to --subfile-name is the name of the subfile inside the H5 volume data file where data is stored. In the above example, --subfile-name would be VolumePsiPiPhiEvery50Slabs. The argument passed to --output is the desired .xmf file name, also without filename extension. Use --help to see a further description of possible arguments.
Open the .xmf file in ParaView and select the Xdmf Reader, not the version 3 readers. On the left hand side of the main ParaView window is a section named Properties, here you must click the highlighted Apply button. ParaView will now render your volume data. If you only wish to visualize a few datasets out of a large set, we recommended unchecking the boxes for the datasets you wish to ignore under Point Arrays before clicking Apply. Also, don't forget to check which dataset is being used for the color under Coloring. The default won't necessarily be the dataset you want.
.xmf files when subfile names have colons in them. Please avoid subfile names like My::Subfile:Name.Here we describe the usage of filters we've found to better visualize our data. Feel free to contribute to this section!
You may notice what appears to be mesh imprinting on the data. The imprinting effect can be removed by applying the Tetrahedralize filter. To apply the filter select the Filters menu item, then Alphabetical and finally Tetrahedralize.
New volume data can be created from the existing volume data using the Calculator filter. In the Calculator's text box, input a numerical expression in terms of existing datasets evaluating to the desired quantity. For example, a vector-valued velocity dataset can be created from three scalar velocity component datasets using the expression velocity_x * iHat + velocity_y * jHat + velocity_z * kHat and hitting the Apply button. By default, this will create a new dataset Result. The name of the new dataset can be changed by changing the name provided to Result Array Name above the Calculator text box.
Use the Calculator filter described above in "Creating Derived Volume Data" to create a new vector-valued dataset. Once this is created, use the Glyph filter and set the Active Attributes to the vector you wish to visualize. Make sure that Scale Mode is set to vector.
Global quantities such as error norms are stored in h5::Dat subfiles in the reduction/global HDF5 file. The command-line endpoint spectre plot dat can be used to plot quantities, and spectre extract-dat can be used to extract them into space-separated text files. These text files can then be read with other plotting programs or viewed in an editor.
Being able to reproduce the results of simulations is important for scientific integrity and accuracy. To this end, SpECTRE encodes the source tree, environment variables, and a record of build information (e.g., the third-party library versions) into the executables. From most executables these can be retrieved using the flags --dump-source-as filename_without_extension, --dump-paths, --dump-build-info, and --dump-environment. The flag --dump-only will cause the program to exit after the information has been dumped. You can use the --help flag to get more details on what flags are available. The flag --dump-paths will print to screen the environment variables PATH, CPATH, LD_LIBRARY_PATH, LIBRARY_PATH, and CMAKE_PREFIX_PATH at link time. The flag --dump-build-info will print to screen the contents of the BuildInfo.txt file generated by CMake. The flag --dump-environment will print to screen the output of printenv at link time. Finally, the flag --dump-source-as filename_without_extension will dump the archive filename_without_extension.tar.gz, which contains the entire source tree at link time. The combined above information should be enough to reproduce simulations results to roundoff error as long as the third party libraries used to build the executable have not changed on the system the simulation was run on.
Often only data from the simulation is retained, not the entire executable. In order to make simulations reproducible from only the HDF5 output, any HDF5 files written with SpECTRE's HDF5 wrappers contain the info about the run, output of printenv at link time, the contents of BuildInfo.txt, and an archive of the source tree. These are inside the header.hdr group with names SimulationInfo, EnvironmentVariablesAtCompilation, BuildInfo, and src.tar.gz. The source archive can be extracted by running the command
You will need to make sure the argument to -b is the same endianness as your computer. In the above example little endian is used.