SMASH-analysis

The SMASH-analysis suite contains various tools and scripts useful to compute different observables from the output generated by the SMASH transport code. Most of this scripts are written in python. The latest results extracted by means of the SMASH-analysis suite, can be found here.

Among the observables and cross-checks that are extracted with the scripts in this repository are:

Angular Distributions
Non-isotropic cross sections of NN ↔ NΔ scatterings.
Cross Sections
Inclusive and exclusive cross sections for different elementary scattering processes.
Densities
Density distribution in an infinite matter simulation (box) and density evolution in the central cell of a Au+Au collision.
Detailed Balance
Detailed balance tests in an infinite matter simulation (box) with different particle species.
Dileptons
Invariant mass spectra for dileptons from different elementary and heavy-ion collisions.
Scattering Rates
Analysis of scattering rates in infinite matter simulations with only elastic collisions.
Energy Scan
Multiplicities, midrapidity yields, transverse momentum and rapidity spectra for the most abundant hadron species over a large range of beam energies.
FOPI Pions
Pion multiplicities and transverse momentum spectra following the FOPI experiment.
Afterburner
Multiplicities, midrapidity yields, transverse momentum and rapidity spectra for the most abundant hadron species for afterburner runs.

Prerequisites

Running SMASH for the above listed observables to generate the necessary data sets is embedded in the SMASH-analysis suite. As such, the SMASH source code (if run on a cluster), or at least the SMASH binary needs to be accessible from where the SMASH-analysis suite is executed. The SMASH source code as well as further instructions how to download and compile it can be found here.

Thus, SMASH-analysis suite has been tested with the following software versions:

SMASH version = (SMASH-analysis version)
cmake version ≥ 3.16.3

The SMASH-analysis suite is regularly tagged (SMASH-ana-1.6 and later), where it is ensured that the SMASH-analysis version is compatible with the same version of the SMASH transport code. Please make sure to use compatible versions.

Furthermore, Python 3.8.x is necessary with the following modules:

numpy version ≥ 1.17.4
scipy version ≥ 1.3.3
matplotlib version ≥ 3.3.1
argparse version ≥ 1.1
pyyaml version ≥ 5.3.1
pandas version ≥ 0.25.3

Python 2 is not supported anymore. The versions of Python and the different modules can be determined by running the script version_check.py located in the smash-analysis/python_scripts directory. The versions listed above have been verified to work. Different versions (newer or slightly older) could work fine as well, but it is not guaranteed. Any missing python modules (e.g. pandas) can be installed via the python package installer pip:

pip install --user pandas

For further details, see here. If the pip module itself is not available, it can be installed via

wget https://bootstrap.pypa.io/get-pip.py
python get-pip.py --user

Then the remaining modules (e.g. pandas) can be installed via:

python -m pip install --user pandas

In order to install a particular version of a module (e.g. version 0.14.1 of pandas), execute:

python -m pip install --user pandas==0.14.1

For plotting, the DejaVu font is used by default. If font problems in plots arise, it may be downloaded here. After downloading, the font needs to be installed on the system. It might also be necessary to delete the local matplotlib font cache.

Note

Should you encounter severe difficulties installing a numpy version 1.17.3 on your machine, you may consider using an older version. The only necessary functionality from numpy 1.17.3 is being used for plotting the detailed balance results. The simulations and analyses of each target can be carried out with numpy 1.6.2 or higher. The detailed balance plots can then be created locally from the generated analysis outputs.
Remember to adjust the required numpy version in the top level CMakeLists.txt to build the SMASH-analysis suite with an older numpy version. Other adjustments regarding the minimal version of the required modules are at the user's own risk.

Building the SMASH-analysis suite

Use the following commands to create all subtargets of the SMASH-analysis suite in a separate build directory:

mkdir build
cd build
cmake .. -DSMASH_PATH=[...]/smash/build

All subtargets of the analysis suite have been created by cmake.

Note though, that most of them are split into simulation targets (sims), analysis targets (analysis) and plotting targets (plots). This bears the advantage that the SMASH simulations need not be run again if the date of last modification has changed because the files where touched unrelated to code modifications. On the other hand, this procedure requires three subtargets to be executed one after the other to yield the final plots. In the example of the scattering rates (called 'elastic_box'), this means the following commands need to be executed in the appropriate order:

make elastic_box_sims -jN
make elastic_box_analysis -jN
make elastic_box_plots -jN

Where N corresponds to the number of cores, if the SMASH-analysis suite is run in parallel.

The above described splitting applies to the following targets:

angular_distributions
cross_sections
detailed_balance
elastic_box
energy_scan

In the case of FOPI Pions, there are only two subtargets:

make FOPI_pions_sims -jN
make FOPI_pions_plots -jN

And for the dileptons there is only one single target:

make dileptons -jN

Note, that all dilepton data files generated during the SMASH run are immediately removed once they have been analyzed, as they are exceptionally large.

In case of the afterburner target, the SMASH-analysis suite has to be compiled with an additional argument -DSAMPLED_LISTS. This has to be an URL from where the sampled lists as an input for the afterburner target can be downloaded. Without this argument, the target is not available. The URL for the lists used in the latest report is http://theory.gsi.de/~smash/data/sampled_lists/.

Running the SMASH-analysis suite on a SLURM cluster

As most of the targets have a significant runtime (some of them up to several days), it is recommended to run the SMASH-analysis on a cluster. An exemplary slurm script, slurm-job.sh, is provided in the smash-analysis directory. It may be used and adapted. Note that in this case, also SMASH is compiled on the computing note.

To submit all the above mentioned targets instantaneously to the cluster, the script submit_to_slurm.sh may be useful. It is currently optimized for the kronos cluster at GSI. It can be executed via

./submit-to-slurm.sh [...]/smash [...]/eigen all [...]/output_dir

where the path to the SMASH directory, the path to the eigen directory, the target (all in this case, such that all targets are being executed) and the path to the output directory are being passed from the command line.

In case SMASH crashes for some reason, there will be stale lock files which prevent SMASH to be run again in the same directory. They can be deleted via

find [...]/output_dir -type f -name smash.lock -exec rm {} \;

Creating the report

To interpret the generated plots at one glance, it has proven to be most suitable to create a report in the shape of an html page. The following tools are required in addition to those listed above:

Python subprocesses version ≥ 16.6.1.
pdftoppm, embedded in poppler

The html page contains all plots created by means of the SMASH-analysis suite, sorted by observables. It can be produced by executing the create_report.sh script located in the smash-analysis directory:

./create_report.sh [...]/output_dir

Where [...]/output_dir is the path to the directory, that contains the results of the SMASH-analysis runs. Note though, that this script relies on a specific folder structure of these results. It might not work, if the output directory was created in a different manner than described above.

Upon execution, two new directories are created in the output_dir directory. The first, PDFs_configs_{smash_version} is a collection of all plots and configuration files associated with the SMASH runs. The second, report_{smash_version}, is the actual html report, containing the plots and references to the configuration files.

Should you encounter difficulties installing poppler on the machine you ran the SMASH simulations and analysis, you may want to consider copying the PDFs_configs_{smash_version} directory to a different machine and execute the generate_report_folder.sh script located in smash-analysis/test/report.

Storing the results for future comparison

Unless disabled, all generated results are compared to those of the previous tagged SMASH version. The data files corresponding to the results of the latest version are located in the old_results directory, sorted by version. This collection can be generated by executing the store_old_results.sh script, located in the test/report directory:

cd test/report
./store_old_results.sh {smash-version} [...]/output_dir

where {smash-version} is the version of the SMASH code that is being tested and output_dir the directory that contains the results. A new directory with all data files, old_results_{smash-version}, is then created within the output_dir. Note though that it is only useful to assemble the old results for a newly-tagged version of SMASH. Release candidate tags are usually not considered.

smash-transport / smash-analysis