Overview of Unit Tests

There are a few unit tests in Microphysics that operate on a generic EOS, reaction network, conductivity, or some smaller component to Microphysics. Many of these tests exercise the main interfaces in Microphysics/interfaces/ and the code that those call.

These tests compile using the AMReX build system, which assumes that main is in C++, so each have a main.cpp driver. The files Microphysics/Make.Microphysics and Microphysics/Make.Microphysics_extern provide the macros necessary to build the executable. Runtime parameters are parsed in the same fashion as in the application codes, using the write_probin.py script.

Note

Most of these tests work with MPI+OpenMP, MPI+CUDA, and MPI+HIP

Tests are divided into three categories:

• comprehensive tests work on a cube of data (usually \(\rho\), \(T\), and composition varying along the three dimensions) and are meant to exercise a wide range of input conditions.

  These are mainly used for regression testing.

• one-zone tests allow you to evaluate the conditions for a particular thermodynamic state.

  These are often used for interactive explorations and within the CI.

• infrastructure tests test small bits of the solver, function interfaces, or runtime infrastructure.

  These are not really meant for exploring the actual thermodynamic state.

Comprehensive tests

Each of these tests sets up a cube of data, \((\rho, T, X_k)\), with the range of \(T\) and \(\rho\), and the species to focus on for \(X_k\) controlled by options in the input file.

• test_aprox_rates :

  call each of the hardcoded rate functions in Microphysics/rates/ on each cell in the data cube and store the output in a plotfile.

• test_conductivity :

  call one of the conductivity routines (set via CONDUCTIVITY_DIR) on each cell in the data cube and store the output in a plotfile.

• test_eos :

  call one of the equations of state (set via EOS_DIR) on each cell in the data cube. We first call it with \((\rho, T, X_k)\) as input (eos_input_rt), and then test each of the other EOS modes (eos_input_rh, eos_input_tp, eos_input_rp, eos_input_re, eos_input_ps, eos_input_ph, eos_input_th) and for each of these modes, we compute the error in the recovered \(T\) and/or \(\rho\) (as appropriate). The full thermodynamic state and errors are stored in a plotfile.

• test_jac :

  for each cell in the data cube, and for a specific network (set via NETWORK_DIR) call the analytic Jacobian provided by the network and compute the numerical Jacobian (via finite differencing) and store the relative difference between each approximation for each Jacobian element in a plotfile.

• test_neutrino_cooling :

  for each cell in the data cube, call the neutrino cooling function and store the output in a plotfile.

• test_react :

  for each cell in the data cube, call the reaction network and integrate to a specified time. Statistics about the number of RHS calls are reported at the end. A lot of options can be set via the inputs file to control the integration.

• test_rhs :

  for each cell in the data cube, call the network’s righthand side and Jacobian functions and store the output in a plotfile. The network is controlled by the NETWORK_DIR variable.

• test_screening_templated :

  for any of the templated-C++ networks, this test will loop over all of the rates and calculate the screening factor (the screening routine can be set via SCREEN_METHOD). The factors for each cell in the data cube are written to a plotfile.

• test_sdc :

  similar to test_react, except now we exercise the SDC integration infrastructure. The conserved state that is input to the burner is chosen to have a Mach number of \(0.1\) (and otherwise use the thermodynamics from the data cube). No advective terms are modeled.

One-zone tests

• burn_cell :

  given a \(\rho\), \(T\), and \(X_k\), integrate a reaction network through a specified time and output the new state.

• burn_cell_primordial_chem :

  similar to burn_cell except specific for the primordial chemistry network.

• burn_cell_sdc :

  similar to burn_cell except this uses the SDC integration code paths.

• eos_cell :

  given a \(\rho\), \(T\), and \(X_k\), call the equation of state and print out the thermodynamic information.

• jac_cell :

  for a single thermodynamic state, compute the analytic Jacobian (using the functions provided by the network) and a numerical finite-difference approximation to the Jacobian and output them, element-by-element, to the display.

• nse_table_cell :

  given a \(\rho\), \(T\), and \(Y_e\), evaluate the NSE state via table interpolation and print it out.

• test_ase :

  for the self-consistent NSE, take a \(\rho\), \(T\), and \(Y_e\), and solve for the NSE state. Then check the NSE condition to see if we are actually satisfying the NSE criteria for the network.

• test_part_func

  exercise the partition function interpolation for a few select nuclei.

Infrastructure tests

• test_linear_algebra :

  create a diagonally dominant matrix, multiply it by a test vector, \(x\), to get \(b = Ax\), and then call the linear algebra routines to see if we we recover \(x\) from \(b\).

• test_nse_interp :

  run various tests of the NSE interpolation routines.

• test_parameters :

  a simple setup that initializes the runtime parameters and can be used to test if we can override them at runtime via inputs or the commandline. This uses both the global data and the struct form of the runtime parameters.

• test_sdc_vode_rhs :

  a simple driver for the SDC RHS routines. Given a thermodynamic state, it outputs the RHS that the integrator will see.