Interacting with Grackle in Your Simulation Code

The majority of this document follows the implementation of Grackle in a C++ simulation code. Full implementation examples for C, C++, and Fortran are also available in the Grackle source. See Example Executables for more information. For a list of all available functions, see the API Reference.

Grackle currently supports two APIs. The Primary API, manages some of Grackle’s data structures (e.g. chemistry_data and chemistry_data_storage) for a downstream simulation code by making use of global variables. In contrast, the Local API, requires that the downstream application explicitly manage pointers to these same data-structures and requires that the pointers are provided as arguments to each function. The latter API is explicitly thread-safe as it involves no global data.

Example Executables

The grackle source code contains examples for C, C++, and Fortran codes. They are located in the src/example directory and provide examples of calling all of grackle’s functions.

c_example.c - C example

c_local_example.c - C example using only Local Functions

cxx_example.C - C++ example

cxx_omp_example.C - C++ example using OpenMP

fortran_example.F - Fortran example

The instructions for building and executing the examples vary based on the build-system:

Once you have already installed the grackle library, you can build the examples by typing make and the name of the file without extension. For example, to build the C++ example, type:

$ make cxx_example

To run the example, make sure to add the path to the directory containing the installed libgrackle.so to your LD_LIBRARY_PATH (or DYLD_LIBRARY_PATH on Mac).

Important

The examples make certain assumptions about the location of the input files. To ensure that the input files can be found, you should execute each example-binary from the same directory where the example binary is produced.

Header Files

Based on the language that your simulation code is written in, you will interact with one of the 2 header files that is shipped with Grackle:

grackle.h - the primary header file, containing declarations for all the available functions and data structures. This is the only header file that needs to be included for C and C++ codes.

grackle.def - the header file to be used in Fortran codes. Only this file needs to be included.

Note

While other header files are also shipped with Grackle we consider them to be implementation details. With that said, we currently maintain the grackle_types.h, grackle_chemistry_data.h, grackle_fortran_types.def, grackle_chemistry_data.h, grackle_fortran_types.def, grackle_fortran_interface.def, and grackle_macros.h because they were previously mentioned in the documentation. However, their usage is now deprecated. We will only maintain these for a couple releases after 3.3. Afterwards, we may alter their contents or delete them entirely.

Note

Earlier versions of Grackle required C++ codes to enclose the #include <grackle.h> include-directive within a C “language-linkage block” (the block starts with extern "C" {). C++ codes should now directly include the header (without the block). The headers internally use a standard idiom to properly handle this case.

Data Types

The grackle library provides special configurable data types that are used to specify the floating-point precision of the baryon fields passed to the grackle functions. For C and C++ codes, this is the gr_float type. For Fortran codes, this is the R_PREC type. Based on how you configure Grackle during compilation, both types will either alias the respective langauges’ 4-byte (32-bit/single-precision) or 8-byte (64-bit/double precision) floating-point type. When using the classic build system, this is controlled by assigning the CONFIG_PRECISION setting to precision-32 or precision-64. When using the CMake build system, this is controlled by setting the GRACKLE_USE_DOUBLE cmake-variable to OFF or ON.

type gr_float: Floating point type used for the baryon fields in C and C++. When Grackle is compiled with precision-32 (under the classic build system) or GRACKLE_USE_DOUBLE=OFF (under the CMake build system), this aliases the float type. When Grackle is compiled with precision-64 or GRACKLE_USE_DOUBLE=ON, this aliases the double type.

type R_PREC: The Fortran analog of gr_float. When Grackle is compiled with precision-32 (under the classic build system) or GRACKLE_USE_DOUBLE=OFF (under the CMake build system), this aliases the real*4 type. When Grackle is compiled with precision-64 or GRACKLE_USE_DOUBLE=ON, this aliases the real*8 type.

Enabling Output

By default, grackle will not print anything but error messages. However, a short summary of the running configuration can be printed by setting the global grackle_verbose variable to 1. In a parallel code, it is recommended that output only be enabled for the root process.

// Enable output
grackle_verbose = 1;

Code Units

Many of the calculations involved in chemical reactions and radiative cooling include multiplications by density squared or even density cubed. With typical gas densities relevant to galaxy formation being of the order of one hydrogren atom per cubic centimeter (~10^-24 g/cm³, give or take a few orders of magnitude), it is easy to end up with significant roundoff or underflow errors when quantities are stored in CGS units.

The code_units structure contains conversions from code units to CGS such that a value passed to Grackle multiplied by the appropriate code unit gives that value in CGS units. Units for density, length, time, and the expansion factor must be set manually. Units for velocity are then set by calling set_velocity_units. When using the proper frame (i.e., setting comoving_coordinates to 0), a_units (units for the expansion factor) must be set to 1.0. See below for recommendations on choosing appropriate units.

type code_units: This structure contains the following members.

int comoving_coordinates: If set to 1, the incoming field data is assumed to be in the comoving frame. If set to 0, the incoming field data is assumed to be in the proper frame.

double density_units: Conversion factor to be multiplied by density fields to return densities in proper g/cm³.

double length_units: Conversion factor to be multiplied by length variables to return lengths in proper cm.

double time_units: Conversion factor to be multiplied by time variables to return times in s.

double velocity_units: Conversion factor to be multiplied by velocities to return proper cm/s. This should be set units the set_velocity_units function. Note, units of specific energy (i.e., conversion to erg/g) are then defined as velocity_units² (velocity units squared).

double a_units: Conversion factor to be multiplied by the expansion factor such that a_true = a_code* a_units. When using proper coordinates, a_units must be set to 1.

double a_value: The current value of the expansion factor in units of a_units. The conversion from redshift to expansion factor in code units is given by a_value = 1 / (1 + z) / a_units. If the simulation is not cosmological, a_value should be set to 1. Note, if a_value is set to something other than 1 in a non-cosmological simulation, all redshift dependent chemistry and cooling terms will be set corresponding to the redshift given.

code_units my_units;
my_units.comoving_coordinates = 0; // 1 if cosmological sim, 0 if not
my_units.density_units = 1.67e-24; // 1 m_H/cc
my_units.length_units = 3.086e21;  // 1 kpc
my_units.time_units = 3.15569e13;  // 1 Myr
my_units.a_units = 1.0;            // units for the expansion factor
my_units.a_value = 1. / (1. + current_redshift) / my_units.a_units;
// set velocity units
set_velocity_units(&my_units);

Choosing Appropriate Units

The main consideration when setting code units is to keep density, length, and time values close to 1. Reasonable values for density, length, and time units are the hydrogen mass in g, 1 kpc to 1 Mpc in cm, and 1 Myr to 1 Gyr in s.

Comoving Coordinates

For cosmological simulations, a comoving unit system is preferred, though not required, since it allows the densities to stay close to 1 as the universe expands. If comoving_coordinates is set to 1, it is assumed that the fields being passed to the solver are in the comoving frame. Hence, the units must convert from code units in the comoving frame to CGS in the proper frame. If comoving_coordinates is set to 0, it is assumed that the fields passed into the solver are in the proper frame. For an example of using comoving units, see the cosmological unit system in the Enzo code.

As the unit system is designed to convert from the comoving to the proper frame, some of the values in the code_units struct are expected to change with expansion factor (or redshift) while some others should remain constant. Units that should remain constant include time_units and a_units. Units that should vary are a_value (obviously), length_units, and density_units. Moving forward in time, length_units should be increasing proportional to a_value and density_units should be decreasing as a_value^-3.

There are two important corollaries of the above behavior. First, the velocity_units should remain constant. In comoving coordinates, velocity units are given by

\[VU = \frac{LU}{a\ TU},\]

where VU is velocity_units, a is a_value, and TU is time_units. Second, the internal unit for the cooling rate (equivalent to [erg s^-1 cm⁺³]) should remain constant. The cooling unit (CU) is given by

\[CU = \frac{VU^2\ m_H^2}{DU\ a^3\ TU},\]

where DU is density_units and m_H is the hydrogen mass. The above definitions also hold for proper coordinates by setting a to 1.

As a convenience, we provide the gr_query_units() function to let the user query the code units at arbitrary values of the expansion factor. This is intended to be used when implementing Grackle support into your simulation (and debugging) in order to let you manually confirm that the code units you are passing in are correct (internally, Grackle simply assumes that the caller is giving the right values). The values returned by this function are based on the the initial values stored in the code_units data structure (that are passed into grackle while setting up chemistry data and the associated storage).

Chemistry Data

Grackle’s behavior is controlled by the chemistry_data type. It is a structure that stores all of the relevant run-time parameters. After an instance of the type is configured, it is used to initialize all of the chemistry and cooling rate arrays (stored by the separate chemistry_data_storage type).

type chemistry_data: This structure holds all grackle run-time parameters, which are listed in Parameters and Data Files.

type chemistry_data_storage: This structure holds all chemistry and cooling rate arrays. The user will not normally need to work directly with its internals. Users calling the Primary Functions will not encounter it (because the Primary Functions make use of an internally stored instance of this type). Users implementing the Local Functions will have to store a pointer to one of these.

There is a 2 step procedure for setting up Grackle’s chemistry data (the precise details vary based on the choosen API flavor):

Configure the solver’s run-time parameters. There are 3 parts to this step. First, allocate memory for a chemistry_data instance. Next, initialize the instance by calling a function to perform the initial setup of the instance and store all the default parameter values. Finally, assign the desired values to the parameters tracked by the instance. See Parameters and Data Files for a full list of the available parameters.

In this case, the set_default_chemistry_parameters() routine for initializing the structure holding the parameters. The function must be handed a pointer to an instance of chemistry_data. That pointer is attached to the global grackle_data variable and the memory is properly initialized. The function returns an integer indicating success (1) or failure (0).

The convention is to assign desired parameter values by accessing grackle_data.

chemistry_data *my_grackle_data;
my_grackle_data = new chemistry_data;
if (set_default_chemistry_parameters(my_grackle_data) == 0) {
  fprintf(stderr, "Error in set_default_chemistry_parameters.\n");
}

// Set parameter values for chemistry.
// Now access the global copy of the chemistry_data struct (grackle_data).
grackle_data->use_grackle = 1;            // chemistry on
grackle_data->with_radiative_cooling = 1; // cooling on
grackle_data->primordial_chemistry = 3;   // molecular network with H, He, D
grackle_data->metal_cooling = 1;          // metal cooling on
grackle_data->UVbackground = 1;           // UV background on
grackle_data->grackle_data_file = "CloudyData_UVB=HM2012.h5"; // data file

In this scenario, the local_initialize_chemistry_parameters() routine is used to initialize the previously allocated memory of a chemistry_data instance. The function returns an integer indicating success (1) or failure (0).

Unlike the other scenario, no global variable is initialized; instead the application is responsible for keeping track of the allocated object.

chemistry_data *my_grackle_data;
my_grackle_data = new chemistry_data;
if (local_initialize_chemistry_parameters(my_grackle_data) == 0) {
  fprintf(stderr, "Error in local_initialize_chemistry_parameters.\n");
}

// Set parameter values for chemistry.
my_grackle_data->use_grackle = 1;            // chemistry on
my_grackle_data->with_radiative_cooling = 1; // cooling on
my_grackle_data->primordial_chemistry = 3;   // molecular network with H, He, D
my_grackle_data->metal_cooling = 1;          // metal cooling on
my_grackle_data->UVbackground = 1;           // UV background on
my_grackle_data->grackle_data_file = "CloudyData_UVB=HM2012.h5"; // data file

2. Once the desired parameters have been set, the chemistry and cooling rates must be initialized by calling with a pointer to the code_units struct created earlier. This function will return an integer indicating success (1) or failure (0).

In this case, the initialization function is called initialize_chemistry_data(). This function internally allocates memory for the type that holds the storage data, chemistry_data_storage, and stores the address in a variable that is tracked behind the scenes. This storage data is automatically passed to other Primary Functions.

// Finally, initialize the chemistry object.
if (initialize_chemistry_data(&my_units) == 0) {
  fprintf(stderr, "Error in initialize_chemistry_data.\n");
  return 0;
}

In this case, the initialization function is called local_initialize_chemistry_data(). The application is responsible for passing the pointer to the (previously configured) chemistry_data() instance into the function. The application is also responsible for allocating the memory for the chemistry_data_storage instance that the function initializes. Note: the initialization function allocates additional memory that is tracked within the chemistry_data_storage for holding chemistry and cooling rates.

Unlike the other scenario, the downstream application is responsible for tracking the initialized object (so that it can be passed to other functions).

// Allocate the chemistry_data_storage type
chemistry_data_storage *my_grackle_rates = new chemistry_data_storage;

// Finally, initialize the chemistry object.
if (local_initialize_chemistry_data(my_chemistry_data, my_grackle_rates,
                                    &my_units) == 0) {
  fprintf(stderr, "Error in local_initialize_chemistry_data.\n");
  return 0;
}

The Grackle is now ready to be used.

As an aside, see Dynamic configuration of Chemistry Data for a description of an alternative approach for configuring a chemistry_data struct. This other approach may provide additional compatability with multiple versions of Grackle, and in some cases may facillitate less-verbose, easier-to-maintain code.

Running with OpenMP

As of version 2.2, Grackle can be run with OpenMP parallelism. To do this, the library must be compiled with OpenMP support. When compiling Grackle

with the classic build system, you should execute make omp-on before compiling (more information about modifying settings in this system are provided here)

with the CMake build system, you should configure the build by assigning the GRACKLE_USE_OPENMP cmake-variable a value of ON (more information about modifying settings in this system are provided here)

For an example of how to compile your code with OpenMP, see the cxx_omp_example.C code example (Example Executables). Once your code has been compiled with OpenMP enabled, the number of threads used can be controlled by setting the omp_nthreads parameter, stored in the grackle_data struct.

// 8 threads per process
grackle_data->omp_nthreads = 8;

If not set, this parameter will be set to the maximum number of threads possible, as determined by the system or as configured by setting the OMP_NUM_THREADS environment variable.

Creating the Necessary Fields

As of version 3.0, the various density and energy fields are passed to Grackle’s functions using a struct of type grackle_field_data. The struct contains information about the size and shape of the field arrays and pointers to all field arrays.

Note on Density Fields

All density fields provided to Grackle should be mass densities, i.e., the number density of a given species multiplied by its mass. The units should be such that the field value multiplied by density_units results in a value with units of g/cm³. See Code Units for further discussion of Grackle unit systems.

Note on the Electron Density Field

For largely historical reasons, the value of the electron density (provided in e_density) should be the true value of the electron mass density multiplied by the ratio of the proton mass to the electron mass, i.e., e_density = \({\rho}\)_e * m_p / m_e, where \({\rho}\)_e is the true electron mass density in density_units (see Note on Density Fields).

type grackle_field_data: This structure is used to pass field data to Grackle’s functions. It contains the following members:

int grid_rank: The active dimensions (not including ignored boundary zones) of the field arrays.

int *grid_dimension: This should point to an array of size grid_rank. This stores the size of the field arrays in each dimension.

int *grid_start: This should point to an array of size grid_rank. This stores the starting value in each dimension for the field data. This can be used to ignore boundary cells in grid data.

int *grid_end: This should point to an array of size grid_rank. This stores the end value in each dimension for the field data. This can be used to ignore boundary cells in grid data.

gr_float grid_dx: This is the grid cell width in length_units. This is currently used only in computing approximate H2 self-shielding when H2 is tracked (primordial_chemistry >= 2) and H2_self_shielding is set to 1. If this can’t be assigned a meaningful value (e.g. the field data does not organized on a grid or the grid cells aren’t perfect cubes), we recommend assigning it a value of -1 (so that error-handling works properly)

gr_float *density: Pointer to the gas density field array.

gr_float *HI_density: Pointer to the HI density field array. Used when primordial_chemistry is set to 1, 2, or 3.

gr_float *HII_density: Pointer to the HII density field array. Used when primordial_chemistry is set to 1, 2, or 3.

gr_float *HM_density: Pointer to the H^- density field array. Used when primordial_chemistry is set to 2 or 3.

gr_float *HeI_density: Pointer to the HeI density field array. Used when primordial_chemistry is set to 1, 2, or 3.

gr_float *HeII_density: Pointer to the HeII density field array. Used when primordial_chemistry is set to 1, 2, or 3.

gr_float *HeIII_density: Pointer to the HeIII density field array. Used when primordial_chemistry is set to 1, 2, or 3.

gr_float *H2I_density: Pointer to the H₂ density field array. Used when primordial_chemistry is set to 2 or 3.

gr_float *H2II_density: Pointer to the H₂⁺ density field array. Used when primordial_chemistry is set to 2 or 3.

gr_float *DI_density: Pointer to the DI density field array. Used when primordial_chemistry is set to 3.

gr_float *DII_density: Pointer to the DII density field array. Used when primordial_chemistry is set to 3.

gr_float *HDI_density: Pointer to the HD density field array. Used when primordial_chemistry is set to 3.

gr_float *e_density: Pointer to the electron density field array. Used when primordial_chemistry is set to 1, 2, or 3. Note, the electron mass density should be scaled by the ratio of the proton mass to the electron mass. See Note on the Electron Density Field for more information.

gr_float *metal_density: Pointer to the metal density field array. Used when metal_cooling is set to 1.

gr_float *dust_density: Pointer to the dust density field array. Used when use_dust_density_field is set to 1.

gr_float *internal_energy: Pointer to the internal energy field array. Internal energies should be in units of velocity_units² (velocity units squared). This can be converted to and from a temperature by using the get_temperature_units function.

gr_float *x_velocity: Pointer to the x-velocity field array. Currently not used.

gr_float *y_velocity: Pointer to the y-velocity field array. Currently not used.

gr_float *z_velocity: Pointer to the z-velocity field array. Currently not used.

gr_float *volumetric_heating_rate: Pointer to values containing volumetric heating rates. Rates should be in units of erg/s/cm³. Used when use_volumetric_heating_rate is set to 1.

gr_float *specific_heating_rate: Pointer to values containing specific heating rates. Rates should be in units of erg/s/g. Used when use_specific_heating_rate is set to 1.

gr_float *temperature_floor: Pointer to values containing a temperature floor for each element in units of K. No chemistry or cooling calculations will be performed on an element with a temperature at or below the specified value. Used when use_temperature_floor is set to 2.

gr_float *RT_heating_rate: Pointer to the radiation transfer heating rate field. Rates should be in units of (erg/s/cm³) / n_HI, where n_HI is the neutral hydrogen number density. Heating rates for additional species are currently not yet supported. Used when use_radiative_transfer is set to 1.

gr_float *RT_HI_ionization_rate: Pointer to the HI photo-ionization rate field used with radiative transfer. Rates should be in units of 1/time_units. Used when use_radiative_transfer is set to 1.

gr_float *RT_HeI_ionization_rate: Pointer to the HeI photo-ionization rate field used with radiative transfer. Rates should be in units of 1/time_units. Used when use_radiative_transfer is set to 1.

gr_float *RT_HeII_ionization_rate: Pointer to the HeII photo-ionization rate field used with radiative transfer. Rates should be in units of 1/time_units. Used when use_radiative_transfer is set to 1.

gr_float *RT_H2_dissociation_rate: Pointer to the H₂ photo-dissociation rate field used with radiative transfer. Rates should be in units of 1/time_units. Used when use_radiative_transfer is set to 1 and primordial_chemistry is either 2 or 3.

gr_float *H2_self_shielding_length: Pointer to a field containing lengths to be used for calculating molecular hydrogen column denisty for H2₂self-shielding. Used when H2_self_shielding is set to 2. Field data should be in length_units.

gr_float *H2_custom_shielding_factor: Pointer to a field containing attenuation factors to be multiplied with the H₂dissociation rate. Used when the H2_custom_shielding flag is set.

gr_float *isrf_habing: Pointer to a field containing values of the strength of the insterstellar radiation field used in the calculation of dust heating. This is used when use_isrf_field is set to 1. The units of this field should be the same as those of the interstellar_radiation_field parameter.

It is not necessary to attach a pointer to any field that you do not intend to use.

// Create struct for storing grackle field data
grackle_field_data my_fields;

// initialize members of my_fields to sensible defaults
gr_initialize_field_data(&my_fields);

// Set grid dimension and size.
// grid_start and grid_end are used to ignore ghost zones.
int field_size = 1;
my_fields.grid_rank = 3;
my_fields.grid_dimension = new int[3];
my_fields.grid_start = new int[3];
my_fields.grid_end = new int[3];
my_fields.grid_dx  = 1.0; // only matters if H2 self-shielding is used
                          // we recommend assigning it a value of -1 if your
                          // simulation doesn't have a meaningful value for
                          // it (so that error-handling works properly)
for (int i = 0;i < 3;i++) {
  my_fields.grid_dimension[i] = 1;
  my_fields.grid_start[i] = 0;
  my_fields.grid_end[i] = 0;
}
my_fields.grid_dimension[0] = field_size;
my_fields.grid_end[0] = field_size - 1;

// Set field arrays.
my_fields.density         = new gr_float[field_size];
my_fields.internal_energy = new gr_float[field_size];
my_fields.x_velocity      = new gr_float[field_size];
my_fields.y_velocity      = new gr_float[field_size];
my_fields.z_velocity      = new gr_float[field_size];
// for primordial_chemistry >= 1
my_fields.HI_density      = new gr_float[field_size];
my_fields.HII_density     = new gr_float[field_size];
my_fields.HeI_density     = new gr_float[field_size];
my_fields.HeII_density    = new gr_float[field_size];
my_fields.HeIII_density   = new gr_float[field_size];
my_fields.e_density       = new gr_float[field_size];
// for primordial_chemistry >= 2
my_fields.HM_density      = new gr_float[field_size];
my_fields.H2I_density     = new gr_float[field_size];
my_fields.H2II_density    = new gr_float[field_size];
// for primordial_chemistry >= 3
my_fields.DI_density      = new gr_float[field_size];
my_fields.DII_density     = new gr_float[field_size];
my_fields.HDI_density     = new gr_float[field_size];
// for metal_cooling = 1
my_fields.metal_density   = new gr_float[field_size];
// volumetric heating rate (provide in units [erg s^-1 cm^-3])
my_fields.volumetric_heating_rate = new gr_float[field_size];
// specific heating rate (provide in units [egs s^-1 g^-1]
my_fields.specific_heating_rate = new gr_float[field_size];
// heating rate from radiative transfer calculations (provide in units [erg s^-1 cm^-3]
my_fields.RT_heating_rate = new gr_float[field_size];
// HI ionization rate from radiative transfer calculations (provide in units of [ 1/time_units ]
my_fields.RT_HI_ionization_rate = new gr_float[field_size];
// HeI ionization rate from radiative transfer calculations (provide in units of [1/time_units]
my_fields.RT_HeI_ionization_rate = new gr_float[field_size];
// HeII ionization rate from radiative transfer calculations (provide in units of [1/time_units]
my_fields.RT_HeII_ionization_rate = new gr_float[field_size];
// H2 dissociation rate from radiative transfer calculations (provide in units of [1/time_units]
my_fields.RT_H2_dissociation_rate = new gr_float[field_size];

Note

The electron mass density should be scaled by the ratio of the proton mass to the electron mass such that the electron density in the code is the electron number density times the proton mass.

Calling the Available Functions

There are six functions available, one to solve the chemistry and cooling and five others to calculate the cooling time, temperature, pressure, ratio of the specific heats (gamma), and dust temperature. The arguments required are the code_units structure and the grackle_field_data struct. For the chemistry solving routine, a timestep must also be given. For the four field calculator routines, the array to be filled with the field values must be created and passed as an argument as well.

The examples below make use of Grackle’s Primary Functions, where the parameters and rate data are stored in instances of the chemistry_data and chemistry_data_storage structs declared in grackle.h. Alternatively, a set of Local Functions require these structs to be provided as arguments, allowing for explicitly thread-safe code.

Solve the Chemistry and Cooling

// some timestep (one million years)
double dt = 3.15e7 * 1e6 / my_units.time_units;

if (solve_chemistry(&my_units, &my_fields, dt) == 0) {
  fprintf(stderr, "Error in solve_chemistry.\n");
  return 0;
}

Calculating the Cooling Time

gr_float *cooling_time;
cooling_time = new gr_float[field_size];
if (calculate_cooling_time(&my_units, &my_fields,
                           cooling_time) == 0) {
  fprintf(stderr, "Error in calculate_cooling_time.\n");
  return 0;
}

Calculating the Temperature Field

gr_float *temperature;
temperature = new gr_float[field_size];
if (calculate_temperature(&my_units, &my_fields,
                          temperature) == 0) {
  fprintf(stderr, "Error in calculate_temperature.\n");
  return EXIT_FAILURE;
}

Calculating the Pressure Field

gr_float *pressure;
pressure = new gr_float[field_size];
if (calculate_pressure(&my_units, &my_fields,
                       pressure) == 0) {
  fprintf(stderr, "Error in calculate_pressure.\n");
  return EXIT_FAILURE;
}

Calculating the Gamma Field

gr_float *gamma;
gamma = new gr_float[field_size];
if (calculate_gamma(&my_units, &my_fields,
                    gamma) == 0) {
  fprintf(stderr, "Error in calculate_gamma.\n");
  return EXIT_FAILURE;
}

Calculating the Dust Temperature Field

gr_float *dust_temperature;
dust_temperature = new gr_float[field_size];
if (calculate_dust_temperature(&my_units, &my_fields,
                          dust_temperature) == 0) {
  fprintf(stderr, "Error in calculate_dust_temperature.\n");
  return EXIT_FAILURE;
}

Clearing the memory

When using the Grackle’s Primary Functions, global structures are used and therefore the global structure grackle_rates needs to be released with

free_chemistry_data();

When using the Local Functions, you should call the counter-part function, local_free_chemistry_data(), to free memory used for storing chemistry and cooling rates.

Querying library version information

A struct of type grackle_version is used to hold version information about the version of Grackle that is being used. The struct contains information about the version number and particular git revision.

type grackle_version: This structure is used to organize version information for the library.

const char *version: Specifies the version of the library using this template: <MAJOR>.<MINOR>(.<MICRO>)(.dev<DEV_NUM>). In this template <MAJOR>, <MINOR>, and <MICRO> correspond to a major, minor, and micro version numbers (the micro version number is omitted if it’s zero). The final section can specify a development version. More details about this version number can be found here.

const char *branch: Specifies the name of the git branch that the library was compiled from.

const char *revision: Specifies the hash identifying the git commit that the library was compiled from.

The get_grackle_version() function is used to retrieve a properly intialized grackle_version object. The following code snippet illustrates how one might query and print this information:

grackle_version gversion = get_grackle_version();
printf ("The Grackle Version: %s\n", gversion.version);
printf ("Git Branch:   %s\n", gversion.branch);
printf ("Git Revision: %s\n", gversion.revision);

Dynamic configuration of Chemistry Data

The functions providing dynamic access to the fields of chemistry_data are useful for maintaining backwards compatibility with older versions of Grackle (that also provide this API) as new fields get added to chemistry_data. This is exemplified in the following scenario.

Suppose Grackle is updated to have a new heating/cooling mechanism, and to allow users to control that mechanism two new fields are added to chemistry_data:

an int field called use_fancy_feature
a double field called fancy_feature_param

Now suppose a downstream simulation code, written in c or c++, wanted to support configuration of this feature. In this scenario, imagine that we have a pointer to a chemistry_data structure called my_grackle_data.

The obvious way to configure this feature is to include the following snippet in the simulation code:

if (configure_fancy_feature) {
  my_grackle_data->use_fancy_feature = 1;
  my_grackle_data->fancy_feature_param = 5.0; // arbitrary value
}

However, inclusion of the above snippet will prevent the simulation code from compiling if the user has a version of Grackle installed in which chemistry_data does not have the use_fancy_feature and fancy_feature_param fields. Consequently, such users will have to update Grackle.

This can be inconvenient when a user has no interest in using this new feature, but needs an unrelated feature/bugfix introduced to the code in a subsequent changeset

This is especially inconvenient if a user is prototying a new feature in a custom Grackle branch in which the chemistry_data struct is missing these fields.

The following snippet shows how the dynamic access API can be used in the same way for versions of Grackle that include these parameters, and don’t set the features in cases

if (configure_fancy_feature) {
  int* use_fancy_feature = local_chemistry_data_access_int(
    my_grackle_data, "use_fancy_feature"
  );
  double* fancy_feature_param = local_chemistry_data_access_double(
    my_grackle_data, "fancy_feature_param"
  );

  if ((use_fancy_feature == NULL) || (fancy_feature_param == NULL)){
    fprintf(stderr, "Update grackle version to use fancy feature\n");
  } else {
    *use_fancy_feature = 1;
    *fancy_feature_param = 5.0;
  }
}

There are a few points worth noting:

As the above snippets show, the dynamic api clearly produces more verbose code when configuring chemistry_data field-by-field. However, in codes where users configure Grackle by specifying the name of fields in the chemistry_data struct and the associated values in a parameter file, the dynamic API can facillitate MUCH less verbose code. Under certain implementations, it may not even be necessary to modify a simulation code to support newly-introduced grackle parameters.

The dynamic API is slower than configuring chemistry_data in the classic approach. However, this shouldn’t be an issue since chemistry_data is usually just configured once when the simulation code starts up.

The highlighted functions can also be used in tandem with other functions described in Dynamic Configuration Functions to simplify (de)serialization of chemistry_data.

For completeness, the dynamic API also provides an analogous function for configuring string parameters.