pynucastro.networks.network_compare module

pynucastro.networks.network_compare module#

Helper functions for comparing the output of different backends (python, C++) for the same network. This will build and run each net and return the ydots, rates, jacobian, and energy generation

class pynucastro.networks.network_compare.NetworkCompare(lib, *, use_screening=False, include_amrex=False, include_simple_cxx=True, python_module_name='compare_net.py', amrex_test_path=None, cxx_test_path=None)[source]#

Bases: object

A simple class to manage building a network with different backends to facilitate comparisons.

Note

For the Jacobian, only the species terms are compared.

Parameters:

  • lib (Library) – pynucastro Library containing the rates

  • use_screening (bool) – do we include screening in the comparison (Chugunov 2007)

  • include_amrex (bool) – do we build and evaluate an AmrexAstroCxxNetwork?

  • include_simple_cxx (bool) – do we build and evaluate a SimpleCxxNetwork?

  • python_module_name (str) – the name of the module file to output the python net to

  • amrex_test_path (pathlib.Path) – the Path to output the AMReX C++ source files and build in

  • cxx_test_path (pathlib.Path) – the Path to output the C++ source files and build in

ydots_py_inline#

The dY/dt terms from the python inline network (RateCollection.evaluate_ydots).

Type:

dict(Nucleus)

ydots_py_module#

The dYdt terms from a PythonNetwork module.

Type:

dict(Nucleus)

ydots_amrex#

The dYdt terms parsed from a build of an AmrexAstroCxxNetwork.

Type:

dict(Nucleus)

ydots_cxx#

The dYdt terms parsed from a build of a SimpleCxxNetwork.

Type:

dict(Nucleus)

jac_py_inline#

The Jacobian terms from the python inline network (RateCollection.evaluate_jacobian).

Type:

dict((Nucleus, Nucleus))

jac_py_module#

The Jacobian terms from a PythonNetwork module.

Type:

dict((Nucleus, Nucleus))

jac_amrex#

The Jacobian terms parsed from a build of an AmrexAstroCxxNetwork.

Type:

dict((Nucleus, Nucleus))

jac_cxx#

The Jacobian terms parsed from a build of a SimpleCxxNetwork.

Type:

dict((Nucleus, Nucleus))

rates_py_inline#

The raw rates (e.g. N_A <σv>) from the python inline network (computed via Rate.eval).

Type:

dict(Rate)

rates_py_module#

The raw rates (e.g. N_A <σv>) from a PythonNetwork module.

Type:

dict(Rate)

rates_amrex#

The raw rates (e.g. N_A <σv>) parsed from a build of an AmrexAstroCxxNetwork.

Type:

dict(Rate)

rates_cxx#

The raw rates (e.g. N_A <σv>) parsed from a build of a SimpleCxxNetwork.

Type:

dict(Rate)

enuc_py_inline#

The total energy generation rate from the mass changes (just strong reactions) in erg/g/s computed via evaluate_energy_generation.

Type:

float

enuc_py_module#

The total energy generation rate from the mass changes (just strong reactions) in erg/g/s evaluated from a PythonNetwork module.

Type:

float

enuc_amrex#

The total energy generation rate from the mass changes (just strong reactions) in erg/g/s parsed from a build of an AmrexAstroCxxNetwork.

Type:

float

enuc_cxx#

The total energy generation rate from the mass changes (just strong reactions) in erg/g/s parsed from a build of a SimpleCxxNetwork.

Type:

float

T_eval#

The temperature used in the evaluation.

Type:

float

rho_eval#

The density used in the evaluation.

Type:

float

compare_results(*, quantity='ydots', rtol=1e-11, atol=1e-30)[source]#

Perform the comparison of quantity across the different network types.

Parameters:

  • quantity (str) – The quantity we are comparing. Should be one of “ydots”, “rates”, “jac”, “enuc”, “enu_weak”

  • rtol (float) – The relative tolerance to use in the comparison.

  • atol (float) – The absolute tolerance to use in the comparison.

Raises:

ValueError –

evaluate(rho=200000000.0, T=1000000000.0)[source]#

Evaluate the ydots from all the backends we are considering. This will compile the C++ networks if they have not already been built. This does not do any comparisons.

Parameters:

  • rho (float) – density to evaluate rates at (CGS)

  • T (float) – temperature to evaluate rates at (K)

print_summary(*, jac_floor=1e-90)[source]#

Print a summary of the dY/dt comparison and errors for each network type run.

Parameters:

jac_floor (float) – Value for abs(J_{i,j}) below which to switch to an absolute error instead of relative error. This helps deal with the different ways nets floor rates.