Skip to content

Castro 20.02
Compare
Choose a tag to compare
@zingale zingale released this 01 Feb 19:38
20.02
4c9a23c

20.02

  • Fixed a bug in the nuclear burning timestep estimator when on GPUs
    (#745)

  • rewrote the 4th order SDC hydro driver in C++ to allow code reuse
    with other solvers (#742), and simplified the 2nd order SDC code
    to do dimensional sweeps to reduce memory (#749)

  • The option radiation.integrate_planck has been removed; it was only
    used by one test. By default we always do the full integral of the
    Planck function. (#740)

  • Most of the radiation test problems have been moved over to a new
    opacity directory, rad_power_law, and all of the parameters that
    controlled the behavior of the power law opacity have been moved
    to the extern probin module. We now always expect you to pick a
    specific opacity implementation, so the parameter
    radiation.use_opacity_table_module has been removed. The "null"
    opacity implementation has been previously moved, and the code
    will fail to compile if you attempt to use it; you will need to
    update to rad_power_law. (See the documentation for information
    about how to use this new implementation.)

    Additionally, the code for the multigroup solver was effectively
    previously setting the Rosseland opacity, kappa_r, equal to the
    Planck opacity, kappa_p, if the latter was set but the former was
    not. There was similar unintuitive behavior for the behavior of
    the scattering parameter. Now you will get exactly what you ask
    for in the probin file, given the defaults in the _parameters file
    for the rad_power_law opacity. By default the constant coefficients
    for both are negative, which is invalid, so both must be set to a
    non-negative value for the code to work. Problems that were previously
    setting const_kappa_p but not const_kappa_r should set the latter
    equal to the former to maintain the same code behavior. The analogous
    thing should be done for the exponents (kappa_p_exp_m, kappa_p_exp_n,
    and kappa_p_exp_p). (#725)

  • The parameter radiation.do_real_eos = 0 has been removed, and its
    functionality is now enabled with a new equation of state called
    rad_power_law. This new EOS is only compatible with the pure
    radiation-diffusion tests, not with castro.do_hydro = 1. (#722)

  • We now default to use_retry = 1, instructing Castro to retry a
    step with a smaller dt if there is a CFL violation, burning
    failure, or negative timestep. For the burning failure, we have
    Castro set the Microphysics parameter abort_on_failure to .false.
    at a high priority (so it overrides the Microphysics default).
    We also check to make sure the combination of parameters makes
    sense at runtime. (#724)

  • The parameter castro.hard_cfl_limit has been removed. (#723)

  • Some unnecessary clean_state calls were removed (#721)

  • Support for neutrino radiation diffusion has been removed.

  • A bug was fixed in the hydro CFL timestep estimator for
    simplified-SDC. The timestep was more restrictive than it needed
    to be. (#727)

  • A bug was fixed in the simplified-SDC nuclear burning timestep
    estimator (#733)

Assets 2
Loading