Releases: AMReX-Astro/Microphysics
Release 20.08
-
Several of the unit tests had separate C++ and Fortran
implementions. These have been unified (#343, #344, #345) -
The VBDF integrator was removed (#348)
-
VODE can now reject an internal timestep that has any abundance
change by more than a factor of 2, or an abundance < 0 or > 1,
as well as timesteps where the temperature ends up negative. (#350)
Microphysics 20.07
Microphysics 20.06
20.06
-
For integration with simplified SDC, we now interpret atol_spec
as an absolute tolerance on X alone instead of (rho X) (#311) -
burn_cell can now use the C++ burner if compiled with
USE_CXX_REACTIONS=TRUE and run with do_cxx = 1. (#313) -
The original burn_cell (which used the F90 BoxLib build system)
is removed and replaced with burn_cell_C (which uses the newer
build system). (#316) -
The analytic Jacobian with simplified SDC now is written in terms
of the conserved fluid state and works for a wide range of
problems (#228)
Microphysics 20.05
20.05
-
We now have an option for using sparse storage for aprox13 in C++
(#307) -
iso7 and aprox13 are now available as a C++ network (#303, 305)
-
species names are available as an enum in network_properties.H (#304)
-
The screening on O16+O16 in iso7 was fixed (#302)
-
The VODE integrator is now available in C++ (#299)
Microphysics 20.04.1
Microphysics 20.04.1
Microphysics 20.04
20.04
-
The wion network property was removed (#294)
-
There are new unit tests for the screening and aprox rates
modules (both C++ and Fortran interfaces). -
The screening routines were ported to C++ (#290) and the screenz
routine was removed in favor of screen5 (#293) -
a new method, is_input_valid, was added to all EOSes (both C++
and Fortran interfaces) that can be used to query whether an EOS
supports a particular input mode (e.g. eos_input_rp). (#291) -
The aprox rates used with iso7, aprox13, aprox19, and aprox21
have been converted to C++ (#288) -
We've rewritten the VODE integrator to remove all "go to"
statements (#275, 276, 278, 280, 281, 282, 283, 284, 285, 286,
287) -
We removed the ability to have nspec_evolve < nspec. This
feature was not widely used and greatly complicated the code
paths.(#279)
Microphysics 20.03
20.03
-
The nuclei information for both Fortran and C++ is now
automatically generated from a network inputs file at compile
time. As part of this change, 1/A is precomputed and stored as a
constant (#253, 258). -
The license for StarKiller Microphyscs was made explicit and
a license.txt file was added (#267) -
A framework for pure C++ EOSes has been created and a pure C++
unit test, test_eos_C, is available to test these. (#246) The
following EOSes have been ported to C++: ztwd (#268), multigamma
(#265), polytrope (#264), gamma_law (#263), helmholtz (#262),
gamma_law_general (#246), rad_power_law (#269), breakout (#270) -
The GPackage.mak files that were a remnant of the old
BoxLib F90 build system have been removed. They were
not maintained. (#212). -
All of the interface files have been collected together
in the interfaces/ dir. (#240) -
The network C++ headers have been renamed network_properties.H
and the nuclei information (aion and zion) have been
added. (#244)
Microphysics 20.02
20.02
-
Added a C++ header file, actual_network.H, that defines the
network size. This is the start of making making the
microphysics routines available in C++. -
regenerated the pynucastro networks with the latest weak rate
formulations from pynucastro.
Microphysics 20.01
20.01
-
The burn_t type no longer includes ydot or jac -- this allows
us to optimize the memory access on GPUs (#220) -
The original VODE integrator was removed and the Fortran 90
version VODE90 was renamed to VODE. (#221) -
The test_react unit tests no longer require a composition inputs
file (xin*). They now create the composition profile at runtime.
(#211)
Microphysics 19.12
-
Simplified SDC integration now uses the same retry strategy
as the default (non-SDC) integration. (#215) -
VODE90 can now participate in the retry strategy that was
previously available to the VODE integrator, where it can
switch to the BS integrator if loosening the tolerances
does not allow the burn to complete. (#201) -
The parameter ode_max_steps was made consistent in VODE and
VODE90; in some places it was being ignored. (#214) -
The helmholtz EOS was restructured, splitting the different
components into different functions and optimizing the memory
accesses. (#200) -
The derivatives with respect to mass fraction (dpdX, dedX, dhdX)
were removed from eos_t and are now available through a new type,
eos_xderivs_t and the composition_derivatives() routine. (#207) -
A bug in the screening of the C12+C12 and O16+O16 rates in iso7
was fixed. -
The test_eos unit test now outputs all of the variables in the
eos_t type.