Propagator base class

Propagator base class.

class struphy.propagators.base.Propagator

Bases: object

Base class for propagators used in StruphyModels.

Note

All Struphy propagators are subclasses of Propagator and must be added under struphy/propagators/.

class Variables

Bases: object

Define variable names and types to be updated by the propagator.

property var1

class Options(opt1: Literal['implicit', 'explicit'] = 'implicit')

Bases: OptionsBase

Template for configuration options of a propagator.

Subclasses should override this to define specific propagator options.

OptsTemplate

alias of Literal[‘implicit’, ‘explicit’]

opt1: Literal['implicit', 'explicit'] = 'implicit'

abstract property options: Options

abstract allocate()

Allocate all data/objects of the instance.

prepare_time_step(dt: float)

Set up the data that depends on the time step (e.g. preconditioners) for dt, before the first call with dt, so that it is part of the setup rather than of the first time step (see run()). Optional; does nothing by default.

Parameters:

dt (float) – Time step size of the first call.

show_options()

Print the options of the propagator.

update_feec_variables(**new_coeffs)

Return max_diff = max(abs(new - old)) for each new_coeffs, update feec coefficients and update ghost regions.

Returns:

diffs – max_diff for all feec variables.

Return type:

dict

kernels() → tuple[Kernel, ...]

The Kernel objects this propagator calls in a time step.

Used by compile_cuda_kernels() to compile the CUDA kernels before the time stepping. The default collects them from the attributes of the propagator (after allocate()): kernels kept as attributes, and the kernels of attributes with a kernels() method (Pusher, KernelSetup, Accumulator, AccumulatorVector), also inside lists, tuples and dicts. A propagator that calls a kernel it does not keep (e.g. a module-level kernel called directly in __call__()) overrides this method and adds it.

property init_kernels: tuple[KernelSetup, ...]

Tuple of initialization kernel setups for evaluation at \(\boldsymbol \eta^n\) in an iterative Pusher.

property eval_kernels: tuple[KernelSetup, ...]

Tuple of evaluation kernel setups for evaluation at \(\alpha_i \eta_{i}^{n+1,k} + (1 - \alpha_i) \eta_{i}^n\) for \(i=1, 2, 3\) and different \(\alpha_i \in [0,1]\), in an iterative Pusher.

property rank

MPI rank, is 0 if no communicator.

property derham: Derham

Derham spaces and projectors.

property domain: Domain

Domain object that characterizes the mapping from the logical to the physical domain.

property mass_ops: WeightedMassOperators

Weighted mass operators.

property basis_ops: BasisProjectionOperators

Basis projection operators.

property projected_equil: ProjectedFluidEquilibriumWithB

Fluid equilibrium projected on 3d Derham sequence with commuting projectors.

property time_state

A pointer to the time variable of the dynamics (‘t’).

add_time_state(time_state)

Add a pointer to the time variable of the dynamics (‘t’).

Parameters:

time_state (ndarray) – Of size 1, holds the current physical time ‘t’.

add_init_kernel(setup: KernelSetup)

Register an evaluation at the start of each push.

The setup names the callable, its arguments, and the exact output marker columns. Its evaluation weights must all be zero.

add_eval_kernel(setup: KernelSetup)

Register an evaluation before each pusher stage/iteration.

The setup’s alpha weights determine both the evaluation state and the preceding MPI sort.