BaseAdaptiveStepController

class cubie.integrators.step_control.adaptive_step_controller.BaseAdaptiveStepController(config: AdaptiveStepControlConfig)

Bases: BaseStepController

Base class for adaptive step-size controllers.

__init__(config: AdaptiveStepControlConfig) → None

Initialise the adaptive controller.

Parameters:

config – Configuration for the controller.

property algorithm_order: int

Return the integration algorithm order assumed by the controller.

property atol: ndarray

Return absolute tolerance.

build() → ControllerCache

Construct the device function implementing the controller.

Returns:

Compiled CUDA device function for adaptive control.

Return type:

Callable

abstractmethod build_controller(precision: type[float16] | type[float32] | type[float64] | dtype[float16] | dtype[float32] | dtype[float64], clamp: Callable, min_gain: float, max_gain: float, dt_min: float, dt_max: float, n: int, atol: ndarray, rtol: ndarray, algorithm_order: int, safety: float) → ControllerCache

Create the device function for the specific controller.

Parameters:

• precision – Precision callable used to coerce values.

• clamp – Callable that limits step updates.

• min_gain – Minimum allowed gain when adapting the step size.

• max_gain – Maximum allowed gain when adapting the step size.

• dt_min – Minimum permissible step size.

• dt_max – Maximum permissible step size.

• n – Number of state variables handled by the controller.

• atol – Absolute tolerance vector.

• rtol – Relative tolerance vector.

• algorithm_order – Order of the integration algorithm.

• safety – Safety factor used when scaling the step size.

Returns:

CUDA device function implementing the controller policy.

Return type:

Callable

property deadband_max: float

Return the upper gain threshold for unity selection.

property deadband_min: float

Return the lower gain threshold for unity selection.

abstract property local_memory_elements: int

Return number of floats required for controller local memory.

property max_gain: float

Return the maximum gain factor.

property min_gain: float

Return the minimum gain factor.

property rtol: ndarray

Return relative tolerance.

property safety: float

Return the safety scaling factor.