gridfire::trigger::solver::CVODE Namespace Reference

Classes
class	ConvergenceFailureTrigger
class	OffDiagonalTrigger
	Triggers when any off-diagonal Jacobian entry magnitude exceeds a threshold. More...
class	SimulationTimeTrigger
	Triggers when the current simulation time advances by at least a fixed interval. More...
class	TimestepCollapseTrigger
	Triggers when the timestep deviates from its recent average beyond a threshold. More...

Functions
std::unique_ptr< Trigger< gridfire::solver::CVODESolverStrategy::TimestepContext > >	makeEnginePartitioningTrigger (const double simulationTimeInterval, const double offDiagonalThreshold, const double timestepCollapseRatio, const size_t maxConvergenceFailures)
	Compose a trigger suitable for deciding engine re-partitioning during CVODE solves.

Function Documentation

makeEnginePartitioningTrigger()

std::unique_ptr< Trigger< gridfire::solver::CVODESolverStrategy::TimestepContext > > gridfire::trigger::solver::CVODE::makeEnginePartitioningTrigger	(	const double	simulationTimeInterval,
		const double	offDiagonalThreshold,
		const double	timestepCollapseRatio,
		const size_t	maxConvergenceFailures
	)

Compose a trigger suitable for deciding engine re-partitioning during CVODE solves.

Policy (as of implementation):

• OR of the following three conditions: 1) Every 1000th firing of SimulationTimeTrigger(simulationTimeInterval) 2) OffDiagonalTrigger(offDiagonalThreshold) 3) Every 10th firing of TimestepCollapseTrigger(timestepGrowthThreshold, timestepGrowthRelative, timestepGrowthWindowSize)

See engine_partitioning_trigger.cpp for construction details using OrTrigger and EveryNthTrigger from trigger_logical.h.

Parameters

simulationTimeInterval	Interval used by SimulationTimeTrigger (> 0).
offDiagonalThreshold	Off-diagonal Jacobian magnitude threshold (>= 0).
timestepCollapseRatio	Threshold for timestep deviation (>= 0, and <= 1 when relative).
maxConvergenceFailures	Window size for timestep averaging (>= 1 recommended).

Returns

A unique_ptr to a composed Trigger<TimestepContext> implementing the policy above.

Note

The exact policy is subject to change; this function centralizes that decision.