A collection of contol data for molecular mechanics simulations, conveying the current step number, progress counters through various work units, the time step, RATTLE tolerance, and other critical parameters to guide calculations. This common struct accommodates both molecular dynamics and molecular mechanics energy minimizations, and can be created from control objects derived from &dynamics or &minimize namelists. More...

#include <mm_controls.h>

Public Member Functions
	MolecularMechanicsControls (const MolecularMechanicsControls &original)
	The copy constructor handles assignment of internal POINTER-kind Hybrid objects.

	MolecularMechanicsControls (MolecularMechanicsControls &&original)
	The move constructor prepared the original object for destruction.

MolecularMechanicsControls &	operator= (const MolecularMechanicsControls &other)
	The copy assignment operator likewise handles assignment of internal POINTER-kind Hybrid objects.

MolecularMechanicsControls &	operator= (MolecularMechanicsControls &&other)
	The move assignment operator looks much like the copy assignment operator.

int	getStepNumber () const
	Get the current step number.

int	getSteepestDescentCycles () const
	Get the number of steepest descent minimization cycles.

int	getTotalCycles () const
	Get the total number of minimization steps, or total MD cycles.

int	getNTWarpMultiplicity () const
	Get the neutral-territory warp multiplicity.

double	getInitialMinimizationStep () const
	Get the initial step for energy minimization.

double	getElectrostaticCutoff () const
	Get the cutoff for non-bonded electrostatic interactions in periodic simulations.

double	getVanDerWaalsCutoff () const
	Get the cutoff for non-bonded van-der Waals interactions in periodic simulations.

int	getValenceWorkUnitProgress (int counter_index, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the value of one of the valence work unit progress counters on the host or the HPC device.

int	getNonbondedWorkUnitProgress (int counter_index, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the value of one of the non-bonded work unit progress counters on the host or the HPC device.

int	getPmeWorkUnitProgress (int counter_index, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the value of one of the Particle-Mesh Ewald work unit progress counters on the host or the HPC device.

int	getReductionWorkUnitProgress (int counter_index, ReductionStage process, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the value of one of the reduction work unit progress counters on the host or the HPC device.

MMControlKit< double >	dpData (HybridTargetLevel tier=HybridTargetLevel::HOST)
	Obtain a double-precision abstract for this object.

MMControlKit< float >	spData (HybridTargetLevel tier=HybridTargetLevel::HOST)
	Obtain a single-precision abstract for this object.

void	incrementStep ()
	Increment the step counter, moving the controls to a different progress counter.


	MolecularMechanicsControls (double initial_step_in=default_minimize_dx0, int sd_cycles_in=default_minimize_ncyc, int max_cycles_in=default_minimize_maxcyc, int nt_warp_multiplicity_in=default_nt_warp_multiplicity, double electrostatic_cutoff_in=default_electrostatic_cutoff, double van_der_waals_cutoff_in=default_van_der_waals_cutoff)
	The constructor can create an empty object with default parameters for the time step and rattle tolerance, accept user-specified values for critical constants, or be built from a namelist-derived object on the CPU. The reason that those namelist-derived control objects don't just get pushed by value to the GPU is that the work unit counters need special arrays, which this object manages.

	MolecularMechanicsControls (const DynamicsControls &user_input)

	MolecularMechanicsControls (const MinimizeControls &user_input)


void	primeWorkUnitCounters (const CoreKlManager &launcher, EvaluateForce eval_frc, EvaluateEnergy eval_nrg, ClashResponse softcore, VwuGoal purpose, PrecisionModel valence_prec, PrecisionModel nonbond_prec, QMapMethod qspread_approach, PrecisionModel acc_prec, size_t image_coord_type, int qspread_order, NeighborListKind nbgr_config, TinyBoxPresence has_tiny_box, const AtomGraphSynthesis &poly_ag)
	Prime the work unit counters based on a particular GPU configuration.

void	primeWorkUnitCounters (const CoreKlManager &launcher, EvaluateForce eval_frc, EvaluateEnergy eval_nrg, const ClashResponse softcore, VwuGoal purpose, PrecisionModel valence_prec, PrecisionModel nonbond_prec, const AtomGraphSynthesis &poly_ag)

void	primeWorkUnitCounters (const CoreKlManager &launcher, EvaluateForce eval_frc, EvaluateEnergy eval_nrg, VwuGoal purpose, PrecisionModel valence_prec, PrecisionModel nonbond_prec, const AtomGraphSynthesis &poly_ag)

void	primeWorkUnitCounters (const CoreKlManager &launcher, EvaluateForce eval_frc, EvaluateEnergy eval_nrg, VwuGoal purpose, PrecisionModel general_prec, const AtomGraphSynthesis &poly_ag)


template<typename T, typename Tacc, typename Tcalc, typename T4>
void	setNTWarpMultiplicity (const CellGrid< T, Tacc, Tcalc, T4 > cg_a, const CellGrid< T, Tacc, Tcalc, T4 > cg_b, const GpuDetails &gpu)
	Set the neutral-territory warp multiplicity based on one or two gell grids, for a given GPU.

template<typename T, typename Tacc, typename Tcalc, typename T4>
void	setNTWarpMultiplicity (const CellGrid< T, Tacc, Tcalc, T4 > *cg_a, const GpuDetails &gpu)

void	setNTWarpMultiplicity (int mult_in)

Detailed Description

A collection of contol data for molecular mechanics simulations, conveying the current step number, progress counters through various work units, the time step, RATTLE tolerance, and other critical parameters to guide calculations. This common struct accommodates both molecular dynamics and molecular mechanics energy minimizations, and can be created from control objects derived from &dynamics or &minimize namelists.

Constructor & Destructor Documentation

◆ MolecularMechanicsControls() [1/3]

stormm::mm::MolecularMechanicsControls::MolecularMechanicsControls	(	double	initial_step_in = default_minimize_dx0,
		int	sd_cycles_in = default_minimize_ncyc,
		int	max_cycles_in = default_minimize_maxcyc,
		int	nt_warp_multiplicity_in = default_nt_warp_multiplicity,
		double	electrostatic_cutoff_in = default_electrostatic_cutoff,
		double	van_der_waals_cutoff_in = default_van_der_waals_cutoff )

The constructor can create an empty object with default parameters for the time step and rattle tolerance, accept user-specified values for critical constants, or be built from a namelist-derived object on the CPU. The reason that those namelist-derived control objects don't just get pushed by value to the GPU is that the work unit counters need special arrays, which this object manages.

Parameters

initial_step_in	The desired initial step (for energy minimization)
user_input	Namelist-derived molecular dynamics or energy minimization controls

◆ MolecularMechanicsControls() [2/3]

stormm::mm::MolecularMechanicsControls::MolecularMechanicsControls ( const MolecularMechanicsControls & original )

The copy constructor handles assignment of internal POINTER-kind Hybrid objects.

Parameters

original The object to copy

◆ MolecularMechanicsControls() [3/3]

stormm::mm::MolecularMechanicsControls::MolecularMechanicsControls ( MolecularMechanicsControls && original )

The move constructor prepared the original object for destruction.

Parameters

original The object from which to take contents

Member Function Documentation

◆ getNonbondedWorkUnitProgress()

int stormm::mm::MolecularMechanicsControls::getNonbondedWorkUnitProgress	(	int	counter_index,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Get the value of one of the non-bonded work unit progress counters on the host or the HPC device.

Parameters

counter_index	Index of the work unit counter to retrieve
tier	Get the result from the CPU host or the HPC device

◆ getPmeWorkUnitProgress()

int stormm::mm::MolecularMechanicsControls::getPmeWorkUnitProgress	(	int	counter_index,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Get the value of one of the Particle-Mesh Ewald work unit progress counters on the host or the HPC device.

Parameters

counter_index	Index of the work unit counter to retrieve
tier	Get the result from the CPU host or the HPC device

◆ getReductionWorkUnitProgress()

int stormm::mm::MolecularMechanicsControls::getReductionWorkUnitProgress	(	int	counter_index,
		ReductionStage	process,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Get the value of one of the reduction work unit progress counters on the host or the HPC device.

Parameters

counter_index	Index of the work unit counter to retrieve
tier	Get the result from the CPU host or the HPC device

◆ getValenceWorkUnitProgress()

int stormm::mm::MolecularMechanicsControls::getValenceWorkUnitProgress	(	int	counter_index,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Get the value of one of the valence work unit progress counters on the host or the HPC device.

Parameters

counter_index	Index of the work unit counter to retrieve
tier	Get the result from the CPU host or the HPC device

◆ operator=() [1/2]

MolecularMechanicsControls & stormm::mm::MolecularMechanicsControls::operator= ( const MolecularMechanicsControls & other )

The copy assignment operator likewise handles assignment of internal POINTER-kind Hybrid objects.

Parameters

other Another way to say original, in a different semantic context

◆ operator=() [2/2]

MolecularMechanicsControls & stormm::mm::MolecularMechanicsControls::operator= ( MolecularMechanicsControls && other )

The move assignment operator looks much like the copy assignment operator.

Parameters

other The object from which to take contents

◆ primeWorkUnitCounters()

void stormm::mm::MolecularMechanicsControls::primeWorkUnitCounters	(	const CoreKlManager &	launcher,
		EvaluateForce	eval_frc,
		EvaluateEnergy	eval_nrg,
		ClashResponse	softcore,
		VwuGoal	purpose,
		PrecisionModel	valence_prec,
		PrecisionModel	nonbond_prec,
		QMapMethod	qspread_approach,
		PrecisionModel	acc_prec,
		size_t	image_coord_type,
		int	qspread_order,
		NeighborListKind	nbgr_config,
		TinyBoxPresence	has_tiny_box,
		const AtomGraphSynthesis &	poly_ag )

Prime the work unit counters based on a particular GPU configuration.

Overloaded:

Provide information about the PME approach
Make provisions for clashes to be forgiven (or not) with softcore potential functions
Provide only basic information for a vacuum-phase MM calculation

Parameters

launcher	Object containing launch parameters for all kernels
eval_frc	Indicate whether forces are to be evaluated–some kernels allocate different numbers of blocks in the launch grid if forces are not required.
eval_nrg	Indicate whether energy is to be evaluated–some kernels allocate different numbers of blocks in the launch grid if the energy is not required.
softcore	Indicate whether the control counters should take clash handling kernel variants into account
valence_prec	Precision model for valence calculations
nonbond_prec	Precision model for non-bonded calculations
general_prec	Precision model for all calculations
qspread_approach	The approach taken by GPU kernels to spread charge density onto the particle-mesh interaction grids
acc_prec	Precision model for accumulations (needed in case the density mapping approach is accumulation in the shared memory space)
image_coord_type	Data type used to represent coordinates in the neighbor list images
qspread_order	The order of B-spline interpolation used to spread charge density to the particle-mesh interaction grids
poly_ag	Compilation of topologies describing the workload (used here for general descriptors such as the non-bonded work unit type)

◆ setNTWarpMultiplicity()

template<typename T, typename Tacc, typename Tcalc, typename T4>

void stormm::mm::MolecularMechanicsControls::setNTWarpMultiplicity	(	const CellGrid< T, Tacc, Tcalc, T4 > *	cg_a,
		const CellGrid< T, Tacc, Tcalc, T4 > *	cg_b,
		const GpuDetails &	gpu )

Set the neutral-territory warp multiplicity based on one or two gell grids, for a given GPU.

Overloaded:

Accept one neighbor list and specifications the GPU that will perform the calculations
Accept two neighbor lists and the specifications of the GPU

Parameters

cg_a	The first cell grid neighbor list
cg_b	The second (optional) cell grid neighbor list
gpu	Details of the GPU that will evaluate pairwise interactions
mult_in	The multiplicity to explicitly set

The documentation for this class was generated from the following files:

src/MolecularMechanics/mm_controls.h
src/MolecularMechanics/mm_controls.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ MolecularMechanicsControls() [1/3]

◆ MolecularMechanicsControls() [2/3]

◆ MolecularMechanicsControls() [3/3]

Member Function Documentation

◆ getNonbondedWorkUnitProgress()

◆ getPmeWorkUnitProgress()

◆ getReductionWorkUnitProgress()

◆ getValenceWorkUnitProgress()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ primeWorkUnitCounters()

◆ setNTWarpMultiplicity()