time_integrator.h File Reference

Time integrator class declarations. More...

#include <functional>
#include <string>
#include <unordered_map>
#include <vector>
#include <Kokkos_Core.hpp>
#include "common.h"
#include "state.h"

Go to the source code of this file.

Classes
class	TimeIntegrator
class	FE
class	RK4
class	SSPRK3
	Three-stage, third-order strong-stability-preserving Runge-Kutta (Shu & Osher 1988). More...

Typedefs
using	RHSFunction = std::function< void(State solution, State rhs, rtype t)>

Enumerations
enum class	TimeIntegratorType { FE , RK4 , SSPRK3 }
enum	DtScale {   DT = 0 , DT_HALF = 1 , DT_SIXTH = 2 , DT_THIRD = 3 ,   N_DT_SCALES = 4 }
	Index of each multiple of dt in the device array of take_step's dt_scales. More...

Functions
void	axpby (const rtype a, const State &x, const rtype b, const State &y, const rtype *a_device=nullptr)
	y = a * x + b * y, on both blocks of the state.
KOKKOS_INLINE_FUNCTION void	dt_scales (const rtype dt, rtype *scales)
	The multiples of dt that the integrators use, as take_step's device dt_scales holds them.
void	lerp (const rtype a, const State &x, const State &y)
	y = (1 - a) y + a x as y + a (x - y), on both blocks of the state: unlike axpby with rounded weights a and 1 - a, it leaves y unchanged where x == y.

Detailed Description

Time integrator class declarations.

Author

Matthew Bonanni(mbona.nosp@m.nni0.nosp@m.01@gm.nosp@m.ail..nosp@m.com)

Version

0.2

Date

2023-12-22

Copyright

Copyright (c) 2023 Matthew Bonanni

Typedef Documentation

RHSFunction

using RHSFunction = std::function<void(State solution, State rhs, rtype t)>

Enumeration Type Documentation

DtScale

enum DtScale

Index of each multiple of dt in the device array of take_step's dt_scales.

Enumerator
DT
DT_HALF
DT_SIXTH
DT_THIRD
N_DT_SCALES

TimeIntegratorType

enum class TimeIntegratorType

strong

Enumerator
FE
RK4
SSPRK3

Function Documentation

axpby()

void axpby	(	const rtype	a,
		const State &	x,
		const rtype	b,
		const State &	y,
		const rtype *	a_device = nullptr
	)

y = a * x + b * y, on both blocks of the state.

y = a x + b y, with a read from device memory instead when a_device is given (graph-captured steps, whose dt lives on the device).

dt_scales()

KOKKOS_INLINE_FUNCTION void dt_scales	(	const rtype	dt,
		rtype *	scales
	)

The multiples of dt that the integrators use, as take_step's device dt_scales holds them.

lerp()

void lerp	(	const rtype	a,
		const State &	x,
		const State &	y
	)

y = (1 - a) y + a x as y + a (x - y), on both blocks of the state: unlike axpby with rounded weights a and 1 - a, it leaves y unchanged where x == y.