Performs the line search algorithm based on the results of a single newton iteration step. More...

#include <LineSearch.h>

Public Member Functions
constexpr	LineSearchImplementation (TInfo info, int maxNumLineSearchSteps)
	ctor More...

template<typename TProblem , typename TX >
bool	operator() (TProblem &&problem, TX r, TX x, TX dx) const
	actual line search implementation More...

Detailed Description

Performs the line search algorithm based on the results of a single newton iteration step.

Constructor & Destructor Documentation

template<typename TInfo>

constexpr NuTo::NewtonRaphson::LineSearchImplementation< TInfo >::LineSearchImplementation	(	TInfo	info,
		int	maxNumLineSearchSteps
	)

inline

ctor

Parameters

maxNumLineSearchSteps maximal number of line search steps

template<typename TInfo>

template<typename TProblem , typename TX >

bool NuTo::NewtonRaphson::LineSearchImplementation< TInfo >::operator()	(	TProblem &&	problem,
		TX *	r,
		TX *	x,
		TX	dx
	)		const

inline

actual line search implementation

Remarks: several return arguments are used to help inlining everything. A lot of time went into actual benchmarking. Dunno why, but the a previous version that avoids the return arguments and returns various values in a std::tuple was significantly slower (10%) in a benchmark for a scalar function

Parameters

problem	class that implements NormFunction, ResidualFunction and mTolerance
r	residual, return argument, see remark
x	value of the argument x, return argument, see remark
dx	dx from the solver

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