Diffpack Documentation

---

              Search  

---

ErrorRate Class Reference

compare pairwise rates for error estimates. More...

#include <ErrorRate.h>

List of all members.

Public Methods
	ErrorRate ()
	~ErrorRate ()
void	setNameOfUnknown (const String &unknown_name)
bool	ok () const
void	add (real h, real error, Norm_type norm, int run_no)
void	add (real h, real L2_error, real energy_error, int run_no)
void	add (real h, real L1_error, real L2_error, real Linf_error, int run_no)
void	add (real h, real L1_error, real L2_error, real Linf_error, real energy_error, int run_no)
void	add (real h, real L1_error, real L2_error, real Linf_error, real energy_error, real H1_error, int run_no)
void	reset ()
void	print (Os os) const
void	writeExtendedResults (MultipleReporter &rep)

---

Detailed Description

compare pairwise rates for error estimates.

NAME: ErrorRate - compare pairwise rates for error estimates

DESCRIPTION:

The class offers some utilities for convenient computation of convergence rates. If the programmer can estimate the error in some norms (the "L1", "L2" and "Linf" norms as defined by "enum Norm_type"), the estimates can be fed into a class "ErrorRate" object. At a later stage, one can ask class "ErrorRate" to write a table of the error estimates and compute the pairwise exponent (rate) of the items in the table. Let the error in two rows be "e1" and "e2", and let the corresponding numerical discretization parameters be "h1" and "h2" (one will typically choose "h1" and "h2" as a characteristic length of a grid cell, for example, one can take this length as "1/n**d), where "n" is the number of unknowns and "d" is the number of space dimensions). Then we can use the model "e1=C*h**r", where "**" means exponentiation. Using a similar formula for relating "e2" to "n2", one can easily compute the "r" constant.

The "run_no" argument to the "add" functions is an identification of the problem that led to the current error norms. For example, in a Diffpack simulator employing multiple loops, one can simply send "getMultipleLoopIndex()" as the "run_no" argument. Otherwise one can send some kind of a loop counter (in a loop over decreasing mesh sizes for example).

---

Constructor & Destructor Documentation

ErrorRate::ErrorRate ( )

No special initialization is needed.

ErrorRate::~ErrorRate ( ) [inline]

---

Member Function Documentation

void ErrorRate::add ( real h, real L1_error, real L2_error, real Linf_error, real energy_error, real H1_error, int run_no )

See documentation of one of the overloaded functions.

void ErrorRate::add ( real h, real L1_error, real L2_error, real Linf_error, real energy_error, int run_no )

See documentation of one of the overloaded functions.

void ErrorRate::add ( real h, real L1_error, real L2_error, real Linf_error, int run_no )

See documentation of one of the overloaded functions.

void ErrorRate::add ( real h, real L2_error, real energy_error, int run_no )

See documentation of one of the overloaded functions.

void ErrorRate::add ( real h, real error, Norm_type norm, int run_no )

adds a new data pair ("h" and one or more error quantities) to the stored data sets.

bool ErrorRate::ok ( ) const

returns true of the object contains a consistent data set.

void ErrorRate::print ( Os os ) const

computes the error rate by use of "exponent" and present the results in tabular form by printing to an "Os" data stream..

void ErrorRate::reset ( )

forgets all recorded data and prepares for a new session.

void ErrorRate::setNameOfUnknown ( const String & unknown_name )

defines the name of error quantityl, used for printouts.

void ErrorRate::writeExtendedResults ( MultipleReporter & rep )

as "print", but sends the results to a "MultipleReporter".

---

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

• ErrorRate.h

---