TBCI Namespace Reference

The TBCI Numerical Library is put into TBCI namespace, if the compiler supports it (egcs-1.1.x and later for example) More...

Classes
class	vec_fill_fn

Typedefs
typedef enum _vararg	vararg

Enumerations
enum	_vararg {   vag0 =0, vag1, vag2, vag3,   vag4, vag5, vag6, vag7,   vag8, vag9, vag10, vag11,   vag12, vag13, vag14, vag15,   vag16, vag_max = 4294967295U }
	This is a helper type to identify and count varargs. More...

Functions
template<typename T >
void	SWAP (T &a, T &b)
	SWAP function Note: We could implement a swap function without temporaries: a -= b b += a a -= b a = -a That's four arithmetic operations to save one temporary and it can be only done for integers. More...
template<typename T >
void	_tbci_copy (const unsigned long sz, T *const res, const T *const v1)
template<typename T >
void	_par_copy (const unsigned long sz, T *v1, const T *v2)
template<typename T >
void	_tbci_fill (const unsigned long sz, T *const res, register typename tbci_traits< T >::loop_const_refval_type f2)
template<typename T >
void	_par_fill (const unsigned long, T *const, typename tbci_traits< T >::loop_const_refval_type)
template<typename T >
void	_tbci_fill_fn (const unsigned long sz, T *vec, vec_fill_fn< T > fn, void *par)
template<typename T >
void	do_vv_comp (const unsigned long sz, const T *const v1, const T *const v2, volatile long &_f2)
	f2 = number of differences vec, vec More...
template<typename T >
int	_par_comp (const unsigned long sz, const T *v1, const T *v2)
template<typename T >
unsigned long	_bin_search (const T *vec, T el, unsigned long start, unsigned long end)
	Search for an element el in a sorted vector between start and end-1, returns (unsigned long)-1 if element is not found. More...
template<typename T >
unsigned long	bin_search (const T *vec, T el, unsigned long start, unsigned long end)
	Search for an element el in a sorted vector between start and end-1, returns (unsigned long)-1 if element is not found. More...
template<typename T >
int	sign (const T &x)
	Signum. More...
double	fabssqr (const double a)
double	fabssqr (const float a)
double	fabssqr (const int a)
double	fabssqr (const unsigned a)
template<typename T >
T	sqr (const T &a)
template<typename T >
T	dot (const T &a1, const T &a2)

Detailed Description

The TBCI Numerical Library is put into TBCI namespace, if the compiler supports it (egcs-1.1.x and later for example)

Typedef Documentation

typedef enum _vararg TBCI::vararg

Definition at line 1276 of file basics.h.

Enumeration Type Documentation

enum TBCI::_vararg

This is a helper type to identify and count varargs.

Enumerator
vag0
vag1
vag2
vag3
vag4
vag5
vag6
vag7
vag8
vag9
vag10
vag11
vag12
vag13
vag14
vag15
vag16
vag_max

Definition at line 1272 of file basics.h.

Function Documentation

template<typename T >

unsigned long TBCI::_bin_search	(	const T *	vec,
		T	el,
		unsigned long	start,
		unsigned long	end
	)

Search for an element el in a sorted vector between start and end-1, returns (unsigned long)-1 if element is not found.

Bound checking (el in range) needs to be done before calling this, also need to make sure that the last element does not match.

Definition at line 992 of file basics.h.

Referenced by CSCMatrix< T >::get().

template<typename T >

int TBCI::_par_comp	(	const unsigned long	sz,
		const T *	v1,
		const T *	v2
	)

template<typename T >

void TBCI::_par_copy	(	const unsigned long	sz,
		T *	v1,
		const T *	v2
	)

template<typename T >

void TBCI::_par_fill	(	const unsigned	long,
		T *	const,
		typename tbci_traits< T >::loop_const_refval_type
	)

template<typename T >

void TBCI::_tbci_copy ( const unsigned long  sz, T *const  res, const T *const  v1  )

inline

Definition at line 891 of file basics.h.

Referenced by HOTDECL(), and job_vv_copy().

template<typename T >

void TBCI::_tbci_fill	(	const unsigned long	sz,
		T *const	res,
		register typename tbci_traits< T >::loop_const_refval_type	f2
	)

Definition at line 907 of file basics.h.

template<typename T >

void TBCI::_tbci_fill_fn	(	const unsigned long	sz,
		T *	vec,
		vec_fill_fn< T >	fn,
		void *	par
	)

Definition at line 928 of file basics.h.

References TBCI::vec_fill_fn< T >::fn, i, and REGISTER.

template<typename T >

unsigned long TBCI::bin_search	(	const T *	vec,
		T	el,
		unsigned long	start,
		unsigned long	end
	)

Search for an element el in a sorted vector between start and end-1, returns (unsigned long)-1 if element is not found.

Does all the necessary sanity checking.

Definition at line 1011 of file basics.h.

template<typename T >

void TBCI::do_vv_comp	(	const unsigned long	sz,
		const T *const	v1,
		const T *const	v2,
		volatile long &	_f2
	)

f2 = number of differences vec, vec

Definition at line 975 of file basics.h.

Referenced by job_vv_comp().

template<typename T >

T TBCI::dot ( const T &  a1, const T &  a2  )

inline

Definition at line 1183 of file basics.h.

References std::conj(), and CPLX__.

Referenced by BiCGSTAB(), CG(), CGS(), dot(), and FS_Vector< dims, T >::fabssqr().

double TBCI::fabssqr ( const double  a )

inline

Speaking of abs(), fabs() and fabssqr() : abs (T) returns type T, fabs() returns double and fabssqr() also a double (the sqr of fabs)

We want to be able to write templated code that can accept both std:: and TBCI:: types. Now, if such a routine uses functions like fabs() or abs(), we are in trouble: What namespace should the template author use for those? std:: or TBCI:: ?

Options are: (1) importing the pre-existing functions from std:: into TBCI:: (2) export the TBCI::functions to namespace std::

The answer may be different for each of those three types, as fabssqr() does not exist at all in std::, fabs() does exist, but only for doubles and abs() does exist for a number of builtin types as well as (std::complex)

In any case, the functions need to be completed, as no fabs() exists for non-double types, e.g. Additional complication is added by the fact that the whole thing should still work if compiled with NO_NS.

Solution (2) is the prefered one, as it does make users less dependent on TBCI. Unfortunately, this means that we may break backwards compatibility (on source level).

Note that the self-invented functions conj(), real(), imag() and fabssqr() remain in namespace TBCI. The same holds for sqr() and dot().

Definition at line 1157 of file basics.h.

Referenced by FS_Vector< dims, T >::abs(), BiCGSTAB(), fabs(), FS_Vector< dims, T >::fabs(), F_TSMatrix< T >::fabs(), F_Matrix< T >::fabs(), TSMatrix< T >::fabssqr(), operator/(), and QMR().

double TBCI::fabssqr ( const float  a )

inline

Definition at line 1158 of file basics.h.

double TBCI::fabssqr ( const int  a )

inline

Definition at line 1159 of file basics.h.

double TBCI::fabssqr ( const unsigned  a )

inline

Definition at line 1160 of file basics.h.

template<typename T >

int TBCI::sign ( const T &  x )

inline

Signum.

Definition at line 1105 of file basics.h.

References CPLX__, LIKELY, and real.

template<typename T >

T TBCI::sqr ( const T &  a )

inline

Definition at line 1174 of file basics.h.

References a.

template<typename T >

void TBCI::SWAP ( T &  a, T &  b  )

inline

SWAP function Note: We could implement a swap function without temporaries: a -= b b += a a -= b a = -a That's four arithmetic operations to save one temporary and it can be only done for integers.

Unlikely that it's worth it.

Definition at line 813 of file basics.h.

References a, ALIGN, b, MIN_ALIGN, REGISTER, and T.

Referenced by tbci_memalloc_cache< T >::_init(), gaussj(), F_TMatrix< T >::herm(), BVector< T >::swap(), F_BandMatrix< T >::swap(), CRMatrix< T >::swap(), CSCMatrix< T >::swap(), F_TMatrix< T >::swap(), TMatrix< T >::swap(), F_TMatrix< T >::trans(), CTensor< T >::transpose(), F_TMatrix< T >::transpose(), and TMatrix< T >::transpose().