lib_ex_pour_HZpp/Partie_2/algebre_lineaire/sparselib++/sp1_5c/include/tmvec.h

1 line
No EOL
8.9 KiB
C
Executable file

/*+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
/* ******** *** SparseLib++ */
/* ******* ** *** *** *** v. 1.5c */
/* ***** *** ******** ******** */
/* ***** *** ******** ******** R. Pozo */
/* ** ******* *** ** *** *** K. Remington */
/* ******** ******** A. Lumsdaine */
/*+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
/* */
/* */
/* SparseLib++ : Sparse Matrix Library */
/* */
/* National Institute of Standards and Technology */
/* University of Notre Dame */
/* Authors: R. Pozo, K. Remington, A. Lumsdaine */
/* */
/* NOTICE */
/* */
/* Permission to use, copy, modify, and distribute this software and */
/* its documentation for any purpose and without fee is hereby granted */
/* provided that the above notice appear in all copies and supporting */
/* documentation. */
/* */
/* Neither the Institutions (National Institute of Standards and Technology, */
/* University of Notre Dame) nor the Authors make any representations about */
/* the suitability of this software for any purpose. This software is */
/* provided ``as is'' without expressed or implied warranty. */
/* */
/*+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++*/
//
// mv_vector_TYPE.h Basic vector class (TYPE precision)
//
#ifndef _MV_VECTOR_TYPE_H
#define _MV_VECTOR_TYPE_H
//
// Key features:
//
// o efficient indexing as fast as native C arrays
// o supports only unit strides (for efficient indexing)
// o copy-by-value semantics
// o optional "share" semantics allows vectors to be constructed as
// "views", or "references" of an existing memory, using
// MV_Vector_::ref modifier in the constuctor. (see note below.)
// o vector views can assign and references sections of vector, but
// cannot modify their *size*.
// o block-range indexing via MV_VecIndex class (e.g. A(I) = B; )
// (note for the above to work, A(I) must return a vector view.)
// o optional range checking (compile switch)
// o fast copying (A=B) via loop unrolling
// o (experimental derived FMV_Vector class) for even faster copying via
// memcpy() for data elements employing simple bit-wise copying (e.g.
// float, complex, et.c)
// o support for both [] and () style indexing ([] not available
// for matrices.)
//
// NOTES:
//
// o O(N) loops for copying and assigning scalars to vectors unroll
// loops to a depth of 4. Thus on some machines, it is faster
// to execute A=scalar, than to manually assign a native C
// array using an explicit for loop:
//
// for (i=0; i<N; d[i++] = scalar);
//
// o function code for the () and [] operators has been
// inlined into the class declaration, for compilers
// (e.g. Turbo C++ v. 3.0) that refuse to inline otherwise.
//
// o The MV_Vector(*TYPE, int len) constructor is now a deep-copy to
// match the MV_Vector(const &MV_Vector) constructor. To create a view
// (share semantics) use
//
// MV_Vector_TYPE A( &d[0], n, MV_Vector_::ref );
//
// This allows one to construct vectors as views of any contiguous C
// array. It will not destruct the memory space when the vector
// is destroyed or goes out of scope.
//
#include <iostream.h> // for formatted printing of matrices
#ifdef MV_VECTOR_BOUNDS_CHECK
# include <assert.h>
#endif
#include "mv_vecindex.h"
// this is really used as a sort of global constant. The reason
// for creating its own type is that so it can be overloaded to perform
// a deep or shallow assignement. (Any variable of type MV_Vector_::ref_type
// has only one possible value: one.)
#include "mv_vector_ref.h"
class MV_Vector_TYPE
{
protected:
TYPE *p_;
unsigned int dim_;
int ref_; // 0 or 1; does this own its own memory space?
public:
/*::::::::::::::::::::::::::*/
/* Constructors/Destructors */
/*::::::::::::::::::::::::::*/
MV_Vector_TYPE();
MV_Vector_TYPE(unsigned int);
MV_Vector_TYPE(unsigned int, const TYPE&); // can't be inlined
//because of 'for'
// statement.
MV_Vector_TYPE(TYPE*, unsigned int); // new copy
MV_Vector_TYPE(const TYPE*, unsigned int); // new copy ???
// reference of an exisiting data structure
//
MV_Vector_TYPE(TYPE*, unsigned int, MV_Vector_::ref_type i);
MV_Vector_TYPE(const MV_Vector_TYPE &);
~MV_Vector_TYPE();
/*::::::::::::::::::::::::::::::::*/
/* Indices and access operations */
/*::::::::::::::::::::::::::::::::*/
// code for operator() is defined here, otherwise some compilers
// (e.g. Turbo C++ v 3.0) cannot inline them properly...
//
TYPE& operator()(unsigned int i)
{
# ifdef MV_VECTOR_BOUNDS_CHECK
assert(i < dim_);
# endif
return p_[i];
}
const TYPE& operator()(unsigned int i) const
{
# ifdef MV_VECTOR_BOUNDS_CHECK
assert(i < dim_);
# endif
return p_[i];
}
TYPE& operator[](unsigned int i)
{
# ifdef MV_VECTOR_BOUNDS_CHECK
assert(i < dim_);
# endif
return p_[i];
}
const TYPE& operator[](unsigned int i) const
{
# ifdef MV_VECTOR_BOUNDS_CHECK
assert(i < dim_);
# endif
return p_[i];
}
MV_Vector_TYPE operator()(const MV_VecIndex &I) ;
MV_Vector_TYPE operator()(void);
const MV_Vector_TYPE operator()(void) const;
const MV_Vector_TYPE operator()(const MV_VecIndex &I) const;
//
// the following line causes ambiguatities with template instantiations
// should be avoid. Used &v(0) explicitly when converting to TYPE*.
//
// inline operator const TYPE*() const {return p_;}
inline unsigned int size() const { return dim_;}
inline int ref() const { return ref_;}
inline int null() const {return dim_== 0;}
//
// Create a new *uninitalized* vector of size N
MV_Vector_TYPE & newsize(unsigned int );
/*::::::::::::::*/
/* Assignment */
/*::::::::::::::*/
MV_Vector_TYPE & operator=(const MV_Vector_TYPE&);
MV_Vector_TYPE & operator=(const TYPE&);
friend ostream& operator<<(ostream &s, const MV_Vector_TYPE &A);
};
#include "mv_blas1_TYPE.h"
#endif