HCL_Vector

class HCL_Vector_d : public HCL_Base

HCL_Vector_d is the base class for all vectors in HCL

Inheritance:

Public

Inherited from HCL_Base:

Public Methods
void IncCount() const
void DecCount() const
int Count() const
virtual ostream& Write(ostream &) const

Documentation

HCL_Vector_d is the base class for all vectors in HCL. Vectors are implemented as concrete classes derived from HCL_Vector_d so that the optimization code need make no assumptions about the data structures used to implement the vectors. This is essential when nonstandard data structures (such as disk files for very large problems) are used.
Note that each class derived from HCL_Vector_d must correspond to a class derived from HCL_VectorSpace_d. That is, if one wishes to create a class MyVector, derived from HCL_Vector_d, one is obligated to first define MyVectorSpace, derived from HCL_VectorSpace_d.

Vector operations (z is the object invoking the method). virtual void Copy( const HCL_Vector_d & x ) z <-- x virtual void Neg() z <-- (-z) virtual void Mul( const double & a ) z <-- a*z virtual void Add( const HCL_Vector_d & x ) z <-- z + x virtual void Mul( const double & a, const HCL_Vector_d & x ) z <-- a*x virtual void Add( const HCL_Vector_d & x, const HCL_Vector_d & y ) z <-- x + y virtual void Sub( const HCL_Vector_d & x ) z <-- z - x virtual void Sub(const HCL_Vector_d & x, const HCL_Vector_d & y ) z <-- x - y virtual void ScaleAdd(const double & a, const HCL_Vector_d & x ) z <-- a*z + x virtual void AddScale(const double & a, const HCL_Vector_d & x ) z <-- z + a*x virtual void AddScale(const double & a, const HCL_Vector_d & x, const HCL_Vector_d & y ) z <-- x + a*y virtual double Inner( const HCL_Vector_d & x ) const Inner product of z with x. virtual double Norm() const Norm of z. virtual double Norm2() const Norm squared of z. virtual HCL_VectorSpace_d& Space() const Returns a reference to the space of which z is a member virtual void Zero() z <-- 0 virtual void Random() Returns a "random" vector. The meaning of random is left up to the implementor of each derived class.

Access to components.

Dim and Data return the dimension of the data and a pointer to the internal data array, respectively. Efficient access to the data of the vector object is usually most easily implemented using these two functions.

virtual int Dim() const

Dim returns dimension of space.

virtual double* Data()

Data returns a pointer to the array of components, which is assumed to exist or to be created for this purpose. The default implementation returns NULL so that the method need not be re-implemented for, e.g., out-of-core classes.

Array operations - ``Matlab'' methods.

In describing these,

denotes the ith component of the current array.

virtual void Fill( double a )

virtual void Add( double a )

- overload of Add with scalar argument a la Matlab

virtual void Add( const HCL_Vector_d & x, double a )

- overload of Add with scalar argument a la Matlab

virtual void DiagScale( const HCL_Vector_d & x )

- matrix multiply by diag(x)

virtual void DiagScale( const HCL_Vector_d & x, const HCL_Vector_d & y )

virtual void DiagRecipScale( const HCL_Vector_d & x, double tol=0.0)

- matrix division by diag(x). Safeguarded against overflow - if tol is less that the smallest safe divisor sfmin as defined by LAPACK::[SD]LAMCH, then tol is replaced by sfmin - thus the default is tol=0.0, which effectively means tol=sfmin.

virtual void DiagRecipScale( const HCL_Vector_d & x, const HCL_Vector_d & y, double tol=0.0)

virtual void DotStar( const HCL_Vector_d & x )