The base class for tensors defines generic capabilities. More...

#include <basetensor.h>

Inheritance diagram for madness::BaseTensor:

Public Member Functions
	BaseTensor ()

virtual	~BaseTensor ()

long	size () const
	Returns the number of elements in the tensor. More...

long	id () const
	Returns the typeid of the tensor (c.f., `TensorTypeData<T>` ) More...

long	ndim () const
	Returns the number of dimensions in the tensor. More...

long	dim (int i) const
	Returns the size of dmension `i`. More...

long	stride (int i) const
	Returns the stride associated with dimension `i`. More...

const long *	dims () const
	Returns the array of tensor dimensions. More...

const long *	strides () const
	Returns the array of tensor strides. More...

bool	conforms (const BaseTensor *t) const
	Returns true if this and *t are the same shape and size. More...

bool	iscontiguous () const
	Returns true if the tensor refers to contiguous memory locations. More...

Static Public Member Functions
static int	get_instance_count ()
	Returns the count of all current instances of tensors & slice tensors of all types. More...

Protected Member Functions
void	set_dims_and_size (long nd, const long d[])

void	reshape_inplace (const std::vector< long > &d)
	Reshapes the tensor inplace. More...

void	reshape_inplace (int ndimnew, const long *d)
	Reshapes the tensor inplace. More...

void	flat_inplace ()
	Reshapes the tensor inplace into 1D. More...

void	splitdim_inplace (long i, long dimi0, long dimi1)
	Splits dimension `i`. More...

void	fusedim_inplace (long i)
	Fuses dimensions `i` and `i+1`. More...

void	swapdim_inplace (long i, long j)
	Swaps the dimensions. More...

void	cycledim_inplace (long shift, long start, long end)
	Cyclic shift of dimensions. More...

void	mapdim_inplace (const std::vector< long > &map)
	General permutation of dimensions. More...

Protected Attributes
long	_size
	Number of elements in the tensor. More...

long	_ndim
	Number of dimensions (-1=invalid; 0=scalar; >0=tensor) More...

long	_id
	Id from TensorTypeData<T> in type_data.h. More...

long	_dim [TENSOR_MAXDIM]
	Size of each dimension. More...

long	_stride [TENSOR_MAXDIM]
	Increment between elements in each dimension. More...

Detailed Description

The base class for tensors defines generic capabilities.

The base class manages the size, dimension and stride information, and provides operations to manipulate them.

It also provides methods for type-safe operation on tensors using just the base class pointers. This interface is primarily useful only to the interface to Python, since Python is largely neutral to (and ignorant of) the type. These are still being re-implemented after the big clean up.

Since the base tensor class is virtual, you cannot have an instance of it. Thus, in addition to methods that return information or perform checks, there are two types of base tensor operations.

Inplace operations change *this , and return void .
Operations that must return a new tensor return a pointer to a tensor allocated with new on the heap. The caller is responsible for eventually freeing the memory using delete .

Constructor & Destructor Documentation

madness::BaseTensor::BaseTensor ( )

inline

virtual madness::BaseTensor::~BaseTensor ( )

inlinevirtual

Member Function Documentation

bool madness::BaseTensor::conforms ( const BaseTensor * t ) const

inline

Returns true if this and *t are the same shape and size.

References _dim, and _ndim.

void madness::BaseTensor::cycledim_inplace	(	long	shift,
		long	start,
		long	end
	)

protected

Cyclic shift of dimensions.

Cyclic shift by nshift places of the inclusive range of dimensions [start,....,end].

References _dim, _ndim, _stride, TENSOR_ASSERT, and TENSOR_MAXDIM.

long madness::BaseTensor::dim ( int i ) const

inline

Returns the size of dmension i.

const long* madness::BaseTensor::dims ( ) const

inline

Returns the array of tensor dimensions.

References _dim.

void madness::BaseTensor::flat_inplace ( )

protected

Reshapes the tensor inplace into 1D.

Reshape the current tensor to be the same size and 1-d. It must be contiguous.

References _size, iscontiguous(), set_dims_and_size(), and TENSOR_ASSERT.

void madness::BaseTensor::fusedim_inplace ( long i )

protected

Fuses dimensions i and i+1.

Fuse the contiguous dimensions i and i+1.

References _dim, _ndim, _stride, and TENSOR_ASSERT.

static int madness::BaseTensor::get_instance_count ( )

inlinestatic

Returns the count of all current instances of tensors & slice tensors of all types.

Referenced by main().

long madness::BaseTensor::id ( ) const

inline

Returns the typeid of the tensor (c.f., TensorTypeData<T> )

References _id.

bool madness::BaseTensor::iscontiguous ( ) const

inline

Returns true if the tensor refers to contiguous memory locations.

Referenced by flat_inplace(), and reshape_inplace().

void madness::BaseTensor::mapdim_inplace ( const std::vector< long > & map )

protected

General permutation of dimensions.

General permuation of the dimensions.

References _dim, _ndim, _stride, TENSOR_ASSERT, and TENSOR_MAXDIM.

long madness::BaseTensor::ndim ( ) const

inline

Returns the number of dimensions in the tensor.

References _ndim.

Referenced by madness::SeparatedConvolution< T, NDIM >::apply2(), madness::SeparatedConvolution< T, NDIM >::apply2_lowdim(), madness::FunctionNode< T, NDIM >::coeff(), and madness::SeparatedConvolution< T, NDIM >::estimate_costs().

void madness::BaseTensor::reshape_inplace ( const std::vector< long > & d )

protected

Reshapes the tensor inplace.

Reshape the size and number of dimensions.

Modifies the current tensor to have the number and size of dimensions as described in the vector d . The total number of elements must be the same before and after, and the current tensor must be contiguous.

void madness::BaseTensor::reshape_inplace	(	int	nd,
		const long *	d
	)

protected

Reshapes the tensor inplace.

Reshape the size and number of dimensions.

Modifies the current tensor to have the number and size of dimensions as described in the vector d . The total number of elements must be the same before and after, and the current tensor must be contiguous.

References _size, iscontiguous(), set_dims_and_size(), and TENSOR_ASSERT.