Tensor-Like Primitives

Primitives that expose a Tensor-like interface, allowing them to be read and possibly written to using a uniform interface. More...

Modules
	Tensor Expressions
	Tensor expressions are TensorLikes that represent a composite operation on other TensorLikes.

Classes
class	sp::BlendingTensor< T, R, Options, BlendingOptions >
	A tensor that implements the "output blending" concept common in public APIs. More...
class	sp::BroadcastingTensor< T, R, Options, BroadcastOptions >
	A Tensor that implements "broadcasting". More...
class	sp::MemTensorLike< Subclass, T, TensorRank, Options >
	CRTP base class for TensorLikes directly backed by memory. More...
class	sp::NomadicTensor< TensorType, Args >
	Represents a TensorLike that can be synchronised between the CPU and various GPUs. More...
class	sp::PackedTriangularMat< Upper, T, Options >
	Represents a packed triangular matrix. More...
class	sp::StaticTensor< T, Dims >
	A Tensor that has constexpr dimensions. More...
class	sp::Tensor< T, R, Options >
	Represents a Tensor- a multidimensional array that can represent a multilinear map. More...
class	sp::TensorLike< Subclass, TensorRank >
	The interface exposed by anything that is "tensor like". More...
struct	sp::TensorProps< P, PT >
	Determines the layout and optimisation flags for a TensorLike. More...
struct	sp::MergeTensorOpts<... >
	Find the "lowest common denominator" of a set of TensorProperties. More...
class	sp::TensorView< TensorType, R >
	A view into another TensorLike. More...
class	sp::TexTensor< T, R, Options >
	A Tensor that's backed by a texture object. More...
struct	sp::IsTexTensor< typename >
	Type-trait to detect if a type is an instantiation of TexTensor. More...
class	sp::TriangularMat< Upper, T, Options >
	Represents a non-packed triangular matrix. More...

Typedefs
template<int Rank, typename TP >
using	sp::CanonTensorOpts = decltype(canonTensorOpts< Rank, TP >())
template<int Rank, PackMode P, bool PT>
using	sp::TensorProperties = CanonTensorOpts< Rank, TensorProps< P, PT > >
using	sp::MostAggressiveTensorOptions = TensorProps< PackMode::FULL, true >
	The most restrictive, but optimised Tensor options. More...
using	sp::DefaultTensorOptions = TensorProps< PackMode::NONE, false >
	The default Tensor options. More...
template<int Rank>
using	sp::DefaultAllocatedTensorOptions = TensorProperties< Rank, PackMode::PARTIAL, true >
	Default Tensor options used by tensors we allocate ourselves. Since we get to choose padding, we can improve layout. More...
template<typename... Ts>
using	sp::MergedTensorOpts = typename MergeTensorOpts< Ts... >::type
	Handy type alias for accessing MergeTensorOpts. More...
using	sp::DefaultTexTensorOptions = TensorProperties< false, PackMode::PARTIAL, true >

Enumerations
enum	sp::PackMode { sp::NONE = 0 , sp::PARTIAL = 1 , sp::FULL = 2 }
	Determines in what way a TensorLike's underlying buffer is packed. More...

Functions
template<int Rank, typename TensorProps >
constexpr auto	sp::canonTensorOpts ()

Detailed Description

Primitives that expose a Tensor-like interface, allowing them to be read and possibly written to using a uniform interface.

Examples include Tensor, TriangularMatrix, and PackedTriangularMatrix.

Typedef Documentation

DefaultAllocatedTensorOptions

template<int Rank>

using sp::DefaultAllocatedTensorOptions = typedef TensorProperties<Rank, PackMode::PARTIAL, true>

Default Tensor options used by tensors we allocate ourselves. Since we get to choose padding, we can improve layout.

DefaultTensorOptions

using sp::DefaultTensorOptions = typedef TensorProps<PackMode::NONE, false>

The default Tensor options.

MergedTensorOpts

template<typename... Ts>

using sp::MergedTensorOpts = typedef typename MergeTensorOpts<Ts...>::type

Handy type alias for accessing MergeTensorOpts.

MostAggressiveTensorOptions

using sp::MostAggressiveTensorOptions = typedef TensorProps<PackMode::FULL, true>

The most restrictive, but optimised Tensor options.

Enumeration Type Documentation

PackMode

enum sp::PackMode

Determines in what way a TensorLike's underlying buffer is packed.

Enumerator
NONE	There are gaps at every dimension.
PARTIAL	The stride at the innermost dimension is 1, but there are gaps between rows (or higher dimensions).
FULL	There are no gaps in the allocation, at all.