tannic::function::Function< Functor, Argument > Class Template Reference

Expression template for mathematical function operations. More...

#include <functions.hpp>

Public Member Functions
constexpr	Function (Functor functor, typename Trait< Argument >::Reference argument)
	Constructs a Function expression.
constexpr type	dtype () const
	Returns the data type of the result.
constexpr Shape const &	shape () const
	Returns the shape of the result.
constexpr Strides const &	strides () const
	Returns the strides of the result.
auto	offset () const
	Returns the offset of the result.
Tensor	forward () const

Public Attributes
Functor	functor
Trait< Argument >::Reference	argument

Detailed Description

template<Functional Functor, Expression Argument>
class tannic::function::Function< Functor, Argument >

Expression template for mathematical function operations.

Represents a lazily evaluated unary function that takes a functor as argument and maps it to an argument. The actual computation is deferred until forward() is called.

Template Parameters

Functor	A mathematical functor satisfying the Functional concept
Argument	An expression type satisfying the Expression concept

Constructor & Destructor Documentation

Function()

template<Functional Functor, Expression Argument>

constexpr tannic::function::Function< Functor, Argument >::Function ( Functor  functor, typename Trait< Argument >::Reference  argument  )

inlineconstexpr

Constructs a Function expression.

Parameters

functor	The functor (e.g., Log, Exp)
argument	The input tensor expression

Member Function Documentation

dtype()

template<Functional Functor, Expression Argument>

constexpr type tannic::function::Function< Functor, Argument >::dtype ( ) const

inlineconstexpr

Returns the data type of the result.

Returns

The data type of the underlying argument.

Functors in this expression template system are type-preserving — applying them element-wise does not change the scalar type (e.g., applying exp() to a float32 tensor produces float32 output). Therefore, we can directly query and return the argument's dtype.

forward()

template<Functional Functor, Expression Argument>

Tensor tannic::function::Function< Functor, Argument >::forward ( ) const

inline

offset()

template<Functional Functor, Expression Argument>

auto tannic::function::Function< Functor, Argument >::offset ( ) const

inline

Returns the offset of the result.

Returns

The offset (in elements) into the underlying tensor storage.

Since element-wise functors do not change the memory location of the data, the offset is taken directly from the argument.

shape()

template<Functional Functor, Expression Argument>

constexpr Shape const & tannic::function::Function< Functor, Argument >::shape ( ) const

inlineconstexpr

Returns the shape of the result.

Returns

A const reference to the shape of the underlying argument.

The shape is returned as const& to avoid copying the shape object, and because element-wise functors do not alter tensor dimensions.

strides()

template<Functional Functor, Expression Argument>

constexpr Strides const & tannic::function::Function< Functor, Argument >::strides ( ) const

inlineconstexpr

Returns the strides of the result.

Returns

A const reference to the strides of the underlying argument.

Strides describe memory layout, and for pure element-wise operations they remain identical to those of the input tensor. Returning const& avoids copying and preserves the original stride information.

Member Data Documentation

argument

template<Functional Functor, Expression Argument>

Trait<Argument>::Reference tannic::function::Function< Functor, Argument >::argument

functor

template<Functional Functor, Expression Argument>

Functor tannic::function::Function< Functor, Argument >::functor

---

The documentation for this class was generated from the following file:

• include/tannic/functions.hpp