PermutedViewConstExpr< Tensor, Perm > Class Template Reference

Compile-time permuted view over a tensor. More...

#include <permuted_view_constexpr.h>

Inheritance diagram for PermutedViewConstExpr< Tensor, Perm >:

Collaboration diagram for PermutedViewConstExpr< Tensor, Perm >:

Public Types
using	value_type = typename Tensor::value_type
using	PadPolicy = typename Tensor::AccessPolicy::PadPolicy
using	Layout = StridedLayoutConstExpr< PadPolicy, Perm... >

Public Member Functions
	PermutedViewConstExpr (const Tensor &t) noexcept
	PermutedViewConstExpr (const PermutedViewConstExpr &)=delete
PermutedViewConstExpr &	operator= (const PermutedViewConstExpr &)=delete
	PermutedViewConstExpr (PermutedViewConstExpr &&)=delete
PermutedViewConstExpr &	operator= (PermutedViewConstExpr &&)=delete
template<typename Output >
bool	may_alias (const Output &output) const noexcept
template<typename... Indices> requires (sizeof...(Indices) == NumDims)
FORCE_INLINE const value_type &	operator() (Indices... indices) const TESSERACT_CONDITIONAL_NOEXCEPT
FORCE_INLINE const value_type &	operator() (my_size_t(&indices)[NumDims]) const TESSERACT_CONDITIONAL_NOEXCEPT
FORCE_INLINE const value_type &	operator() (const my_size_t *indices) const TESSERACT_CONDITIONAL_NOEXCEPT
template<typename T , my_size_t Bits, typename Arch >
FORCE_INLINE Microkernel< T, Bits, Arch >::VecType	evalu (my_size_t logical_flat) const noexcept
	SIMD EVALUATION — logical flat, K::gather.
template<typename T , my_size_t Bits, typename Arch >
FORCE_INLINE Microkernel< T, Bits, Arch >::VecType	logical_evalu (my_size_t logical_flat) const noexcept
void	printLayoutInfo () const
FORCE_INLINE const Tensor &	transpose () const noexcept
std::string	getShape () const
FORCE_INLINE constexpr const value_type *	data () const noexcept
FORCE_INLINE constexpr value_type *	data () noexcept
Public Member Functions inherited from BaseExpr< PermutedViewConstExpr< Tensor, Perm... > >
const PermutedViewConstExpr< Tensor, Perm... > &	derived () const

Static Public Member Functions
static FORCE_INLINE constexpr my_size_t	getDim (my_size_t i) TESSERACT_CONDITIONAL_NOEXCEPT
static FORCE_INLINE constexpr my_size_t	getStride (my_size_t i) TESSERACT_CONDITIONAL_NOEXCEPT
static FORCE_INLINE constexpr my_size_t	getNumDims () noexcept
static FORCE_INLINE constexpr my_size_t	getTotalSize () noexcept

Static Public Attributes
static constexpr my_size_t	NumDims = Layout::NumDims
static constexpr my_size_t	Dim [] = {Tensor::Dim[Perm]...}
static constexpr my_size_t	TotalSize = Tensor::TotalSize

Detailed Description

template<typename Tensor, my_size_t... Perm>
class PermutedViewConstExpr< Tensor, Perm >

Compile-time permuted view over a tensor.

Does not own or copy data — references the underlying tensor's physical buffer. Permutation is applied entirely at compile time through StridedLayoutConstExpr.

The permuted layout reinterprets the same physical memory with permuted logical dimensions and strides:

Source A[2,3] padded to [2,4]: Physical: [a b c 0 | d e f 0] Layout: LogicalDims=[2,3], Strides=[4,1]

Transposed view (Perm = 1,0): Same physical buffer Layout: LogicalDims=[3,2], Strides=[1,4]

view(0,0) → 0*1 + 0*4 = offset 0 → 'a' view(0,1) → 0*1 + 1*4 = offset 4 → 'd' view(1,0) → 1*1 + 0*4 = offset 1 → 'b' view(2,1) → 2*1 + 1*4 = offset 6 → 'f'

Template Parameters

Tensor	The underlying tensor type (e.g., FusedTensorND<double, 2, 3>)
Perm	Compile-time permutation indices

Member Typedef Documentation

Layout

template<typename Tensor , my_size_t... Perm>

using PermutedViewConstExpr< Tensor, Perm >::Layout = StridedLayoutConstExpr<PadPolicy, Perm...>

PadPolicy

template<typename Tensor , my_size_t... Perm>

using PermutedViewConstExpr< Tensor, Perm >::PadPolicy = typename Tensor::AccessPolicy::PadPolicy

value_type

template<typename Tensor , my_size_t... Perm>

using PermutedViewConstExpr< Tensor, Perm >::value_type = typename Tensor::value_type

Constructor & Destructor Documentation

PermutedViewConstExpr() [1/3]

template<typename Tensor , my_size_t... Perm>

PermutedViewConstExpr< Tensor, Perm >::PermutedViewConstExpr ( const Tensor &  t )

inlineexplicitnoexcept

PermutedViewConstExpr() [2/3]

template<typename Tensor , my_size_t... Perm>

PermutedViewConstExpr< Tensor, Perm >::PermutedViewConstExpr ( const PermutedViewConstExpr< Tensor, Perm > &  )

delete

PermutedViewConstExpr() [3/3]

template<typename Tensor , my_size_t... Perm>

PermutedViewConstExpr< Tensor, Perm >::PermutedViewConstExpr ( PermutedViewConstExpr< Tensor, Perm > &&  )

delete

Member Function Documentation

data() [1/2]

template<typename Tensor , my_size_t... Perm>

FORCE_INLINE constexpr const value_type * PermutedViewConstExpr< Tensor, Perm >::data ( ) const

inlineconstexprnoexcept

data() [2/2]

template<typename Tensor , my_size_t... Perm>

FORCE_INLINE constexpr value_type * PermutedViewConstExpr< Tensor, Perm >::data ( )

inlineconstexprnoexcept

evalu()

template<typename Tensor , my_size_t... Perm>

template<typename T , my_size_t Bits, typename Arch >

FORCE_INLINE Microkernel< T, Bits, Arch >::VecType PermutedViewConstExpr< Tensor, Perm >::evalu ( my_size_t  logical_flat ) const

inlinenoexcept

SIMD EVALUATION — logical flat, K::gather.

Consecutive logical flat indices map to non-contiguous physical offsets through the permuted layout, so gather is required.

Example: transposed [3,2] view of [2,3] source: logical_flat 0 → coords(0,0) → physical 0 logical_flat 1 → coords(0,1) → physical 4 logical_flat 2 → coords(1,0) → physical 1 logical_flat 3 → coords(1,1) → physical 5 → gather from offsets [0, 4, 1, 5]

Template Parameters

T	The value type for evaluation (e.g., float, double)
Bits	Number of bits for the microkernel (e.g., 256 for AVX2)
Arch	The target architecture for the microkernel (e.g., AVX2, AVX-512)

Parameters

logical_flat

The logical flat index to evaluate from

Returns

Microkernel vector type containing the evaluated values for the SIMD width

Here is the call graph for this function:

getDim()

template<typename Tensor , my_size_t... Perm>

static FORCE_INLINE constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::getDim ( my_size_t  i )

inlinestaticconstexpr

Here is the call graph for this function:

getNumDims()

template<typename Tensor , my_size_t... Perm>

static FORCE_INLINE constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::getNumDims ( )

inlinestaticconstexprnoexcept

getShape()

template<typename Tensor , my_size_t... Perm>

std::string PermutedViewConstExpr< Tensor, Perm >::getShape ( ) const

inline

Here is the call graph for this function:

getStride()

template<typename Tensor , my_size_t... Perm>

static FORCE_INLINE constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::getStride ( my_size_t  i )

inlinestaticconstexpr

Here is the call graph for this function:

getTotalSize()

template<typename Tensor , my_size_t... Perm>

static FORCE_INLINE constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::getTotalSize ( )

inlinestaticconstexprnoexcept

logical_evalu()

template<typename Tensor , my_size_t... Perm>

template<typename T , my_size_t Bits, typename Arch >

FORCE_INLINE Microkernel< T, Bits, Arch >::VecType PermutedViewConstExpr< Tensor, Perm >::logical_evalu ( my_size_t  logical_flat ) const

inlinenoexcept

may_alias()

template<typename Tensor , my_size_t... Perm>

template<typename Output >

bool PermutedViewConstExpr< Tensor, Perm >::may_alias ( const Output &  output ) const

inlinenoexcept

operator()() [1/3]

template<typename Tensor , my_size_t... Perm>

FORCE_INLINE const value_type & PermutedViewConstExpr< Tensor, Perm >::operator() ( const my_size_t *  indices ) const

inline

Here is the call graph for this function:

operator()() [2/3]

template<typename Tensor , my_size_t... Perm>

template<typename... Indices>
requires (sizeof...(Indices) == NumDims)

FORCE_INLINE const value_type & PermutedViewConstExpr< Tensor, Perm >::operator() ( Indices...  indices ) const

inline

Here is the call graph for this function:

operator()() [3/3]

template<typename Tensor , my_size_t... Perm>

FORCE_INLINE const value_type & PermutedViewConstExpr< Tensor, Perm >::operator() ( my_size_t(&)  indices[NumDims] ) const

inline

Here is the call graph for this function:

operator=() [1/2]

template<typename Tensor , my_size_t... Perm>

PermutedViewConstExpr & PermutedViewConstExpr< Tensor, Perm >::operator= ( const PermutedViewConstExpr< Tensor, Perm > &  )

delete

operator=() [2/2]

template<typename Tensor , my_size_t... Perm>

PermutedViewConstExpr & PermutedViewConstExpr< Tensor, Perm >::operator= ( PermutedViewConstExpr< Tensor, Perm > &&  )

delete

printLayoutInfo()

template<typename Tensor , my_size_t... Perm>

void PermutedViewConstExpr< Tensor, Perm >::printLayoutInfo ( ) const

inline

Here is the call graph for this function:

transpose()

template<typename Tensor , my_size_t... Perm>

FORCE_INLINE const Tensor & PermutedViewConstExpr< Tensor, Perm >::transpose ( ) const

inlinenoexcept

Member Data Documentation

Dim

template<typename Tensor , my_size_t... Perm>

constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::Dim[] = {Tensor::Dim[Perm]...}

staticconstexpr

NumDims

template<typename Tensor , my_size_t... Perm>

constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::NumDims = Layout::NumDims

staticconstexpr

TotalSize

template<typename Tensor , my_size_t... Perm>

constexpr my_size_t PermutedViewConstExpr< Tensor, Perm >::TotalSize = Tensor::TotalSize

staticconstexpr

---

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

• core/include/algebra/permuted_view_constexpr_algebraic_traits.h
• core/include/fused/views/permuted_view_constexpr.h