seqan3::detail::view_to_simd< urng_t, simd_t > Class Template Reference

Transforms a range of ranges into chunks of seqan3::simd vectors. More...

#include <seqan3/utility/simd/views/to_simd.hpp>

Inheritance diagram for seqan3::detail::view_to_simd< urng_t, simd_t >:

Classes
class	iterator_type
	Iterator that transposes the underlying range of ranges and transforms that to SIMD types. More...

Public Member Functions
constexpr bool	empty () const noexcept
	Checks whether the range is empty.
constexpr size_t	size () const noexcept
	Returns the size of this range.
Constructors, destructor and assignment
constexpr	view_to_simd ()=default
	Defaulted.
constexpr	view_to_simd (view_to_simd const &)=default
	Defaulted.
constexpr	view_to_simd (view_to_simd &&)=default
	Defaulted.
constexpr view_to_simd &	operator= (view_to_simd const &)=default
	Defaulted.
constexpr view_to_simd &	operator= (view_to_simd &&)=default
	Defaulted.
	~view_to_simd ()=default
	Defaulted.
constexpr	view_to_simd (urng_t urng, scalar_type const padding_value=alphabet_size)
	Construction from the underlying range.
template<typename other_urng_t > requires (!std::same_as<std::remove_cvref_t<other_urng_t>, view_to_simd>) && (!std::same_as<other_urng_t, urng_t>) && std::ranges::viewable_range<other_urng_t>
constexpr	view_to_simd (other_urng_t &&urng, scalar_type const padding_value=alphabet_size)
	This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.
Iterators
constexpr iterator_type	begin () noexcept
	The iterator to begin of this range.
constexpr void	begin () const noexcept=delete
	Const iteration is disabled.
constexpr std::default_sentinel_t	end () noexcept
	A sentinel representing the end of this range.
constexpr void	end () const noexcept=delete
	Const iteration is disabled.

Private Types
Auxiliary types
using	inner_range_type = std::ranges::range_value_t< urng_t >
	The inner range type.
using	chunk_type = std::array< simd_t, simd_traits< simd_t >::length >
	The underlying type to hold the chunks.
using	scalar_type = typename simd_traits< simd_t >::scalar_type
using	max_simd_type = simd_type_t< uint8_t, simd_traits< simd_t >::max_length >
	The SIMD type with maximal number of lanes for the current arch.

Private Attributes
std::array< chunk_type, total_chunks >	cached_simd_chunks {}
	The cached chunks of transformed simd vectors.
simd_t	padding_simd_vector {}
	A cached simd vector with the padding symbol.
scalar_type	padding_value {}
	The padding value used to fill the corresponding simd vector element.
urng_t	urng {}
	The underlying range.

Static Private Attributes
Auxiliary variables
static constexpr bool	fast_load
	Check if fast load is enabled.
static constexpr uint8_t	chunk_size = simd_traits<simd_t>::length
	The size of one chunk. Equals the number of elements in the simd vector.
static constexpr uint8_t	chunks_per_load = simd_traits<simd_t>::max_length / chunk_size
	The number of chunks that can be gathered with a single load.
static constexpr uint8_t	total_chunks = fast_load ? (chunks_per_load * chunks_per_load) : 1
	The total number of chunks that can be cached.
static constexpr auto	alphabet_size = seqan3::alphabet_size<std::ranges::range_value_t<inner_range_type>>
	The alphabet size.

Detailed Description

template<std::ranges::view urng_t, simd::simd_concept simd_t>
class seqan3::detail::view_to_simd< urng_t, simd_t >

Transforms a range of ranges into chunks of seqan3::simd vectors.

Template Parameters

urng_t	The underlying range type; must model std::ranges::forward_range.
simd_t	The simd type to convert to.

This view applies an Array-of-Structure to Structure-of-Array transformation on a set of sequences. The typical application is to transform the memory layout of sequences such that they can be efficiently used in vectorised algorithms. Accordingly, this view transforms the memory already into chunks of simd vectors. The number of sequences must correspond to the length of the target simd vector, i.e. there are exactly as many sequences as their are simd vector elements. The output range will be a range over chunks, where each chunk represents a quadratic std::array over simd vectors (the size of the chunk is equal to the number of elements in the target simd vector).

Depending on the types of the input ranges a more efficient transformation using simd instructions is used. The following requirements must be fulfilled by the inner range type of the underlying range:

• they must model std::ranges::contiguous_range
• the iterator and sentinel types must model std::sized_sentinel_for
• the size of the rank type of the underlying alphabet must be 1.

If one of these requirements is not fulfilled a standard fallback algorithm is used which might be slower to transform the sequences, depending on the auto-vectorisation capabilities of the used compiler.

Member Typedef Documentation

scalar_type

template<std::ranges::view urng_t, simd::simd_concept simd_t>

using seqan3::detail::view_to_simd< urng_t, simd_t >::scalar_type = typename simd_traits<simd_t>::scalar_type

private

The scalar type.

Constructor & Destructor Documentation

view_to_simd()

template<std::ranges::view urng_t, simd::simd_concept simd_t>

constexpr seqan3::detail::view_to_simd< urng_t, simd_t >::view_to_simd ( urng_t  urng, scalar_type const  padding_value = alphabet_size  )

inlineconstexpr

Construction from the underlying range.

Parameters

[in]	urng	The underlying range.
[in]	padding_value	The value used to fill up smaller sequences.

Member Function Documentation

size()

template<std::ranges::view urng_t, simd::simd_concept simd_t>

constexpr size_t seqan3::detail::view_to_simd< urng_t, simd_t >::size ( ) const

inlineconstexprnoexcept

Returns the size of this range.

Only available if the inner range types model std::ranges::sized_range.

Member Data Documentation

fast_load

template<std::ranges::view urng_t, simd::simd_concept simd_t>

constexpr bool seqan3::detail::view_to_simd< urng_t, simd_t >::fast_load

staticconstexprprivate

Initial value:

=
        std::ranges::contiguous_range<inner_range_type>
        && std::sized_sentinel_for<std::ranges::iterator_t<inner_range_type>, std::ranges::sentinel_t<inner_range_type>>
        && sizeof(alphabet_rank_t<std::ranges::range_value_t<inner_range_type>>) == 1

Check if fast load is enabled.

---

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

• to_simd.hpp