SIMD

The simd module contains a unified interface to provide simd types and functions used in seqan3. More...

Collaboration diagram for SIMD:

Modules
	Views
	SIMD related views.

Classes
struct	seqan3::detail::builtin_simd< scalar_t, length >
	A class that holds the type of a simd implementation called [vector extension] (https://gcc.gnu.org/onlinedocs/gcc/Vector-Extensions.html) (formerly known as "seqan simd" in seqan2). More...
struct	seqan3::detail::builtin_simd< scalar_t, length >
	A class that holds the type of a simd implementation called [vector extension] (https://gcc.gnu.org/onlinedocs/gcc/Vector-Extensions.html) (formerly known as "seqan simd" in seqan2). More...
struct	seqan3::detail::builtin_simd_traits_helper< builtin_simd_t >
	Helper struct for seqan3::detail::is_builtin_simd. More...
struct	seqan3::detail::builtin_simd_traits_helper< builtin_simd_t >
	Helper struct for seqan3::detail::is_builtin_simd. More...
struct	seqan3::detail::is_builtin_simd< builtin_simd_t >
	This class inherits from std::true_type, iff seqan3::detail::builtin_simd<scalar_t, length>::type is a builtin simd type. More...
struct	seqan3::detail::is_native_builtin_simd< builtin_simd_t >
	This class inherits from std::true_type, iff the builtin simd type is supported by the current architecture. More...
interface	simd_concept
	The generic simd concept. More...
interface	simd_index
	Refines the seqan3::simd::simd_concept requiring the underlying scalar type to model std::integral. More...
struct	seqan3::simd::simd_traits< simd_t >
	seqan3::simd::simd_traits is the trait class that provides uniform interface to the properties of simd_t types. More...
struct	seqan3::simd::simd_traits< builtin_simd_t >
	This class specializes seqan3::simd::simd_traits for seqan3::detail::builtin_simd types. More...
struct	seqan3::simd::simd_type< scalar_t, length, simd_backend >
	seqan3::simd::simd_type encapsulates simd vector types, which can be manipulated by simd operations. More...

Typedefs
template<typename scalar_t , size_t length>
using	seqan3::detail::default_simd_backend = builtin_simd< scalar_t, length >
	seqan3::detail::default_simd_backend is the default used implementation of seqan3::simd::simd_type.
template<typename scalar_t , size_t length = detail::default_simd_length<scalar_t, detail::default_simd_backend>, template< typename scalar_t_, size_t length_ > typename simd_backend = detail::default_simd_backend>
using	seqan3::simd::simd_type_t = typename simd_type< scalar_t, length, simd_backend >::type
	Helper type of seqan3::simd::simd_type.

Functions
template<uint8_t index, simd::simd_concept simd_t>
constexpr simd_t	seqan3::detail::extract_eighth (simd_t const &src)
	Extracts one eighth of the given simd vector and stores it in the lower eighth of the target vector.
template<uint8_t index, simd::simd_concept simd_t>
constexpr simd_t	seqan3::detail::extract_half (simd_t const &src)
	Extracts one half of the given simd vector and stores the result in the lower half of the target vector.
template<size_t divisor, simd_concept simd_t>
constexpr simd_t	seqan3::detail::extract_impl (simd_t const &src, uint8_t const mask)
	Helper function to extract a part of the given simd vector.
template<uint8_t index, simd::simd_concept simd_t>
constexpr simd_t	seqan3::detail::extract_quarter (simd_t const &src)
	Extracts one quarter of the given simd vector and stores it in the lower quarter of the target vector.
template<simd::simd_concept simd_t>
constexpr simd_t	seqan3::simd::fill (typename simd_traits< simd_t >::scalar_type const scalar) noexcept
	Fills a seqan3::simd::simd_type vector with a scalar value.
template<simd::simd_concept simd_t, size_t... I>
constexpr simd_t	seqan3::detail::fill_impl (typename simd_traits< simd_t >::scalar_type const scalar, std::index_sequence< I... >) noexcept
	Helper function for seqan3::simd::fill.
template<simd::simd_concept simd_t>
constexpr simd_t	seqan3::simd::iota (typename simd_traits< simd_t >::scalar_type const offset)
	Fills a seqan3::simd::simd_type vector with the scalar values offset, offset+1, offset+2, ...
template<simd::simd_concept simd_t, typename scalar_t , scalar_t... I>
constexpr simd_t	seqan3::detail::iota_impl (scalar_t const offset, std::integer_sequence< scalar_t, I... >)
	Helper function for seqan3::simd::iota.
template<simd::simd_concept simd_t>
constexpr simd_t	seqan3::simd::load (void const *mem_addr)
	Load simd_t size bits of integral data from memory.
template<typename char_t , typename simd_t > requires simd::simd_concept<std::remove_cvref_t<simd_t>>
debug_stream_type< char_t > &	seqan3::operator<< (debug_stream_type< char_t > &s, simd_t &&simd)
	Overload for debug_stream for simd types.
template<simd::simd_concept simd_t>
constexpr void	seqan3::simd::store (void *mem_addr, simd_t const &simd_vec)
	Store simd_t size bits of integral data into memory.
template<simd::simd_concept simd_t>
constexpr void	seqan3::simd::transpose (std::array< simd_t, simd_traits< simd_t >::length > &matrix)
	Transposes the given simd vector matrix.
template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>
constexpr target_simd_t	seqan3::simd::upcast (source_simd_t const &src)
	Upcasts the given vector into the target vector using sign extension of packed values.
template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>
constexpr target_simd_t	seqan3::detail::upcast_signed (source_simd_t const &src)
	Upcasts the given vector into the target vector using signed extension of packed values.
template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>
constexpr target_simd_t	seqan3::detail::upcast_unsigned (source_simd_t const &src)
	Upcasts the given vector into the target vector using unsigned extension of packed values.

Variables
template<typename scalar_t , template< typename, size_t > typename simd_backend_t>
constexpr auto	seqan3::detail::default_simd_length
	seqan3::detail::default_simd_length returns the default length depending on the given scalar_t type, which is used in seqan3::simd::simd_type.
template<template< typename, size_t > typename simd_backend_t>
constexpr auto	seqan3::detail::default_simd_max_length = 0u
	seqan3 auto-detects the maximum number of packable [u]int8_t types.
template<>
constexpr auto	seqan3::detail::default_simd_max_length< builtin_simd >
	This function specializes seqan3::detail::default_simd_max_length for seqan3::detail::builtin_simd.
template<typename builtin_simd_t >
constexpr bool	seqan3::detail::is_builtin_simd_v = is_builtin_simd<builtin_simd_t>::value
	Helper variable to test whether a type is a simd builtin type.
template<typename builtin_simd_t >
constexpr bool	seqan3::detail::is_native_builtin_simd_v = is_native_builtin_simd<builtin_simd_t>::value
	Helper variable to test whether a type is a native simd builtin type.

Detailed Description

The simd module contains a unified interface to provide simd types and functions used in seqan3.

See also

Utility

https://en.wikipedia.org/wiki/SIMD

There are different simd implementations (backends), which are auto-selected by seqan3::simd::simd_type_t. Namely seqan3::detail::builtin_simd.

Todo:

Make this public again. We made this documentation internal-documentation only for the 3.0.0 release, because the API was not in shape yet. Remove /utility/simd/ exclusion from EXCLUDE_PATTERNS.

Todo:

Describe more what SIMD is and how to use it.

Typedef Documentation

default_simd_backend

template<typename scalar_t , size_t length>

using seqan3::detail::default_simd_backend = typedef builtin_simd<scalar_t, length>

seqan3::detail::default_simd_backend is the default used implementation of seqan3::simd::simd_type.

Template Parameters

scalar_t	The underlying type of a simd vector
length	The number of packed values in a simd vector

Function Documentation

extract_eighth()

template<uint8_t index, simd::simd_concept simd_t>

constexpr simd_t seqan3::detail::extract_eighth ( simd_t const &  src )

constexpr

Extracts one eighth of the given simd vector and stores it in the lower eighth of the target vector.

Template Parameters

index	An index value in the range of [0, 1, 2, 3, 4, 5, 6, 7].
simd_t	The simd type.

Parameters

src	The source to extract the eighth from.

Returns

A simd vector with the lower bits set with the respective eighth from src If the simd vector contains less than 8 elements, the unchanged source will be returned.

Only the first simd length / 8 elements are defined. The value of the remaining elements is implementation defined.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using int16x8_t = seqan3::simd::simd_type_t<int16_t, 8>;
 
int main()
{
    int16x8_t a{0, -1, 2, -3, 4, -5, 6, -7};
 
    int16x8_t b = seqan3::detail::extract_half<1>(a);
    int16x8_t c = seqan3::detail::extract_quarter<1>(a);
    int16x8_t d = seqan3::detail::extract_eighth<7>(a);
 
    seqan3::debug_stream << b << '\n'; // [4,-5,6,-7,0,0,0,0]
    seqan3::debug_stream << c << '\n'; // [2,-3,0,0,0,0,0,0]
    seqan3::debug_stream << d << '\n'; // [-7,0,0,0,0,0,0,0]
    return 0;
}

Example operation for SSE4:

i := index * 16
dst[15:0] := src[i+15:i]

extract_half()

template<uint8_t index, simd::simd_concept simd_t>

constexpr simd_t seqan3::detail::extract_half ( simd_t const &  src )

constexpr

Extracts one half of the given simd vector and stores the result in the lower half of the target vector.

Template Parameters

index	An index value in the range of [0, 1].
simd_t	The simd type.

Parameters

src	The source to extract the half from.

Returns

A simd vector with the lower bits set with the respective half from src If the simd vector contains less than 2 elements, the unchanged source will be returned.

Only the first simd length / 2 elements are defined. The value of the remaining elements is implementation defined.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using int16x8_t = seqan3::simd::simd_type_t<int16_t, 8>;
 
int main()
{
    int16x8_t a{0, -1, 2, -3, 4, -5, 6, -7};
 
    int16x8_t b = seqan3::detail::extract_half<1>(a);
    int16x8_t c = seqan3::detail::extract_quarter<1>(a);
    int16x8_t d = seqan3::detail::extract_eighth<7>(a);
 
    seqan3::debug_stream << b << '\n'; // [4,-5,6,-7,0,0,0,0]
    seqan3::debug_stream << c << '\n'; // [2,-3,0,0,0,0,0,0]
    seqan3::debug_stream << d << '\n'; // [-7,0,0,0,0,0,0,0]
    return 0;
}

Example operation for SSE4:

i := index * 64
dst[63:0] := src[i+63:i]

extract_impl()

template<size_t divisor, simd_concept simd_t>

constexpr simd_t seqan3::detail::extract_impl ( simd_t const &  src, uint8_t const  mask  )

constexpr

Helper function to extract a part of the given simd vector.

Template Parameters

divisor	The divisor to select the chunk size.
simd_t	The simd type; must model seqan3::simd::simd_concept.

Parameters

[in]	src	The source vector to extract from.
[in]	mask	The control mask to select which chunk is extracted.

Returns

The destination vector containing the extracted part.

Extracts the specified part of the source simd vector and stores it in the first chunk starting at offset 0 in the destination vector.

extract_quarter()

template<uint8_t index, simd::simd_concept simd_t>

constexpr simd_t seqan3::detail::extract_quarter ( simd_t const &  src )

constexpr

Extracts one quarter of the given simd vector and stores it in the lower quarter of the target vector.

Template Parameters

index	An index value in the range of [0, 1, 2, 3].
simd_t	The simd type.

Parameters

src	The source to extract the quarter from.

Returns

A simd vector with the lower bits set with the respective quarter from src. If the simd vector contains less than 4 elements, the unchanged source will be returned.

Only the first simd length / 4 elements are defined. The value of the remaining elements is implementation defined.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using int16x8_t = seqan3::simd::simd_type_t<int16_t, 8>;
 
int main()
{
    int16x8_t a{0, -1, 2, -3, 4, -5, 6, -7};
 
    int16x8_t b = seqan3::detail::extract_half<1>(a);
    int16x8_t c = seqan3::detail::extract_quarter<1>(a);
    int16x8_t d = seqan3::detail::extract_eighth<7>(a);
 
    seqan3::debug_stream << b << '\n'; // [4,-5,6,-7,0,0,0,0]
    seqan3::debug_stream << c << '\n'; // [2,-3,0,0,0,0,0,0]
    seqan3::debug_stream << d << '\n'; // [-7,0,0,0,0,0,0,0]
    return 0;
}

Example operation for SSE4:

i := index * 32
dst[31:0] := src[i+31:i]

fill()

template<simd::simd_concept simd_t>

constexpr simd_t seqan3::simd::fill ( typename simd_traits< simd_t >::scalar_type const  scalar )

constexprnoexcept

Fills a seqan3::simd::simd_type vector with a scalar value.

Template Parameters

simd_t

The simd type which satisfies seqan3::simd::simd_concept.

Parameters

[in]

scalar

The scalar value to fill the seqan3::simd::simd_type vector.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using uint16x8_t = seqan3::simd::simd_type_t<uint16_t, 8>;
 
int main()
{
    uint16x8_t a = seqan3::fill<uint16x8_t>(4);
    seqan3::debug_stream << a << "\n"; // [4,4,4,4,4,4,4,4]
 
    // or:
 
    uint16x8_t b = seqan3::simd::fill<uint16x8_t>(4);
    seqan3::debug_stream << b << "\n"; // [4,4,4,4,4,4,4,4]
    return 0;
}

iota()

template<simd::simd_concept simd_t>

constexpr simd_t seqan3::simd::iota ( typename simd_traits< simd_t >::scalar_type const  offset )

constexpr

Fills a seqan3::simd::simd_type vector with the scalar values offset, offset+1, offset+2, ...

Template Parameters

simd_t

The simd type which satisfies seqan3::simd::simd_concept.

Parameters

[in]

offset

The scalar offset to fill the seqan3::simd::simd_type vector

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using uint16x8_t = seqan3::simd::simd_type_t<uint16_t, 8>;
 
int main()
{
    uint16x8_t a = seqan3::iota<uint16x8_t>(1);
    seqan3::debug_stream << a << "\n"; // [1,2,3,4,5,6,7,8]
 
    // or:
 
    uint16x8_t b = seqan3::simd::iota<uint16x8_t>(1);
    seqan3::debug_stream << b << "\n"; // [1,2,3,4,5,6,7,8]
    return 0;
}

load()

template<simd::simd_concept simd_t>

constexpr simd_t seqan3::simd::load ( void const *  mem_addr )

constexpr

Load simd_t size bits of integral data from memory.

Template Parameters

simd_t

The simd type; must model seqan3::simd::simd_concept.

Parameters

[in]

mem_addr

The memory address to load from. Does not need to be aligned on any particular boundary.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using uint16x8_t = seqan3::simd::simd_type_t<uint16_t, 8>;
 
int main()
{
    std::vector<uint16_t> memory{0, 1, 2, 3, 4, 5, 6, 7};
    uint16x8_t a = seqan3::simd::load<uint16x8_t>(memory.data());
    seqan3::debug_stream << a << '\n'; // [0,1,2,3,4,5,6,7]
    return 0;
}

store()

template<simd::simd_concept simd_t>

constexpr void seqan3::simd::store ( void *  mem_addr, simd_t const &  simd_vec  )

constexpr

Store simd_t size bits of integral data into memory.

Template Parameters

simd_t

The simd type; must model seqan3::simd::simd_concept.

Parameters

[in]	mem_addr	The memory address to write to. Does not need to be aligned on any particular boundary.
[in]	simd_vec	The simd vector to read the data from.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
 
using uint16x8_t = seqan3::simd::simd_type_t<uint16_t, 8>;
 
int main()
{
    uint16x8_t a = seqan3::simd::iota<uint16x8_t>(0);
    std::vector<uint16_t> memory(seqan3::simd::simd_traits<uint16x8_t>::length);
    seqan3::simd::store(memory.data(), a);
    seqan3::debug_stream << memory << '\n'; // [0,1,2,3,4,5,6,7]
    return 0;
}

transpose()

template<simd::simd_concept simd_t>

constexpr void seqan3::simd::transpose ( std::array< simd_t, simd_traits< simd_t >::length > &  matrix )

constexpr

Transposes the given simd vector matrix.

Template Parameters

simd_t

The simd vector type; must model seqan3::simd::simd_concept and must be a simd built-in type.

Parameters

[in,out]

matrix

The matrix that is transposed in place.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <array>
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using uint8x4_t = seqan3::simd::simd_type_t<uint8_t, 4>;
 
int main()
{
    std::array<uint8x4_t, 4> matrix{
        {uint8x4_t{0, 1, 2, 3}, uint8x4_t{0, 1, 2, 3}, uint8x4_t{0, 1, 2, 3}, uint8x4_t{0, 1, 2, 3}}};
    seqan3::simd::transpose(matrix);
 
    seqan3::debug_stream << matrix[0] << '\n'; // [0,0,0,0]
    seqan3::debug_stream << matrix[1] << '\n'; // [1,1,1,1]
    seqan3::debug_stream << matrix[2] << '\n'; // [2,2,2,2]
    seqan3::debug_stream << matrix[3] << '\n'; // [3,3,3,3]
    return 0;
}

Exception

Strong exception guarantee.

Complexity

Quadratic.

upcast()

template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>

constexpr target_simd_t seqan3::simd::upcast ( source_simd_t const &  src )

constexpr

Upcasts the given vector into the target vector using sign extension of packed values.

Template Parameters

simd_t

The simd type; must model seqan3::simd::simd_concept and must be a builtin simd type.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
using int16x8_t = seqan3::simd::simd_type_t<int16_t, 8>;
using int32x4_t = seqan3::simd::simd_type_t<int32_t, 4>;
 
int main()
{
    int16x8_t a{0, -1, 2, -3, 4, -5, 6, -7};
    int32x4_t b = seqan3::simd::upcast<int32x4_t>(a);
 
    seqan3::debug_stream << b << '\n'; // [0,-1,2,-3]
    return 0;
}

upcast_signed()

template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>

constexpr target_simd_t seqan3::detail::upcast_signed ( source_simd_t const &  src )

constexpr

Upcasts the given vector into the target vector using signed extension of packed values.

Template Parameters

target_simd_t	The target simd type; must model seqan3::simd::simd_concept and must be a native builtin simd type.
source_simd_t	The source simd type; must model seqan3::simd::simd_concept and must be a native builtin simd type.

Parameters

[in]

src

The source to upcast into target_simd_t.

upcast_unsigned()

template<simd::simd_concept target_simd_t, simd::simd_concept source_simd_t>

constexpr target_simd_t seqan3::detail::upcast_unsigned ( source_simd_t const &  src )

constexpr

Upcasts the given vector into the target vector using unsigned extension of packed values.

Template Parameters

target_simd_t	The target simd type; must model seqan3::simd::simd_concept and must be a native builtin simd type.
source_simd_t	The source simd type; must model seqan3::simd::simd_concept and must be a native builtin simd type.

Parameters

[in]

src

The source to upcast into target_simd_t.

Variable Documentation

default_simd_length

template<typename scalar_t , template< typename, size_t > typename simd_backend_t>

constexpr auto seqan3::detail::default_simd_length

constexpr

Initial value:

= []
{
    constexpr auto max_length = default_simd_max_length<simd_backend_t>;
    if constexpr (max_length == 0)
        return min_viable_uint_v<1u>;
    else
        return min_viable_uint_v<max_length / sizeof(scalar_t)>;
}()

seqan3::detail::default_simd_length returns the default length depending on the given scalar_t type, which is used in seqan3::simd::simd_type.

Template Parameters

scalar_t	The underlying type of a simd vector
simd_backend_t	The name of the simd backend.

default_simd_max_length

template<template< typename, size_t > typename simd_backend_t>

constexpr auto seqan3::detail::default_simd_max_length = 0u

constexpr

seqan3 auto-detects the maximum number of packable [u]int8_t types.

Template Parameters

simd_backend_t

The name of the simd backend.

// SPDX-FileCopyrightText: 2006-2024 Knut Reinert & Freie Universität Berlin
// SPDX-FileCopyrightText: 2016-2024 Knut Reinert & MPI für molekulare Genetik
// SPDX-License-Identifier: CC0-1.0
 
#include <seqan3/core/debug_stream.hpp>
#include <seqan3/utility/simd/all.hpp>
#include <seqan3/utility/simd/detail/debug_stream_simd.hpp>
 
int main()
{
    constexpr auto max_length = seqan3::detail::default_simd_max_length<seqan3::detail::default_simd_backend>;
    using uint8_simd_t = seqan3::simd::simd_type_t<uint8_t, max_length == 0 ? 1 : max_length>;
 
    uint8_simd_t a = seqan3::fill<uint8_simd_t>(4);
    uint8_simd_t b = seqan3::fill<uint8_simd_t>(5);
    uint8_simd_t c = a + b;
 
    seqan3::debug_stream << c << "\n"; // [9,9,9,9,9,9,9,9]
    return 0;
}

This value is influenced by compiler flags like -march=native, -msse4, -mavx2, etc and sets this value accordingly.

See also

https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html#x86-Options for simd instruction sets and their flags.

default_simd_max_length< builtin_simd >

template<>

constexpr auto seqan3::detail::default_simd_max_length< builtin_simd >

inlineconstexpr

Initial value:

= []()
{
 
 
 
 
 
 
 
    return min_viable_uint_v<0u>;
 
}()

This function specializes seqan3::detail::default_simd_max_length for seqan3::detail::builtin_simd.

The redefinition of default_simd_max_length influences the default length (i.e., seqan3::detail::default_simd_length) of seqan3::simd::simd_type for seqan3::detail::builtin_simd types.

is_builtin_simd_v

template<typename builtin_simd_t >

constexpr bool seqan3::detail::is_builtin_simd_v = is_builtin_simd<builtin_simd_t>::value

constexpr

Helper variable to test whether a type is a simd builtin type.

Template Parameters

builtin_simd_t

The type to check.

See also

seqan3::detail::is_builtin_simd

is_native_builtin_simd_v

template<typename builtin_simd_t >

constexpr bool seqan3::detail::is_native_builtin_simd_v = is_native_builtin_simd<builtin_simd_t>::value

constexpr

Helper variable to test whether a type is a native simd builtin type.

Template Parameters

builtin_simd_t

The type to check.

See also

seqan3::detail::is_native_builtin_simd_v