chromium/third_party/libc++/src/include/__algorithm/simd_utils.h

//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//

#ifndef _LIBCPP___ALGORITHM_SIMD_UTILS_H
#define _LIBCPP___ALGORITHM_SIMD_UTILS_H

#include <__algorithm/min.h>
#include <__bit/bit_cast.h>
#include <__bit/countl.h>
#include <__bit/countr.h>
#include <__config>
#include <__type_traits/is_arithmetic.h>
#include <__type_traits/is_same.h>
#include <__utility/integer_sequence.h>
#include <cstddef>
#include <cstdint>

#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
#  pragma GCC system_header
#endif

_LIBCPP_PUSH_MACROS
#include <__undef_macros>

// TODO: Find out how altivec changes things and allow vectorizations there too.
#if _LIBCPP_STD_VER >= 14 && defined(_LIBCPP_CLANG_VER) && !defined(__ALTIVEC__)
#define _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS …
#else
#define _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS …
#endif

#if _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS && !defined(__OPTIMIZE_SIZE__)
#define _LIBCPP_VECTORIZE_ALGORITHMS …
#else
#define _LIBCPP_VECTORIZE_ALGORITHMS …
#endif

#if _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS

_LIBCPP_BEGIN_NAMESPACE_STD

__can_map_to_integer_v;

template <size_t _TypeSize>
struct __get_as_integer_type_impl;

template <>
struct __get_as_integer_type_impl<1> { … };

template <>
struct __get_as_integer_type_impl<2> { … };
template <>
struct __get_as_integer_type_impl<4> { … };
template <>
struct __get_as_integer_type_impl<8> { … };

__get_as_integer_type_t;

// This isn't specialized for 64 byte vectors on purpose. They have the potential to significantly reduce performance
// in mixed simd/non-simd workloads and don't provide any performance improvement for currently vectorized algorithms
// as far as benchmarks are concerned.
#  if defined(__AVX__) || defined(__MVS__)
template <class _Tp>
inline constexpr size_t __native_vector_size = 32 / sizeof(_Tp);
#  elif defined(__SSE__) || defined(__ARM_NEON__)
__native_vector_size;
#  elif defined(__MMX__)
template <class _Tp>
inline constexpr size_t __native_vector_size = 8 / sizeof(_Tp);
#  else
template <class _Tp>
inline constexpr size_t __native_vector_size = 1;
#  endif

__simd_vector;

__simd_vector_size_v;

__simd_vector_size_v;

template <class _Tp, size_t _Np>
_LIBCPP_HIDE_FROM_ABI _Tp __simd_vector_underlying_type_impl(__simd_vector<_Tp, _Np>) { … }

__simd_vector_underlying_type_t;

// This isn't inlined without always_inline when loading chars.
template <class _VecT, class _Iter>
_LIBCPP_NODISCARD _LIBCPP_ALWAYS_INLINE _LIBCPP_HIDE_FROM_ABI _VecT __load_vector(_Iter __iter) noexcept { … }

template <class _Tp, size_t _Np>
_LIBCPP_NODISCARD _LIBCPP_HIDE_FROM_ABI bool __all_of(__simd_vector<_Tp, _Np> __vec) noexcept { … }

template <class _Tp, size_t _Np>
_LIBCPP_NODISCARD _LIBCPP_HIDE_FROM_ABI size_t __find_first_set(__simd_vector<_Tp, _Np> __vec) noexcept { … }

template <class _Tp, size_t _Np>
_LIBCPP_NODISCARD _LIBCPP_HIDE_FROM_ABI size_t __find_first_not_set(__simd_vector<_Tp, _Np> __vec) noexcept { … }

_LIBCPP_END_NAMESPACE_STD

#endif // _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS

_LIBCPP_POP_MACROS

#endif // _LIBCPP___ALGORITHM_SIMD_UTILS_H