xsimd_avx512f.hpp

/***************************************************************************
 * Copyright (c) Johan Mabille, Sylvain Corlay, Wolf Vollprecht and         *
 * Martin Renou                                                             *
 * Copyright (c) QuantStack                                                 *
 * Copyright (c) Serge Guelton                                              *
 *                                                                          *
 * Distributed under the terms of the BSD 3-Clause License.                 *
 *                                                                          *
 * The full license is in the file LICENSE, distributed with this software. *
 ****************************************************************************/

#ifndef XSIMD_AVX512F_HPP
#define XSIMD_AVX512F_HPP

#include <complex>
#include <limits>
#include <type_traits>

#include "../types/xsimd_avx512f_register.hpp"

namespace xsimd
{

    namespace kernel
    {
        using namespace types;

        // fwd
        template <class A, class T, class Mask>
        XSIMD_INLINE batch<T, A> decr_if(batch<T, A> const& self, Mask const& mask, requires_arch<generic>) noexcept;
        template <class A, class T, class Mask>
        XSIMD_INLINE batch<T, A> incr_if(batch<T, A> const& self, Mask const& mask, requires_arch<generic>) noexcept;
        template <class A, class T, size_t I>
        XSIMD_INLINE batch<T, A> insert(batch<T, A> const& self, T val, index<I>, requires_arch<generic>) noexcept;
        template <class A>
        XSIMD_INLINE void transpose(batch<uint16_t, A>* matrix_begin, batch<uint16_t, A>* matrix_end, requires_arch<generic>) noexcept;
        template <class A>
        XSIMD_INLINE void transpose(batch<uint8_t, A>* matrix_begin, batch<uint8_t, A>* matrix_end, requires_arch<generic>) noexcept;

        namespace detail
        {
            XSIMD_INLINE void split_avx512(__m512 val, __m256& low, __m256& high) noexcept
            {
                low = _mm512_castps512_ps256(val);
                high = _mm512_extractf32x8_ps(val, 1);
            }
            XSIMD_INLINE void split_avx512(__m512d val, __m256d& low, __m256d& high) noexcept
            {
                low = _mm512_castpd512_pd256(val);
                high = _mm512_extractf64x4_pd(val, 1);
            }
            XSIMD_INLINE void split_avx512(__m512i val, __m256i& low, __m256i& high) noexcept
            {
                low = _mm512_castsi512_si256(val);
                high = _mm512_extracti64x4_epi64(val, 1);
            }
            XSIMD_INLINE __m512i merge_avx(__m256i low, __m256i high) noexcept
            {
                return _mm512_inserti64x4(_mm512_castsi256_si512(low), high, 1);
            }
            XSIMD_INLINE __m512 merge_avx(__m256 low, __m256 high) noexcept
            {
                return _mm512_castpd_ps(_mm512_insertf64x4(_mm512_castpd256_pd512(_mm256_castps_pd(low)), _mm256_castps_pd(high), 1));
            }
            XSIMD_INLINE __m512d merge_avx(__m256d low, __m256d high) noexcept
            {
                return _mm512_insertf64x4(_mm512_castpd256_pd512(low), high, 1);
            }
            template <class F>
            __m512i fwd_to_avx(F f, __m512i self)
            {
                __m256i self_low, self_high;
                split_avx512(self, self_low, self_high);
                __m256i res_low = f(self_low);
                __m256i res_high = f(self_high);
                return merge_avx(res_low, res_high);
            }
            template <class F>
            __m512i fwd_to_avx(F f, __m512i self, __m512i other)
            {
                __m256i self_low, self_high, other_low, other_high;
                split_avx512(self, self_low, self_high);
                split_avx512(other, other_low, other_high);
                __m256i res_low = f(self_low, other_low);
                __m256i res_high = f(self_high, other_high);
                return merge_avx(res_low, res_high);
            }
            template <class F>
            __m512i fwd_to_avx(F f, __m512i self, int32_t other)
            {
                __m256i self_low, self_high;
                split_avx512(self, self_low, self_high);
                __m256i res_low = f(self_low, other);
                __m256i res_high = f(self_high, other);
                return merge_avx(res_low, res_high);
            }
        }
        namespace detail
        {

            XSIMD_INLINE uint32_t morton(uint16_t x, uint16_t y) noexcept
            {

                static const unsigned short MortonTable256[256] = {
                    0x0000, 0x0001, 0x0004, 0x0005, 0x0010, 0x0011, 0x0014, 0x0015,
                    0x0040, 0x0041, 0x0044, 0x0045, 0x0050, 0x0051, 0x0054, 0x0055,
                    0x0100, 0x0101, 0x0104, 0x0105, 0x0110, 0x0111, 0x0114, 0x0115,
                    0x0140, 0x0141, 0x0144, 0x0145, 0x0150, 0x0151, 0x0154, 0x0155,
                    0x0400, 0x0401, 0x0404, 0x0405, 0x0410, 0x0411, 0x0414, 0x0415,
                    0x0440, 0x0441, 0x0444, 0x0445, 0x0450, 0x0451, 0x0454, 0x0455,
                    0x0500, 0x0501, 0x0504, 0x0505, 0x0510, 0x0511, 0x0514, 0x0515,
                    0x0540, 0x0541, 0x0544, 0x0545, 0x0550, 0x0551, 0x0554, 0x0555,
                    0x1000, 0x1001, 0x1004, 0x1005, 0x1010, 0x1011, 0x1014, 0x1015,
                    0x1040, 0x1041, 0x1044, 0x1045, 0x1050, 0x1051, 0x1054, 0x1055,
                    0x1100, 0x1101, 0x1104, 0x1105, 0x1110, 0x1111, 0x1114, 0x1115,
                    0x1140, 0x1141, 0x1144, 0x1145, 0x1150, 0x1151, 0x1154, 0x1155,
                    0x1400, 0x1401, 0x1404, 0x1405, 0x1410, 0x1411, 0x1414, 0x1415,
                    0x1440, 0x1441, 0x1444, 0x1445, 0x1450, 0x1451, 0x1454, 0x1455,
                    0x1500, 0x1501, 0x1504, 0x1505, 0x1510, 0x1511, 0x1514, 0x1515,
                    0x1540, 0x1541, 0x1544, 0x1545, 0x1550, 0x1551, 0x1554, 0x1555,
                    0x4000, 0x4001, 0x4004, 0x4005, 0x4010, 0x4011, 0x4014, 0x4015,
                    0x4040, 0x4041, 0x4044, 0x4045, 0x4050, 0x4051, 0x4054, 0x4055,
                    0x4100, 0x4101, 0x4104, 0x4105, 0x4110, 0x4111, 0x4114, 0x4115,
                    0x4140, 0x4141, 0x4144, 0x4145, 0x4150, 0x4151, 0x4154, 0x4155,
                    0x4400, 0x4401, 0x4404, 0x4405, 0x4410, 0x4411, 0x4414, 0x4415,
                    0x4440, 0x4441, 0x4444, 0x4445, 0x4450, 0x4451, 0x4454, 0x4455,
                    0x4500, 0x4501, 0x4504, 0x4505, 0x4510, 0x4511, 0x4514, 0x4515,
                    0x4540, 0x4541, 0x4544, 0x4545, 0x4550, 0x4551, 0x4554, 0x4555,
                    0x5000, 0x5001, 0x5004, 0x5005, 0x5010, 0x5011, 0x5014, 0x5015,
                    0x5040, 0x5041, 0x5044, 0x5045, 0x5050, 0x5051, 0x5054, 0x5055,
                    0x5100, 0x5101, 0x5104, 0x5105, 0x5110, 0x5111, 0x5114, 0x5115,
                    0x5140, 0x5141, 0x5144, 0x5145, 0x5150, 0x5151, 0x5154, 0x5155,
                    0x5400, 0x5401, 0x5404, 0x5405, 0x5410, 0x5411, 0x5414, 0x5415,
                    0x5440, 0x5441, 0x5444, 0x5445, 0x5450, 0x5451, 0x5454, 0x5455,
                    0x5500, 0x5501, 0x5504, 0x5505, 0x5510, 0x5511, 0x5514, 0x5515,
                    0x5540, 0x5541, 0x5544, 0x5545, 0x5550, 0x5551, 0x5554, 0x5555
                };

                uint32_t z = MortonTable256[y >> 8] << 17 | MortonTable256[x >> 8] << 16 | MortonTable256[y & 0xFF] << 1 | MortonTable256[x & 0xFF];
                return z;
            }

            template <class A, class T, int Cmp>
            XSIMD_INLINE batch_bool<T, A> compare_int_avx512f(batch<T, A> const& self, batch<T, A> const& other) noexcept
            {
                using register_type = typename batch_bool<T, A>::register_type;
                if (std::is_signed<T>::value)
                {
                    XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
                    {
                        // shifting to take sign into account
                        uint64_t mask_low0 = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0x000000FF)) << 24,
                                                                   (batch<int32_t, A>(other.data) & batch<int32_t, A>(0x000000FF)) << 24,
                                                                   Cmp);
                        uint64_t mask_low1 = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0x0000FF00)) << 16,
                                                                   (batch<int32_t, A>(other.data) & batch<int32_t, A>(0x0000FF00)) << 16,
                                                                   Cmp);
                        uint64_t mask_high0 = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0x00FF0000)) << 8,
                                                                    (batch<int32_t, A>(other.data) & batch<int32_t, A>(0x00FF0000)) << 8,
                                                                    Cmp);
                        uint64_t mask_high1 = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0xFF000000)),
                                                                    (batch<int32_t, A>(other.data) & batch<int32_t, A>(0xFF000000)),
                                                                    Cmp);
                        uint64_t mask = 0;
                        for (unsigned i = 0; i < 16; ++i)
                        {
                            mask |= (mask_low0 & (uint64_t(1) << i)) << (3 * i + 0);
                            mask |= (mask_low1 & (uint64_t(1) << i)) << (3 * i + 1);
                            mask |= (mask_high0 & (uint64_t(1) << i)) << (3 * i + 2);
                            mask |= (mask_high1 & (uint64_t(1) << i)) << (3 * i + 3);
                        }
                        return (register_type)mask;
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
                    {
                        // shifting to take sign into account
                        uint16_t mask_low = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0x0000FFFF)) << 16,
                                                                  (batch<int32_t, A>(other.data) & batch<int32_t, A>(0x0000FFFF)) << 16,
                                                                  Cmp);
                        uint16_t mask_high = _mm512_cmp_epi32_mask((batch<int32_t, A>(self.data) & batch<int32_t, A>(0xFFFF0000)),
                                                                   (batch<int32_t, A>(other.data) & batch<int32_t, A>(0xFFFF0000)),
                                                                   Cmp);
                        return static_cast<register_type>(morton(mask_low, mask_high));
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                    {
                        return (register_type)_mm512_cmp_epi32_mask(self, other, Cmp);
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                    {
                        return (register_type)_mm512_cmp_epi64_mask(self, other, Cmp);
                    }
                }
                else
                {
                    XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
                    {
                        uint64_t mask_low0 = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0x000000FF)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0x000000FF)), Cmp);
                        uint64_t mask_low1 = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0x0000FF00)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0x0000FF00)), Cmp);
                        uint64_t mask_high0 = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0x00FF0000)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0x00FF0000)), Cmp);
                        uint64_t mask_high1 = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0xFF000000)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0xFF000000)), Cmp);
                        uint64_t mask = 0;
                        for (unsigned i = 0; i < 16; ++i)
                        {
                            mask |= (mask_low0 & (uint64_t(1) << i)) << (3 * i + 0);
                            mask |= (mask_low1 & (uint64_t(1) << i)) << (3 * i + 1);
                            mask |= (mask_high0 & (uint64_t(1) << i)) << (3 * i + 2);
                            mask |= (mask_high1 & (uint64_t(1) << i)) << (3 * i + 3);
                        }
                        return (register_type)mask;
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
                    {
                        uint16_t mask_low = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0x0000FFFF)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0x0000FFFF)), Cmp);
                        uint16_t mask_high = _mm512_cmp_epu32_mask((batch<uint32_t, A>(self.data) & batch<uint32_t, A>(0xFFFF0000)), (batch<uint32_t, A>(other.data) & batch<uint32_t, A>(0xFFFF0000)), Cmp);
                        return static_cast<register_type>(morton(mask_low, mask_high));
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                    {
                        return (register_type)_mm512_cmp_epu32_mask(self, other, Cmp);
                    }
                    else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                    {
                        return (register_type)_mm512_cmp_epu64_mask(self, other, Cmp);
                    }
                }
            }
        }

        // abs
        template <class A>
        XSIMD_INLINE batch<float, A> abs(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            __m512 self_asf = (__m512)self;
            __m512i self_asi = *reinterpret_cast<__m512i*>(&self_asf);
            __m512i res_asi = _mm512_and_epi32(_mm512_set1_epi32(0x7FFFFFFF), self_asi);
            return *reinterpret_cast<__m512*>(&res_asi);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> abs(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            __m512d self_asd = (__m512d)self;
            __m512i self_asi = *reinterpret_cast<__m512i*>(&self_asd);
            __m512i res_asi = _mm512_and_epi64(_mm512_set1_epi64(0x7FFFFFFFFFFFFFFF),
                                               self_asi);
            return *reinterpret_cast<__m512d*>(&res_asi);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> abs(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            if (std::is_unsigned<T>::value)
            {
                return self;
            }

            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                return detail::fwd_to_avx([](__m256i s) noexcept
                                          { return abs(batch<T, avx2>(s)); },
                                          self);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                return detail::fwd_to_avx([](__m256i s) noexcept
                                          { return abs(batch<T, avx2>(s)); },
                                          self);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_abs_epi32(self);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_abs_epi64(self);
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
        }

        // add
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> add(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                          { return add(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                          self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                          { return add(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                          self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_add_epi32(self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_add_epi64(self, other);
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
        }
        template <class A>
        XSIMD_INLINE batch<float, A> add(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_add_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> add(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_add_pd(self, other);
        }

        // all
        template <class A, class T>
        XSIMD_INLINE bool all(batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return self.data == register_type(-1);
        }

        // any
        template <class A, class T>
        XSIMD_INLINE bool any(batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return self.data != register_type(0);
        }

        // batch_bool_cast
        template <class A, class T_out, class T_in>
        XSIMD_INLINE batch_bool<T_out, A> batch_bool_cast(batch_bool<T_in, A> const& self, batch_bool<T_out, A> const&, requires_arch<avx512f>) noexcept
        {
            return self.data;
        }

        // bitwise_and
        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_and(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
#if defined(_MSC_VER)
            return _mm512_and_ps(self, other);
#else
            return _mm512_castsi512_ps(_mm512_and_si512(_mm512_castps_si512(self), _mm512_castps_si512(other)));
#endif
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_and(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(_mm512_and_si512(_mm512_castpd_si512(self), _mm512_castpd_si512(other)));
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_and(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_and_si512(self, other);
        }

        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> bitwise_and(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(self.data & other.data);
        }

        // bitwise_andnot
        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_andnot(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_ps(_mm512_andnot_si512(_mm512_castps_si512(other), _mm512_castps_si512(self)));
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_andnot(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(_mm512_andnot_si512(_mm512_castpd_si512(other), _mm512_castpd_si512(self)));
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_andnot(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_andnot_si512(other, self);
        }

        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> bitwise_andnot(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(self.data & ~other.data);
        }

        // bitwise_lshift
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_lshift(batch<T, A> const& self, int32_t other, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
#if defined(XSIMD_AVX512_SHIFT_INTRINSICS_IMM_ONLY)
                __m512i tmp = _mm512_sllv_epi32(self, _mm512_set1_epi32(other));
#else
                __m512i tmp = _mm512_slli_epi32(self, other);
#endif
                return _mm512_and_si512(_mm512_set1_epi8(0xFF << other), tmp);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                return detail::fwd_to_avx([](__m256i s, int32_t o) noexcept
                                          { return bitwise_lshift(batch<T, avx2>(s), o, avx2 {}); },
                                          self, other);
#if defined(XSIMD_AVX512_SHIFT_INTRINSICS_IMM_ONLY)
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_sllv_epi32(self, _mm512_set1_epi32(other));
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_sllv_epi64(self, _mm512_set1_epi64(other));
#else
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_slli_epi32(self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_slli_epi64(self, other);
#endif
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
        }

        // bitwise_not
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_not(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_xor_si512(self, _mm512_set1_epi32(-1));
        }
        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> bitwise_not(batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(~self.data);
        }

        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_not(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_ps(_mm512_xor_si512(_mm512_castps_si512(self), _mm512_set1_epi32(-1)));
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_not(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(_mm512_xor_si512(_mm512_castpd_si512(self), _mm512_set1_epi32(-1)));
        }

        // bitwise_or
        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_or(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_ps(_mm512_or_si512(_mm512_castps_si512(self), _mm512_castps_si512(other)));
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_or(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(_mm512_or_si512(_mm512_castpd_si512(self), _mm512_castpd_si512(other)));
        }

        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> bitwise_or(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(self.data | other.data);
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_or(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_or_si512(self, other);
        }

        // bitwise_rshift
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_rshift(batch<T, A> const& self, int32_t other, requires_arch<avx512f>) noexcept
        {
            if (std::is_signed<T>::value)
            {
#if defined(XSIMD_AVX512_SHIFT_INTRINSICS_IMM_ONLY)
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_srav_epi32(self, _mm512_set1_epi32(other));
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_srav_epi64(self, _mm512_set1_epi64(other));
#else
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_srai_epi32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_srai_epi64(self, other);
#endif
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, int32_t o) noexcept
                                              { return bitwise_rshift(batch<T, avx2>(s), o, avx2 {}); },
                                              self, other);
                }
            }
            else
            {
                XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
                {
#if defined(XSIMD_AVX512_SHIFT_INTRINSICS_IMM_ONLY)
                    __m512i tmp = _mm512_srlv_epi32(self, _mm512_set1_epi32(other));
#else
                    __m512i tmp = _mm512_srli_epi32(self, other);
#endif
                    return _mm512_and_si512(_mm512_set1_epi8(0xFF >> other), tmp);
#if defined(XSIMD_AVX512_SHIFT_INTRINSICS_IMM_ONLY)
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_srlv_epi32(self, _mm512_set1_epi32(other));
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_srlv_epi64(self, _mm512_set1_epi64(other));
#else
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_srli_epi32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_srli_epi64(self, other);
#endif
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, int32_t o) noexcept
                                              { return bitwise_rshift(batch<T, avx2>(s), o, avx2 {}); },
                                              self, other);
                }
            }
        }

        // bitwise_xor
        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_xor(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_ps(_mm512_xor_si512(_mm512_castps_si512(self), _mm512_castps_si512(other)));
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_xor(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(_mm512_xor_si512(_mm512_castpd_si512(self), _mm512_castpd_si512(other)));
        }

        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> bitwise_xor(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(self.data | other.data);
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_xor(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_xor_si512(self, other);
        }

        // bitwise_cast
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<float, A> bitwise_cast(batch<T, A> const& self, batch<float, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_ps(self);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<double, A> bitwise_cast(batch<T, A> const& self, batch<double, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castsi512_pd(self);
        }
        template <class A, class T, class Tp, class = typename std::enable_if<std::is_integral<typename std::common_type<T, Tp>::type>::value, void>::type>
        XSIMD_INLINE batch<Tp, A> bitwise_cast(batch<T, A> const& self, batch<Tp, A> const&, requires_arch<avx512f>) noexcept
        {
            return batch<Tp, A>(self.data);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> bitwise_cast(batch<float, A> const& self, batch<double, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castps_pd(self);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_cast(batch<float, A> const& self, batch<T, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castps_si512(self);
        }
        template <class A>
        XSIMD_INLINE batch<float, A> bitwise_cast(batch<double, A> const& self, batch<float, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castpd_ps(self);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> bitwise_cast(batch<double, A> const& self, batch<T, A> const&, requires_arch<avx512f>) noexcept
        {
            return _mm512_castpd_si512(self);
        }

        // broadcast
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> broadcast(T val, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                return _mm512_set1_epi8(val);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                return _mm512_set1_epi16(val);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_set1_epi32(val);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_set1_epi64(val);
            }
            else
            {
                assert(false && "unsupported");
                return {};
            }
        }
        template <class A>
        XSIMD_INLINE batch<float, A> broadcast(float val, requires_arch<avx512f>) noexcept
        {
            return _mm512_set1_ps(val);
        }
        template <class A>
        batch<double, A> XSIMD_INLINE broadcast(double val, requires_arch<avx512f>) noexcept
        {
            return _mm512_set1_pd(val);
        }

        // ceil
        template <class A>
        XSIMD_INLINE batch<float, A> ceil(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_ps(self, _MM_FROUND_TO_POS_INF);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> ceil(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_pd(self, _MM_FROUND_TO_POS_INF);
        }

        // compress
        template <class A>
        XSIMD_INLINE batch<float, A> compress(batch<float, A> const& self, batch_bool<float, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_ps(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> compress(batch<double, A> const& self, batch_bool<double, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_pd(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<int32_t, A> compress(batch<int32_t, A> const& self, batch_bool<int32_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_epi32(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<uint32_t, A> compress(batch<uint32_t, A> const& self, batch_bool<uint32_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_epi32(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<int64_t, A> compress(batch<int64_t, A> const& self, batch_bool<int64_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_epi64(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<uint64_t, A> compress(batch<uint64_t, A> const& self, batch_bool<uint64_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_compress_epi64(mask.mask(), self);
        }

        // convert
        namespace detail
        {
            template <class A>
            XSIMD_INLINE batch<float, A> fast_cast(batch<int32_t, A> const& self, batch<float, A> const&, requires_arch<avx512f>) noexcept
            {
                return _mm512_cvtepi32_ps(self);
            }

            template <class A>
            XSIMD_INLINE batch<int32_t, A> fast_cast(batch<float, A> const& self, batch<int32_t, A> const&, requires_arch<avx512f>) noexcept
            {
                return _mm512_cvttps_epi32(self);
            }

            template <class A>
            XSIMD_INLINE batch<float, A> fast_cast(batch<uint32_t, A> const& self, batch<float, A> const&, requires_arch<avx512f>) noexcept
            {
                return _mm512_cvtepu32_ps(self);
            }

            template <class A>
            batch<uint32_t, A> fast_cast(batch<float, A> const& self, batch<uint32_t, A> const&, requires_arch<avx512f>)
            {
                return _mm512_cvttps_epu32(self);
            }
        }

        namespace detail
        {
            // complex_low
            template <class A>
            XSIMD_INLINE batch<float, A> complex_low(batch<std::complex<float>, A> const& self, requires_arch<avx512f>) noexcept
            {
                __m512i idx = _mm512_setr_epi32(0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23);
                return _mm512_permutex2var_ps(self.real(), idx, self.imag());
            }
            template <class A>
            XSIMD_INLINE batch<double, A> complex_low(batch<std::complex<double>, A> const& self, requires_arch<avx512f>) noexcept
            {
                __m512i idx = _mm512_setr_epi64(0, 8, 1, 9, 2, 10, 3, 11);
                return _mm512_permutex2var_pd(self.real(), idx, self.imag());
            }

            // complex_high
            template <class A>
            XSIMD_INLINE batch<float, A> complex_high(batch<std::complex<float>, A> const& self, requires_arch<avx512f>) noexcept
            {
                __m512i idx = _mm512_setr_epi32(8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31);
                return _mm512_permutex2var_ps(self.real(), idx, self.imag());
            }
            template <class A>
            XSIMD_INLINE batch<double, A> complex_high(batch<std::complex<double>, A> const& self, requires_arch<avx512f>) noexcept
            {
                __m512i idx = _mm512_setr_epi64(4, 12, 5, 13, 6, 14, 7, 15);
                return _mm512_permutex2var_pd(self.real(), idx, self.imag());
            }
        }
        // incr_if
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> decr_if(batch<T, A> const& self, batch_bool<T, A> const& mask, requires_arch<avx512f>) noexcept
        {

            XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_mask_sub_epi32(self, mask.data, self, _mm512_set1_epi32(1));
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_mask_sub_epi64(self, mask.data, self, _mm512_set1_epi64(1));
            }
            else
            {
                return decr_if(self, mask, generic {});
            }
        }

        // div
        template <class A>
        XSIMD_INLINE batch<float, A> div(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_div_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> div(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_div_pd(self, other);
        }

        // eq
        template <class A>
        XSIMD_INLINE batch_bool<float, A> eq(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_EQ_OQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> eq(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_EQ_OQ);
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> eq(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return detail::compare_int_avx512f<A, T, _MM_CMPINT_EQ>(self, other);
        }
        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> eq(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(~self.data ^ other.data);
        }

        // expand
        template <class A>
        XSIMD_INLINE batch<float, A> expand(batch<float, A> const& self, batch_bool<float, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_ps(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> expand(batch<double, A> const& self, batch_bool<double, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_pd(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<int32_t, A> expand(batch<int32_t, A> const& self, batch_bool<int32_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_epi32(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<uint32_t, A> expand(batch<uint32_t, A> const& self, batch_bool<uint32_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_epi32(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<int64_t, A> expand(batch<int64_t, A> const& self, batch_bool<int64_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_epi64(mask.mask(), self);
        }
        template <class A>
        XSIMD_INLINE batch<uint64_t, A> expand(batch<uint64_t, A> const& self, batch_bool<uint64_t, A> const& mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_maskz_expand_epi64(mask.mask(), self);
        }

        // floor
        template <class A>
        XSIMD_INLINE batch<float, A> floor(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_ps(self, _MM_FROUND_TO_NEG_INF);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> floor(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_pd(self, _MM_FROUND_TO_NEG_INF);
        }

        // fnma
        template <class A>
        XSIMD_INLINE batch<float, A> fnma(batch<float, A> const& x, batch<float, A> const& y, batch<float, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fnmadd_ps(x, y, z);
        }

        template <class A>
        XSIMD_INLINE batch<double, A> fnma(batch<double, A> const& x, batch<double, A> const& y, batch<double, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fnmadd_pd(x, y, z);
        }

        // fma
        template <class A>
        XSIMD_INLINE batch<float, A> fma(batch<float, A> const& x, batch<float, A> const& y, batch<float, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fmadd_ps(x, y, z);
        }

        template <class A>
        XSIMD_INLINE batch<double, A> fma(batch<double, A> const& x, batch<double, A> const& y, batch<double, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fmadd_pd(x, y, z);
        }

        // fms
        template <class A>
        XSIMD_INLINE batch<float, A> fms(batch<float, A> const& x, batch<float, A> const& y, batch<float, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fmsub_ps(x, y, z);
        }

        template <class A>
        XSIMD_INLINE batch<double, A> fms(batch<double, A> const& x, batch<double, A> const& y, batch<double, A> const& z, requires_arch<avx512f>) noexcept
        {
            return _mm512_fmsub_pd(x, y, z);
        }

        // from bool
        template <class A, class T>
        XSIMD_INLINE batch<T, A> from_bool(batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return select(self, batch<T, A>(1), batch<T, A>(0));
        }

        // from_mask
        template <class T, class A>
        XSIMD_INLINE batch_bool<T, A> from_mask(batch_bool<T, A> const&, uint64_t mask, requires_arch<avx512f>) noexcept
        {
            return static_cast<typename batch_bool<T, A>::register_type>(mask);
        }

        // gather
        template <class T, class A, class U, detail::enable_sized_integral_t<T, 4> = 0, detail::enable_sized_integral_t<U, 4> = 0>
        XSIMD_INLINE batch<T, A> gather(batch<T, A> const&, T const* src, batch<U, A> const& index,
                                        kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_i32gather_epi32(index, static_cast<const void*>(src), sizeof(T));
        }

        template <class T, class A, class U, detail::enable_sized_integral_t<T, 8> = 0, detail::enable_sized_integral_t<U, 8> = 0>
        XSIMD_INLINE batch<T, A> gather(batch<T, A> const&, T const* src, batch<U, A> const& index,
                                        kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_i64gather_epi64(index, static_cast<const void*>(src), sizeof(T));
        }

        template <class A, class U, detail::enable_sized_integral_t<U, 4> = 0>
        XSIMD_INLINE batch<float, A> gather(batch<float, A> const&, float const* src,
                                            batch<U, A> const& index,
                                            kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_i32gather_ps(index, src, sizeof(float));
        }

        template <class A, class U, detail::enable_sized_integral_t<U, 8> = 0>
        XSIMD_INLINE batch<double, A>
        gather(batch<double, A> const&, double const* src, batch<U, A> const& index,
               kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_i64gather_pd(index, src, sizeof(double));
        }

        // gather: handmade conversions
        template <class A, class V, detail::enable_sized_integral_t<V, 4> = 0>
        XSIMD_INLINE batch<float, A> gather(batch<float, A> const&, double const* src,
                                            batch<V, A> const& index,
                                            requires_arch<avx512f>) noexcept
        {
            const batch<double, A> low(_mm512_i32gather_pd(_mm512_castsi512_si256(index.data), src, sizeof(double)));
            const batch<double, A> high(_mm512_i32gather_pd(_mm256_castpd_si256(_mm512_extractf64x4_pd(_mm512_castsi512_pd(index.data), 1)), src, sizeof(double)));
            return detail::merge_avx(_mm512_cvtpd_ps(low.data), _mm512_cvtpd_ps(high.data));
        }

        template <class A, class V, detail::enable_sized_integral_t<V, 4> = 0>
        XSIMD_INLINE batch<int32_t, A> gather(batch<int32_t, A> const&, double const* src,
                                              batch<V, A> const& index,
                                              requires_arch<avx512f>) noexcept
        {
            const batch<double, A> low(_mm512_i32gather_pd(_mm512_castsi512_si256(index.data), src, sizeof(double)));
            const batch<double, A> high(_mm512_i32gather_pd(_mm256_castpd_si256(_mm512_extractf64x4_pd(_mm512_castsi512_pd(index.data), 1)), src, sizeof(double)));
            return detail::merge_avx(_mm512_cvtpd_epi32(low.data), _mm512_cvtpd_epi32(high.data));
        }

        // ge
        template <class A>
        XSIMD_INLINE batch_bool<float, A> ge(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_GE_OQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> ge(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_GE_OQ);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> ge(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return detail::compare_int_avx512f<A, T, _MM_CMPINT_GE>(self, other);
        }

        // gt
        template <class A>
        XSIMD_INLINE batch_bool<float, A> gt(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_GT_OQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> gt(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_GT_OQ);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> gt(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return detail::compare_int_avx512f<A, T, _MM_CMPINT_GT>(self, other);
        }

        // haddp
        template <class A>
        XSIMD_INLINE batch<float, A> haddp(batch<float, A> const* row, requires_arch<avx512f>) noexcept
        {
            // The following folds over the vector once:
            // tmp1 = [a0..8, b0..8]
            // tmp2 = [a8..f, b8..f]
#define XSIMD_AVX512_HADDP_STEP1(I, a, b)                                \
    batch<float, avx512f> res##I;                                        \
    {                                                                    \
        auto tmp1 = _mm512_shuffle_f32x4(a, b, _MM_SHUFFLE(1, 0, 1, 0)); \
        auto tmp2 = _mm512_shuffle_f32x4(a, b, _MM_SHUFFLE(3, 2, 3, 2)); \
        res##I = _mm512_add_ps(tmp1, tmp2);                              \
    }

            XSIMD_AVX512_HADDP_STEP1(0, row[0], row[2]);
            XSIMD_AVX512_HADDP_STEP1(1, row[4], row[6]);
            XSIMD_AVX512_HADDP_STEP1(2, row[1], row[3]);
            XSIMD_AVX512_HADDP_STEP1(3, row[5], row[7]);
            XSIMD_AVX512_HADDP_STEP1(4, row[8], row[10]);
            XSIMD_AVX512_HADDP_STEP1(5, row[12], row[14]);
            XSIMD_AVX512_HADDP_STEP1(6, row[9], row[11]);
            XSIMD_AVX512_HADDP_STEP1(7, row[13], row[15]);

#undef XSIMD_AVX512_HADDP_STEP1

            // The following flds the code and shuffles so that hadd_ps produces the correct result
            // tmp1 = [a0..4,  a8..12,  b0..4,  b8..12] (same for tmp3)
            // tmp2 = [a5..8, a12..16, b5..8, b12..16]  (same for tmp4)
            // tmp5 = [r1[0], r1[2], r2[0], r2[2], r1[4], r1[6] ...
#define XSIMD_AVX512_HADDP_STEP2(I, a, b, c, d)                                                                                                         \
    batch<float, avx2> halfx##I;                                                                                                                        \
    {                                                                                                                                                   \
        auto tmp1 = _mm512_shuffle_f32x4(a, b, _MM_SHUFFLE(2, 0, 2, 0));                                                                                \
        auto tmp2 = _mm512_shuffle_f32x4(a, b, _MM_SHUFFLE(3, 1, 3, 1));                                                                                \
                                                                                                                                                        \
        auto resx1 = _mm512_add_ps(tmp1, tmp2);                                                                                                         \
                                                                                                                                                        \
        auto tmp3 = _mm512_shuffle_f32x4(c, d, _MM_SHUFFLE(2, 0, 2, 0));                                                                                \
        auto tmp4 = _mm512_shuffle_f32x4(c, d, _MM_SHUFFLE(3, 1, 3, 1));                                                                                \
                                                                                                                                                        \
        auto resx2 = _mm512_add_ps(tmp3, tmp4);                                                                                                         \
                                                                                                                                                        \
        auto tmp5 = _mm512_shuffle_ps(resx1, resx2, _MM_SHUFFLE(2, 0, 2, 0));                                                                           \
        auto tmp6 = _mm512_shuffle_ps(resx1, resx2, _MM_SHUFFLE(3, 1, 3, 1));                                                                           \
                                                                                                                                                        \
        auto resx3 = _mm512_add_ps(tmp5, tmp6);                                                                                                         \
                                                                                                                                                        \
        halfx##I = _mm256_hadd_ps(_mm256_insertf128_ps(_mm256_castps128_ps256(_mm512_extractf32x4_ps(resx3, 0)), _mm512_extractf32x4_ps(resx3, 1), 1),  \
                                  _mm256_insertf128_ps(_mm256_castps128_ps256(_mm512_extractf32x4_ps(resx3, 2)), _mm512_extractf32x4_ps(resx3, 3), 1)); \
    }

            XSIMD_AVX512_HADDP_STEP2(0, res0, res1, res2, res3);
            XSIMD_AVX512_HADDP_STEP2(1, res4, res5, res6, res7);

#undef XSIMD_AVX512_HADDP_STEP2

            auto concat = _mm512_castps256_ps512(halfx0);
            concat = _mm512_castpd_ps(_mm512_insertf64x4(_mm512_castps_pd(concat), _mm256_castps_pd(halfx1), 1));
            return concat;
        }

        template <class A>
        XSIMD_INLINE batch<double, A> haddp(batch<double, A> const* row, requires_arch<avx512f>) noexcept
        {
#define step1(I, a, b)                                                   \
    batch<double, avx512f> res##I;                                       \
    {                                                                    \
        auto tmp1 = _mm512_shuffle_f64x2(a, b, _MM_SHUFFLE(1, 0, 1, 0)); \
        auto tmp2 = _mm512_shuffle_f64x2(a, b, _MM_SHUFFLE(3, 2, 3, 2)); \
        res##I = _mm512_add_pd(tmp1, tmp2);                              \
    }

            step1(1, row[0], row[2]);
            step1(2, row[4], row[6]);
            step1(3, row[1], row[3]);
            step1(4, row[5], row[7]);

#undef step1

            auto tmp5 = _mm512_shuffle_f64x2(res1, res2, _MM_SHUFFLE(2, 0, 2, 0));
            auto tmp6 = _mm512_shuffle_f64x2(res1, res2, _MM_SHUFFLE(3, 1, 3, 1));

            auto resx1 = _mm512_add_pd(tmp5, tmp6);

            auto tmp7 = _mm512_shuffle_f64x2(res3, res4, _MM_SHUFFLE(2, 0, 2, 0));
            auto tmp8 = _mm512_shuffle_f64x2(res3, res4, _MM_SHUFFLE(3, 1, 3, 1));

            auto resx2 = _mm512_add_pd(tmp7, tmp8);

            auto tmpx = _mm512_shuffle_pd(resx1, resx2, 0b00000000);
            auto tmpy = _mm512_shuffle_pd(resx1, resx2, 0b11111111);

            return _mm512_add_pd(tmpx, tmpy);
        }

        // incr_if
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> incr_if(batch<T, A> const& self, batch_bool<T, A> const& mask, requires_arch<avx512f>) noexcept
        {

            XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_mask_add_epi32(self, mask.data, self, _mm512_set1_epi32(1));
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_mask_add_epi64(self, mask.data, self, _mm512_set1_epi64(1));
            }
            else
            {
                return incr_if(self, mask, generic {});
            }
        }

        // insert
        template <class A, size_t I>
        XSIMD_INLINE batch<float, A> insert(batch<float, A> const& self, float val, index<I>, requires_arch<avx512f>) noexcept
        {

            int32_t tmp = bit_cast<int32_t>(val);
            return _mm512_castsi512_ps(_mm512_mask_set1_epi32(_mm512_castps_si512(self), __mmask16(1 << (I & 15)), tmp));
        }

        template <class A, size_t I>
        XSIMD_INLINE batch<double, A> insert(batch<double, A> const& self, double val, index<I>, requires_arch<avx512f>) noexcept
        {
            int64_t tmp = bit_cast<int64_t>(val);
            return _mm512_castsi512_pd(_mm512_mask_set1_epi64(_mm512_castpd_si512(self), __mmask8(1 << (I & 7)), tmp));
        }
        template <class A, class T, size_t I, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> insert(batch<T, A> const& self, T val, index<I> pos, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_mask_set1_epi32(self, __mmask16(1 << (I & 15)), val);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_mask_set1_epi64(self, __mmask8(1 << (I & 7)), val);
            }
            else
            {
                return insert(self, val, pos, generic {});
            }
        }

        // isnan
        template <class A>
        XSIMD_INLINE batch_bool<float, A> isnan(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, self, _CMP_UNORD_Q);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> isnan(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, self, _CMP_UNORD_Q);
        }

        // ldexp
        template <class A>
        XSIMD_INLINE batch<float, A> ldexp(const batch<float, A>& self, const batch<as_integer_t<float>, A>& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_scalef_ps(self, _mm512_cvtepi32_ps(other));
        }

        template <class A>
        XSIMD_INLINE batch<double, A> ldexp(const batch<double, A>& self, const batch<as_integer_t<double>, A>& other, requires_arch<avx512f>) noexcept
        {
            // FIXME: potential data loss here when converting other elements to
            // int32 before converting them back to double.
            __m512d adjusted_index = _mm512_cvtepi32_pd(_mm512_cvtepi64_epi32(other));
            return _mm512_scalef_pd(self, adjusted_index);
        }

        // le
        template <class A>
        XSIMD_INLINE batch_bool<float, A> le(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_LE_OQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> le(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_LE_OQ);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> le(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return detail::compare_int_avx512f<A, T, _MM_CMPINT_LE>(self, other);
        }

        namespace detail
        {
            // Adapted from https://github.com/serge-sans-paille/fast-bitset-from-bool-array
            // Generate a bitset from an array of boolean.
            XSIMD_INLINE unsigned char tobitset(unsigned char unpacked[8])
            {
                uint64_t data;
                memcpy(&data, unpacked, sizeof(uint64_t));

                const uint64_t magic = (0x80 + 0x4000 + 0x200000 + 0x10000000 + 0x0800000000 + 0x040000000000 + 0x02000000000000 + 0x0100000000000000);

                unsigned char res = ((data * magic) >> 56) & 0xFF;
                return res;
            }
        }

        // load mask
        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            constexpr auto size = batch_bool<T, A>::size;
            constexpr auto iter = size / 8;
            static_assert((size % 8) == 0, "incorrect size of bool batch");
            register_type mask = 0;
            for (std::size_t i = 0; i < iter; ++i)
            {
                unsigned char block = detail::tobitset((unsigned char*)mem + i * 8);
                mask |= (register_type(block) << (i * 8));
            }
            return mask;
        }

        // load_aligned
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> load_aligned(T const* mem, convert<T>, requires_arch<avx512f>) noexcept
        {
            return _mm512_load_si512((__m512i const*)mem);
        }
        template <class A>
        XSIMD_INLINE batch<float, A> load_aligned(float const* mem, convert<float>, requires_arch<avx512f>) noexcept
        {
            return _mm512_load_ps(mem);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> load_aligned(double const* mem, convert<double>, requires_arch<avx512f>) noexcept
        {
            return _mm512_load_pd(mem);
        }

        // load_complex
        namespace detail
        {
            template <class A>
            XSIMD_INLINE batch<std::complex<float>, A> load_complex(batch<float, A> const& hi, batch<float, A> const& lo, requires_arch<avx512f>) noexcept
            {
                __m512i real_idx = _mm512_setr_epi32(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
                __m512i imag_idx = _mm512_setr_epi32(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
                auto real = _mm512_permutex2var_ps(hi, real_idx, lo);
                auto imag = _mm512_permutex2var_ps(hi, imag_idx, lo);
                return { real, imag };
            }
            template <class A>
            XSIMD_INLINE batch<std::complex<double>, A> load_complex(batch<double, A> const& hi, batch<double, A> const& lo, requires_arch<avx512f>) noexcept
            {
                __m512i real_idx = _mm512_setr_epi64(0, 2, 4, 6, 8, 10, 12, 14);
                __m512i imag_idx = _mm512_setr_epi64(1, 3, 5, 7, 9, 11, 13, 15);
                auto real = _mm512_permutex2var_pd(hi, real_idx, lo);
                auto imag = _mm512_permutex2var_pd(hi, imag_idx, lo);
                return { real, imag };
            }
        }

        // load_unaligned
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> load_unaligned(T const* mem, convert<T>, requires_arch<avx512f>) noexcept
        {
            return _mm512_loadu_si512((__m512i const*)mem);
        }
        template <class A>
        XSIMD_INLINE batch<float, A> load_unaligned(float const* mem, convert<float>, requires_arch<avx512f>) noexcept
        {
            return _mm512_loadu_ps(mem);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> load_unaligned(double const* mem, convert<double>, requires_arch<avx512f>) noexcept
        {
            return _mm512_loadu_pd(mem);
        }

        // lt
        template <class A>
        XSIMD_INLINE batch_bool<float, A> lt(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_LT_OQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> lt(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_LT_OQ);
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> lt(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return detail::compare_int_avx512f<A, T, _MM_CMPINT_LT>(self, other);
        }

        // mask
        template <class A, class T>
        XSIMD_INLINE uint64_t mask(batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return self.data;
        }

        // max
        template <class A>
        XSIMD_INLINE batch<float, A> max(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_max_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> max(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_max_pd(self, other);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> max(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            if (std::is_signed<T>::value)
            {
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_max_epi32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_max_epi64(self, other);
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                              { return max(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                              self, other);
                }
            }
            else
            {
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_max_epu32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_max_epu64(self, other);
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                              { return max(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                              self, other);
                }
            }
        }

        // min
        template <class A>
        XSIMD_INLINE batch<float, A> min(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_min_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> min(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_min_pd(self, other);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> min(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            if (std::is_signed<T>::value)
            {
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_min_epi32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_min_epi64(self, other);
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                              { return min(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                              self, other);
                }
            }
            else
            {
                XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
                {
                    return _mm512_min_epu32(self, other);
                }
                else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
                {
                    return _mm512_min_epu64(self, other);
                }
                else
                {
                    return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                              { return min(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                              self, other);
                }
            }
        }

        // mul
        template <class A>
        XSIMD_INLINE batch<float, A> mul(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_mul_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> mul(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_mul_pd(self, other);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> mul(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_mullo_epi32(self, other);
            }
            else
            {
                return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                          { return mul(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                          self, other);
            }
        }

        // nearbyint
        template <class A>
        XSIMD_INLINE batch<float, A> nearbyint(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_round_ps(self, _MM_FROUND_TO_NEAREST_INT, _MM_FROUND_CUR_DIRECTION);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> nearbyint(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_round_pd(self, _MM_FROUND_TO_NEAREST_INT, _MM_FROUND_CUR_DIRECTION);
        }

        // nearbyint_as_int
        template <class A>
        XSIMD_INLINE batch<int32_t, A> nearbyint_as_int(batch<float, A> const& self,
                                                        requires_arch<avx512f>) noexcept
        {
            return _mm512_cvtps_epi32(self);
        }

        // neg
        template <class A, class T>
        XSIMD_INLINE batch<T, A> neg(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return 0 - self;
        }

        // neq
        template <class A>
        XSIMD_INLINE batch_bool<float, A> neq(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_ps_mask(self, other, _CMP_NEQ_UQ);
        }
        template <class A>
        XSIMD_INLINE batch_bool<double, A> neq(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_cmp_pd_mask(self, other, _CMP_NEQ_UQ);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch_bool<T, A> neq(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            return ~(self == other);
        }

        template <class A, class T>
        XSIMD_INLINE batch_bool<T, A> neq(batch_bool<T, A> const& self, batch_bool<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            return register_type(self.data ^ other.data);
        }

        // reciprocal
        template <class A>
        XSIMD_INLINE batch<float, A>
        reciprocal(batch<float, A> const& self,
                   kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_rcp14_ps(self);
        }

        template <class A>
        XSIMD_INLINE batch<double, A>
        reciprocal(batch<double, A> const& self,
                   kernel::requires_arch<avx512f>) noexcept
        {
            return _mm512_rcp14_pd(self);
        }

        // reduce_add
        template <class A>
        XSIMD_INLINE float reduce_add(batch<float, A> const& rhs, requires_arch<avx512f>) noexcept
        {
            __m128 tmp1 = _mm512_extractf32x4_ps(rhs, 0);
            __m128 tmp2 = _mm512_extractf32x4_ps(rhs, 1);
            __m128 tmp3 = _mm512_extractf32x4_ps(rhs, 2);
            __m128 tmp4 = _mm512_extractf32x4_ps(rhs, 3);
            __m128 res1 = _mm_add_ps(tmp1, tmp2);
            __m128 res2 = _mm_add_ps(tmp3, tmp4);
            __m128 res3 = _mm_add_ps(res1, res2);
            return reduce_add(batch<float, sse4_2>(res3), sse4_2 {});
        }
        template <class A>
        XSIMD_INLINE double reduce_add(batch<double, A> const& rhs, requires_arch<avx512f>) noexcept
        {
            __m256d tmp1 = _mm512_extractf64x4_pd(rhs, 1);
            __m256d tmp2 = _mm512_extractf64x4_pd(rhs, 0);
            __m256d res1 = _mm256_add_pd(tmp1, tmp2);
            return reduce_add(batch<double, avx2>(res1), avx2 {});
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE T reduce_add(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            __m256i low, high;
            detail::split_avx512(self, low, high);
            batch<T, avx2> blow(low), bhigh(high);
            return reduce_add(blow, avx2 {}) + reduce_add(bhigh, avx2 {});
        }

        // reduce_max
        template <class A, class T, class _ = typename std::enable_if<(sizeof(T) == 1), void>::type>
        XSIMD_INLINE T reduce_max(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            constexpr batch_constant<uint64_t, A, 5, 6, 7, 8, 0, 0, 0, 0> mask;
            batch<T, A> step = _mm512_permutexvar_epi64(mask.as_batch(), self);
            batch<T, A> acc = max(self, step);
            __m256i low = _mm512_castsi512_si256(acc);
            return reduce_max(batch<T, avx2>(low));
        }

        // reduce_min
        template <class A, class T, class _ = typename std::enable_if<(sizeof(T) == 1), void>::type>
        XSIMD_INLINE T reduce_min(batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            constexpr batch_constant<uint64_t, A, 5, 6, 7, 8, 0, 0, 0, 0> mask;
            batch<T, A> step = _mm512_permutexvar_epi64(mask.as_batch(), self);
            batch<T, A> acc = min(self, step);
            __m256i low = _mm512_castsi512_si256(acc);
            return reduce_min(batch<T, avx2>(low));
        }

        // rsqrt
        template <class A>
        XSIMD_INLINE batch<float, A> rsqrt(batch<float, A> const& val, requires_arch<avx512f>) noexcept
        {
            return _mm512_rsqrt14_ps(val);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> rsqrt(batch<double, A> const& val, requires_arch<avx512f>) noexcept
        {
            return _mm512_rsqrt14_pd(val);
        }

        // sadd
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> sadd(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            if (std::is_signed<T>::value)
            {
                auto mask = other < 0;
                auto self_pos_branch = min(std::numeric_limits<T>::max() - other, self);
                auto self_neg_branch = max(std::numeric_limits<T>::min() - other, self);
                return other + select(mask, self_neg_branch, self_pos_branch);
            }
            else
            {
                const auto diffmax = std::numeric_limits<T>::max() - self;
                const auto mindiff = min(diffmax, other);
                return self + mindiff;
            }
        }

        // scatter
        template <class A, class T,
                  class = typename std::enable_if<std::is_same<uint32_t, T>::value || std::is_same<int32_t, T>::value, void>::type>
        XSIMD_INLINE void scatter(batch<T, A> const& src, T* dst,
                                  batch<int32_t, A> const& index,
                                  kernel::requires_arch<avx512f>) noexcept
        {
            _mm512_i32scatter_epi32(dst, index, src, sizeof(T));
        }

        template <class A, class T,
                  class = typename std::enable_if<std::is_same<uint64_t, T>::value || std::is_same<int64_t, T>::value, void>::type>
        XSIMD_INLINE void scatter(batch<T, A> const& src, T* dst,
                                  batch<int64_t, A> const& index,
                                  kernel::requires_arch<avx512f>) noexcept
        {
            _mm512_i64scatter_epi64(dst, index, src, sizeof(T));
        }

        template <class A>
        XSIMD_INLINE void scatter(batch<float, A> const& src, float* dst,
                                  batch<int32_t, A> const& index,
                                  kernel::requires_arch<avx512f>) noexcept
        {
            _mm512_i32scatter_ps(dst, index, src, sizeof(float));
        }

        template <class A>
        XSIMD_INLINE void scatter(batch<double, A> const& src, double* dst,
                                  batch<int64_t, A> const& index,
                                  kernel::requires_arch<avx512f>) noexcept
        {
            _mm512_i64scatter_pd(dst, index, src, sizeof(double));
        }

        // select
        template <class A>
        XSIMD_INLINE batch<float, A> select(batch_bool<float, A> const& cond, batch<float, A> const& true_br, batch<float, A> const& false_br, requires_arch<avx512f>) noexcept
        {
            return _mm512_mask_blend_ps(cond, false_br, true_br);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> select(batch_bool<double, A> const& cond, batch<double, A> const& true_br, batch<double, A> const& false_br, requires_arch<avx512f>) noexcept
        {
            return _mm512_mask_blend_pd(cond, false_br, true_br);
        }

        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> select(batch_bool<T, A> const& cond, batch<T, A> const& true_br, batch<T, A> const& false_br, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                alignas(avx2::alignment()) uint8_t buffer[64];
                // FIXME: ultra inefficient
                for (int i = 0; i < 64; ++i)
                    buffer[i] = cond.data & (1ull << i) ? 0xFF : 0;
                __m256i cond_low = batch<uint8_t, avx2>::load_aligned(&buffer[0]);
                __m256i cond_hi = batch<uint8_t, avx2>::load_aligned(&buffer[32]);

                __m256i true_low, true_hi;
                detail::split_avx512(true_br, true_low, true_hi);

                __m256i false_low, false_hi;
                detail::split_avx512(false_br, false_low, false_hi);

                __m256i res_low = select(batch_bool<T, avx2>(cond_low), batch<T, avx2>(true_low), batch<T, avx2>(false_low), avx2 {});
                __m256i res_hi = select(batch_bool<T, avx2>(cond_hi), batch<T, avx2>(true_hi), batch<T, avx2>(false_hi), avx2 {});
                return detail::merge_avx(res_low, res_hi);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                __m256i cond_low = _mm512_maskz_cvtepi32_epi16((uint64_t)cond.data & 0xFFFF, _mm512_set1_epi32(~0));
                __m256i cond_hi = _mm512_maskz_cvtepi32_epi16((uint64_t)cond.data >> 16, _mm512_set1_epi32(~0));

                __m256i true_low, true_hi;
                detail::split_avx512(true_br, true_low, true_hi);

                __m256i false_low, false_hi;
                detail::split_avx512(false_br, false_low, false_hi);

                __m256i res_low = select(batch_bool<T, avx2>(cond_low), batch<T, avx2>(true_low), batch<T, avx2>(false_low), avx2 {});
                __m256i res_hi = select(batch_bool<T, avx2>(cond_hi), batch<T, avx2>(true_hi), batch<T, avx2>(false_hi), avx2 {});
                return detail::merge_avx(res_low, res_hi);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_mask_blend_epi32(cond, false_br, true_br);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_mask_blend_epi64(cond, false_br, true_br);
            }
            else
            {
                assert(false && "unsupported arch/type combination");
                return {};
            }
        }

        template <class A, class T, bool... Values, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> select(batch_bool_constant<T, A, Values...> const&, batch<T, A> const& true_br, batch<T, A> const& false_br, requires_arch<avx512f>) noexcept
        {
            return select(batch_bool<T, A> { Values... }, true_br, false_br, avx512f {});
        }

        namespace detail
        {
            template <class T>
            using enable_signed_integer_t = typename std::enable_if<std::is_integral<T>::value && std::is_signed<T>::value,
                                                                    int>::type;

            template <class T>
            using enable_unsigned_integer_t = typename std::enable_if<std::is_integral<T>::value && std::is_unsigned<T>::value,
                                                                      int>::type;
        }

        // set
        template <class A>
        XSIMD_INLINE batch<float, A> set(batch<float, A> const&, requires_arch<avx512f>, float v0, float v1, float v2, float v3, float v4, float v5, float v6, float v7, float v8, float v9, float v10, float v11, float v12, float v13, float v14, float v15) noexcept
        {
            return _mm512_setr_ps(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15);
        }

        template <class A>
        XSIMD_INLINE batch<double, A> set(batch<double, A> const&, requires_arch<avx512f>, double v0, double v1, double v2, double v3, double v4, double v5, double v6, double v7) noexcept
        {
            return _mm512_setr_pd(v0, v1, v2, v3, v4, v5, v6, v7);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7) noexcept
        {
            return _mm512_set_epi64(v7, v6, v5, v4, v3, v2, v1, v0);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7,
                                     T v8, T v9, T v10, T v11, T v12, T v13, T v14, T v15) noexcept
        {
            return _mm512_setr_epi32(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15);
        }
        template <class A, class T, detail::enable_signed_integer_t<T> = 0>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7,
                                     T v8, T v9, T v10, T v11, T v12, T v13, T v14, T v15,
                                     T v16, T v17, T v18, T v19, T v20, T v21, T v22, T v23,
                                     T v24, T v25, T v26, T v27, T v28, T v29, T v30, T v31) noexcept
        {
#if defined(__clang__) || __GNUC__
            return __extension__(__m512i)(__v32hi) {
                v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31
            };
#else
            return _mm512_set_epi16(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                                    v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31);
#endif
        }

        template <class A, class T, detail::enable_unsigned_integer_t<T> = 0>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7,
                                     T v8, T v9, T v10, T v11, T v12, T v13, T v14, T v15,
                                     T v16, T v17, T v18, T v19, T v20, T v21, T v22, T v23,
                                     T v24, T v25, T v26, T v27, T v28, T v29, T v30, T v31) noexcept
        {
#if defined(__clang__) || __GNUC__
            return __extension__(__m512i)(__v32hu) {
                v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31
            };
#else
            return _mm512_set_epi16(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                                    v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31);
#endif
        }

        template <class A, class T, detail::enable_signed_integer_t<T> = 0>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7,
                                     T v8, T v9, T v10, T v11, T v12, T v13, T v14, T v15,
                                     T v16, T v17, T v18, T v19, T v20, T v21, T v22, T v23,
                                     T v24, T v25, T v26, T v27, T v28, T v29, T v30, T v31,
                                     T v32, T v33, T v34, T v35, T v36, T v37, T v38, T v39,
                                     T v40, T v41, T v42, T v43, T v44, T v45, T v46, T v47,
                                     T v48, T v49, T v50, T v51, T v52, T v53, T v54, T v55,
                                     T v56, T v57, T v58, T v59, T v60, T v61, T v62, T v63) noexcept
        {

#if defined(__clang__) || __GNUC__
            return __extension__(__m512i)(__v64qi) {
                v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31,
                v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47,
                v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63
            };
#else
            return _mm512_set_epi8(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                                   v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31,
                                   v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47,
                                   v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63);
#endif
        }
        template <class A, class T, detail::enable_unsigned_integer_t<T> = 0>
        XSIMD_INLINE batch<T, A> set(batch<T, A> const&, requires_arch<avx512f>, T v0, T v1, T v2, T v3, T v4, T v5, T v6, T v7,
                                     T v8, T v9, T v10, T v11, T v12, T v13, T v14, T v15,
                                     T v16, T v17, T v18, T v19, T v20, T v21, T v22, T v23,
                                     T v24, T v25, T v26, T v27, T v28, T v29, T v30, T v31,
                                     T v32, T v33, T v34, T v35, T v36, T v37, T v38, T v39,
                                     T v40, T v41, T v42, T v43, T v44, T v45, T v46, T v47,
                                     T v48, T v49, T v50, T v51, T v52, T v53, T v54, T v55,
                                     T v56, T v57, T v58, T v59, T v60, T v61, T v62, T v63) noexcept
        {

#if defined(__clang__) || __GNUC__
            return __extension__(__m512i)(__v64qu) {
                v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31,
                v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47,
                v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63
            };
#else
            return _mm512_set_epi8(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15,
                                   v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31,
                                   v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47,
                                   v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63);
#endif
        }

        template <class A, class T, class... Values>
        XSIMD_INLINE batch_bool<T, A> set(batch_bool<T, A> const&, requires_arch<avx512f>, Values... values) noexcept
        {
            static_assert(sizeof...(Values) == batch_bool<T, A>::size, "consistent init");
            using register_type = typename batch_bool<T, A>::register_type;
            register_type r = 0;
            unsigned shift = 0;
            (void)std::initializer_list<register_type> { (r |= register_type(values ? 1 : 0) << (shift++))... };
            return r;
        }

        // shuffle
        template <class A, class ITy, ITy I0, ITy I1, ITy I2, ITy I3, ITy I4, ITy I5, ITy I6, ITy I7, ITy I8, ITy I9, ITy I10, ITy I11, ITy I12, ITy I13, ITy I14, ITy I15>
        XSIMD_INLINE batch<float, A> shuffle(batch<float, A> const& x, batch<float, A> const& y,
                                             batch_constant<ITy, A, I0, I1, I2, I3, I4, I5, I6, I7, I8, I9, I10, I11, I12, I13, I14, I15> mask,
                                             requires_arch<avx512f>) noexcept
        {
            constexpr uint32_t smask = (I0 & 0x3) | ((I1 & 0x3) << 2) | ((I2 & 0x3) << 4) | ((I3 & 0x3) << 6);

            // shuffle within lane
            if ((I4 == I0 + 4) && (I5 == I1 + 4) && (I6 == I2 + 4) && (I7 == I3 + 4) && (I8 == I0 + 8) && (I9 == I1 + 8) && (I10 == I2 + 8) && (I11 == I3 + 8) && (I12 == I0 + 12) && (I13 == I1 + 12) && (I14 == I2 + 12) && (I15 == I3 + 12) && I0 < 4 && I1 < 4 && I2 >= 16 && I2 < 20 && I3 >= 16 && I3 < 20)
                return _mm512_shuffle_ps(x, y, smask);

            // shuffle within opposite lane
            if ((I4 == I0 + 4) && (I5 == I1 + 4) && (I6 == I2 + 4) && (I7 == I3 + 4) && (I8 == I0 + 8) && (I9 == I1 + 8) && (I10 == I2 + 8) && (I11 == I3 + 8) && (I12 == I0 + 12) && (I13 == I1 + 12) && (I14 == I2 + 12) && (I15 == I3 + 12) && I2 < 4 && I3 < 4 && I0 >= 16 && I0 < 20 && I1 >= 16 && I1 < 20)
                return _mm512_shuffle_ps(y, x, smask);

            return shuffle(x, y, mask, generic {});
        }

        template <class A, class ITy, ITy I0, ITy I1, ITy I2, ITy I3, ITy I4, ITy I5, ITy I6, ITy I7>
        XSIMD_INLINE batch<double, A> shuffle(batch<double, A> const& x, batch<double, A> const& y, batch_constant<ITy, A, I0, I1, I2, I3, I4, I5, I6, I7> mask, requires_arch<avx512f>) noexcept
        {
            constexpr uint32_t smask = (I0 & 0x1) | ((I1 & 0x1) << 1) | ((I2 & 0x1) << 2) | ((I3 & 0x1) << 3) | ((I4 & 0x1) << 4) | ((I5 & 0x1) << 5) | ((I6 & 0x1) << 6) | ((I7 & 0x1) << 7);
            // shuffle within lane
            if (I0 < 2 && I1 >= 8 && I1 < 10 && I2 >= 2 && I2 < 4 && I3 >= 10 && I3 < 12 && I4 >= 4 && I4 < 6 && I5 >= 12 && I5 < 14 && I6 >= 6 && I6 < 8 && I7 >= 14)
                return _mm512_shuffle_pd(x, y, smask);

            // shuffle within opposite lane
            if (I1 < 2 && I0 >= 8 && I0 < 10 && I3 >= 2 && I3 < 4 && I2 >= 10 && I2 < 12 && I5 >= 4 && I5 < 6 && I4 >= 12 && I4 < 14 && I7 >= 6 && I7 < 8 && I6 >= 14)
                return _mm512_shuffle_pd(y, x, smask);

            return shuffle(x, y, mask, generic {});
        }

        // slide_left
        template <size_t N, class A, class T>
        XSIMD_INLINE batch<T, A> slide_left(batch<T, A> const&, requires_arch<avx512f>) noexcept
        {
            static_assert(N == 0xDEAD, "not implemented yet");
            return {};
        }

        // slide_right
        template <size_t N, class A, class T>
        XSIMD_INLINE batch<T, A> slide_right(batch<T, A> const&, requires_arch<avx512f>) noexcept
        {
            static_assert(N == 0xDEAD, "not implemented yet");
            return {};
        }

        // sqrt
        template <class A>
        XSIMD_INLINE batch<float, A> sqrt(batch<float, A> const& val, requires_arch<avx512f>) noexcept
        {
            return _mm512_sqrt_ps(val);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> sqrt(batch<double, A> const& val, requires_arch<avx512f>) noexcept
        {
            return _mm512_sqrt_pd(val);
        }

        // ssub
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> ssub(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            if (std::is_signed<T>::value)
            {
                return sadd(self, -other);
            }
            else
            {
                const auto diff = min(self, other);
                return self - diff;
            }
        }

        // store
        template <class T, class A>
        XSIMD_INLINE void store(batch_bool<T, A> const& self, bool* mem, requires_arch<avx512f>) noexcept
        {
            using register_type = typename batch_bool<T, A>::register_type;
            constexpr auto size = batch_bool<T, A>::size;
            for (std::size_t i = 0; i < size; ++i)
                mem[i] = self.data & (register_type(1) << i);
        }

        // store_aligned
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE void store_aligned(T* mem, batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_store_si512((__m512i*)mem, self);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE void store_aligned(T* mem, batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_store_si512((__m512i*)mem, self);
        }
        template <class A>
        XSIMD_INLINE void store_aligned(float* mem, batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_store_ps(mem, self);
        }
        template <class A>
        XSIMD_INLINE void store_aligned(double* mem, batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_store_pd(mem, self);
        }

        // store_unaligned
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE void store_unaligned(T* mem, batch<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_storeu_si512((__m512i*)mem, self);
        }
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE void store_unaligned(T* mem, batch_bool<T, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_storeu_si512((__m512i*)mem, self);
        }
        template <class A>
        XSIMD_INLINE void store_unaligned(float* mem, batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_storeu_ps(mem, self);
        }
        template <class A>
        XSIMD_INLINE void store_unaligned(double* mem, batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_storeu_pd(mem, self);
        }

        // sub
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A> sub(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                          { return sub(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                          self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                return detail::fwd_to_avx([](__m256i s, __m256i o) noexcept
                                          { return sub(batch<T, avx2>(s), batch<T, avx2>(o)); },
                                          self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                return _mm512_sub_epi32(self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                return _mm512_sub_epi64(self, other);
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
        }
        template <class A>
        XSIMD_INLINE batch<float, A> sub(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_sub_ps(self, other);
        }
        template <class A>
        XSIMD_INLINE batch<double, A> sub(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            return _mm512_sub_pd(self, other);
        }

        // swizzle (dynamic version)
        template <class A>
        XSIMD_INLINE batch<float, A> swizzle(batch<float, A> const& self, batch<uint32_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_permutexvar_ps(mask, self);
        }

        template <class A>
        XSIMD_INLINE batch<double, A> swizzle(batch<double, A> const& self, batch<uint64_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_permutexvar_pd(mask, self);
        }

        template <class A>
        XSIMD_INLINE batch<uint64_t, A> swizzle(batch<uint64_t, A> const& self, batch<uint64_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_permutexvar_epi64(mask, self);
        }

        template <class A>
        XSIMD_INLINE batch<int64_t, A> swizzle(batch<int64_t, A> const& self, batch<uint64_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return bitwise_cast<int64_t>(swizzle(bitwise_cast<uint64_t>(self), mask, avx512f {}));
        }

        template <class A>
        XSIMD_INLINE batch<uint32_t, A> swizzle(batch<uint32_t, A> const& self, batch<uint32_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return _mm512_permutexvar_epi32(mask, self);
        }

        template <class A>
        XSIMD_INLINE batch<int32_t, A> swizzle(batch<int32_t, A> const& self, batch<uint32_t, A> mask, requires_arch<avx512f>) noexcept
        {
            return bitwise_cast<int32_t>(swizzle(bitwise_cast<uint32_t>(self), mask, avx512f {}));
        }

        // swizzle (constant version)
        template <class A, uint32_t... Vs>
        XSIMD_INLINE batch<float, A> swizzle(batch<float, A> const& self, batch_constant<uint32_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        template <class A, uint64_t... Vs>
        XSIMD_INLINE batch<double, A> swizzle(batch<double, A> const& self, batch_constant<uint64_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        template <class A, uint64_t... Vs>
        XSIMD_INLINE batch<uint64_t, A> swizzle(batch<uint64_t, A> const& self, batch_constant<uint64_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        template <class A, uint64_t... Vs>
        XSIMD_INLINE batch<int64_t, A> swizzle(batch<int64_t, A> const& self, batch_constant<uint64_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        template <class A, uint32_t... Vs>
        XSIMD_INLINE batch<uint32_t, A> swizzle(batch<uint32_t, A> const& self, batch_constant<uint32_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        template <class A, uint32_t... Vs>
        XSIMD_INLINE batch<int32_t, A> swizzle(batch<int32_t, A> const& self, batch_constant<uint32_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return swizzle(self, mask.as_batch(), avx512f {});
        }

        namespace detail
        {
            template <class T, class A, T... Idx>
            struct is_pair_of_contiguous_indices;

            template <class T, class A>
            struct is_pair_of_contiguous_indices<T, A> : std::true_type
            {
            };

            template <class T, class A, T Idx0, T Idx1, T... Idx>
            struct is_pair_of_contiguous_indices<T, A, Idx0, Idx1, Idx...> : std::conditional<(Idx0 % 2 == 0) && (Idx0 + 1 == Idx1), is_pair_of_contiguous_indices<T, A, Idx...>, std::false_type>::type
            {
            };

            template <class A, uint16_t I0, uint16_t I1, uint16_t I2, uint16_t I3, uint16_t I4, uint16_t I5, uint16_t I6, uint16_t I7,
                      uint16_t I8, uint16_t I9, uint16_t I10, uint16_t I11, uint16_t I12, uint16_t I13, uint16_t I14, uint16_t I15,
                      uint16_t I16, uint16_t I17, uint16_t I18, uint16_t I19, uint16_t I20, uint16_t I21, uint16_t I22, uint16_t I23,
                      uint16_t I24, uint16_t I25, uint16_t I26, uint16_t I27, uint16_t I28, uint16_t I29, uint16_t I30, uint16_t I31>
            struct fold_batch_constant
            {
                using type = batch_constant<uint32_t, A, I0 / 2, I2 / 2, I4 / 2, I6 / 2, I8 / 2, I10 / 2, I12 / 2, I14 / 2,
                                            I16 / 2, I18 / 2, I20 / 2, I22 / 2, I24 / 2, I26 / 2, I28 / 2, I30 / 2>;
            };

        }

        template <class A, uint16_t... Idx, class _ = typename std::enable_if<detail::is_pair_of_contiguous_indices<uint16_t, A, Idx...>::value, void>::type>
        XSIMD_INLINE batch<uint16_t, A> swizzle(batch<uint16_t, A> const& self, batch_constant<uint16_t, A, Idx...>, requires_arch<avx512f>) noexcept
        {
            constexpr typename detail::fold_batch_constant<A, Idx...>::type mask32;
            return _mm512_permutexvar_epi32(static_cast<batch<uint32_t, A>>(mask32), self);
        }

        template <class A>
        XSIMD_INLINE batch<uint16_t, A>
        swizzle(batch<uint16_t, A> const& self, batch_constant<uint16_t, A, (uint16_t)1, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1, (uint16_t)0, (uint16_t)1>, requires_arch<avx512f>) noexcept
        {
            // FIXME: this sequence is very inefficient, but it's here to catch
            // a pattern generated by detail::reduce from xsimd_generic_math.hpp.
            // The whole pattern is actually decently folded by GCC and Clang,
            // so bare with it.
            constexpr batch_constant<uint32_t, A, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0> mask32;
            auto tmp = _mm512_permutexvar_epi32(static_cast<batch<uint32_t, A>>(mask32), self);

            alignas(A::alignment()) uint16_t buffer[32];
            _mm512_store_si512((__m512i*)&buffer[0], tmp);
            buffer[0] = buffer[1];
            return _mm512_load_si512(&buffer[0]);
        }

        template <class A, uint16_t... Vs>
        XSIMD_INLINE batch<int16_t, A>
        swizzle(batch<int16_t, A> const& self, batch_constant<uint16_t, A, Vs...> mask, requires_arch<avx512f>) noexcept
        {
            return bitwise_cast<int16_t>(swizzle(bitwise_cast<uint16_t>(self), mask, avx512f {}));
        }

        // transpose
        template <class A>
        XSIMD_INLINE void transpose(batch<uint16_t, A>* matrix_begin, batch<uint16_t, A>* matrix_end, requires_arch<avx512f>) noexcept
        {
            assert((matrix_end - matrix_begin == batch<uint16_t, A>::size) && "correctly sized matrix");
            (void)matrix_end;
            batch<uint16_t, avx2> tmp_lo0[16];
            for (int i = 0; i < 16; ++i)
                tmp_lo0[i] = _mm512_castsi512_si256(matrix_begin[i]);
            transpose(tmp_lo0 + 0, tmp_lo0 + 16, avx2 {});

            batch<uint16_t, avx2> tmp_hi0[16];
            for (int i = 0; i < 16; ++i)
                tmp_hi0[i] = _mm512_castsi512_si256(matrix_begin[16 + i]);
            transpose(tmp_hi0 + 0, tmp_hi0 + 16, avx2 {});

            batch<uint16_t, avx2> tmp_lo1[16];
            for (int i = 0; i < 16; ++i)
                tmp_lo1[i] = _mm512_extracti64x4_epi64(matrix_begin[i], 1);
            transpose(tmp_lo1 + 0, tmp_lo1 + 16, avx2 {});

            batch<uint16_t, avx2> tmp_hi1[16];
            for (int i = 0; i < 16; ++i)
                tmp_hi1[i] = _mm512_extracti64x4_epi64(matrix_begin[16 + i], 1);
            transpose(tmp_hi1 + 0, tmp_hi1 + 16, avx2 {});

            for (int i = 0; i < 16; ++i)
                matrix_begin[i] = detail::merge_avx(tmp_lo0[i], tmp_hi0[i]);
            for (int i = 0; i < 16; ++i)
                matrix_begin[i + 16] = detail::merge_avx(tmp_lo1[i], tmp_hi1[i]);
        }
        template <class A>
        XSIMD_INLINE void transpose(batch<int16_t, A>* matrix_begin, batch<int16_t, A>* matrix_end, requires_arch<avx512f>) noexcept
        {
            return transpose(reinterpret_cast<batch<uint16_t, A>*>(matrix_begin), reinterpret_cast<batch<uint16_t, A>*>(matrix_end), A {});
        }

        template <class A>
        XSIMD_INLINE void transpose(batch<uint8_t, A>* matrix_begin, batch<uint8_t, A>* matrix_end, requires_arch<avx512f>) noexcept
        {
            assert((matrix_end - matrix_begin == batch<uint8_t, A>::size) && "correctly sized matrix");
            (void)matrix_end;
            batch<uint8_t, avx2> tmp_lo0[32];
            for (int i = 0; i < 32; ++i)
                tmp_lo0[i] = _mm512_castsi512_si256(matrix_begin[i]);
            transpose(tmp_lo0 + 0, tmp_lo0 + 32, avx2 {});

            batch<uint8_t, avx2> tmp_hi0[32];
            for (int i = 0; i < 32; ++i)
                tmp_hi0[i] = _mm512_castsi512_si256(matrix_begin[32 + i]);
            transpose(tmp_hi0 + 0, tmp_hi0 + 32, avx2 {});

            batch<uint8_t, avx2> tmp_lo1[32];
            for (int i = 0; i < 32; ++i)
                tmp_lo1[i] = _mm512_extracti64x4_epi64(matrix_begin[i], 1);
            transpose(tmp_lo1 + 0, tmp_lo1 + 32, avx2 {});

            batch<uint8_t, avx2> tmp_hi1[32];
            for (int i = 0; i < 32; ++i)
                tmp_hi1[i] = _mm512_extracti64x4_epi64(matrix_begin[32 + i], 1);
            transpose(tmp_hi1 + 0, tmp_hi1 + 32, avx2 {});

            for (int i = 0; i < 32; ++i)
                matrix_begin[i] = detail::merge_avx(tmp_lo0[i], tmp_hi0[i]);
            for (int i = 0; i < 32; ++i)
                matrix_begin[i + 32] = detail::merge_avx(tmp_lo1[i], tmp_hi1[i]);
        }
        template <class A>
        XSIMD_INLINE void transpose(batch<int8_t, A>* matrix_begin, batch<int8_t, A>* matrix_end, requires_arch<avx512f>) noexcept
        {
            return transpose(reinterpret_cast<batch<uint8_t, A>*>(matrix_begin), reinterpret_cast<batch<uint8_t, A>*>(matrix_end), A {});
        }

        // trunc
        template <class A>
        XSIMD_INLINE batch<float, A>
        trunc(batch<float, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_round_ps(self, _MM_FROUND_TO_ZERO, _MM_FROUND_CUR_DIRECTION);
        }
        template <class A>
        XSIMD_INLINE batch<double, A>
        trunc(batch<double, A> const& self, requires_arch<avx512f>) noexcept
        {
            return _mm512_roundscale_round_pd(self, _MM_FROUND_TO_ZERO, _MM_FROUND_CUR_DIRECTION);
        }

        // zip_hi
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A>
        zip_hi(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            __m512i lo, hi;
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                assert(false && "not implemented yet");
                return {};
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                assert(false && "not implemented yet");
                return {};
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                lo = _mm512_unpacklo_epi32(self, other);
                hi = _mm512_unpackhi_epi32(self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                lo = _mm512_unpacklo_epi64(self, other);
                hi = _mm512_unpackhi_epi64(self, other);
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
            return _mm512_inserti32x4(
                _mm512_inserti32x4(
                    _mm512_inserti32x4(hi, _mm512_extracti32x4_epi32(lo, 2), 0),
                    _mm512_extracti32x4_epi32(lo, 3),
                    2),
                _mm512_extracti32x4_epi32(hi, 2),
                1);
        }
        template <class A>
        XSIMD_INLINE batch<float, A>
        zip_hi(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            auto lo = _mm512_unpacklo_ps(self, other);
            auto hi = _mm512_unpackhi_ps(self, other);
            return _mm512_insertf32x4(
                _mm512_insertf32x4(
                    _mm512_insertf32x4(hi, _mm512_extractf32x4_ps(lo, 2), 0),
                    _mm512_extractf32x4_ps(lo, 3),
                    2),
                _mm512_extractf32x4_ps(hi, 2),
                1);
        }
        template <class A>
        XSIMD_INLINE batch<double, A>
        zip_hi(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            auto lo = _mm512_castpd_ps(_mm512_unpacklo_pd(self, other));
            auto hi = _mm512_castpd_ps(_mm512_unpackhi_pd(self, other));
            return _mm512_castps_pd(_mm512_insertf32x4(
                _mm512_insertf32x4(
                    _mm512_insertf32x4(hi, _mm512_extractf32x4_ps(lo, 2), 0),
                    _mm512_extractf32x4_ps(lo, 3),
                    2),
                _mm512_extractf32x4_ps(hi, 2),
                1));
        }

        // zip_lo
        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
        XSIMD_INLINE batch<T, A>
        zip_lo(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512f>) noexcept
        {
            __m512i lo, hi;
            XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
            {
                assert(false && "not implemented yet");
                return {};
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
            {
                assert(false && "not implemented yet");
                return {};
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
            {
                lo = _mm512_unpacklo_epi32(self, other);
                hi = _mm512_unpackhi_epi32(self, other);
            }
            else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
            {
                lo = _mm512_unpacklo_epi64(self, other);
                hi = _mm512_unpackhi_epi64(self, other);
            }
            else
            {
                assert(false && "unsupported arch/op combination");
                return {};
            }
            return _mm512_inserti32x4(
                _mm512_inserti32x4(
                    _mm512_inserti32x4(lo, _mm512_extracti32x4_epi32(hi, 0), 1),
                    _mm512_extracti32x4_epi32(hi, 1),
                    3),
                _mm512_extracti32x4_epi32(lo, 1),
                2);
        }
        template <class A>
        XSIMD_INLINE batch<float, A>
        zip_lo(batch<float, A> const& self, batch<float, A> const& other, requires_arch<avx512f>) noexcept
        {
            auto lo = _mm512_unpacklo_ps(self, other);
            auto hi = _mm512_unpackhi_ps(self, other);
            return _mm512_insertf32x4(
                _mm512_insertf32x4(
                    _mm512_insertf32x4(lo, _mm512_extractf32x4_ps(hi, 0), 1),
                    _mm512_extractf32x4_ps(hi, 1),
                    3),
                _mm512_extractf32x4_ps(lo, 1),
                2);
        }
        template <class A>
        XSIMD_INLINE batch<double, A>
        zip_lo(batch<double, A> const& self, batch<double, A> const& other, requires_arch<avx512f>) noexcept
        {
            auto lo = _mm512_castpd_ps(_mm512_unpacklo_pd(self, other));
            auto hi = _mm512_castpd_ps(_mm512_unpackhi_pd(self, other));
            return _mm512_castps_pd(_mm512_insertf32x4(
                _mm512_insertf32x4(
                    _mm512_insertf32x4(lo, _mm512_extractf32x4_ps(hi, 0), 1),
                    _mm512_extractf32x4_ps(hi, 1),
                    3),
                _mm512_extractf32x4_ps(lo, 1),
                2));
        }

    }

}

#endif