#pragma once #define IS_LE_MACHINE union u128 { u64 _u64[2]; s64 _s64[2]; class u64_reversed_array_2 { u64 data[2]; public: u64& operator [] (s32 index) { return data[1 - index]; } const u64& operator [] (s32 index) const { return data[1 - index]; } } u64r; u32 _u32[4]; s32 _s32[4]; class u32_reversed_array_4 { u32 data[4]; public: u32& operator [] (s32 index) { return data[3 - index]; } const u32& operator [] (s32 index) const { return data[3 - index]; } } u32r; u16 _u16[8]; s16 _s16[8]; class u16_reversed_array_8 { u16 data[8]; public: u16& operator [] (s32 index) { return data[7 - index]; } const u16& operator [] (s32 index) const { return data[7 - index]; } } u16r; u8 _u8[16]; s8 _s8[16]; class u8_reversed_array_16 { u8 data[16]; public: u8& operator [] (s32 index) { return data[15 - index]; } const u8& operator [] (s32 index) const { return data[15 - index]; } } u8r; float _f[4]; double _d[2]; __m128 vf; __m128i vi; __m128d vd; class bit_array_128 { u64 data[2]; public: class bit_element { u64& data; const u64 mask; public: bit_element(u64& data, const u64 mask) : data(data) , mask(mask) { } force_inline operator bool() const { return (data & mask) != 0; } force_inline bit_element& operator = (const bool right) { if (right) { data |= mask; } else { data &= ~mask; } return *this; } force_inline bit_element& operator = (const bit_element& right) { if (right) { data |= mask; } else { data &= ~mask; } return *this; } }; // Index 0 returns the MSB and index 127 returns the LSB bit_element operator [] (u32 index) { assert(index < 128); #ifdef IS_LE_MACHINE return bit_element(data[1 - (index >> 6)], 0x8000000000000000ull >> (index & 0x3F)); #else return bit_element(data[index >> 6], 0x8000000000000000ull >> (index & 0x3F)); #endif } // Index 0 returns the MSB and index 127 returns the LSB const bool operator [] (u32 index) const { assert(index < 128); #ifdef IS_LE_MACHINE return (data[1 - (index >> 6)] & (0x8000000000000000ull >> (index & 0x3F))) != 0; #else return (data[index >> 6] & (0x8000000000000000ull >> (index & 0x3F))) != 0; #endif } } _bit; //operator u64() const { return _u64[0]; } //operator u32() const { return _u32[0]; } //operator u16() const { return _u16[0]; } //operator u8() const { return _u8[0]; } //operator bool() const { return _u64[0] != 0 || _u64[1] != 0; } static u128 from64(u64 _0, u64 _1 = 0) { u128 ret; ret._u64[0] = _0; ret._u64[1] = _1; return ret; } static u128 from64r(u64 _1, u64 _0 = 0) { return from64(_0, _1); } static u128 from32(u32 _0, u32 _1 = 0, u32 _2 = 0, u32 _3 = 0) { u128 ret; ret._u32[0] = _0; ret._u32[1] = _1; ret._u32[2] = _2; ret._u32[3] = _3; return ret; } static u128 from32r(u32 _3, u32 _2 = 0, u32 _1 = 0, u32 _0 = 0) { return from32(_0, _1, _2, _3); } static u128 from32p(u32 value) { u128 ret; ret.vi = _mm_set1_epi32(static_cast(value)); return ret; } static u128 from16p(u16 value) { u128 ret; ret.vi = _mm_set1_epi16(static_cast(value)); return ret; } static u128 from8p(u8 value) { u128 ret; ret.vi = _mm_set1_epi8(static_cast(value)); return ret; } static u128 fromBit(u32 bit) { u128 ret = {}; ret._bit[bit] = true; return ret; } static u128 fromV(__m128i value) { u128 ret; ret.vi = value; return ret; } static u128 fromF(__m128 value) { u128 ret; ret.vf = value; return ret; } static u128 fromD(__m128d value) { u128 ret; ret.vd = value; return ret; } static force_inline u128 add8(const u128& left, const u128& right) { return fromV(_mm_add_epi8(left.vi, right.vi)); } static force_inline u128 add16(const u128& left, const u128& right) { return fromV(_mm_add_epi16(left.vi, right.vi)); } static force_inline u128 add32(const u128& left, const u128& right) { return fromV(_mm_add_epi32(left.vi, right.vi)); } static force_inline u128 addfs(const u128& left, const u128& right) { return fromF(_mm_add_ps(left.vf, right.vf)); } static force_inline u128 addfd(const u128& left, const u128& right) { return fromD(_mm_add_pd(left.vd, right.vd)); } static force_inline u128 sub8(const u128& left, const u128& right) { return fromV(_mm_sub_epi8(left.vi, right.vi)); } static force_inline u128 sub16(const u128& left, const u128& right) { return fromV(_mm_sub_epi16(left.vi, right.vi)); } static force_inline u128 sub32(const u128& left, const u128& right) { return fromV(_mm_sub_epi32(left.vi, right.vi)); } static force_inline u128 subfs(const u128& left, const u128& right) { return fromF(_mm_sub_ps(left.vf, right.vf)); } static force_inline u128 subfd(const u128& left, const u128& right) { return fromD(_mm_sub_pd(left.vd, right.vd)); } static force_inline u128 maxu8(const u128& left, const u128& right) { return fromV(_mm_max_epu8(left.vi, right.vi)); } static force_inline u128 minu8(const u128& left, const u128& right) { return fromV(_mm_min_epu8(left.vi, right.vi)); } static force_inline u128 eq8(const u128& left, const u128& right) { return fromV(_mm_cmpeq_epi8(left.vi, right.vi)); } static force_inline u128 eq16(const u128& left, const u128& right) { return fromV(_mm_cmpeq_epi16(left.vi, right.vi)); } static force_inline u128 eq32(const u128& left, const u128& right) { return fromV(_mm_cmpeq_epi32(left.vi, right.vi)); } bool operator == (const u128& right) const { return (_u64[0] == right._u64[0]) && (_u64[1] == right._u64[1]); } bool operator != (const u128& right) const { return (_u64[0] != right._u64[0]) || (_u64[1] != right._u64[1]); } force_inline u128 operator | (const u128& right) const { return fromV(_mm_or_si128(vi, right.vi)); } force_inline u128 operator & (const u128& right) const { return fromV(_mm_and_si128(vi, right.vi)); } force_inline u128 operator ^ (const u128& right) const { return fromV(_mm_xor_si128(vi, right.vi)); } u128 operator ~ () const { return from64(~_u64[0], ~_u64[1]); } force_inline bool is_any_1() const // check if any bit is 1 { return _u64[0] || _u64[1]; } force_inline bool is_any_0() const // check if any bit is 0 { return ~_u64[0] || ~_u64[1]; } // result = (~left) & (right) static force_inline u128 andnot(const u128& left, const u128& right) { return fromV(_mm_andnot_si128(left.vi, right.vi)); } void clear() { _u64[1] = _u64[0] = 0; } std::string to_hex() const; std::string to_xyzw() const; static force_inline u128 byteswap(const u128 val) { u128 ret; ret._u64[0] = _byteswap_uint64(val._u64[1]); ret._u64[1] = _byteswap_uint64(val._u64[0]); return ret; } }; static_assert(__alignof(u128) == 16 && sizeof(u128) == 16, "Wrong u128 size or alignment"); static force_inline u128 sync_val_compare_and_swap(volatile u128* dest, u128 comp, u128 exch) { #if !defined(_MSC_VER) auto res = __sync_val_compare_and_swap((volatile __int128_t*)dest, (__int128_t&)comp, (__int128_t&)exch); return (u128&)res; #else _InterlockedCompareExchange128((volatile long long*)dest, exch._u64[1], exch._u64[0], (long long*)&comp); return comp; #endif } static force_inline bool sync_bool_compare_and_swap(volatile u128* dest, u128 comp, u128 exch) { #if !defined(_MSC_VER) return __sync_bool_compare_and_swap((volatile __int128_t*)dest, (__int128_t&)comp, (__int128_t&)exch); #else return _InterlockedCompareExchange128((volatile long long*)dest, exch._u64[1], exch._u64[0], (long long*)&comp) != 0; #endif } static force_inline u128 sync_lock_test_and_set(volatile u128* dest, u128 value) { while (true) { const u128 old = *(u128*)dest; if (sync_bool_compare_and_swap(dest, old, value)) return old; } } static force_inline u128 sync_fetch_and_or(volatile u128* dest, u128 value) { while (true) { const u128 old = *(u128*)dest; if (sync_bool_compare_and_swap(dest, old, value | old)) return old; } } static force_inline u128 sync_fetch_and_and(volatile u128* dest, u128 value) { while (true) { const u128 old = *(u128*)dest; if (sync_bool_compare_and_swap(dest, old, value & old)) return old; } } static force_inline u128 sync_fetch_and_xor(volatile u128* dest, u128 value) { while (true) { const u128 old = *(u128*)dest; if (sync_bool_compare_and_swap(dest, old, value ^ old)) return old; } } #define re16(val) _byteswap_ushort(val) #define re32(val) _byteswap_ulong(val) #define re64(val) _byteswap_uint64(val) #define re128(val) u128::byteswap(val) template struct se_t; template struct se_t { static force_inline u8 to_be(const T& src) { return (u8&)src; } static force_inline T from_be(const u8 src) { return (T&)src; } }; template struct se_t { static force_inline u16 to_be(const T& src) { return _byteswap_ushort((u16&)src); } static force_inline T from_be(const u16 src) { const u16 res = _byteswap_ushort(src); return (T&)res; } }; template struct se_t { static force_inline u32 to_be(const T& src) { return _byteswap_ulong((u32&)src); } static force_inline T from_be(const u32 src) { const u32 res = _byteswap_ulong(src); return (T&)res; } }; template struct se_t { static force_inline u64 to_be(const T& src) { return _byteswap_uint64((u64&)src); } static force_inline T from_be(const u64 src) { const u64 res = _byteswap_uint64(src); return (T&)res; } }; template struct se_t { static force_inline u128 to_be(const T& src) { return u128::byteswap((u128&)src); } static force_inline T from_be(const u128& src) { const u128 res = u128::byteswap(src); return (T&)res; } }; template struct const_se_t; template struct const_se_t { static const u8 value = _value; }; template struct const_se_t { static const u16 value = ((_value >> 8) & 0xff) | ((_value << 8) & 0xff00); }; template struct const_se_t { static const u32 value = ((_value >> 24) & 0x000000ff) | ((_value >> 8) & 0x0000ff00) | ((_value << 8) & 0x00ff0000) | ((_value << 24) & 0xff000000); }; template struct const_se_t { static const u64 value = ((_value >> 56) & 0x00000000000000ff) | ((_value >> 40) & 0x000000000000ff00) | ((_value >> 24) & 0x0000000000ff0000) | ((_value >> 8) & 0x00000000ff000000) | ((_value << 8) & 0x000000ff00000000) | ((_value << 24) & 0x0000ff0000000000) | ((_value << 40) & 0x00ff000000000000) | ((_value << 56) & 0xff00000000000000); }; template struct be_storage { static_assert(!size, "Bad be_storage_t<> type"); }; template struct be_storage { typedef u16 type; }; template struct be_storage { typedef u32 type; }; template struct be_storage { typedef u64 type; }; template struct be_storage { typedef u128 type; }; template using be_storage_t = typename be_storage::type; template struct be_t { using type = std::remove_cv_t; using stype = be_storage_t; stype m_data; static_assert(!std::is_class::value, "be_t<> error: invalid type (class or structure)"); static_assert(!std::is_union::value || std::is_same::value, "be_t<> error: invalid type (union)"); static_assert(!std::is_pointer::value, "be_t<> error: invalid type (pointer)"); static_assert(!std::is_reference::value, "be_t<> error: invalid type (reference)"); static_assert(!std::is_array::value, "be_t<> error: invalid type (array)"); static_assert(__alignof(type) == __alignof(stype), "be_t<> error: unexpected alignment"); private: template struct _convert { static force_inline be_t& func(Tfrom& be_value) { Tto res = be_value; return (be_t&)res; } }; template struct _convert { static force_inline be_t& func(Tfrom& be_value) { Tto res = se_t::func(se_t::func(be_value)); return (be_t&)res; } }; template struct _convert { static force_inline be_t& func(Tfrom& be_value) { Tto res = be_value >> ((sizeof(Tfrom)-sizeof(Tto)) * 8); return (be_t&)res; } }; const stype& ToBE() const { return m_data; } type ToLE() const { return se_t::from_be(m_data); } void FromBE(const stype& value) { m_data = value; } void FromLE(const type& value) { m_data = se_t::to_be(value); } static be_t MakeFromLE(const type& value) { stype data = se_t::to_be(value); return (be_t&)data; } static be_t MakeFromBE(const stype& value) { return (be_t&)value; } public: //make be_t from current machine byte ordering static be_t make(const type& value) { #ifdef IS_LE_MACHINE return MakeFromLE(value); #else return MakeFromBE(value); #endif } //get value in current machine byte ordering force_inline type value() const { #ifdef IS_LE_MACHINE return ToLE(); #else return ToBE(); #endif } const stype& data() const { return ToBE(); } be_t& operator =(const be_t& value) = default; template std::enable_if_t::value, be_t&> operator =(const CT& value) { m_data = se_t::to_be(value); return *this; } //template operator std::enable_if_t::value, CT>() const //{ // return value(); //} operator type() const { return value(); } // conversion to another be_t type template operator be_t() const { return be_t::make(value()); // TODO (complicated cases like int-float conversions are not handled correctly) //return _convert sizeof(T)) ? 1 : (sizeof(T1) < sizeof(T) ? 2 : 0))>::func(m_data); } template be_t& operator += (T1 right) { return *this = value() + right; } template be_t& operator -= (T1 right) { return *this = value() - right; } template be_t& operator *= (T1 right) { return *this = value() * right; } template be_t& operator /= (T1 right) { return *this = value() / right; } template be_t& operator %= (T1 right) { return *this = value() % right; } template be_t& operator &= (T1 right) { return *this = value() & right; } template be_t& operator |= (T1 right) { return *this = value() | right; } template be_t& operator ^= (T1 right) { return *this = value() ^ right; } template be_t& operator <<= (T1 right) { return *this = value() << right; } template be_t& operator >>= (T1 right) { return *this = value() >> right; } template be_t& operator += (const be_t& right) { return *this = ToLE() + right.ToLE(); } template be_t& operator -= (const be_t& right) { return *this = ToLE() - right.ToLE(); } template be_t& operator *= (const be_t& right) { return *this = ToLE() * right.ToLE(); } template be_t& operator /= (const be_t& right) { return *this = ToLE() / right.ToLE(); } template be_t& operator %= (const be_t& right) { return *this = ToLE() % right.ToLE(); } template be_t& operator &= (const be_t& right) { return *this = ToBE() & right.ToBE(); } template be_t& operator |= (const be_t& right) { return *this = ToBE() | right.ToBE(); } template be_t& operator ^= (const be_t& right) { return *this = ToBE() ^ right.ToBE(); } template be_t operator & (const be_t& right) const { be_t res; res.FromBE(ToBE() & right.ToBE()); return res; } template be_t operator | (const be_t& right) const { be_t res; res.FromBE(ToBE() | right.ToBE()); return res; } template be_t operator ^ (const be_t& right) const { be_t res; res.FromBE(ToBE() ^ right.ToBE()); return res; } template bool operator == (T1 right) const { return (T1)ToLE() == right; } template bool operator != (T1 right) const { return !(*this == right); } template bool operator > (T1 right) const { return (T1)ToLE() > right; } template bool operator < (T1 right) const { return (T1)ToLE() < right; } template bool operator >= (T1 right) const { return (T1)ToLE() >= right; } template bool operator <= (T1 right) const { return (T1)ToLE() <= right; } template bool operator == (const be_t& right) const { return ToBE() == right.ToBE(); } template bool operator != (const be_t& right) const { return !(*this == right); } template bool operator > (const be_t& right) const { return (T1)ToLE() > right.ToLE(); } template bool operator < (const be_t& right) const { return (T1)ToLE() < right.ToLE(); } template bool operator >= (const be_t& right) const { return (T1)ToLE() >= right.ToLE(); } template bool operator <= (const be_t& right) const { return (T1)ToLE() <= right.ToLE(); } be_t operator++ (int) { be_t res = *this; *this += 1; return res; } be_t operator-- (int) { be_t res = *this; *this -= 1; return res; } be_t& operator++ () { *this += 1; return *this; } be_t& operator-- () { *this -= 1; return *this; } }; template struct is_be_t : public std::integral_constant { }; template struct is_be_t> : public std::integral_constant { }; // to_be_t helper struct template struct to_be { using type = std::conditional_t::value || std::is_enum::value, be_t, T>; }; // be_t if possible, T otherwise template using to_be_t = typename to_be::type; template struct to_be { // move const qualifier using type = const to_be_t>; }; template struct to_be { // move volatile qualifier using type = volatile to_be_t>; }; template<> struct to_be { using type = u128; }; template<> struct to_be { using type = void; }; template<> struct to_be { using type = bool; }; template<> struct to_be { using type = char; }; template<> struct to_be { using type = u8; }; template<> struct to_be { using type = s8; }; template struct _se : public const_se_t {}; template struct _se, T1, value> : public const_se_t {}; //#define se(t, x) _se::value #define se16(x) _se::value #define se32(x) _se::value #define se64(x) _se::value template struct convert_le_be_t { static Tto func(Tfrom value) { return (Tto)value; } }; template struct convert_le_be_t, Tfrom> { static be_t func(Tfrom value) { return be_t::make(value); } }; template struct convert_le_be_t, be_t> { static be_t func(be_t value) { return value; } }; template struct convert_le_be_t> { static Tto func(be_t value) { return value.value(); } }; template force_inline Tto convert_le_be(Tfrom value) { return convert_le_be_t::func(value); } template force_inline void convert_le_be(Tto& dst, Tfrom src) { dst = convert_le_be_t::func(src); } template struct le_t { using type = std::remove_cv_t; using stype = be_storage_t; stype m_data; type value() const { return reinterpret_cast(m_data); } le_t& operator =(const le_t& value) = default; template std::enable_if_t::value, le_t&> operator =(const CT& value) { m_data = reinterpret_cast(value); return *this; } //template operator std::enable_if_t::value, CT>() const //{ // return value(); //} operator type() const { return value(); } // conversion to another le_t type //template operator const le_t() const //{ // return le_t::make(value()); //} }; template struct to_le { using type = std::conditional_t::value || std::is_enum::value, le_t, T>; }; // le_t if possible, T otherwise template using to_le_t = typename to_le::type; template struct to_le { // move const qualifier using type = const to_le_t>; }; template struct to_le { // move volatile qualifier using type = volatile to_le_t>; }; template<> struct to_le { using type = u128; }; template<> struct to_le { using type = void; }; template<> struct to_le { using type = bool; }; template<> struct to_le { using type = char; }; template<> struct to_le { using type = u8; }; template<> struct to_le { using type = s8; }; // to_ne_t helper struct template struct to_ne { using type = T; }; // restore native endianness for T: returns T for be_t or le_t, T otherwise template using to_ne_t = typename to_ne::type; template struct to_ne> { using type = T; }; template struct to_ne> { using type = T; };