dvec.h

C语言库函数的原型,有用的拿去

        Is32vec4& operator -=(const I32vec4 &a) { return *this = (Is32vec4)_mm_sub_epi32(vec,a); }

        /* Shift Logical Operators */
        Is32vec4 operator<<(const M128 &a)              { return _mm_sll_epi32(vec,a); }
        Is32vec4 operator<<(int count)                  { return _mm_slli_epi32(vec,count); }
        Is32vec4& operator<<=(const M128 &a)    { return *this = (Is32vec4)_mm_sll_epi32(vec,a); }
        Is32vec4& operator<<=(int count)                { return *this = (Is32vec4)_mm_slli_epi32(vec,count); }
        /* Shift Arithmetic Operations */
        Is32vec4 operator>>(const M128 &a)              { return _mm_sra_epi32(vec,a); }
        Is32vec4 operator>>(int count)                  { return _mm_srai_epi32(vec,count); }
        Is32vec4& operator>>=(const M128 &a)    { return *this = (Is32vec4) _mm_sra_epi32(vec,a); }
        Is32vec4& operator>>=(int count)                { return *this = (Is32vec4) _mm_srai_epi32(vec,count); }

#if defined (_ENABLE_VEC_DEBUG)
        /* Output for Debug */
        friend std::ostream& operator<< (std::ostream &os, const Is32vec4 &a)
        {
                os << "[3]:" << _MM_4DW(3,a)
                        << " [2]:" << _MM_4DW(2,a)
                        << " [1]:" << _MM_4DW(1,a)
                        << " [0]:" << _MM_4DW(0,a);
                return os;
        }
#endif  /* defined (_ENABLE_VEC_DEBUG) */

        /* Element Access for Debug, No data modified */
        const int& operator[](int i)const
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 4);  /* Only 4 elements to access */
                return _MM_4DW(i,vec);
        }

        /* Element Access for Debug */
        int& operator[](int i)
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 4);  /* Only 4 elements to access */
                return _MM_4DW(i,vec);
        }
};

/* Compares */
inline Is32vec4 cmpeq(const Is32vec4 &a, const Is32vec4 &b)             { return _mm_cmpeq_epi32(a,b); }
inline Is32vec4 cmpneq(const Is32vec4 &a, const Is32vec4 &b)                    { return _mm_andnot_si128(_mm_cmpeq_epi32(a,b), get_mask128()); }
inline Is32vec4 cmpgt(const Is32vec4 &a, const Is32vec4 &b)                             { return _mm_cmpgt_epi32(a,b); }
inline Is32vec4 cmplt(const Is32vec4 &a, const Is32vec4 &b)                             { return _mm_cmpgt_epi32(b,a); }

/* Unpacks */
inline Is32vec4 unpack_low(const Is32vec4 &a, const Is32vec4 &b)                { return _mm_unpacklo_epi32(a,b); }
inline Is32vec4 unpack_high(const Is32vec4 &a, const Is32vec4 &b)               { return _mm_unpackhi_epi32(a,b); }



/* Iu32vec4 Class:
 * 4 elements, each element unsigned int
 */
class Iu32vec4 : public I32vec4
{
public:
        Iu32vec4() { }
        Iu32vec4(__m128i mm) : I32vec4(mm) { }
        Iu32vec4(unsigned int ui3, unsigned int ui2, unsigned int ui1, unsigned int ui0)
        {
                _MM_4UDW(0,vec) = ui0;
                _MM_4UDW(1,vec) = ui1;
                _MM_4UDW(2,vec) = ui2;
                _MM_4UDW(3,vec) = ui3;
        }

        /* Assignment Operator */
        Iu32vec4& operator= (const M128 &a)             { return *this = (Iu32vec4) a; }

        /* Logical Assignment Operators */
        Iu32vec4& operator&=(const M128 &a)             { return *this = (Iu32vec4) _mm_and_si128(vec,a); }
        Iu32vec4& operator|=(const M128 &a)             { return *this = (Iu32vec4) _mm_or_si128(vec,a); }
        Iu32vec4& operator^=(const M128 &a)             { return *this = (Iu32vec4) _mm_xor_si128(vec,a); }

        /* Addition & Subtraction Assignment Operators */
        Iu32vec4& operator +=(const I32vec4 &a) { return *this = (Iu32vec4)_mm_add_epi32(vec,a); }
        Iu32vec4& operator -=(const I32vec4 &a) { return *this = (Iu32vec4)_mm_sub_epi32(vec,a); }

        /* Shift Logical Operators */
        Iu32vec4 operator<<(const M128 &a)                              { return _mm_sll_epi32(vec,a); }
        Iu32vec4 operator<<(int count)                                  { return _mm_slli_epi32(vec,count); }
        Iu32vec4& operator<<=(const M128 &a)                    { return *this = (Iu32vec4)_mm_sll_epi32(vec,a); }
        Iu32vec4& operator<<=(int count)                                { return *this = (Iu32vec4)_mm_slli_epi32(vec,count); }
        Iu32vec4 operator>>(const M128 &a)                              { return _mm_srl_epi32(vec,a); }
        Iu32vec4 operator>>(int count)                                  { return _mm_srli_epi32(vec,count); }
        Iu32vec4& operator>>=(const M128 &a)                    { return *this = (Iu32vec4) _mm_srl_epi32(vec,a); }
        Iu32vec4& operator>>=(int count)                                { return *this = (Iu32vec4) _mm_srli_epi32(vec,count); }

#if defined (_ENABLE_VEC_DEBUG)
        /* Output for Debug */
        friend std::ostream& operator<< (std::ostream &os, const Iu32vec4 &a)
        {
                os << "[3]:" << _MM_4UDW(3,a)
                        << " [2]:" << _MM_4UDW(2,a)
                        << " [1]:" << _MM_4UDW(1,a)
                        << " [0]:" << _MM_4UDW(0,a);
                return os;
        }
#endif  /* defined (_ENABLE_VEC_DEBUG) */

        /* Element Access for Debug, No data modified */
        const unsigned int& operator[](int i)const
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 4);  /* Only 4 elements to access */
                return _MM_4UDW(i,vec);
        }

        /* Element Access and Assignment for Debug */
        unsigned int& operator[](int i)
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 4);  /* Only 4 elements to access */
                return _MM_4UDW(i,vec);
        }
};

inline I64vec2 operator*(const Iu32vec4 &a, const Iu32vec4 &b) { return _mm_mul_epu32(a,b); }
inline Iu32vec4 cmpeq(const Iu32vec4 &a, const Iu32vec4 &b)     { return _mm_cmpeq_epi32(a,b); }
inline Iu32vec4 cmpneq(const Iu32vec4 &a, const Iu32vec4 &b)    { return _mm_andnot_si128(_mm_cmpeq_epi32(a,b), get_mask128()); }

inline Iu32vec4 unpack_low(const Iu32vec4 &a, const Iu32vec4 &b)        { return _mm_unpacklo_epi32(a,b); }
inline Iu32vec4 unpack_high(const Iu32vec4 &a, const Iu32vec4 &b)       { return _mm_unpackhi_epi32(a,b); }

/* I16vec8 Class:
 * 8 elements, each element either unsigned or signed short
 */
class I16vec8 : public M128
{
public:
        I16vec8() { }
        I16vec8(__m128i mm) : M128(mm) { }

        /* Assignment Operator */
        I16vec8& operator= (const M128 &a)              { return *this = (I16vec8) a; }

        /* Logical Assignment Operators */
        I16vec8& operator&=(const M128 &a)              { return *this = (I16vec8) _mm_and_si128(vec,a); }
        I16vec8& operator|=(const M128 &a)              { return *this = (I16vec8) _mm_or_si128(vec,a); }
        I16vec8& operator^=(const M128 &a)              { return *this = (I16vec8) _mm_xor_si128(vec,a); }

        /* Addition & Subtraction Assignment Operators */
        I16vec8& operator +=(const I16vec8 &a)  { return *this = (I16vec8) _mm_add_epi16(vec,a); }
        I16vec8& operator -=(const I16vec8 &a)  { return *this = (I16vec8) _mm_sub_epi16(vec,a); }
        I16vec8& operator *=(const I16vec8 &a)  { return *this = (I16vec8) _mm_mullo_epi16(vec,a); }

        /* Shift Logical Operators */
        I16vec8 operator<<(const M128 &a)                               { return _mm_sll_epi16(vec,a); }
        I16vec8 operator<<(int count)                           { return _mm_slli_epi16(vec,count); }
        I16vec8& operator<<=(const M128 &a)                             { return *this = (I16vec8)_mm_sll_epi16(vec,a); }
        I16vec8& operator<<=(int count)                                 { return *this = (I16vec8)_mm_slli_epi16(vec,count); }

};


inline I16vec8 operator*(const I16vec8 &a, const I16vec8 &b)    { return _mm_mullo_epi16(a,b); }

inline I16vec8 cmpeq(const I16vec8 &a, const I16vec8 &b)                { return _mm_cmpeq_epi16(a,b); }
inline I16vec8 cmpneq(const I16vec8 &a, const I16vec8 &b)               { return _mm_andnot_si128(_mm_cmpeq_epi16(a,b), get_mask128()); }

inline I16vec8 unpack_low(const I16vec8 &a, const I16vec8 &b)   { return _mm_unpacklo_epi16(a,b); }
inline I16vec8 unpack_high(const I16vec8 &a, const I16vec8 &b)  { return _mm_unpackhi_epi16(a,b); }

/* Is16vec8 Class:
 * 8 elements, each element signed short
 */
class Is16vec8 : public I16vec8
{
public:
        Is16vec8() { }
        Is16vec8(__m128i mm) : I16vec8(mm) { }
        Is16vec8(signed short s7,signed short s6,signed short s5,signed short s4,signed short s3,signed short s2,signed short s1,signed short s0)
         {
                _MM_8W(0,vec) = s0;
                _MM_8W(1,vec) = s1;
                _MM_8W(2,vec) = s2;
                _MM_8W(3,vec) = s3;
                _MM_8W(4,vec) = s4;
                _MM_8W(5,vec) = s5;
                _MM_8W(6,vec) = s6;
                _MM_8W(7,vec) = s7;
        }

        /* Assignment Operator */
        Is16vec8& operator= (const M128 &a)             { return *this = (Is16vec8) a; }

        /* Logical Assignment Operators */
        Is16vec8& operator&=(const M128 &a)             { return *this = (Is16vec8) _mm_and_si128(vec,a); }
        Is16vec8& operator|=(const M128 &a)             { return *this = (Is16vec8) _mm_or_si128(vec,a); }
        Is16vec8& operator^=(const M128 &a)             { return *this = (Is16vec8) _mm_xor_si128(vec,a); }

        /* Addition & Subtraction Assignment Operators */
        Is16vec8& operator +=(const I16vec8 &a) { return *this = (Is16vec8) _mm_add_epi16(vec,a); }
        Is16vec8& operator -=(const I16vec8 &a) { return *this = (Is16vec8) _mm_sub_epi16(vec,a); }
        Is16vec8& operator *=(const I16vec8 &a) { return *this = (Is16vec8) _mm_mullo_epi16(vec,a); }

        /* Shift Logical Operators */
        Is16vec8 operator<<(const M128 &a)                              { return _mm_sll_epi16(vec,a); }
        Is16vec8 operator<<(int count)                          { return _mm_slli_epi16(vec,count); }
        Is16vec8& operator<<=(const M128 &a)                    { return *this = (Is16vec8)_mm_sll_epi16(vec,a); }
        Is16vec8& operator<<=(int count)                                { return *this = (Is16vec8)_mm_slli_epi16(vec,count); }
        /* Shift Arithmetic Operators */
        Is16vec8 operator>>(const M128 &a)                              { return _mm_sra_epi16(vec,a); }
        Is16vec8 operator>>(int count)                          { return _mm_srai_epi16(vec,count); }
        Is16vec8& operator>>=(const M128 &a)                    { return *this = (Is16vec8)_mm_sra_epi16(vec,a); }
        Is16vec8& operator>>=(int count)                                { return *this = (Is16vec8)_mm_srai_epi16(vec,count); }

#if defined (_ENABLE_VEC_DEBUG)
        /* Output for Debug */
        friend std::ostream& operator<< (std::ostream &os, const Is16vec8 &a)
        {
                os << "[7]:" << _MM_8W(7,a)
                        << " [6]:" << _MM_8W(6,a)
                        << " [5]:" << _MM_8W(5,a)
                        << " [4]:" << _MM_8W(4,a)
                        << " [3]:" << _MM_8W(3,a)
                        << " [2]:" << _MM_8W(2,a)
                        << " [1]:" << _MM_8W(1,a)
                        << " [0]:" << _MM_8W(0,a);
                return os;
        }
#endif  /* defined (_ENABLE_VEC_DEBUG) */

        /* Element Access for Debug, No data modified */
        const signed short& operator[](int i)const
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 8);  /* Only 8 elements to access */
                return _MM_8W(i,vec);
        }

        /* Element Access and Assignment for Debug */
        signed short& operator[](int i)
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 8);  /* Only 8 elements to access */
                return _MM_8W(i,vec);
        }
};

inline Is16vec8 operator*(const Is16vec8 &a, const Is16vec8 &b) { return _mm_mullo_epi16(a,b); }


/* Additional Is16vec8 functions: compares, unpacks, sat add/sub */
inline Is16vec8 cmpeq(const Is16vec8 &a, const Is16vec8 &b)     { return _mm_cmpeq_epi16(a,b); }
inline Is16vec8 cmpneq(const Is16vec8 &a, const Is16vec8 &b)    { return _mm_andnot_si128(_mm_cmpeq_epi16(a,b), get_mask128()); }
inline Is16vec8 cmpgt(const Is16vec8 &a, const Is16vec8 &b)             { return _mm_cmpgt_epi16(a,b); }
inline Is16vec8 cmplt(const Is16vec8 &a, const Is16vec8 &b)             { return _mm_cmpgt_epi16(b,a); }

inline Is16vec8 unpack_low(const Is16vec8 &a, const Is16vec8 &b)        { return _mm_unpacklo_epi16(a,b); }
inline Is16vec8 unpack_high(const Is16vec8 &a, const Is16vec8 &b)       { return _mm_unpackhi_epi16(a,b); }

inline Is16vec8 mul_high(const Is16vec8 &a, const Is16vec8 &b)  { return _mm_mulhi_epi16(a,b); }
inline Is32vec4 mul_add(const Is16vec8 &a, const Is16vec8 &b)   { return _mm_madd_epi16(a,b);}

inline Is16vec8 sat_add(const Is16vec8 &a, const Is16vec8 &b)   { return _mm_adds_epi16(a,b); }
inline Is16vec8 sat_sub(const Is16vec8 &a, const Is16vec8 &b)   { return _mm_subs_epi16(a,b); }

inline Is16vec8 simd_max(const Is16vec8 &a, const Is16vec8 &b)  { return _mm_max_epi16(a,b); }
inline Is16vec8 simd_min(const Is16vec8 &a, const Is16vec8 &b)  { return _mm_min_epi16(a,b); }


/* Iu16vec8 Class:
 * 8 elements, each element unsigned short
 */
class Iu16vec8 : public I16vec8
{
public:
        Iu16vec8() { }
        Iu16vec8(__m128i mm) : I16vec8(mm) { }
        Iu16vec8(unsigned short s7,unsigned short s6,unsigned short s5,unsigned short s4, unsigned short s3,unsigned short s2,unsigned short s1,unsigned short s0)
        {
                _MM_8UW(0,vec) = s0;
                _MM_8UW(1,vec) = s1;
                _MM_8UW(2,vec) = s2;
                _MM_8UW(3,vec) = s3;
                _MM_8UW(4,vec) = s4;
                _MM_8UW(5,vec) = s5;
                _MM_8UW(6,vec) = s6;
                _MM_8UW(7,vec) = s7;
        }