dvec.h

C语言库函数的原型,有用的拿去

/***
*** Copyright (C) 1985-1999 Intel Corporation.  All rights reserved.
***
*** The information and source code contained herein is the exclusive
*** property of Intel Corporation and may not be disclosed, examined
*** or reproduced in whole or in part without explicit written authorization
*** from the company.
***
****/

/*
 *  Definition of a C++ class interface to Willamette New Instruction intrinsics.
 *
 *      File name : dvec.h  class definitions
 *
 *      Concept: A C++ abstraction of Willamette new intructions designed to improve
 *  programmer productivity.  Speed and accuracy are sacrificed for utility.
 *      Facilitates an easy transition to compiler intrinsics
 *      or assembly language.
 *
 */

#ifndef _DVEC_H_INCLUDED
#define _DVEC_H_INCLUDED
#ifndef RC_INVOKED

#if !defined __cplusplus
        #error ERROR: This file is only supported in C++ compilations!
#endif  /* !defined __cplusplus */

#if defined (_M_CEE_PURE)
        #error ERROR: This file is not supported in the pure mode!
#else  /* defined (_M_CEE_PURE) */

#include <emmintrin.h> /* Willamette New Instructions Intrinsics include file */
#include <fvec.h>
#include <crtdefs.h>

#ifndef _VEC_ASSERT
#ifdef NDEBUG
        #define _VEC_ASSERT(_Expression) ((void)0)
#else  /* NDEBUG */
#ifdef __cplusplus
            extern "C" {
#endif  /* __cplusplus */

        _CRTIMP void __cdecl _wassert(_In_z_ const wchar_t * _Message, _In_z_ const wchar_t *_File, _In_ unsigned _Line);

#ifdef __cplusplus
            }
#endif  /* __cplusplus */

        #define _VEC_ASSERT(_Expression) (void)( (!!(_Expression)) || (_wassert(_CRT_WIDE(#_Expression), _CRT_WIDE(__FILE__), __LINE__), 0) )
#endif  /* NDEBUG */
#endif  /* _VEC_ASSERT */

#ifdef _MSC_VER
#pragma pack(push,_CRT_PACKING)
#endif  /* _MSC_VER */

/* Define _ENABLE_VEC_DEBUG to enable std::ostream inserters for debug output */
#if defined (_ENABLE_VEC_DEBUG)
        #include <iostream>
#endif  /* defined (_ENABLE_VEC_DEBUG) */

#pragma pack(push,16) /* Must ensure class & union 16-B aligned */

/* EMM Functionality Intrinsics */

class I8vec16;                  /* 16 elements, each element a signed or unsigned char data type */
class Is8vec16;                 /* 16 elements, each element a signed char data type */
class Iu8vec16;                 /* 16 elements, each element an unsigned char data type */
class I16vec8;                  /* 8 elements, each element a signed or unsigned short */
class Is16vec8;                 /* 8 elements, each element a signed short */
class Iu16vec8;                 /* 8 elements, each element an unsigned short */
class I32vec4;                  /* 4 elements, each element a signed or unsigned long */
class Is32vec4;                 /* 4 elements, each element a signed long */
class Iu32vec4;                 /* 4 elements, each element a unsigned long */
class I64vec2;                  /* 2 element, each a __m64 data type */
class I128vec1;                 /* 1 element, a __m128i data type */

#define _MM_16UB(element,vector) (*((unsigned char*)&##vector + ##element))
#define _MM_16B(element,vector) (*((signed char*)&##vector + ##element))

#define _MM_8UW(element,vector) (*((unsigned short*)&##vector + ##element))
#define _MM_8W(element,vector) (*((short*)&##vector + ##element))

#define _MM_4UDW(element,vector) (*((unsigned int*)&##vector + ##element))
#define _MM_4DW(element,vector) (*((int*)&##vector + ##element))

#define _MM_2QW(element,vector) (*((__int64*)&##vector + ##element))


/* We need a m128i constant, keeping performance in mind*/

#pragma warning(push)
#pragma warning(disable : 4640)
inline const __m128i get_mask128()
{
        static const __m128i mask128 = _mm_set1_epi64(M64(0xffffffffffffffffi64));
        return mask128;
}
#pragma warning(pop)


//DEVDIV Remove alais created in public\sdk\inc\winnt.h
#ifdef M128
#undef M128
#endif  /* M128 */
#ifdef PM128
#undef PM128
#endif  /* PM128 */
//end DEVDIV

/* M128 Class:
 * 1 element, a __m128i data type
 * Contructors & Logical Operations
 */

class M128
{
protected:
                __m128i vec;

public:
        M128()                                                                  { }
        M128(__m128i mm)                                                { vec = mm; }

        operator __m128i() const                                        { return vec; }

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

};

inline M128 operator&(const M128 &a, const M128 &b)     { return _mm_and_si128(a,b); }
inline M128 operator|(const M128 &a, const M128 &b)     { return _mm_or_si128(a,b); }
inline M128 operator^(const M128 &a, const M128 &b)     { return _mm_xor_si128(a,b); }
inline M128 andnot(const M128 &a, const M128 &b)        { return _mm_andnot_si128(a,b); }

/* I128vec1 Class:
 * 1 element, a __m128i data type
 * Contains Operations which can operate on any __m6128i data type
 */

class I128vec1 : public M128
{
public:
        I128vec1()                                                              { }
        I128vec1(__m128i mm) : M128(mm)                         { }

        I128vec1& operator= (const M128 &a) { return *this = (I128vec1) a; }
        I128vec1& operator&=(const M128 &a) { return *this = (I128vec1) _mm_and_si128(vec,a); }
        I128vec1& operator|=(const M128 &a) { return *this = (I128vec1) _mm_or_si128(vec,a); }
        I128vec1& operator^=(const M128 &a) { return *this = (I128vec1) _mm_xor_si128(vec,a); }

};

/* I64vec2 Class:
 * 2 elements, each element signed or unsigned 64-bit integer
 */
class I64vec2 : public M128
{
public:
        I64vec2() { }
        I64vec2(__m128i mm) : M128(mm) { }

        I64vec2(__m64 q1, __m64 q0)
        {
                _MM_2QW(0,vec) = *(__int64*)&q0;
                _MM_2QW(1,vec) = *(__int64*)&q1;
        }

        /* Assignment Operator */
        I64vec2& operator= (const M128 &a) { return *this = (I64vec2) a; }

        /* Logical Assignment Operators */
        I64vec2& operator&=(const M128 &a) { return *this = (I64vec2) _mm_and_si128(vec,a); }
        I64vec2& operator|=(const M128 &a) { return *this = (I64vec2) _mm_or_si128(vec,a); }
        I64vec2& operator^=(const M128 &a) { return *this = (I64vec2) _mm_xor_si128(vec,a); }

        /* Addition & Subtraction Assignment Operators */
        I64vec2& operator +=(const I64vec2 &a)                  { return *this = (I64vec2) _mm_add_epi64(vec,a); }
        I64vec2& operator -=(const I64vec2 &a)                  { return *this = (I64vec2) _mm_sub_epi64(vec,a); }

        /* Shift Logical Operators */
        I64vec2 operator<<(const I64vec2 &a)                    { return _mm_sll_epi64(vec,a); }
        I64vec2 operator<<(int count)                                   { return _mm_slli_epi64(vec,count); }
        I64vec2& operator<<=(const I64vec2 &a)                  { return *this = (I64vec2) _mm_sll_epi64(vec,a); }
        I64vec2& operator<<=(int count)                                 { return *this = (I64vec2) _mm_slli_epi64(vec,count); }
        I64vec2 operator>>(const I64vec2 &a)                    { return _mm_srl_epi64(vec,a); }
        I64vec2 operator>>(int count)                                   { return _mm_srli_epi64(vec,count); }
        I64vec2& operator>>=(const I64vec2 &a)                  { return *this = (I64vec2) _mm_srl_epi64(vec,a); }
        I64vec2& operator>>=(int count)                                 { return *this = (I64vec2) _mm_srli_epi64(vec,count); }

        /* Element Access for Debug, No data modified */
        const __int64& operator[](int i)const
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 2);  /* Only 2 elements to access */
                return _MM_2QW(i,vec);
        }

        /* Element Access and Assignment for Debug */
        __int64& operator[](int i)
        {
                _VEC_ASSERT(static_cast<unsigned int>(i) < 2);  /* Only 2 elements to access */
                return _MM_2QW(i,vec);
        }


};

/* Unpacks */
inline I64vec2 unpack_low(const I64vec2 &a, const I64vec2 &b)   {return _mm_unpacklo_epi64(a,b); }
inline I64vec2 unpack_high(const I64vec2 &a, const I64vec2 &b)  {return _mm_unpackhi_epi64(a,b); }

/* I32vec4 Class:
 * 4 elements, each element either a signed or unsigned int
 */
class I32vec4 : public M128
{
public:
        I32vec4() { }
        I32vec4(__m128i mm) : M128(mm) { }

        /* Assignment Operator */
        I32vec4& operator= (const M128 &a)                              { return *this = (I32vec4) a; }

        /* Logicals Operators */
        I32vec4& operator&=(const M128 &a)                              { return *this = (I32vec4) _mm_and_si128(vec,a); }
        I32vec4& operator|=(const M128 &a)                              { return *this = (I32vec4) _mm_or_si128(vec,a); }
        I32vec4& operator^=(const M128 &a)                              { return *this = (I32vec4) _mm_xor_si128(vec,a); }

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

        /* Shift Logical Operators */
        I32vec4 operator<<(const I32vec4 &a)                    { return _mm_sll_epi32(vec,a); }
        I32vec4 operator<<(int count)                                   { return _mm_slli_epi32(vec,count); }
        I32vec4& operator<<=(const I32vec4 &a)                  { return *this = (I32vec4)_mm_sll_epi32(vec,a); }
        I32vec4& operator<<=(int count)                                 { return *this = (I32vec4)_mm_slli_epi32(vec,count); }

};

inline I32vec4 cmpeq(const I32vec4 &a, const I32vec4 &b)            { return _mm_cmpeq_epi32(a,b); }
inline I32vec4 cmpneq(const I32vec4 &a, const I32vec4 &b)           { return _mm_andnot_si128(_mm_cmpeq_epi32(a,b), get_mask128()); }

inline I32vec4 unpack_low(const I32vec4 &a, const I32vec4 &b)   { return _mm_unpacklo_epi32(a,b); }
inline I32vec4 unpack_high(const I32vec4 &a, const I32vec4 &b)  { return _mm_unpackhi_epi32(a,b); }

/* Is32vec4 Class:
 * 4 elements, each element signed integer
 */
class Is32vec4 : public I32vec4
{
public:
        Is32vec4() { }
        Is32vec4(__m128i mm) : I32vec4(mm) { }
        Is32vec4(int i3, int i2, int i1, int i0)
        {
                _MM_4DW(0,vec) = i0;
                _MM_4DW(1,vec) = i1;
                _MM_4DW(2,vec) = i2;
                _MM_4DW(3,vec) = i3;
        }

        /* Assignment Operator */
        Is32vec4& operator= (const M128 &a)             { return *this = (Is32vec4) a; }

        /* Logical Operators */
        Is32vec4& operator&=(const M128 &a)             { return *this = (Is32vec4) _mm_and_si128(vec,a); }
        Is32vec4& operator|=(const M128 &a)             { return *this = (Is32vec4) _mm_or_si128(vec,a); }
        Is32vec4& operator^=(const M128 &a)             { return *this = (Is32vec4) _mm_xor_si128(vec,a); }

        /* Addition & Subtraction Assignment Operators */
        Is32vec4& operator +=(const I32vec4 &a) { return *this = (Is32vec4)_mm_add_epi32(vec,a); }