/*
 *
 *  Copyright (C) Microsoft Corporation.  All Rights Reserved.
 *
 *  File:       d3dx8math.h
 *  Content:    D3DX math types and functions
 *
 */

#include "d3dx8.h"

#ifndef __D3DX8MATH_H__
#define __D3DX8MATH_H__

#include <math.h>
#pragma warning(disable:4201) /* anonymous unions warning */

/*
 *
 * General purpose utilities
 *
 */
#define D3DX_PI    ((FLOAT)  3.141592654f)
#define D3DX_1BYPI ((FLOAT)  0.318309886f)

#define D3DXToRadian( degree ) ((degree) * (D3DX_PI / 180.0f))
#define D3DXToDegree( radian ) ((radian) * (180.0f / D3DX_PI))



/*
 *
 * Vectors
 *
 */

/*
 * 2D Vector
 */
typedef struct D3DXVECTOR2
{
#ifdef __cplusplus
public:
    D3DXVECTOR2() {};
    D3DXVECTOR2( CONST FLOAT * );
    D3DXVECTOR2( FLOAT x, FLOAT y );

    /* casting */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* assignment operators */
    D3DXVECTOR2& operator += ( CONST D3DXVECTOR2& );
    D3DXVECTOR2& operator -= ( CONST D3DXVECTOR2& );
    D3DXVECTOR2& operator *= ( FLOAT );
    D3DXVECTOR2& operator /= ( FLOAT );

    /* unary operators */
    D3DXVECTOR2 operator + () const;
    D3DXVECTOR2 operator - () const;

    /* binary operators */
    D3DXVECTOR2 operator + ( CONST D3DXVECTOR2& ) const;
    D3DXVECTOR2 operator - ( CONST D3DXVECTOR2& ) const;
    D3DXVECTOR2 operator * ( FLOAT ) const;
    D3DXVECTOR2 operator / ( FLOAT ) const;

    friend D3DXVECTOR2 operator * ( FLOAT, CONST D3DXVECTOR2& );

    BOOL operator == ( CONST D3DXVECTOR2& ) const;
    BOOL operator != ( CONST D3DXVECTOR2& ) const;


public:
#endif /* __cplusplus */
    FLOAT x, y;
} D3DXVECTOR2, *LPD3DXVECTOR2;


/*
 * 3D Vector
 */
#ifdef __cplusplus
typedef struct D3DXVECTOR3 : public D3DVECTOR
{
public:
    D3DXVECTOR3() {};
    D3DXVECTOR3( CONST FLOAT * );
    D3DXVECTOR3( CONST D3DVECTOR& );
    D3DXVECTOR3( FLOAT x, FLOAT y, FLOAT z );

    /* casting */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* assignment operators */
    D3DXVECTOR3& operator += ( CONST D3DXVECTOR3& );
    D3DXVECTOR3& operator -= ( CONST D3DXVECTOR3& );
    D3DXVECTOR3& operator *= ( FLOAT );
    D3DXVECTOR3& operator /= ( FLOAT );

    /* unary operators */
    D3DXVECTOR3 operator + () const;
    D3DXVECTOR3 operator - () const;

    /* binary operators */
    D3DXVECTOR3 operator + ( CONST D3DXVECTOR3& ) const;
    D3DXVECTOR3 operator - ( CONST D3DXVECTOR3& ) const;
    D3DXVECTOR3 operator * ( FLOAT ) const;
    D3DXVECTOR3 operator / ( FLOAT ) const;

    friend D3DXVECTOR3 operator * ( FLOAT, CONST struct D3DXVECTOR3& );

    BOOL operator == ( CONST D3DXVECTOR3& ) const;
    BOOL operator != ( CONST D3DXVECTOR3& ) const;

} D3DXVECTOR3, *LPD3DXVECTOR3;

#else /* !__cplusplus */
typedef struct _D3DVECTOR D3DXVECTOR3, *LPD3DXVECTOR3;
#endif /* !__cplusplus */


/*
 * 4D Vector
 */
typedef struct D3DXVECTOR4
{
#ifdef __cplusplus
public:
    D3DXVECTOR4() {};
    D3DXVECTOR4( CONST FLOAT* );
    D3DXVECTOR4( FLOAT x, FLOAT y, FLOAT z, FLOAT w );

    /* casting */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* assignment operators */
    D3DXVECTOR4& operator += ( CONST D3DXVECTOR4& );
    D3DXVECTOR4& operator -= ( CONST D3DXVECTOR4& );
    D3DXVECTOR4& operator *= ( FLOAT );
    D3DXVECTOR4& operator /= ( FLOAT );

    /* unary operators */
    D3DXVECTOR4 operator + () const;
    D3DXVECTOR4 operator - () const;

    /* binary operators */
    D3DXVECTOR4 operator + ( CONST D3DXVECTOR4& ) const;
    D3DXVECTOR4 operator - ( CONST D3DXVECTOR4& ) const;
    D3DXVECTOR4 operator * ( FLOAT ) const;
    D3DXVECTOR4 operator / ( FLOAT ) const;

    friend D3DXVECTOR4 operator * ( FLOAT, CONST D3DXVECTOR4& );

    BOOL operator == ( CONST D3DXVECTOR4& ) const;
    BOOL operator != ( CONST D3DXVECTOR4& ) const;

public:
#endif /* __cplusplus */
    FLOAT x, y, z, w;
} D3DXVECTOR4, *LPD3DXVECTOR4;


/*
 *
 * Matrices
 *
 */
#ifdef __cplusplus
typedef struct D3DXMATRIX : public D3DMATRIX
{
public:
    D3DXMATRIX() {};
    D3DXMATRIX( CONST FLOAT * );
    D3DXMATRIX( CONST D3DMATRIX& );
    D3DXMATRIX( FLOAT _11, FLOAT _12, FLOAT _13, FLOAT _14,
                FLOAT _21, FLOAT _22, FLOAT _23, FLOAT _24,
                FLOAT _31, FLOAT _32, FLOAT _33, FLOAT _34,
                FLOAT _41, FLOAT _42, FLOAT _43, FLOAT _44 );


    /* access grants */
    FLOAT& operator () ( UINT Row, UINT Col );
    FLOAT  operator () ( UINT Row, UINT Col ) const;

    /* casting operators */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* assignment operators */
    D3DXMATRIX& operator *= ( CONST D3DXMATRIX& );
    D3DXMATRIX& operator += ( CONST D3DXMATRIX& );
    D3DXMATRIX& operator -= ( CONST D3DXMATRIX& );
    D3DXMATRIX& operator *= ( FLOAT );
    D3DXMATRIX& operator /= ( FLOAT );

    /* unary operators */
    D3DXMATRIX operator + () const;
    D3DXMATRIX operator - () const;

    /* binary operators */
    D3DXMATRIX operator * ( CONST D3DXMATRIX& ) const;
    D3DXMATRIX operator + ( CONST D3DXMATRIX& ) const;
    D3DXMATRIX operator - ( CONST D3DXMATRIX& ) const;
    D3DXMATRIX operator * ( FLOAT ) const;
    D3DXMATRIX operator / ( FLOAT ) const;

    friend D3DXMATRIX operator * ( FLOAT, CONST D3DXMATRIX& );

    BOOL operator == ( CONST D3DXMATRIX& ) const;
    BOOL operator != ( CONST D3DXMATRIX& ) const;

} D3DXMATRIX, *LPD3DXMATRIX;

#else /* !__cplusplus */
typedef struct _D3DMATRIX D3DXMATRIX, *LPD3DXMATRIX;
#endif /* !__cplusplus */

/*
 *
 * Aligned Matrices
 *
 * This class helps keep matrices 16-byte aligned as preferred by P4 cpus.
 * It aligns matrices on the stack and on the heap or in global scope.
 * It does this using __declspec(align(16)) which works on VC7 and on VC 6
 * with the processor pack. Unfortunately there is no way to detect the
 * latter so this is turned on only on VC7. On other compilers this is the
 * the same as D3DXMATRIX.
 * Using this class on a compiler that does not actually do the alignment
 * can be dangerous since it will not expose bugs that ignore alignment.
 * E.g if an object of this class in inside a struct or class, and some code
 * memcopys data in it assuming tight packing. This could break on a compiler
 * that eventually start aligning the matrix.
 *
 */
#ifdef __cplusplus
typedef struct _D3DXMATRIXA16 : public D3DXMATRIX
{
    _D3DXMATRIXA16() {}
    _D3DXMATRIXA16( CONST FLOAT * f): D3DXMATRIX(f) {}
    _D3DXMATRIXA16( CONST D3DMATRIX& m): D3DXMATRIX(m) {}
    _D3DXMATRIXA16( FLOAT _11, FLOAT _12, FLOAT _13, FLOAT _14,
                    FLOAT _21, FLOAT _22, FLOAT _23, FLOAT _24,
                    FLOAT _31, FLOAT _32, FLOAT _33, FLOAT _34,
                    FLOAT _41, FLOAT _42, FLOAT _43, FLOAT _44 ) :
                D3DXMATRIX(_11, _12, _13, _14,
                           _21, _22, _23, _24,
                           _31, _32, _33, _34,
                           _41, _42, _43, _44) {}
    void* operator new(size_t s)
    {
        LPBYTE p = ::new BYTE[s + 16];
        if (p)
        {
            BYTE offset = (BYTE)(16 - ((UINT_PTR)p & 15));
            p += offset;
            p[-1] = offset;
        }
        return p;
    };

    void* operator new[](size_t s)
    {
        LPBYTE p = ::new BYTE[s + 16];
        if (p)
        {
            BYTE offset = (BYTE)(16 - ((UINT_PTR)p & 15));
            p += offset;
            p[-1] = offset;
        }
        return p;
    };

    /* This is NOT a virtual operator. If you cast
     * to D3DXMATRIX, do not delete using that */
    void operator delete(void* p)
    {
        if(p)
        {
            BYTE* pb = static_cast<BYTE*>(p);
            pb -= pb[-1];
            ::delete [] pb;
        }
    };

    /* This is NOT a virtual operator. If you cast
     * to D3DXMATRIX, do not delete using that */
    void operator delete[](void* p)
    {
        if(p)
        {
            BYTE* pb = static_cast<BYTE*>(p);
            pb -= pb[-1];
            ::delete [] pb;
        }
    };

    struct _D3DXMATRIXA16& operator=(CONST D3DXMATRIX& rhs)
    {
        memcpy(&_11, &rhs, sizeof(D3DXMATRIX));
        return *this;
    };
} _D3DXMATRIXA16;

#else /* !__cplusplus */
typedef D3DXMATRIX  _D3DXMATRIXA16;
#endif /* !__cplusplus */

#if _MSC_VER >= 1300        /* VC7 */
#define _ALIGN_16 __declspec(align(16))
#else
#define _ALIGN_16                   /* Earlier compiler may not understand this, do nothing. */
#endif

#define D3DXMATRIXA16 _ALIGN_16 _D3DXMATRIXA16

typedef D3DXMATRIXA16 *LPD3DXMATRIXA16;

/*
 *
 *    Quaternions
 *
 */
typedef struct D3DXQUATERNION
{
#ifdef __cplusplus
public:
    D3DXQUATERNION() {}
    D3DXQUATERNION( CONST FLOAT * );
    D3DXQUATERNION( FLOAT x, FLOAT y, FLOAT z, FLOAT w );

    /* casting */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* assignment operators */
    D3DXQUATERNION& operator += ( CONST D3DXQUATERNION& );
    D3DXQUATERNION& operator -= ( CONST D3DXQUATERNION& );
    D3DXQUATERNION& operator *= ( CONST D3DXQUATERNION& );
    D3DXQUATERNION& operator *= ( FLOAT );
    D3DXQUATERNION& operator /= ( FLOAT );

    /* unary operators */
    D3DXQUATERNION  operator + () const;
    D3DXQUATERNION  operator - () const;

    /* binary operators */
    D3DXQUATERNION operator + ( CONST D3DXQUATERNION& ) const;
    D3DXQUATERNION operator - ( CONST D3DXQUATERNION& ) const;
    D3DXQUATERNION operator * ( CONST D3DXQUATERNION& ) const;
    D3DXQUATERNION operator * ( FLOAT ) const;
    D3DXQUATERNION operator / ( FLOAT ) const;

    friend D3DXQUATERNION operator * (FLOAT, CONST D3DXQUATERNION& );

    BOOL operator == ( CONST D3DXQUATERNION& ) const;
    BOOL operator != ( CONST D3DXQUATERNION& ) const;

#endif /*__cplusplus */
    FLOAT x, y, z, w;
} D3DXQUATERNION, *LPD3DXQUATERNION;


/*
 *
 * Planes
 *
 */
typedef struct D3DXPLANE
{
#ifdef __cplusplus
public:
    D3DXPLANE() {}
    D3DXPLANE( CONST FLOAT* );
    D3DXPLANE( FLOAT a, FLOAT b, FLOAT c, FLOAT d );

    /* casting */
    operator FLOAT* ();
    operator CONST FLOAT* () const;

    /* unary operators */
    D3DXPLANE operator + () const;
    D3DXPLANE operator - () const;

    /* binary operators */
    BOOL operator == ( CONST D3DXPLANE& ) const;
    BOOL operator != ( CONST D3DXPLANE& ) const;

#endif /* __cplusplus */
    FLOAT a, b, c, d;
} D3DXPLANE, *LPD3DXPLANE;


/*
 *
 * Colors
 *
 */

typedef struct D3DXCOLOR
{
#ifdef __cplusplus
public:
    D3DXCOLOR() {}
    D3DXCOLOR( DWORD argb );
    D3DXCOLOR( CONST FLOAT * );
    D3DXCOLOR( CONST D3DCOLORVALUE& );
    D3DXCOLOR( FLOAT r, FLOAT g, FLOAT b, FLOAT a );

    /* casting */
    operator DWORD () const;

    operator FLOAT* ();
    operator CONST FLOAT* () const;

    operator D3DCOLORVALUE* ();
    operator CONST D3DCOLORVALUE* () const;

    operator D3DCOLORVALUE& ();
    operator CONST D3DCOLORVALUE& () const;

    /* assignment operators */
    D3DXCOLOR& operator += ( CONST D3DXCOLOR& );
    D3DXCOLOR& operator -= ( CONST D3DXCOLOR& );
    D3DXCOLOR& operator *= ( FLOAT );
    D3DXCOLOR& operator /= ( FLOAT );

    /* unary operators */
    D3DXCOLOR operator + () const;
    D3DXCOLOR operator - () const;

    /* binary operators */
    D3DXCOLOR operator + ( CONST D3DXCOLOR& ) const;
    D3DXCOLOR operator - ( CONST D3DXCOLOR& ) const;
    D3DXCOLOR operator * ( FLOAT ) const;
    D3DXCOLOR operator / ( FLOAT ) const;

    friend D3DXCOLOR operator * (FLOAT, CONST D3DXCOLOR& );

    BOOL operator == ( CONST D3DXCOLOR& ) const;
    BOOL operator != ( CONST D3DXCOLOR& ) const;

#endif /* __cplusplus */
    FLOAT r, g, b, a;
} D3DXCOLOR, *LPD3DXCOLOR;



/*
 *
 * D3DX math functions:
 *
 * NOTE:
 *  * All these functions can take the same object as in and out parameters.
 *
 *  * Out parameters are typically also returned as return values, so that
 *    the output of one function may be used as a parameter to another.
 *
 */

/*
 * 2D Vector
 */

/* inline */

FLOAT D3DXVec2Length
    ( CONST D3DXVECTOR2 *pV );

FLOAT D3DXVec2LengthSq
    ( CONST D3DXVECTOR2 *pV );

FLOAT D3DXVec2Dot
    ( CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

/* Z component of ((x1,y1,0) cross (x2,y2,0)) */
FLOAT D3DXVec2CCW
    ( CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

D3DXVECTOR2* D3DXVec2Add
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

D3DXVECTOR2* D3DXVec2Subtract
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

/* Minimize each component.  x = min(x1, x2), y = min(y1, y2) */
D3DXVECTOR2* D3DXVec2Minimize
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

/* Maximize each component.  x = max(x1, x2), y = max(y1, y2) */
D3DXVECTOR2* D3DXVec2Maximize
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2 );

D3DXVECTOR2* D3DXVec2Scale
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV, FLOAT s );

/* Linear interpolation. V1 + s(V2-V1) */
D3DXVECTOR2* D3DXVec2Lerp
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2,
      FLOAT s );

/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

D3DXVECTOR2* WINAPI D3DXVec2Normalize
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV );

/* Hermite interpolation between position V1, tangent T1 (when s == 0)
 * and position V2, tangent T2 (when s == 1). */
D3DXVECTOR2* WINAPI D3DXVec2Hermite
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pT1,
      CONST D3DXVECTOR2 *pV2, CONST D3DXVECTOR2 *pT2, FLOAT s );

/* CatmullRom interpolation between V1 (when s == 0) and V2 (when s == 1) */
D3DXVECTOR2* WINAPI D3DXVec2CatmullRom
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV0, CONST D3DXVECTOR2 *pV1,
      CONST D3DXVECTOR2 *pV2, CONST D3DXVECTOR2 *pV3, FLOAT s );

/* Barycentric coordinates.  V1 + f(V2-V1) + g(V3-V1) */
D3DXVECTOR2* WINAPI D3DXVec2BaryCentric
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV1, CONST D3DXVECTOR2 *pV2,
      CONST D3DXVECTOR2 *pV3, FLOAT f, FLOAT g);

/* Transform (x, y, 0, 1) by matrix. */
D3DXVECTOR4* WINAPI D3DXVec2Transform
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR2 *pV, CONST D3DXMATRIX *pM );

/* Transform (x, y, 0, 1) by matrix, project result back into w=1. */
D3DXVECTOR2* WINAPI D3DXVec2TransformCoord
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV, CONST D3DXMATRIX *pM );

/* Transform (x, y, 0, 0) by matrix. */
D3DXVECTOR2* WINAPI D3DXVec2TransformNormal
    ( D3DXVECTOR2 *pOut, CONST D3DXVECTOR2 *pV, CONST D3DXMATRIX *pM );

#ifdef __cplusplus
}
#endif

/*
 * 3D Vector
 */

/* inline */

FLOAT D3DXVec3Length
    ( CONST D3DXVECTOR3 *pV );

FLOAT D3DXVec3LengthSq
    ( CONST D3DXVECTOR3 *pV );

FLOAT D3DXVec3Dot
    ( CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

D3DXVECTOR3* D3DXVec3Cross
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

D3DXVECTOR3* D3DXVec3Add
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

D3DXVECTOR3* D3DXVec3Subtract
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

/* Minimize each component.  x = min(x1, x2), y = min(y1, y2), ... */
D3DXVECTOR3* D3DXVec3Minimize
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

/* Maximize each component.  x = max(x1, x2), y = max(y1, y2), ... */
D3DXVECTOR3* D3DXVec3Maximize
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2 );

D3DXVECTOR3* D3DXVec3Scale
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV, FLOAT s);

/* Linear interpolation. V1 + s(V2-V1) */
D3DXVECTOR3* D3DXVec3Lerp
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2,
      FLOAT s );

/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

D3DXVECTOR3* WINAPI D3DXVec3Normalize
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV );

/* Hermite interpolation between position V1, tangent T1 (when s == 0)
 * and position V2, tangent T2 (when s == 1). */
D3DXVECTOR3* WINAPI D3DXVec3Hermite
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pT1,
      CONST D3DXVECTOR3 *pV2, CONST D3DXVECTOR3 *pT2, FLOAT s );

/* CatmullRom interpolation between V1 (when s == 0) and V2 (when s == 1) */
D3DXVECTOR3* WINAPI D3DXVec3CatmullRom
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV0, CONST D3DXVECTOR3 *pV1,
      CONST D3DXVECTOR3 *pV2, CONST D3DXVECTOR3 *pV3, FLOAT s );

/* Barycentric coordinates.  V1 + f(V2-V1) + g(V3-V1) */
D3DXVECTOR3* WINAPI D3DXVec3BaryCentric
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2,
      CONST D3DXVECTOR3 *pV3, FLOAT f, FLOAT g);

/* Transform (x, y, z, 1) by matrix. */
D3DXVECTOR4* WINAPI D3DXVec3Transform
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR3 *pV, CONST D3DXMATRIX *pM );

/* Transform (x, y, z, 1) by matrix, project result back into w=1. */
D3DXVECTOR3* WINAPI D3DXVec3TransformCoord
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV, CONST D3DXMATRIX *pM );

/* Transform (x, y, z, 0) by matrix.  If you transforming a normal by a
 * non-affine matrix, the matrix you pass to this function should be the
 * transpose of the inverse of the matrix you would use to transform a coord. */
D3DXVECTOR3* WINAPI D3DXVec3TransformNormal
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV, CONST D3DXMATRIX *pM );

/* Project vector from object space into screen space */
D3DXVECTOR3* WINAPI D3DXVec3Project
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV, CONST D3DVIEWPORT8 *pViewport,
      CONST D3DXMATRIX *pProjection, CONST D3DXMATRIX *pView, CONST D3DXMATRIX *pWorld);

/* Project vector from screen space into object space */
D3DXVECTOR3* WINAPI D3DXVec3Unproject
    ( D3DXVECTOR3 *pOut, CONST D3DXVECTOR3 *pV, CONST D3DVIEWPORT8 *pViewport,
      CONST D3DXMATRIX *pProjection, CONST D3DXMATRIX *pView, CONST D3DXMATRIX *pWorld);

#ifdef __cplusplus
}
#endif

/*
 * 4D Vector
 */

/* inline */

FLOAT D3DXVec4Length
    ( CONST D3DXVECTOR4 *pV );

FLOAT D3DXVec4LengthSq
    ( CONST D3DXVECTOR4 *pV );

FLOAT D3DXVec4Dot
    ( CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2 );

D3DXVECTOR4* D3DXVec4Add
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2);

D3DXVECTOR4* D3DXVec4Subtract
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2);

/* Minimize each component.  x = min(x1, x2), y = min(y1, y2), ... */
D3DXVECTOR4* D3DXVec4Minimize
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2);

/* Maximize each component.  x = max(x1, x2), y = max(y1, y2), ... */
D3DXVECTOR4* D3DXVec4Maximize
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2);

D3DXVECTOR4* D3DXVec4Scale
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV, FLOAT s);

/* Linear interpolation. V1 + s(V2-V1) */
D3DXVECTOR4* D3DXVec4Lerp
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2,
      FLOAT s );

/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

/* Cross-product in 4 dimensions. */
D3DXVECTOR4* WINAPI D3DXVec4Cross
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2,
      CONST D3DXVECTOR4 *pV3);

D3DXVECTOR4* WINAPI D3DXVec4Normalize
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV );

/* Hermite interpolation between position V1, tangent T1 (when s == 0)
 * and position V2, tangent T2 (when s == 1). */
D3DXVECTOR4* WINAPI D3DXVec4Hermite
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pT1,
      CONST D3DXVECTOR4 *pV2, CONST D3DXVECTOR4 *pT2, FLOAT s );

/* CatmullRom interpolation between V1 (when s == 0) and V2 (when s == 1) */
D3DXVECTOR4* WINAPI D3DXVec4CatmullRom
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV0, CONST D3DXVECTOR4 *pV1,
      CONST D3DXVECTOR4 *pV2, CONST D3DXVECTOR4 *pV3, FLOAT s );

/* Barycentric coordinates.  V1 + f(V2-V1) + g(V3-V1) */
D3DXVECTOR4* WINAPI D3DXVec4BaryCentric
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV1, CONST D3DXVECTOR4 *pV2,
      CONST D3DXVECTOR4 *pV3, FLOAT f, FLOAT g);

/* Transform vector by matrix. */
D3DXVECTOR4* WINAPI D3DXVec4Transform
    ( D3DXVECTOR4 *pOut, CONST D3DXVECTOR4 *pV, CONST D3DXMATRIX *pM );

#ifdef __cplusplus
}
#endif


/*
 * 4D Matrix
 */

/* inline */

D3DXMATRIX* D3DXMatrixIdentity
    ( D3DXMATRIX *pOut );

BOOL D3DXMatrixIsIdentity
    ( CONST D3DXMATRIX *pM );


/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

FLOAT WINAPI D3DXMatrixfDeterminant
    ( CONST D3DXMATRIX *pM );

D3DXMATRIX* WINAPI D3DXMatrixTranspose
    ( D3DXMATRIX *pOut, CONST D3DXMATRIX *pM );

/* Matrix multiplication.  The result represents the transformation M2
 * followed by the transformation M1.  (Out = M1 * M2) */
D3DXMATRIX* WINAPI D3DXMatrixMultiply
    ( D3DXMATRIX *pOut, CONST D3DXMATRIX *pM1, CONST D3DXMATRIX *pM2 );

/* Matrix multiplication, followed by a transpose. (Out = T(M1 * M2)) */
D3DXMATRIX* WINAPI D3DXMatrixMultiplyTranspose
    ( D3DXMATRIX *pOut, CONST D3DXMATRIX *pM1, CONST D3DXMATRIX *pM2 );

/* Calculate inverse of matrix.  Inversion my fail, in which case NULL will
 * be returned.  The determinant of pM is also returned it pfDeterminant
 * is non-NULL. */
D3DXMATRIX* WINAPI D3DXMatrixInverse
    ( D3DXMATRIX *pOut, FLOAT *pDeterminant, CONST D3DXMATRIX *pM );

/* Build a matrix which scales by (sx, sy, sz) */
D3DXMATRIX* WINAPI D3DXMatrixScaling
    ( D3DXMATRIX *pOut, FLOAT sx, FLOAT sy, FLOAT sz );

/* Build a matrix which translates by (x, y, z) */
D3DXMATRIX* WINAPI D3DXMatrixTranslation
    ( D3DXMATRIX *pOut, FLOAT x, FLOAT y, FLOAT z );

/* Build a matrix which rotates around the X axis */
D3DXMATRIX* WINAPI D3DXMatrixRotationX
    ( D3DXMATRIX *pOut, FLOAT Angle );

/* Build a matrix which rotates around the Y axis */
D3DXMATRIX* WINAPI D3DXMatrixRotationY
    ( D3DXMATRIX *pOut, FLOAT Angle );

/* Build a matrix which rotates around the Z axis */
D3DXMATRIX* WINAPI D3DXMatrixRotationZ
    ( D3DXMATRIX *pOut, FLOAT Angle );

/* Build a matrix which rotates around an arbitrary axis */
D3DXMATRIX* WINAPI D3DXMatrixRotationAxis
    ( D3DXMATRIX *pOut, CONST D3DXVECTOR3 *pV, FLOAT Angle );

/* Build a matrix from a quaternion */
D3DXMATRIX* WINAPI D3DXMatrixRotationQuaternion
    ( D3DXMATRIX *pOut, CONST D3DXQUATERNION *pQ);

/* Yaw around the Y axis, a pitch around the X axis,
 * and a roll around the Z axis. */
D3DXMATRIX* WINAPI D3DXMatrixRotationYawPitchRoll
    ( D3DXMATRIX *pOut, FLOAT Yaw, FLOAT Pitch, FLOAT Roll );


/* Build transformation matrix.  NULL arguments are treated as identity.
 * Mout = Msc-1 * Msr-1 * Ms * Msr * Msc * Mrc-1 * Mr * Mrc * Mt */
D3DXMATRIX* WINAPI D3DXMatrixTransformation
    ( D3DXMATRIX *pOut, CONST D3DXVECTOR3 *pScalingCenter,
      CONST D3DXQUATERNION *pScalingRotation, CONST D3DXVECTOR3 *pScaling,
      CONST D3DXVECTOR3 *pRotationCenter, CONST D3DXQUATERNION *pRotation,
      CONST D3DXVECTOR3 *pTranslation);

/* Build affine transformation matrix.  NULL arguments are treated as identity.
 * Mout = Ms * Mrc-1 * Mr * Mrc * Mt */
D3DXMATRIX* WINAPI D3DXMatrixAffineTransformation
    ( D3DXMATRIX *pOut, FLOAT Scaling, CONST D3DXVECTOR3 *pRotationCenter,
      CONST D3DXQUATERNION *pRotation, CONST D3DXVECTOR3 *pTranslation);

/* Build a lookat matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixLookAtRH
    ( D3DXMATRIX *pOut, CONST D3DXVECTOR3 *pEye, CONST D3DXVECTOR3 *pAt,
      CONST D3DXVECTOR3 *pUp );

/* Build a lookat matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixLookAtLH
    ( D3DXMATRIX *pOut, CONST D3DXVECTOR3 *pEye, CONST D3DXVECTOR3 *pAt,
      CONST D3DXVECTOR3 *pUp );

/* Build a perspective projection matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveRH
    ( D3DXMATRIX *pOut, FLOAT w, FLOAT h, FLOAT zn, FLOAT zf );

/* Build a perspective projection matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveLH
    ( D3DXMATRIX *pOut, FLOAT w, FLOAT h, FLOAT zn, FLOAT zf );

/* Build a perspective projection matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveFovRH
    ( D3DXMATRIX *pOut, FLOAT fovy, FLOAT Aspect, FLOAT zn, FLOAT zf );

/* Build a perspective projection matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveFovLH
    ( D3DXMATRIX *pOut, FLOAT fovy, FLOAT Aspect, FLOAT zn, FLOAT zf );

/* Build a perspective projection matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveOffCenterRH
    ( D3DXMATRIX *pOut, FLOAT l, FLOAT r, FLOAT b, FLOAT t, FLOAT zn,
      FLOAT zf );

/* Build a perspective projection matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixPerspectiveOffCenterLH
    ( D3DXMATRIX *pOut, FLOAT l, FLOAT r, FLOAT b, FLOAT t, FLOAT zn,
      FLOAT zf );

/* Build an ortho projection matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixOrthoRH
    ( D3DXMATRIX *pOut, FLOAT w, FLOAT h, FLOAT zn, FLOAT zf );

/* Build an ortho projection matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixOrthoLH
    ( D3DXMATRIX *pOut, FLOAT w, FLOAT h, FLOAT zn, FLOAT zf );

/* Build an ortho projection matrix. (right-handed) */
D3DXMATRIX* WINAPI D3DXMatrixOrthoOffCenterRH
    ( D3DXMATRIX *pOut, FLOAT l, FLOAT r, FLOAT b, FLOAT t, FLOAT zn,
      FLOAT zf );

/* Build an ortho projection matrix. (left-handed) */
D3DXMATRIX* WINAPI D3DXMatrixOrthoOffCenterLH
    ( D3DXMATRIX *pOut, FLOAT l, FLOAT r, FLOAT b, FLOAT t, FLOAT zn,
      FLOAT zf );

/* Build a matrix which flattens geometry into a plane, as if casting
 * a shadow from a light. */
D3DXMATRIX* WINAPI D3DXMatrixShadow
    ( D3DXMATRIX *pOut, CONST D3DXVECTOR4 *pLight,
      CONST D3DXPLANE *pPlane );

/* Build a matrix which reflects the coordinate system about a plane */
D3DXMATRIX* WINAPI D3DXMatrixReflect
    ( D3DXMATRIX *pOut, CONST D3DXPLANE *pPlane );

#ifdef __cplusplus
}
#endif


/*
 * Quaternion
 */

/* inline */

FLOAT D3DXQuaternionLength
    ( CONST D3DXQUATERNION *pQ );

/* Length squared, or "norm" */
FLOAT D3DXQuaternionLengthSq
    ( CONST D3DXQUATERNION *pQ );

FLOAT D3DXQuaternionDot
    ( CONST D3DXQUATERNION *pQ1, CONST D3DXQUATERNION *pQ2 );

/* (0, 0, 0, 1) */
D3DXQUATERNION* D3DXQuaternionIdentity
    ( D3DXQUATERNION *pOut );

BOOL D3DXQuaternionIsIdentity
    ( CONST D3DXQUATERNION *pQ );

/* (-x, -y, -z, w) */
D3DXQUATERNION* D3DXQuaternionConjugate
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ );


/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

/* Compute a quaternin's axis and angle of rotation. Expects unit quaternions. */
void WINAPI D3DXQuaternionToAxisAngle
    ( CONST D3DXQUATERNION *pQ, D3DXVECTOR3 *pAxis, FLOAT *pAngle );

/* Build a quaternion from a rotation matrix. */
D3DXQUATERNION* WINAPI D3DXQuaternionRotationMatrix
    ( D3DXQUATERNION *pOut, CONST D3DXMATRIX *pM);

/* Rotation about arbitrary axis. */
D3DXQUATERNION* WINAPI D3DXQuaternionRotationAxis
    ( D3DXQUATERNION *pOut, CONST D3DXVECTOR3 *pV, FLOAT Angle );

/* Yaw around the Y axis, a pitch around the X axis,
 * and a roll around the Z axis. */
D3DXQUATERNION* WINAPI D3DXQuaternionRotationYawPitchRoll
    ( D3DXQUATERNION *pOut, FLOAT Yaw, FLOAT Pitch, FLOAT Roll );

/* Quaternion multiplication.  The result represents the rotation Q2
 * followed by the rotation Q1.  (Out = Q2 * Q1) */
D3DXQUATERNION* WINAPI D3DXQuaternionMultiply
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ1,
      CONST D3DXQUATERNION *pQ2 );

D3DXQUATERNION* WINAPI D3DXQuaternionNormalize
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ );

/* Conjugate and re-norm */
D3DXQUATERNION* WINAPI D3DXQuaternionInverse
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ );

/* Expects unit quaternions.
 * if q = (cos(theta), sin(theta) * v); ln(q) = (0, theta * v) */
D3DXQUATERNION* WINAPI D3DXQuaternionLn
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ );

/* Expects pure quaternions. (w == 0)  w is ignored in calculation.
 * if q = (0, theta * v); exp(q) = (cos(theta), sin(theta) * v) */
D3DXQUATERNION* WINAPI D3DXQuaternionExp
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ );

/* Spherical linear interpolation between Q1 (t == 0) and Q2 (t == 1).
 * Expects unit quaternions. */
D3DXQUATERNION* WINAPI D3DXQuaternionSlerp
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ1,
      CONST D3DXQUATERNION *pQ2, FLOAT t );

/* Spherical quadrangle interpolation.
 * Slerp(Slerp(Q1, C, t), Slerp(A, B, t), 2t(1-t)) */
D3DXQUATERNION* WINAPI D3DXQuaternionSquad
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ1,
      CONST D3DXQUATERNION *pA, CONST D3DXQUATERNION *pB,
      CONST D3DXQUATERNION *pC, FLOAT t );

/* Setup control points for spherical quadrangle interpolation
 * from Q1 to Q2.  The control points are chosen in such a way
 * to ensure the continuity of tangents with adjacent segments. */
void WINAPI D3DXQuaternionSquadSetup
    ( D3DXQUATERNION *pAOut, D3DXQUATERNION *pBOut, D3DXQUATERNION *pCOut,
      CONST D3DXQUATERNION *pQ0, CONST D3DXQUATERNION *pQ1,
      CONST D3DXQUATERNION *pQ2, CONST D3DXQUATERNION *pQ3 );

/* Barycentric interpolation.
 * Slerp(Slerp(Q1, Q2, f+g), Slerp(Q1, Q3, f+g), g/(f+g)) */
D3DXQUATERNION* WINAPI D3DXQuaternionBaryCentric
    ( D3DXQUATERNION *pOut, CONST D3DXQUATERNION *pQ1,
      CONST D3DXQUATERNION *pQ2, CONST D3DXQUATERNION *pQ3,
      FLOAT f, FLOAT g );

#ifdef __cplusplus
}
#endif


/*
 * Plane
 */

/* inline */

/* ax + by + cz + dw */
FLOAT D3DXPlaneDot
    ( CONST D3DXPLANE *pP, CONST D3DXVECTOR4 *pV);

/* ax + by + cz + d */
FLOAT D3DXPlaneDotCoord
    ( CONST D3DXPLANE *pP, CONST D3DXVECTOR3 *pV);

/* ax + by + cz */
FLOAT D3DXPlaneDotNormal
    ( CONST D3DXPLANE *pP, CONST D3DXVECTOR3 *pV);

/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

/* Normalize plane (so that |a,b,c| == 1) */
D3DXPLANE* WINAPI D3DXPlaneNormalize
    ( D3DXPLANE *pOut, CONST D3DXPLANE *pP);

/* Find the intersection between a plane and a line.  If the line is
 * parallel to the plane, NULL is returned. */
D3DXVECTOR3* WINAPI D3DXPlaneIntersectLine
    ( D3DXVECTOR3 *pOut, CONST D3DXPLANE *pP, CONST D3DXVECTOR3 *pV1,
      CONST D3DXVECTOR3 *pV2);

/* Construct a plane from a point and a normal */
D3DXPLANE* WINAPI D3DXPlaneFromPointNormal
    ( D3DXPLANE *pOut, CONST D3DXVECTOR3 *pPoint, CONST D3DXVECTOR3 *pNormal);

/* Construct a plane from 3 points */
D3DXPLANE* WINAPI D3DXPlaneFromPoints
    ( D3DXPLANE *pOut, CONST D3DXVECTOR3 *pV1, CONST D3DXVECTOR3 *pV2,
      CONST D3DXVECTOR3 *pV3);

/* Transform a plane by a matrix.  The vector (a,b,c) must be normal.
 * M should be the inverse transpose of the transformation desired. */
D3DXPLANE* WINAPI D3DXPlaneTransform
    ( D3DXPLANE *pOut, CONST D3DXPLANE *pP, CONST D3DXMATRIX *pM );

#ifdef __cplusplus
}
#endif


/*
 * Color
 */

/* inline */

/* (1-r, 1-g, 1-b, a) */
D3DXCOLOR* D3DXColorNegative
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC);

D3DXCOLOR* D3DXColorAdd
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC1, CONST D3DXCOLOR *pC2);

D3DXCOLOR* D3DXColorSubtract
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC1, CONST D3DXCOLOR *pC2);

D3DXCOLOR* D3DXColorScale
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC, FLOAT s);

/* (r1*r2, g1*g2, b1*b2, a1*a2) */
D3DXCOLOR* D3DXColorModulate
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC1, CONST D3DXCOLOR *pC2);

/* Linear interpolation of r,g,b, and a. C1 + s(C2-C1) */
D3DXCOLOR* D3DXColorLerp
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC1, CONST D3DXCOLOR *pC2, FLOAT s);

/* non-inline */
#ifdef __cplusplus
extern "C" {
#endif

/* Interpolate r,g,b between desaturated color and color.
 * DesaturatedColor + s(Color - DesaturatedColor) */
D3DXCOLOR* WINAPI D3DXColorAdjustSaturation
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC, FLOAT s);

/* Interpolate r,g,b between 50% grey and color.  Grey + s(Color - Grey) */
D3DXCOLOR* WINAPI D3DXColorAdjustContrast
    (D3DXCOLOR *pOut, CONST D3DXCOLOR *pC, FLOAT c);

#ifdef __cplusplus
}
#endif

/*
 * Misc
 */

#ifdef __cplusplus
extern "C" {
#endif

/* Calculate Fresnel term given the cosine of theta (likely obtained by
 * taking the dot of two normals), and the refraction index of the material. */
FLOAT WINAPI D3DXFresnelTerm
    (FLOAT CosTheta, FLOAT RefractionIndex);

#ifdef __cplusplus
}
#endif

/*
 *
 *    Matrix Stack
 *
 */

typedef interface ID3DXMatrixStack ID3DXMatrixStack;
typedef interface ID3DXMatrixStack *LPD3DXMATRIXSTACK;

/* {E3357330-CC5E-11d2-A434-00A0C90629A8} */
DEFINE_GUID( IID_ID3DXMatrixStack,
0xe3357330, 0xcc5e, 0x11d2, 0xa4, 0x34, 0x0, 0xa0, 0xc9, 0x6, 0x29, 0xa8);


#undef INTERFACE
#define INTERFACE ID3DXMatrixStack

DECLARE_INTERFACE_(ID3DXMatrixStack, IUnknown)
{
    /* IUnknown methods */
    STDMETHOD(QueryInterface)(THIS_ REFIID riid, LPVOID * ppvObj) PURE;
    STDMETHOD_(ULONG,AddRef)(THIS) PURE;
    STDMETHOD_(ULONG,Release)(THIS) PURE;

    /* ID3DXMatrixStack methods */

    /* Pops the top of the stack, returns the current top
     * *after* popping the top. */
    STDMETHOD(Pop)(THIS) PURE;

    /* Pushes the stack by one, duplicating the current matrix. */
    STDMETHOD(Push)(THIS) PURE;

    /* Loads identity in the current matrix. */
    STDMETHOD(LoadIdentity)(THIS) PURE;

    /* Loads the given matrix into the current matrix */
    STDMETHOD(LoadMatrix)(THIS_ CONST D3DXMATRIX* pM ) PURE;

    /* Right-Multiplies the given matrix to the current matrix.
     * (transformation is about the current world origin) */
    STDMETHOD(MultMatrix)(THIS_ CONST D3DXMATRIX* pM ) PURE;

    /* Left-Multiplies the given matrix to the current matrix
     * (transformation is about the local origin of the object) */
    STDMETHOD(MultMatrixLocal)(THIS_ CONST D3DXMATRIX* pM ) PURE;

    /* Right multiply the current matrix with the computed rotation
     * matrix, counterclockwise about the given axis with the given angle.
     * (rotation is about the current world origin) */
    STDMETHOD(RotateAxis)
        (THIS_ CONST D3DXVECTOR3* pV, FLOAT Angle) PURE;

    /* Left multiply the current matrix with the computed rotation
     * matrix, counterclockwise about the given axis with the given angle.
     * (rotation is about the local origin of the object) */
    STDMETHOD(RotateAxisLocal)
        (THIS_ CONST D3DXVECTOR3* pV, FLOAT Angle) PURE;

    /* Right multiply the current matrix with the computed rotation
     * matrix. All angles are counterclockwise. (rotation is about the
     * current world origin)

     * The rotation is composed of a yaw around the Y axis, a pitch around
     * the X axis, and a roll around the Z axis. */
    STDMETHOD(RotateYawPitchRoll)
        (THIS_ FLOAT Yaw, FLOAT Pitch, FLOAT Roll) PURE;

    /* Left multiply the current matrix with the computed rotation
     * matrix. All angles are counterclockwise. (rotation is about the
     * local origin of the object)

     * The rotation is composed of a yaw around the Y axis, a pitch around
     * the X axis, and a roll around the Z axis. */
    STDMETHOD(RotateYawPitchRollLocal)
        (THIS_ FLOAT Yaw, FLOAT Pitch, FLOAT Roll) PURE;

    /* Right multiply the current matrix with the computed scale
     * matrix. (transformation is about the current world origin) */
    STDMETHOD(Scale)(THIS_ FLOAT x, FLOAT y, FLOAT z) PURE;

    /* Left multiply the current matrix with the computed scale
     * matrix. (transformation is about the local origin of the object) */
    STDMETHOD(ScaleLocal)(THIS_ FLOAT x, FLOAT y, FLOAT z) PURE;

    /* Right multiply the current matrix with the computed translation
     * matrix. (transformation is about the current world origin) */
    STDMETHOD(Translate)(THIS_ FLOAT x, FLOAT y, FLOAT z ) PURE;

    /* Left multiply the current matrix with the computed translation
     * matrix. (transformation is about the local origin of the object) */
    STDMETHOD(TranslateLocal)(THIS_ FLOAT x, FLOAT y, FLOAT z) PURE;

    /* Obtain the current matrix at the top of the stack */
    STDMETHOD_(D3DXMATRIX*, GetTop)(THIS) PURE;
};

#ifdef __cplusplus
extern "C" {
#endif

HRESULT WINAPI
    D3DXCreateMatrixStack(
        DWORD               Flags,
        LPD3DXMATRIXSTACK*  ppStack);

#ifdef __cplusplus
}
#endif

#include "d3dx8math.inl"

#pragma warning(default:4201)

#endif /* __D3DX8MATH_H__ */