368 lines
9.1 KiB
C++
368 lines
9.1 KiB
C++
#ifndef HGL_ALGORITHM_MATH_VECTOR_INCLUDE
|
||
#define HGL_ALGORITHM_MATH_VECTOR_INCLUDE
|
||
|
||
#ifdef _MSC_VER
|
||
#pragma warning(disable:4244) // double -> int 精度丢失警告
|
||
#pragma warning(disable:4996) // sprintf may be unsafe, Consider using sprintf_s instead
|
||
#endif//_MSC_VER
|
||
|
||
#include<hgl/math/FastTriangle.h>
|
||
#include<MathGeoLib.h>
|
||
|
||
/**
|
||
* MathGeoLib
|
||
* Game Math and Geometry Library
|
||
*
|
||
* My C++ library for 3D mathematics and geometry manipulation.
|
||
* Jukka Jylänki
|
||
*
|
||
* offical web: http://clb.demon.fi/MathGeoLib/nightly/
|
||
*
|
||
* License:
|
||
*
|
||
* This library is licensed under the Apache 2 license. I am not a lawyer, but to me that
|
||
* license means that you can use this code for any purpose, both commercial and closed source.
|
||
* You are however restricted from claiming you wrote it yourself, and cannot hold me liable
|
||
* for anything over this code.
|
||
* I acknowledge that most of the non-trivial math routines are taken off a book or a
|
||
* research paper. In all places, I have tried to be diligent to properly attribute the original
|
||
* source. Please contact me if you feel I have misattributed something.
|
||
*/
|
||
|
||
namespace hgl
|
||
{
|
||
using Vector2f=float2;
|
||
using Vector3f=float3;
|
||
using Vector4f=float4;
|
||
|
||
inline bool operator == (const Vector2f &lhs,const Vector2f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(false);
|
||
if(lhs.y!=rhs.y)return(false);
|
||
return(true);
|
||
}
|
||
|
||
inline bool operator != (const Vector2f &lhs,const Vector2f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(true);
|
||
if(lhs.y!=rhs.y)return(true);
|
||
return(false);
|
||
}
|
||
|
||
inline bool operator == (const Vector3f &lhs,const Vector3f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(false);
|
||
if(lhs.y!=rhs.y)return(false);
|
||
if(lhs.z!=rhs.z)return(false);
|
||
return(true);
|
||
}
|
||
|
||
inline bool operator != (const Vector3f &lhs,const Vector3f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(true);
|
||
if(lhs.y!=rhs.y)return(true);
|
||
if(lhs.z!=rhs.z)return(true);
|
||
return(false);
|
||
}
|
||
|
||
inline bool operator == (const Vector4f &lhs,const Vector4f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(false);
|
||
if(lhs.y!=rhs.y)return(false);
|
||
if(lhs.z!=rhs.z)return(false);
|
||
if(lhs.w!=rhs.w)return(false);
|
||
return(true);
|
||
}
|
||
|
||
inline bool operator != (const Vector4f &lhs,const Vector4f &rhs)
|
||
{
|
||
if(lhs.x!=rhs.x)return(true);
|
||
if(lhs.y!=rhs.y)return(true);
|
||
if(lhs.z!=rhs.z)return(true);
|
||
if(lhs.w!=rhs.w)return(true);
|
||
return(false);
|
||
}
|
||
|
||
inline void vec3to2(Vector2f &dst,const Vector3f &src)
|
||
{
|
||
dst.x=src.x;
|
||
dst.y=src.y;
|
||
}
|
||
|
||
inline Vector2f vec3to2(const Vector3f &src)
|
||
{
|
||
return Vector2f(src.x,src.y);
|
||
}
|
||
|
||
inline void vec2to3(Vector3f &dst,const Vector2f &src,const float z)
|
||
{
|
||
dst.x=src.x;
|
||
dst.y=src.y;
|
||
dst.z=z;
|
||
}
|
||
|
||
inline Vector3f vec2to3(const Vector2f &src,const float z)
|
||
{
|
||
return Vector3f(src.x,src.y,z);
|
||
}
|
||
|
||
template<typename T>
|
||
inline T clamp(const T &v,const T &min_v,const T &max_v)
|
||
{
|
||
if(v<min_v)return min_v;
|
||
if(v>max_v)return max_v;
|
||
return v;
|
||
}
|
||
|
||
template<typename T>
|
||
inline T normalized(const T &v)
|
||
{
|
||
return v.Normalized();
|
||
}
|
||
|
||
template<typename T>
|
||
inline void normalize(T &v)
|
||
{
|
||
v.Normalize();
|
||
}
|
||
|
||
template<typename T>
|
||
inline T cross(const T &v1,const T &v2)
|
||
{
|
||
return v1.Cross(v2);
|
||
}
|
||
|
||
template<typename T>
|
||
inline float dot(const T &v1,const T &v2)
|
||
{
|
||
return v1.Dot(v2);
|
||
}
|
||
|
||
template<typename T>
|
||
inline float dot2(const T &v)
|
||
{
|
||
return v.Dot(v);
|
||
}
|
||
|
||
inline float ray_angle_cos(const Ray &ray,const vec &pos)
|
||
{
|
||
return ray.dir.Dot((pos-ray.pos).Normalized());
|
||
}
|
||
|
||
inline float length_squared(const Vector2f &v)
|
||
{
|
||
return (v.x*v.x) + (v.y*v.y);
|
||
}
|
||
|
||
inline float length_squared_2d(const Vector3f &v)
|
||
{
|
||
return (v.x*v.x) + (v.y*v.y);
|
||
}
|
||
|
||
inline float length_squared(const Vector3f &v)
|
||
{
|
||
return (v.x*v.x) + (v.y*v.y) + (v.z*v.z);
|
||
}
|
||
|
||
inline float length_squared(const Vector4f &v)
|
||
{
|
||
return (v.x*v.x) + (v.y*v.y) + (v.z*v.z);
|
||
}
|
||
|
||
template<typename T>
|
||
inline float length(const T &v)
|
||
{
|
||
return sqrt(length_squared(v));
|
||
}
|
||
|
||
inline float length_2d(const Vector3f &v)
|
||
{
|
||
return sqrt(length_squared_2d(v));
|
||
}
|
||
|
||
template<typename T1, typename T2>
|
||
inline float length_squared(const T1 &v1, const T2 &v2)
|
||
{
|
||
const float x = (v1.x - v2.x);
|
||
const float y = (v1.y - v2.y);
|
||
|
||
return x*x + y*y;
|
||
}
|
||
|
||
template<typename T1, typename T2>
|
||
inline float length(const T1 &v1, const T2 &v2)
|
||
{
|
||
return sqrt(length_squared(v1, v2));
|
||
}
|
||
|
||
inline float length_squared(const Vector3f &v1, const Vector3f &v2)
|
||
{
|
||
const float x = (v1.x - v2.x);
|
||
const float y = (v1.y - v2.y);
|
||
const float z = (v1.z - v2.z);
|
||
|
||
return x*x + y*y + z*z;
|
||
}
|
||
|
||
template<typename T1, typename T2>
|
||
inline float length_squared_2d(const T1 &v1, const T2 &v2)
|
||
{
|
||
const float x = (v1.x - v2.x);
|
||
const float y = (v1.y - v2.y);
|
||
|
||
return x*x + y*y;
|
||
}
|
||
|
||
inline float length(const Vector3f &v1, const Vector3f &v2)
|
||
{
|
||
return sqrt(length_squared(v1, v2));
|
||
}
|
||
|
||
template<typename T1, typename T2>
|
||
inline float length_2d(const T1 &v1, const T2 &v2)
|
||
{
|
||
return sqrt(length_squared_2d(v1, v2));
|
||
}
|
||
|
||
inline Vector2f to(const Vector2f &start, const Vector2f &end, float pos)
|
||
{
|
||
return Vector2f(start.x + (end.x - start.x)*pos,
|
||
start.y + (end.y - start.y)*pos);
|
||
}
|
||
|
||
inline Vector3f to(const Vector3f &start, const Vector3f &end, float pos)
|
||
{
|
||
return Vector3f(start.x + (end.x - start.x)*pos,
|
||
start.y + (end.y - start.y)*pos,
|
||
start.z + (end.z - start.z)*pos);
|
||
}
|
||
|
||
template<typename T>
|
||
inline void to_2d(T &result, const T &start, const T &end, float pos)
|
||
{
|
||
result.x = start.x + (end.x - start.x)*pos;
|
||
result.y = start.y + (end.y - start.y)*pos;
|
||
}
|
||
|
||
inline float ray_angle_cos(const Vector3f &ray_dir, const Vector3f &ray_pos, const Vector3f &pos)
|
||
{
|
||
return dot(ray_dir, normalized(pos - ray_pos));
|
||
}
|
||
|
||
/**
|
||
* 做一个2D旋转计算
|
||
* @param result 结果
|
||
* @param source 原始点坐标
|
||
* @param center 圆心坐标
|
||
* @param ang 旋转角度
|
||
*/
|
||
template<typename T1, typename T2, typename T3>
|
||
inline void rotate2d(T1 &result, const T2 &source, const T3 ¢er, const double ang)
|
||
{
|
||
double as, ac;
|
||
// double nx,ny;
|
||
|
||
// as=sin(ang*(HGL_PI/180.0f));
|
||
// ac=cos(ang*(HGL_PI/180.0f));
|
||
//sincos(ang*(HGL_PI/180.0f),&as,&ac); //在80x87指令上,sin/cos是一个指令同时得出sin和cos,所以可以这样做
|
||
Lsincos(ang, as, ac); //低精度sin/cos计算
|
||
|
||
result.x = center.x + ((source.x - center.x)*ac - (source.y - center.y)*as);
|
||
result.y = center.y + ((source.x - center.x)*as + (source.y - center.y)*ac);
|
||
}
|
||
|
||
template<typename T> union vec2
|
||
{
|
||
struct { T x,y; };
|
||
struct { T r,g; };
|
||
struct { T u,v; };
|
||
T data[2];
|
||
|
||
public:
|
||
|
||
vec2(){x=y=0;}
|
||
vec2(T v1,T v2):x(v1),y(v2){}
|
||
vec2(const vec2 &v2)
|
||
{
|
||
x=v2.x;
|
||
y=v2.y;
|
||
}
|
||
|
||
vec2(const Vector2f &v2f)
|
||
{
|
||
x=v2f.x;
|
||
y=v2f.y;
|
||
}
|
||
|
||
operator const Vector2f()const{return Vector2f(x,y);}
|
||
};
|
||
|
||
template<typename T> union vec3
|
||
{
|
||
struct { T x,y,z; };
|
||
struct { T r,g,b; };
|
||
struct { T u,v,w; };
|
||
T data[3];
|
||
|
||
public:
|
||
|
||
vec3(){x=y=z=0;}
|
||
vec3(T v1,T v2,T v3):x(v1),y(v2),z(v3){}
|
||
vec3(const vec3 &v3)
|
||
{
|
||
x=v3.x;
|
||
y=v3.y;
|
||
z=v3.z;
|
||
}
|
||
|
||
vec3(const Vector3f &v3f)
|
||
{
|
||
x=v3f.x;
|
||
y=v3f.y;
|
||
z=v3f.z;
|
||
}
|
||
|
||
operator const Vector3f()const{return Vector3f(x,y,z);}
|
||
};
|
||
|
||
template<typename T> union vec4
|
||
{
|
||
struct { T x,y,z,w; };
|
||
struct { T r,g,b,a; };
|
||
T data[4];
|
||
|
||
public:
|
||
|
||
vec4(){x=y=z=w=0;}
|
||
vec4(T v1,T v2,T v3,T v4):x(v1),y(v2),z(v3),w(v4){}
|
||
vec4(const vec4 &v4)
|
||
{
|
||
x=v4.x;
|
||
y=v4.y;
|
||
z=v4.z;
|
||
w=v4.w;
|
||
}
|
||
|
||
vec4(const Vector4f &v4f)
|
||
{
|
||
x=v4f.x;
|
||
y=v4f.y;
|
||
z=v4f.z;
|
||
w=v4f.w;
|
||
}
|
||
|
||
operator const Vector4f()const{return Vector4f(x,y,z,w);}
|
||
};
|
||
|
||
inline const Vector3f PolarToVector3f(float yaw,float pitch)
|
||
{
|
||
return Vector3f(sinf(yaw) * cosf(pitch), sinf(pitch), cosf(yaw) * cosf(pitch));
|
||
}
|
||
|
||
inline const Vector4f PolarToVector4f(float yaw,float pitch)
|
||
{
|
||
return Vector4f(sinf(yaw) * cosf(pitch), sinf(pitch), cosf(yaw) * cosf(pitch), 0);
|
||
}
|
||
}//namespace hgl
|
||
#endif//HGL_ALGORITHM_MATH_VECTOR_INCLUDE
|