mirror of
https://github.com/DrBeef/Raze.git
synced 2024-11-22 20:21:20 +00:00
5eb9af1e00
IQM model support and a few bugfixes.
213 lines
No EOL
5.9 KiB
C++
213 lines
No EOL
5.9 KiB
C++
|
|
// Matrix class based on code from VSML:
|
|
|
|
/** ----------------------------------------------------------
|
|
* \class VSMathLib
|
|
*
|
|
* Lighthouse3D
|
|
*
|
|
* VSMathLib - Very Simple Matrix Library
|
|
*
|
|
* Full documentation at
|
|
* http://www.lighthouse3d.com/very-simple-libs
|
|
*
|
|
* This class aims at easing geometric transforms, camera
|
|
* placement and projection definition for programmers
|
|
* working with OpenGL core versions.
|
|
*
|
|
*
|
|
---------------------------------------------------------------*/
|
|
#ifndef __VSMatrix__
|
|
#define __VSMatrix__
|
|
|
|
#include <stdlib.h>
|
|
#include "vectors.h"
|
|
|
|
#ifdef USE_DOUBLE
|
|
typedef double FLOATTYPE;
|
|
#else
|
|
typedef float FLOATTYPE;
|
|
#endif
|
|
|
|
class VSMatrix {
|
|
|
|
public:
|
|
|
|
VSMatrix() = default;
|
|
|
|
VSMatrix(int)
|
|
{
|
|
loadIdentity();
|
|
}
|
|
|
|
void translate(FLOATTYPE x, FLOATTYPE y, FLOATTYPE z);
|
|
void scale(FLOATTYPE x, FLOATTYPE y, FLOATTYPE z);
|
|
void rotate(FLOATTYPE angle, FLOATTYPE x, FLOATTYPE y, FLOATTYPE z);
|
|
void loadIdentity();
|
|
#ifdef USE_DOUBLE
|
|
void multMatrix(const float *aMatrix);
|
|
#endif
|
|
void multVector(FLOATTYPE *aVector);
|
|
void multMatrix(const FLOATTYPE *aMatrix);
|
|
void multMatrix(const VSMatrix &aMatrix)
|
|
{
|
|
multMatrix(aMatrix.mMatrix);
|
|
}
|
|
void multQuaternion(const TVector4<FLOATTYPE>& q);
|
|
void loadMatrix(const FLOATTYPE *aMatrix);
|
|
#ifdef USE_DOUBLE
|
|
void loadMatrix(const float *aMatrix);
|
|
#endif
|
|
void lookAt(FLOATTYPE xPos, FLOATTYPE yPos, FLOATTYPE zPos, FLOATTYPE xLook, FLOATTYPE yLook, FLOATTYPE zLook, FLOATTYPE xUp, FLOATTYPE yUp, FLOATTYPE zUp);
|
|
void perspective(FLOATTYPE fov, FLOATTYPE ratio, FLOATTYPE nearp, FLOATTYPE farp);
|
|
void ortho(FLOATTYPE left, FLOATTYPE right, FLOATTYPE bottom, FLOATTYPE top, FLOATTYPE nearp=-1.0f, FLOATTYPE farp=1.0f);
|
|
void frustum(FLOATTYPE left, FLOATTYPE right, FLOATTYPE bottom, FLOATTYPE top, FLOATTYPE nearp, FLOATTYPE farp);
|
|
void copy(FLOATTYPE * pDest)
|
|
{
|
|
memcpy(pDest, mMatrix, 16 * sizeof(FLOATTYPE));
|
|
}
|
|
|
|
#ifdef USE_DOUBLE
|
|
void copy(float * pDest)
|
|
{
|
|
for (int i = 0; i < 16; i++)
|
|
{
|
|
pDest[i] = (float)mMatrix[i];
|
|
}
|
|
}
|
|
#endif
|
|
|
|
const FLOATTYPE *get() const
|
|
{
|
|
return mMatrix;
|
|
}
|
|
|
|
void multMatrixPoint(const FLOATTYPE *point, FLOATTYPE *res);
|
|
|
|
#ifdef USE_DOUBLE
|
|
void computeNormalMatrix(const float *aMatrix);
|
|
#endif
|
|
void computeNormalMatrix(const FLOATTYPE *aMatrix);
|
|
void computeNormalMatrix(const VSMatrix &aMatrix)
|
|
{
|
|
computeNormalMatrix(aMatrix.mMatrix);
|
|
}
|
|
bool inverseMatrix(VSMatrix &result);
|
|
void transpose();
|
|
|
|
protected:
|
|
static void crossProduct(const FLOATTYPE *a, const FLOATTYPE *b, FLOATTYPE *res);
|
|
static FLOATTYPE dotProduct(const FLOATTYPE *a, const FLOATTYPE * b);
|
|
static void normalize(FLOATTYPE *a);
|
|
static void subtract(const FLOATTYPE *a, const FLOATTYPE *b, FLOATTYPE *res);
|
|
static void add(const FLOATTYPE *a, const FLOATTYPE *b, FLOATTYPE *res);
|
|
static FLOATTYPE length(const FLOATTYPE *a);
|
|
static void multMatrix(FLOATTYPE *resMatrix, const FLOATTYPE *aMatrix);
|
|
|
|
static void setIdentityMatrix(FLOATTYPE *mat, int size = 4);
|
|
|
|
/// The storage for matrices
|
|
FLOATTYPE mMatrix[16];
|
|
|
|
};
|
|
|
|
|
|
class Matrix3x4 // used like a 4x4 matrix with the last row always being (0,0,0,1)
|
|
{
|
|
float m[3][4];
|
|
|
|
public:
|
|
|
|
void MakeIdentity()
|
|
{
|
|
memset(m, 0, sizeof(m));
|
|
m[0][0] = m[1][1] = m[2][2] = 1.f;
|
|
}
|
|
|
|
void Translate(float x, float y, float z)
|
|
{
|
|
m[0][3] = m[0][0]*x + m[0][1]*y + m[0][2]*z + m[0][3];
|
|
m[1][3] = m[1][0]*x + m[1][1]*y + m[1][2]*z + m[1][3];
|
|
m[2][3] = m[2][0]*x + m[2][1]*y + m[2][2]*z + m[2][3];
|
|
}
|
|
|
|
void Scale(float x, float y, float z)
|
|
{
|
|
m[0][0] *=x;
|
|
m[1][0] *=x;
|
|
m[2][0] *=x;
|
|
|
|
m[0][1] *=y;
|
|
m[1][1] *=y;
|
|
m[2][1] *=y;
|
|
|
|
m[0][2] *=z;
|
|
m[1][2] *=z;
|
|
m[2][2] *=z;
|
|
}
|
|
|
|
void Rotate(float ax, float ay, float az, float angle)
|
|
{
|
|
Matrix3x4 m1;
|
|
|
|
FVector3 axis(ax, ay, az);
|
|
axis.MakeUnit();
|
|
double c = cos(angle * pi::pi()/180.), s = sin(angle * pi::pi()/180.), t = 1 - c;
|
|
double sx = s*axis.X, sy = s*axis.Y, sz = s*axis.Z;
|
|
double tx, ty, txx, tyy, u, v;
|
|
|
|
tx = t*axis.X;
|
|
m1.m[0][0] = float( (txx=tx*axis.X) + c );
|
|
m1.m[0][1] = float( (u=tx*axis.Y) - sz);
|
|
m1.m[0][2] = float( (v=tx*axis.Z) + sy);
|
|
|
|
ty = t*axis.Y;
|
|
m1.m[1][0] = float( u + sz);
|
|
m1.m[1][1] = float( (tyy=ty*axis.Y) + c );
|
|
m1.m[1][2] = float( (u=ty*axis.Z) - sx);
|
|
|
|
m1.m[2][0] = float( v - sy);
|
|
m1.m[2][1] = float( u + sx);
|
|
m1.m[2][2] = float( (t-txx-tyy) + c );
|
|
|
|
m1.m[0][3] = 0.f;
|
|
m1.m[1][3] = 0.f;
|
|
m1.m[2][3] = 0.f;
|
|
|
|
*this = (*this) * m1;
|
|
}
|
|
|
|
Matrix3x4 operator *(const Matrix3x4 &other)
|
|
{
|
|
Matrix3x4 result;
|
|
|
|
result.m[0][0] = m[0][0]*other.m[0][0] + m[0][1]*other.m[1][0] + m[0][2]*other.m[2][0];
|
|
result.m[0][1] = m[0][0]*other.m[0][1] + m[0][1]*other.m[1][1] + m[0][2]*other.m[2][1];
|
|
result.m[0][2] = m[0][0]*other.m[0][2] + m[0][1]*other.m[1][2] + m[0][2]*other.m[2][2];
|
|
result.m[0][3] = m[0][0]*other.m[0][3] + m[0][1]*other.m[1][3] + m[0][2]*other.m[2][3] + m[0][3];
|
|
|
|
result.m[1][0] = m[1][0]*other.m[0][0] + m[1][1]*other.m[1][0] + m[1][2]*other.m[2][0];
|
|
result.m[1][1] = m[1][0]*other.m[0][1] + m[1][1]*other.m[1][1] + m[1][2]*other.m[2][1];
|
|
result.m[1][2] = m[1][0]*other.m[0][2] + m[1][1]*other.m[1][2] + m[1][2]*other.m[2][2];
|
|
result.m[1][3] = m[1][0]*other.m[0][3] + m[1][1]*other.m[1][3] + m[1][2]*other.m[2][3] + m[1][3];
|
|
|
|
result.m[2][0] = m[2][0]*other.m[0][0] + m[2][1]*other.m[1][0] + m[2][2]*other.m[2][0];
|
|
result.m[2][1] = m[2][0]*other.m[0][1] + m[2][1]*other.m[1][1] + m[2][2]*other.m[2][1];
|
|
result.m[2][2] = m[2][0]*other.m[0][2] + m[2][1]*other.m[1][2] + m[2][2]*other.m[2][2];
|
|
result.m[2][3] = m[2][0]*other.m[0][3] + m[2][1]*other.m[1][3] + m[2][2]*other.m[2][3] + m[2][3];
|
|
|
|
return result;
|
|
}
|
|
|
|
FVector3 operator *(const FVector3 &vec)
|
|
{
|
|
FVector3 result;
|
|
|
|
result.X = vec.X*m[0][0] + vec.Y*m[0][1] + vec.Z*m[0][2] + m[0][3];
|
|
result.Y = vec.X*m[1][0] + vec.Y*m[1][1] + vec.Z*m[1][2] + m[1][3];
|
|
result.Z = vec.X*m[2][0] + vec.Y*m[2][1] + vec.Z*m[2][2] + m[2][3];
|
|
return result;
|
|
}
|
|
};
|
|
|
|
#endif |