Browse Source

WIP: move vmatrix functions to header and inline them

mathlib-optimize
nillerusr 1 year ago
parent
commit
7ceba77616
  1. 58
      mathlib/mathlib_base.cpp
  2. 1264
      mathlib/vmatrix.cpp
  3. 12
      public/mathlib/math_pfns.h
  4. 31
      public/mathlib/mathlib.h
  5. 2
      public/mathlib/vector4d.h
  6. 1798
      public/mathlib/vmatrix.h

58
mathlib/mathlib_base.cpp

@ -420,13 +420,6 @@ void MatrixGetColumn( const matrix3x4_t& in, int column, Vector &out ) @@ -420,13 +420,6 @@ void MatrixGetColumn( const matrix3x4_t& in, int column, Vector &out )
out.z = in[2][column];
}
void MatrixSetColumn( const Vector &in, int column, matrix3x4_t& out )
{
out[0][column] = in.x;
out[1][column] = in.y;
out[2][column] = in.z;
}
void MatrixScaleBy ( const float flScale, matrix3x4_t &out )
{
out[0][0] *= flScale;
@ -1092,57 +1085,6 @@ void SetScaleMatrix( float x, float y, float z, matrix3x4_t &dst ) @@ -1092,57 +1085,6 @@ void SetScaleMatrix( float x, float y, float z, matrix3x4_t &dst )
dst[2][0] = 0.0f; dst[2][1] = 0.0f; dst[2][2] = z; dst[2][3] = 0.0f;
}
//-----------------------------------------------------------------------------
// Purpose: Builds the matrix for a counterclockwise rotation about an arbitrary axis.
//
// | ax2 + (1 - ax2)cosQ axay(1 - cosQ) - azsinQ azax(1 - cosQ) + aysinQ |
// Ra(Q) = | axay(1 - cosQ) + azsinQ ay2 + (1 - ay2)cosQ ayaz(1 - cosQ) - axsinQ |
// | azax(1 - cosQ) - aysinQ ayaz(1 - cosQ) + axsinQ az2 + (1 - az2)cosQ |
//
// Input : mat -
// vAxisOrRot -
// angle -
//-----------------------------------------------------------------------------
void MatrixBuildRotationAboutAxis( const Vector &vAxisOfRot, float angleDegrees, matrix3x4_t &dst )
{
float radians;
float axisXSquared;
float axisYSquared;
float axisZSquared;
float fSin;
float fCos;
radians = angleDegrees * ( M_PI / 180.0 );
fSin = sin( radians );
fCos = cos( radians );
axisXSquared = vAxisOfRot[0] * vAxisOfRot[0];
axisYSquared = vAxisOfRot[1] * vAxisOfRot[1];
axisZSquared = vAxisOfRot[2] * vAxisOfRot[2];
// Column 0:
dst[0][0] = axisXSquared + (1 - axisXSquared) * fCos;
dst[1][0] = vAxisOfRot[0] * vAxisOfRot[1] * (1 - fCos) + vAxisOfRot[2] * fSin;
dst[2][0] = vAxisOfRot[2] * vAxisOfRot[0] * (1 - fCos) - vAxisOfRot[1] * fSin;
// Column 1:
dst[0][1] = vAxisOfRot[0] * vAxisOfRot[1] * (1 - fCos) - vAxisOfRot[2] * fSin;
dst[1][1] = axisYSquared + (1 - axisYSquared) * fCos;
dst[2][1] = vAxisOfRot[1] * vAxisOfRot[2] * (1 - fCos) + vAxisOfRot[0] * fSin;
// Column 2:
dst[0][2] = vAxisOfRot[2] * vAxisOfRot[0] * (1 - fCos) + vAxisOfRot[1] * fSin;
dst[1][2] = vAxisOfRot[1] * vAxisOfRot[2] * (1 - fCos) - vAxisOfRot[0] * fSin;
dst[2][2] = axisZSquared + (1 - axisZSquared) * fCos;
// Column 3:
dst[0][3] = 0;
dst[1][3] = 0;
dst[2][3] = 0;
}
//-----------------------------------------------------------------------------
// Computes the transpose
//-----------------------------------------------------------------------------

1264
mathlib/vmatrix.cpp

File diff suppressed because it is too large Load Diff

12
public/mathlib/math_pfns.h

@ -22,10 +22,16 @@ extern float (*pfFastCos)(float x); @@ -22,10 +22,16 @@ extern float (*pfFastCos)(float x);
// The following are not declared as macros because they are often used in limiting situations,
// and sometimes the compiler simply refuses to inline them for some reason
#define FastSqrt(x) (*pfSqrt)(x)
#define FastRSqrt(x) (*pfRSqrt)(x)
#define FastRSqrtFast(x) (*pfRSqrtFast)(x)
#define FastSqrt(x) sqrtf(x)
#define FastRSqrt(x) (1.f/sqrtf(x))
#define FastRSqrtFast(x) (1.f/sqrtf(x))
#ifdef _WIN32
#define FastSinCos(x,s,c) (*pfFastSinCos)(x,s,c)
#else
#define FastSinCos(x,s,c) sincosf(x,s,c)
#endif
#define FastCos(x) (*pfFastCos)(x)
#if defined(__i386__) || defined(_M_IX86)

31
public/mathlib/mathlib.h

@ -237,8 +237,8 @@ bool R_CullBoxSkipNear( const Vector& mins, const Vector& maxs, const Frustum_t @@ -237,8 +237,8 @@ bool R_CullBoxSkipNear( const Vector& mins, const Vector& maxs, const Frustum_t
struct matrix3x4_t
{
matrix3x4_t() = default;
matrix3x4_t(
inline matrix3x4_t() = default;
inline matrix3x4_t(
float m00, float m01, float m02, float m03,
float m10, float m11, float m12, float m13,
float m20, float m21, float m22, float m23 )
@ -252,7 +252,7 @@ struct matrix3x4_t @@ -252,7 +252,7 @@ struct matrix3x4_t
// Creates a matrix where the X axis = forward
// the Y axis = left, and the Z axis = up
//-----------------------------------------------------------------------------
void Init( const Vector& xAxis, const Vector& yAxis, const Vector& zAxis, const Vector &vecOrigin )
inline void Init( const Vector& xAxis, const Vector& yAxis, const Vector& zAxis, const Vector &vecOrigin )
{
m_flMatVal[0][0] = xAxis.x; m_flMatVal[0][1] = yAxis.x; m_flMatVal[0][2] = zAxis.x; m_flMatVal[0][3] = vecOrigin.x;
m_flMatVal[1][0] = xAxis.y; m_flMatVal[1][1] = yAxis.y; m_flMatVal[1][2] = zAxis.y; m_flMatVal[1][3] = vecOrigin.y;
@ -263,26 +263,23 @@ struct matrix3x4_t @@ -263,26 +263,23 @@ struct matrix3x4_t
// Creates a matrix where the X axis = forward
// the Y axis = left, and the Z axis = up
//-----------------------------------------------------------------------------
matrix3x4_t( const Vector& xAxis, const Vector& yAxis, const Vector& zAxis, const Vector &vecOrigin )
inline matrix3x4_t( const Vector& xAxis, const Vector& yAxis, const Vector& zAxis, const Vector &vecOrigin )
{
Init( xAxis, yAxis, zAxis, vecOrigin );
}
inline void Invalidate( void )
{
for (int i = 0; i < 3; i++)
for( int i=0; i < 12; i++ )
{
for (int j = 0; j < 4; j++)
{
m_flMatVal[i][j] = VEC_T_NAN;
}
((float*)m_flMatVal)[i] = VEC_T_NAN;
}
}
float *operator[]( int i ) { Assert(( i >= 0 ) && ( i < 3 )); return m_flMatVal[i]; }
const float *operator[]( int i ) const { Assert(( i >= 0 ) && ( i < 3 )); return m_flMatVal[i]; }
float *Base() { return &m_flMatVal[0][0]; }
const float *Base() const { return &m_flMatVal[0][0]; }
inline float *operator[]( int i ) { Assert(( i >= 0 ) && ( i < 3 )); return m_flMatVal[i]; }
inline const float *operator[]( int i ) const { Assert(( i >= 0 ) && ( i < 3 )); return m_flMatVal[i]; }
inline float *Base() { return &m_flMatVal[0][0]; }
inline const float *Base() const { return &m_flMatVal[0][0]; }
float m_flMatVal[3][4];
};
@ -565,7 +562,13 @@ void MatrixInvert( const matrix3x4_t &in, matrix3x4_t &out ); @@ -565,7 +562,13 @@ void MatrixInvert( const matrix3x4_t &in, matrix3x4_t &out );
bool MatricesAreEqual( const matrix3x4_t &src1, const matrix3x4_t &src2, float flTolerance = 1e-5 );
void MatrixGetColumn( const matrix3x4_t &in, int column, Vector &out );
void MatrixSetColumn( const Vector &in, int column, matrix3x4_t &out );
inline void MatrixSetColumn( const Vector &in, int column, matrix3x4_t& out )
{
out[0][column] = in.x;
out[1][column] = in.y;
out[2][column] = in.z;
}
inline void MatrixGetTranslation( const matrix3x4_t &in, Vector &out )
{

2
public/mathlib/vector4d.h

@ -35,7 +35,7 @@ class Vector2D; @@ -35,7 +35,7 @@ class Vector2D;
// 4D Vector4D
//=========================================================
class Vector4D
class alignas(16) Vector4D
{
public:
// Members

1798
public/mathlib/vmatrix.h

File diff suppressed because it is too large Load Diff
Loading…
Cancel
Save