Split up mathlib.h into several smaller headers.

This should make maintaining them a little easier.

The copyright block in most of the new headers (execpt vector.h) reflect
when the functions in the relevant header were first created.
This commit is contained in:
Bill Currie 2012-08-19 17:33:46 +09:00
parent a2f2d4d949
commit 54b4d589aa
8 changed files with 896 additions and 624 deletions

View file

@ -20,5 +20,8 @@ nobase_pkginclude_HEADERS = \
GLSL/qf_funcs_list.h GLSL/qf_iqm.h GLSL/qf_lightmap.h GLSL/qf_particles.h \
GLSL/qf_textures.h GLSL/qf_vid.h GLSL/types.h \
\
math/dual.h math/half.h math/matrix3.h math/matrix4.h math/quaternion.h \
math/vector.h \
\
plugin/cd.h plugin/console.h plugin/general.h plugin/input.h \
plugin/snd_output.h plugin/snd_render.h plugin/vid_render.h

194
include/QF/math/dual.h Normal file
View file

@ -0,0 +1,194 @@
/*
dual.h
Dual and dual-quaternion functions
Copyright (C) 2012 Bill Currie <bill@taniwha.org>
Author: Bill Currie <bill@taniwha.org>
Date: 2012/5/1
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_dual_h
#define __QF_math_dual_h
/** \defgroup mathlib_dual Dual and dual quaternion functions
\ingroup utils
*/
//@{
#include "QF/qtypes.h"
#define DualAdd(a,b,c) \
do { \
(c).r = (a).r + (b).r; \
(c).e = (a).e + (b).e; \
} while (0)
#define DualSubtract(a,b,c) \
do { \
(c).r = (a).r - (b).r; \
(c).e = (a).e - (b).e; \
} while (0)
#define DualNegate(a,b) \
do { \
(b).r = -(a).r; \
(b).e = -(a).e; \
} while (0)
#define DualConj(a,b) \
do { \
(b).r = (a).r; \
(b).e = -(a).e; \
} while (0)
#define DualMult(a,b,c) \
do { \
(c).e = (a).r * (b).e + (a).e * (b).r; \
(c).r = (a).r * (b).r; \
} while (0)
#define DualMultAdd(a,s,b,c) \
do { \
(c).r = (a).r + (s) * (b).r; \
(c).e = (a).e + (s) * (b).e; \
} while (0)
#define DualMultSub(a,s,b,c) \
do { \
(c).r = (a).r - (s) * (b).r; \
(c).e = (a).e - (s) * (b).e; \
} while (0)
#define DualNorm(a) ((a).r)
#define DualScale(a,b,c) \
do { \
(c).r = (a).r * (b); \
(c).e = (a).e * (b); \
} while (0)
#define DualCompCompare(x, op, y) ((x).r op (y).r) && ((x).e op (y).e)
#define DualCompare(x, y) DualCompCompare (x, ==, y)
#define DualIsZero(a) ((a).r == 0 && (a).e == 0)
#define DualIsUnit(a) (((a).r - 1) * ((a).r - 1) < 1e-6 && (a).e * (a).e < 1e-6)
#define DualSet(ar,ae,a) \
do { \
(a).ar = r; \
(a).er = r; \
} while (0)
#define DualZero(a) \
do { \
(a).e = (a).r = 0; \
} while (0)
#define DualBlend(d1,d2,b,d) \
do { \
(d).r = (d1).r * (1 - (b)) + (d2).r * (b); \
(d).e = (d1).e * (1 - (b)) + (d2).e * (b); \
} while (0)
#define DualExpand(d) (d).r, (d).e
#define DualQuatAdd(a,b,c) \
do { \
QuatAdd ((a).q0.q, (b).q0.q, (c).q0.q); \
QuatAdd ((a).qe.q, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatSubtract(a,b,c) \
do { \
QuatSub ((a).q0.q, (b).q0.q, (c).q0.q); \
QuatSub ((a).qe.q, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatNegate(a,b) \
do { \
QuatNegate ((a).q0.q, (b).q0.q); \
QuatNegate ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjQ(a,b) \
do { \
QuatConj ((a).q0.q, (b).q0.q); \
QuatConj ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjE(a,b) \
do { \
(b).q0 = (a).q0; \
QuatNegate ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjQE(a,b) \
do { \
QuatConj ((a).q0.q, (b).q0.q); \
(b).qe.sv.s = -(a).qe.sv.s; \
VectorCopy ((a).qe.sv.v, (b).qe.sv.v); \
} while (0)
#define DualQuatMult(a,b,c) \
do { \
Quat_t t; \
QuatMult ((a).q0.q, (b).qe.q, t.q); \
QuatMult ((a).qe.q, (b).q0.q, (c).qe.q); \
QuatAdd (t.q, (c).qe.q, (c).qe.q); \
QuatMult ((a).q0.q, (b).q0.q, (c).q0.q); \
} while (0);
#define DualQuatMultAdd(a,s,b,c) \
do { \
QuatMultAdd ((a).q0.q, s, (b).q0.q, (c).q0.q); \
QuatMultAdd ((a).qe.q, s, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatMultSub(a,s,b,c) \
do { \
QuatMultSub ((a).q0.q, s, (b).q0.q, (c).q0.q); \
QuatMultSub ((a).qe.q, s, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatNorm(a,b) \
do { \
(b).r = QuatLength ((a).q0.q); \
(b).e = 2 * QDotProduct ((a).q0.q, (a).qe.q); \
} while (0)
#define DualQuatScale(a,b,c) \
do { \
QuatSub ((a).q0.q, (b), (c).q0.q); \
QuatSub ((a).qe.q, (b), (c).qe.q); \
} while (0)
#define DualQuatCompCompare(x, op, y) \
(QuatCompCompare ((x).q0.q, op, (y).q0.q) \
&&QuatCompCompare ((x).qe.q, op, (y).qe.q))
#define DualQuatCompare(x, y) DualQuatCompCompare (x, ==, y)
#define DualQuatIsZero(a) (QuatIsZero ((a).q0.q) && QuatIsZero ((a).qe.q))
#define DualQuatSetVect(vec, a) \
do { \
(a).q0.sv.s = 1; \
VectorZero ((a).q0.sv.v); \
(a).qe.sv.s = 0; \
VectorCopy (vec, (a).qe.sv.v); \
} while (0)
#define DualQuatRotTrans(rot, trans, dq) \
do { \
QuatCopy (rot, (dq).q0.q); \
(dq).qe.sv.s = 0; \
VectorScale (trans, 0.5, (dq).qe.sv.v); \
QuatMult ((dq).qe.q, (dq).q0.q, (dq).qe.q); \
} while (0)
#define DualQuatZero(a) \
do { \
QuatZero ((a).q0.q); \
QuatZero ((a).qe.q); \
} while (0)
#define DualQuatBlend(dq1,dq2,b,dq) \
do { \
QuatBlend ((dq1).q0.q, (dq2).q0.q, (b), (dq).q0.q); \
QuatBlend ((dq1).qe.q, (dq2).qe.q, (b), (dq).qe.q); \
} while (0)
#define DualQuatExpand(dq) QuatExpand ((dq).q0.q), QuatExpand ((dq).qe.q)
//@}
#endif // __QF_math_dual_h

47
include/QF/math/half.h Normal file
View file

@ -0,0 +1,47 @@
/*
half.h
Half-float functions
Copyright (C) 2012 Bill Currie <bill@taniwha.org>
Author: Bill Currie <bill@taniwha.org>
Date: 2012/4/26
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_half_h
#define __QF_math_half_h
/** \defgroup mathlib_half Half-float functions
\ingroup utils
*/
//@{
#include "QF/qtypes.h"
int16_t FloatToHalf (float x);
float HalfToFloat (int16_t x);
//@}
#endif // __QF_math_half_h

123
include/QF/math/matrix3.h Normal file
View file

@ -0,0 +1,123 @@
/*
matrix3.h
3x3 Matrix functions
Copyright (C) 2012 Bill Currie <bill@taniwha.org>
Author: Bill Currie <bill@taniwha.org>
Date: 2012/8/18
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_matrix3_h
#define __QF_math_matrix3_h
/** \defgroup mathlib_matrix3 3x3 matrix functions
\ingroup utils
*/
//@{
#include "QF/qtypes.h"
#define Mat3Copy(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); \
VectorCopy ((a) + 3, (b) + 3); \
VectorCopy ((a) + 6, (b) + 6); \
} while (0)
#define Mat3Add(a, b, c) \
do { \
VectorAdd ((a) + 0, (b) + 0, (c) + 0); \
VectorAdd ((a) + 3, (b) + 3, (c) + 3); \
VectorAdd ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Subtract(a, b, c) \
do { \
VectorSubtract ((a) + 0, (b) + 0, (c) + 0); \
VectorSubtract ((a) + 3, (b) + 3, (c) + 3); \
VectorSubtract ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Scale(a, b, c) \
do { \
VectorScale ((a) + 0, (b), (c) + 0); \
VectorScale ((a) + 3, (b), (c) + 3); \
VectorScale ((a) + 6, (b), (c) + 6); \
} while (0)
#define Mat3CompMult(a, b, c) \
do { \
VectorCompMult ((a) + 0, (b) + 0, (c) + 0); \
VectorCompMult ((a) + 3, (b) + 3, (c) + 3); \
VectorCompMult ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Zero(a) \
memset ((a), 0, 9 * sizeof (a)[0])
#define Mat3Identity(a) \
do { \
Mat3Zero (a); \
(a)[8] = (a)[4] = (a)[0] = 1; \
} while (0)
#define Mat3Expand(a) \
VectorExpand ((a) + 0), \
VectorExpand ((a) + 3), \
VectorExpand ((a) + 6)
#define Mat3Blend(m1,m2,b,m) \
do { \
VectorBlend ((m1) + 0, (m2) + 0, (b), (m) + 0); \
VectorBlend ((m1) + 3, (m2) + 3, (b), (m) + 3); \
VectorBlend ((m1) + 6, (m2) + 6, (b), (m) + 6); \
} while (0)
#define Mat3MultAdd(a,s,b,c) \
do { \
VectorMultAdd ((a) + 0, s, (b) + 0, (c) + 0); \
VectorMultAdd ((a) + 3, s, (b) + 3, (c) + 3); \
VectorMultAdd ((a) + 6, s, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Trace(a) ((a)[0] + (a)[4] + (a)[8])
#define Mat4toMat3(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); \
VectorCopy ((a) + 4, (b) + 3); \
VectorCopy ((a) + 8, (b) + 6); \
} while (0)
void Mat3Init (const quat_t rot, const vec3_t scale, mat3_t mat);
void Mat3Transpose (const mat3_t a, mat3_t b);
vec_t Mat3Determinant (const mat3_t m);
int Mat3Inverse (const mat3_t a, mat3_t b);
void Mat3Mult (const mat3_t a, const mat3_t b, mat3_t c);
void Mat3MultVec (const mat3_t a, const vec3_t b, vec3_t c);
void Mat3SymEigen (const mat3_t m, vec3_t e);
/** Decompose a 3x3 column major matrix into its component transformations.
This gives the matrix's rotation as a quaternion, shear (XY, XZ, YZ),
and scale. Using the following sequence will give the same result as
multiplying \a v by \a mat.
QuatMultVec (rot, v, v);
VectorShear (shear, v, v);
VectorCompMult (scale, v, v);
*/
int Mat3Decompose (const mat4_t mat, quat_t rot, vec3_t shear, vec3_t scale);
//@}
#endif // __QF_math_matrix3_h

135
include/QF/math/matrix4.h Normal file
View file

@ -0,0 +1,135 @@
/*
matrix4.h
4x4 Matrix functions
Copyright (C) 2011 Bill Currie <bill@taniwha.org>
Author: Bill Currie <bill@taniwha.org>
Date: 2011/12/30
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_matrix4_h
#define __QF_math_matrix4_h
/** \defgroup mathlib_matrix4 4x4 matrix functions
\ingroup utils
*/
//@{
#include "QF/qtypes.h"
#define Mat4Copy(a, b) \
do { \
QuatCopy ((a) + 0, (b) + 0); \
QuatCopy ((a) + 4, (b) + 4); \
QuatCopy ((a) + 8, (b) + 8); \
QuatCopy ((a) + 12, (b) + 12); \
} while (0)
#define Mat4Add(a, b, c) \
do { \
QuatAdd ((a) + 0, (b) + 0, (c) + 0); \
QuatAdd ((a) + 4, (b) + 4, (c) + 4); \
QuatAdd ((a) + 8, (b) + 8, (c) + 8); \
QuatAdd ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Subtract(a, b, c) \
do { \
QuatSubtract ((a) + 0, (b) + 0, (c) + 0); \
QuatSubtract ((a) + 4, (b) + 4, (c) + 4); \
QuatSubtract ((a) + 8, (b) + 8, (c) + 8); \
QuatSubtract ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Scale(a, b, c) \
do { \
QuatScale ((a) + 0, (b), (c) + 0); \
QuatScale ((a) + 4, (b), (c) + 4); \
QuatScale ((a) + 8, (b), (c) + 8); \
QuatScale ((a) + 12, (b), (c) + 12); \
} while (0)
#define Mat4CompMult(a, b, c) \
do { \
QuatCompMult ((a) + 0, (b) + 0, (c) + 0); \
QuatCompMult ((a) + 4, (b) + 4, (c) + 4); \
QuatCompMult ((a) + 8, (b) + 8, (c) + 8); \
QuatCompMult ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Zero(a) \
memset ((a), 0, 16 * sizeof (a)[0])
#define Mat4Identity(a) \
do { \
Mat4Zero (a); \
(a)[15] = (a)[10] = (a)[5] = (a)[0] = 1; \
} while (0)
#define Mat4Expand(a) \
QuatExpand ((a) + 0), \
QuatExpand ((a) + 4), \
QuatExpand ((a) + 8), \
QuatExpand ((a) + 12)
#define Mat4Blend(m1,m2,b,m) \
do { \
QuatBlend ((m1) + 0, (m2) + 0, (b), (m) + 0); \
QuatBlend ((m1) + 4, (m2) + 4, (b), (m) + 4); \
QuatBlend ((m1) + 8, (m2) + 8, (b), (m) + 8); \
QuatBlend ((m1) + 12, (m2) + 12, (b), (m) + 12); \
} while (0)
#define Mat4MultAdd(a,s,b,c) \
do { \
QuatMultAdd ((a) + 0, s, (b) + 0, (c) + 0); \
QuatMultAdd ((a) + 4, s, (b) + 4, (c) + 4); \
QuatMultAdd ((a) + 8, s, (b) + 8, (c) + 8); \
QuatMultAdd ((a) + 12, s, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Trace(a) ((a)[0] + (a)[5] + (a)[10] + (a)[15])
#define Mat3toMat4(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); (b)[3] = 0; \
VectorCopy ((a) + 3, (b) + 4); (b)[7] = 0; \
VectorCopy ((a) + 6, (b) + 8); (b)[11] = 0; \
QuatZero ((b) + 12); \
} while (0)
void Mat4Init (const quat_t rot, const vec3_t scale, const vec3_t trans,
mat4_t mat);
void Mat4Transpose (const mat4_t a, mat4_t b);
int Mat4Inverse (const mat4_t a, mat4_t b);
void Mat4Mult (const mat4_t a, const mat4_t b, mat4_t c);
void Mat4MultVec (const mat4_t a, const vec3_t b, vec3_t c);
void Mat4as3MultVec (const mat4_t a, const vec3_t b, vec3_t c);
/** Decompose a 4x4 column major matrix into its component transformations.
This gives the matrix's rotation as a quaternion, shear (XY, XZ, YZ),
scale, and translation. Using the following sequence will give the
same result as multiplying \a v by \a mat.
QuatMultVec (rot, v, v);
VectorShear (shear, v, v);
VectorCompMult (scale, v, v);
VectorAdd (trans, v, v);
*/
int Mat4Decompose (const mat4_t mat, quat_t rot, vec3_t shear, vec3_t scale,
vec3_t trans);
//@}
#endif // __QF_math_matrix4_h

View file

@ -0,0 +1,173 @@
/*
quaternion.h
Quaternion functions
Copyright (C) 2004 Bill Currie <bill@taniwha.org>
Author: Bill Currie <bill@taniwha.org>
Date: 2004/4/7
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_quaternion_h
#define __QF_math_quaternion_h
/** \defgroup mathlib_quaternion Quaternion functions
\ingroup utils
*/
//@{
#include "QF/qtypes.h"
extern const vec_t *const quat_origin;
#define QDotProduct(a,b) ((a)[0] * (b)[0] + (a)[1] * (b)[1] \
+ (a)[2] * (b)[2] + (a)[3] * (b)[3])
#define QuatSubtract(a,b,c) \
do { \
(c)[0] = (a)[0] - (b)[0]; \
(c)[1] = (a)[1] - (b)[1]; \
(c)[2] = (a)[2] - (b)[2]; \
(c)[3] = (a)[3] - (b)[3]; \
} while (0)
#define QuatNegate(a,b) \
do { \
(b)[0] = -(a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
(b)[3] = -(a)[3]; \
} while (0)
#define QuatConj(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
(b)[3] = -(a)[3]; \
} while (0)
#define QuatAdd(a,b,c) \
do { \
(c)[0] = (a)[0] + (b)[0]; \
(c)[1] = (a)[1] + (b)[1]; \
(c)[2] = (a)[2] + (b)[2]; \
(c)[3] = (a)[3] + (b)[3]; \
} while (0)
#define QuatCopy(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = (a)[1]; \
(b)[2] = (a)[2]; \
(b)[3] = (a)[3]; \
} while (0)
#define QuatMultAdd(a,s,b,c) \
do { \
(c)[0] = (a)[0] + (s) * (b)[0]; \
(c)[1] = (a)[1] + (s) * (b)[1]; \
(c)[2] = (a)[2] + (s) * (b)[2]; \
(c)[3] = (a)[3] + (s) * (b)[3]; \
} while (0)
#define QuatMultSub(a,s,b,c) \
do { \
(c)[0] = (a)[0] - (s) * (b)[0]; \
(c)[1] = (a)[1] - (s) * (b)[1]; \
(c)[2] = (a)[2] - (s) * (b)[2]; \
(c)[3] = (a)[3] - (s) * (b)[3]; \
} while (0)
#define QuatLength(a) sqrt(QDotProduct(a, a))
#define QuatScale(a,b,c) \
do { \
(c)[0] = (a)[0] * (b); \
(c)[1] = (a)[1] * (b); \
(c)[2] = (a)[2] * (b); \
(c)[3] = (a)[3] * (b); \
} while (0)
#define QuatCompMult(a,b,c) \
do { \
(c)[0] = (a)[0] * (b)[0]; \
(c)[1] = (a)[1] * (b)[1]; \
(c)[2] = (a)[2] * (b)[2]; \
(c)[3] = (a)[3] * (b)[3]; \
} while (0)
#define QuatCompDiv(a,b,c) \
do { \
(c)[0] = (a)[0] / (b)[0]; \
(c)[1] = (a)[1] / (b)[1]; \
(c)[2] = (a)[2] / (b)[2]; \
(c)[3] = (a)[3] / (b)[3]; \
} while (0)
#define QuatCompCompare(x, op, y) \
(((x)[0] op (y)[0]) && ((x)[1] op (y)[1]) \
&& ((x)[2] op (y)[2]) && ((x)[3] op (y)[3]))
#define QuatCompare(x, y) QuatCompCompare (x, ==, y)
#define QuatCompMin(a, b, c) \
do { \
(c)[0] = min ((a)[0], (b)[0]); \
(c)[1] = min ((a)[1], (b)[1]); \
(c)[2] = min ((a)[2], (b)[2]); \
(c)[3] = min ((a)[3], (b)[3]); \
} while (0)
#define QuatCompMax(a, b, c) \
do { \
(c)[0] = max ((a)[0], (b)[0]); \
(c)[1] = max ((a)[1], (b)[1]); \
(c)[2] = max ((a)[2], (b)[2]); \
(c)[3] = max ((a)[3], (b)[3]); \
} while (0)
#define QuatCompBound(a, b, c, d) \
do { \
(d)[0] = bound ((a)[0], (b)[0], (c)[0]); \
(d)[1] = bound ((a)[1], (b)[1], (c)[1]); \
(d)[2] = bound ((a)[2], (b)[2], (c)[2]); \
(d)[3] = bound ((a)[3], (b)[3], (c)[3]); \
} while (0)
#define QuatIsZero(a) (!(a)[0] && !(a)[1] && !(a)[2] && !(a)[3])
#define QuatZero(a) ((a)[3] = (a)[2] = (a)[1] = (a)[0] = 0);
#define QuatSet(a,b,c,d,e) \
do { \
(e)[0] = a; \
(e)[1] = b; \
(e)[2] = c; \
(e)[3] = d; \
} while (0)
#define QuatBlend(q1,q2,b,q) \
do { \
(q)[0] = (q1)[0] * (1 - (b)) + (q2)[0] * (b); \
(q)[1] = (q1)[1] * (1 - (b)) + (q2)[1] * (b); \
(q)[2] = (q1)[2] * (1 - (b)) + (q2)[2] * (b); \
(q)[3] = (q1)[3] * (1 - (b)) + (q2)[3] * (b); \
} while (0)
//For printf etc
#define QuatExpand(q) (q)[0], (q)[1], (q)[2], (q)[3]
void QuatMult (const quat_t q1, const quat_t q2, quat_t out);
void QuatMultVec (const quat_t q, const vec3_t v, vec3_t out);
void QuatInverse (const quat_t in, quat_t out);
void QuatExp (const quat_t a, quat_t b);
void QuatToMatrix (const quat_t q, vec_t *m, int homogenous, int vertical);
//@}
#endif // __QF_math_quaternion_h

214
include/QF/math/vector.h Normal file
View file

@ -0,0 +1,214 @@
/*
mathlib.h
Vector math library
Copyright (C) 1996-1997 Id Software, Inc.
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to:
Free Software Foundation, Inc.
59 Temple Place - Suite 330
Boston, MA 02111-1307, USA
*/
#ifndef __QF_math_vector_h
#define __QF_math_vector_h
/** \defgroup mathlib_vector Vector functions
\ingroup mathlib
*/
//@{
#include "QF/qtypes.h"
extern const vec_t *const vec3_origin;
#define DotProduct(a,b) ((a)[0] * (b)[0] + (a)[1] * (b)[1] + (a)[2] * (b)[2])
#define VectorSubtract(a,b,c) \
do { \
(c)[0] = (a)[0] - (b)[0]; \
(c)[1] = (a)[1] - (b)[1]; \
(c)[2] = (a)[2] - (b)[2]; \
} while (0)
#define VectorNegate(a,b) \
do { \
(b)[0] = -(a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
} while (0)
#define VectorAdd(a,b,c) \
do { \
(c)[0] = (a)[0] + (b)[0]; \
(c)[1] = (a)[1] + (b)[1]; \
(c)[2] = (a)[2] + (b)[2]; \
} while (0)
#define VectorCopy(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = (a)[1]; \
(b)[2] = (a)[2]; \
} while (0)
#define VectorMultAdd(a,s,b,c) \
do { \
(c)[0] = (a)[0] + (s) * (b)[0]; \
(c)[1] = (a)[1] + (s) * (b)[1]; \
(c)[2] = (a)[2] + (s) * (b)[2]; \
} while (0)
#define VectorMultSub(a,s,b,c) \
do { \
(c)[0] = (a)[0] - (s) * (b)[0]; \
(c)[1] = (a)[1] - (s) * (b)[1]; \
(c)[2] = (a)[2] - (s) * (b)[2]; \
} while (0)
#define VectorLength(a) sqrt(DotProduct(a, a))
#define VectorScale(a,b,c) \
do { \
(c)[0] = (a)[0] * (b); \
(c)[1] = (a)[1] * (b); \
(c)[2] = (a)[2] * (b); \
} while (0)
/** Shear vector \a b by vector \a a.
Vector a represents the shear factors XY, XZ, YZ, ie in matrix form:
[ 1 0 0 ] [ b0 ]
[ a0 1 0 ] [ b1 ]
[ a1 a2 1 ] [ b2 ]
The reason for this particular scheme is that is how Mat4Decompose
calculates the shear from a matrix.
\note The order of calculations is important for when b and c refer to
the same vector.
*/
#define VectorShear(a,b,c) \
do { \
(c)[2] = (b)[0] * (a)[1] + (b)[1] * (a)[2] + (b)[2]; \
(c)[1] = (b)[0] * (a)[0] + (b)[1]; \
(c)[0] = (b)[0]; \
} while (0)
#define VectorUnshear(a,b,c) \
do { \
(c)[2] = (b)[2] - (b)[1] * (a)[2] - (b)[0] * ((a)[1]-(a)[0]*(a)[2]); \
(c)[1] = (b)[1] - (b)[0] * (a)[0]; \
(c)[0] = (b)[0]; \
} while (0)
#define VectorCompMult(a,b,c) \
do { \
(c)[0] = (a)[0] * (b)[0]; \
(c)[1] = (a)[1] * (b)[1]; \
(c)[2] = (a)[2] * (b)[2]; \
} while (0)
#define VectorCompDiv(a,b,c) \
do { \
(c)[0] = (a)[0] / (b)[0]; \
(c)[1] = (a)[1] / (b)[1]; \
(c)[2] = (a)[2] / (b)[2]; \
} while (0)
#define VectorCompCompare(x, op, y) \
(((x)[0] op (y)[0]) && ((x)[1] op (y)[1]) && ((x)[2] op (y)[2]))
#define VectorCompare(x, y) VectorCompCompare (x, ==, y)
#define VectorCompMin(a, b, c) \
do { \
(c)[0] = min ((a)[0], (b)[0]); \
(c)[1] = min ((a)[1], (b)[1]); \
(c)[2] = min ((a)[2], (b)[2]); \
} while (0)
#define VectorCompMax(a, b, c) \
do { \
(c)[0] = max ((a)[0], (b)[0]); \
(c)[1] = max ((a)[1], (b)[1]); \
(c)[2] = max ((a)[2], (b)[2]); \
} while (0)
#define VectorCompBound(a, b, c, d) \
do { \
(d)[0] = bound ((a)[0], (b)[0], (c)[0]); \
(d)[1] = bound ((a)[1], (b)[1], (c)[1]); \
(d)[2] = bound ((a)[2], (b)[2], (c)[2]); \
} while (0)
#define VectorIsZero(a) (!(a)[0] && !(a)[1] && !(a)[2])
#define VectorZero(a) ((a)[2] = (a)[1] = (a)[0] = 0);
#define VectorSet(a,b,c,d) \
do { \
(d)[0] = a; \
(d)[1] = b; \
(d)[2] = c; \
} while (0)
#define VectorBlend(v1,v2,b,v) \
do { \
(v)[0] = (v1)[0] * (1 - (b)) + (v2)[0] * (b); \
(v)[1] = (v1)[1] * (1 - (b)) + (v2)[1] * (b); \
(v)[2] = (v1)[2] * (1 - (b)) + (v2)[2] * (b); \
} while (0)
//For printf etc
#define VectorExpand(v) (v)[0], (v)[1], (v)[2]
/*
* VectorDistance, the distance between two points.
* Yes, this is the same as sqrt(VectorSubtract then DotProduct),
* however that way would involve more vars, this is cheaper.
*/
#define VectorDistance_fast(a, b) \
((((a)[0] - (b)[0]) * ((a)[0] - (b)[0])) + \
(((a)[1] - (b)[1]) * ((a)[1] - (b)[1])) + \
(((a)[2] - (b)[2]) * ((a)[2] - (b)[2])))
#define VectorDistance(a, b) sqrt(VectorDistance_fast(a, b))
vec_t _DotProduct (const vec3_t v1, const vec3_t v2);
void _VectorAdd (const vec3_t veca, const vec3_t vecb, vec3_t out);
void _VectorCopy (const vec3_t in, vec3_t out);
int _VectorCompare (const vec3_t v1, const vec3_t v2); // uses EQUAL_EPSILON
vec_t _VectorLength (const vec3_t v);
void _VectorMA (const vec3_t veca, float scale, const vec3_t vecb,
vec3_t vecc);
void _VectorScale (const vec3_t in, vec_t scale, vec3_t out);
void _VectorSubtract (const vec3_t veca, const vec3_t vecb, vec3_t out);
void CrossProduct (const vec3_t v1, const vec3_t v2, vec3_t cross);
vec_t _VectorNormalize (vec3_t v); // returns vector length
extern inline float VectorNormalize (vec3_t v); // returns vector length
#ifndef IMPLEMENT_VectorNormalize
extern inline
#else
VISIBLE
#endif
float
VectorNormalize (vec3_t v)
{
float length;
length = DotProduct (v, v);
if (length) {
float ilength;
length = sqrt (length);
ilength = 1.0 / length;
v[0] *= ilength;
v[1] *= ilength;
v[2] *= ilength;
}
return length;
}
//@}
#endif // __QF_math_vector_h

View file

@ -51,552 +51,18 @@
#endif
extern int nanmask;
extern const vec_t *const vec3_origin;
extern const vec_t *const quat_origin;
#define EQUAL_EPSILON 0.001
#define RINT(x) (floor ((x) + 0.5))
#define IS_NAN(x) (((*(int *) (char *) &x) & nanmask) == nanmask)
#define DotProduct(a,b) ((a)[0] * (b)[0] + (a)[1] * (b)[1] + (a)[2] * (b)[2])
#define VectorSubtract(a,b,c) \
do { \
(c)[0] = (a)[0] - (b)[0]; \
(c)[1] = (a)[1] - (b)[1]; \
(c)[2] = (a)[2] - (b)[2]; \
} while (0)
#define VectorNegate(a,b) \
do { \
(b)[0] = -(a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
} while (0)
#define VectorAdd(a,b,c) \
do { \
(c)[0] = (a)[0] + (b)[0]; \
(c)[1] = (a)[1] + (b)[1]; \
(c)[2] = (a)[2] + (b)[2]; \
} while (0)
#define VectorCopy(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = (a)[1]; \
(b)[2] = (a)[2]; \
} while (0)
#define VectorMultAdd(a,s,b,c) \
do { \
(c)[0] = (a)[0] + (s) * (b)[0]; \
(c)[1] = (a)[1] + (s) * (b)[1]; \
(c)[2] = (a)[2] + (s) * (b)[2]; \
} while (0)
#define VectorMultSub(a,s,b,c) \
do { \
(c)[0] = (a)[0] - (s) * (b)[0]; \
(c)[1] = (a)[1] - (s) * (b)[1]; \
(c)[2] = (a)[2] - (s) * (b)[2]; \
} while (0)
#define VectorLength(a) sqrt(DotProduct(a, a))
#define VectorScale(a,b,c) \
do { \
(c)[0] = (a)[0] * (b); \
(c)[1] = (a)[1] * (b); \
(c)[2] = (a)[2] * (b); \
} while (0)
/** Shear vector \a b by vector \a a.
Vector a represents the shear factors XY, XZ, YZ, ie in matrix form:
[ 1 0 0 ] [ b0 ]
[ a0 1 0 ] [ b1 ]
[ a1 a2 1 ] [ b2 ]
The reason for this particular scheme is that is how Mat4Decompose
calculates the shear from a matrix.
\note The order of calculations is important for when b and c refer to
the same vector.
*/
#define VectorShear(a,b,c) \
do { \
(c)[2] = (b)[0] * (a)[1] + (b)[1] * (a)[2] + (b)[2]; \
(c)[1] = (b)[0] * (a)[0] + (b)[1]; \
(c)[0] = (b)[0]; \
} while (0)
#define VectorUnshear(a,b,c) \
do { \
(c)[2] = (b)[2] - (b)[1] * (a)[2] - (b)[0] * ((a)[1]-(a)[0]*(a)[2]); \
(c)[1] = (b)[1] - (b)[0] * (a)[0]; \
(c)[0] = (b)[0]; \
} while (0)
#define VectorCompMult(a,b,c) \
do { \
(c)[0] = (a)[0] * (b)[0]; \
(c)[1] = (a)[1] * (b)[1]; \
(c)[2] = (a)[2] * (b)[2]; \
} while (0)
#define VectorCompDiv(a,b,c) \
do { \
(c)[0] = (a)[0] / (b)[0]; \
(c)[1] = (a)[1] / (b)[1]; \
(c)[2] = (a)[2] / (b)[2]; \
} while (0)
#define VectorCompCompare(x, op, y) \
(((x)[0] op (y)[0]) && ((x)[1] op (y)[1]) && ((x)[2] op (y)[2]))
#define VectorCompare(x, y) VectorCompCompare (x, ==, y)
#define VectorCompMin(a, b, c) \
do { \
(c)[0] = min ((a)[0], (b)[0]); \
(c)[1] = min ((a)[1], (b)[1]); \
(c)[2] = min ((a)[2], (b)[2]); \
} while (0)
#define VectorCompMax(a, b, c) \
do { \
(c)[0] = max ((a)[0], (b)[0]); \
(c)[1] = max ((a)[1], (b)[1]); \
(c)[2] = max ((a)[2], (b)[2]); \
} while (0)
#define VectorCompBound(a, b, c, d) \
do { \
(d)[0] = bound ((a)[0], (b)[0], (c)[0]); \
(d)[1] = bound ((a)[1], (b)[1], (c)[1]); \
(d)[2] = bound ((a)[2], (b)[2], (c)[2]); \
} while (0)
#define VectorIsZero(a) (!(a)[0] && !(a)[1] && !(a)[2])
#define VectorZero(a) ((a)[2] = (a)[1] = (a)[0] = 0);
#define VectorSet(a,b,c,d) \
do { \
(d)[0] = a; \
(d)[1] = b; \
(d)[2] = c; \
} while (0)
#define VectorBlend(v1,v2,b,v) \
do { \
(v)[0] = (v1)[0] * (1 - (b)) + (v2)[0] * (b); \
(v)[1] = (v1)[1] * (1 - (b)) + (v2)[1] * (b); \
(v)[2] = (v1)[2] * (1 - (b)) + (v2)[2] * (b); \
} while (0)
//For printf etc
#define VectorExpand(v) (v)[0], (v)[1], (v)[2]
/*
* VectorDistance, the distance between two points.
* Yes, this is the same as sqrt(VectorSubtract then DotProduct),
* however that way would involve more vars, this is cheaper.
*/
#define VectorDistance_fast(a, b) \
((((a)[0] - (b)[0]) * ((a)[0] - (b)[0])) + \
(((a)[1] - (b)[1]) * ((a)[1] - (b)[1])) + \
(((a)[2] - (b)[2]) * ((a)[2] - (b)[2])))
#define VectorDistance(a, b) sqrt(VectorDistance_fast(a, b))
#define QDotProduct(a,b) ((a)[0] * (b)[0] + (a)[1] * (b)[1] \
+ (a)[2] * (b)[2] + (a)[3] * (b)[3])
#define QuatSubtract(a,b,c) \
do { \
(c)[0] = (a)[0] - (b)[0]; \
(c)[1] = (a)[1] - (b)[1]; \
(c)[2] = (a)[2] - (b)[2]; \
(c)[3] = (a)[3] - (b)[3]; \
} while (0)
#define QuatNegate(a,b) \
do { \
(b)[0] = -(a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
(b)[3] = -(a)[3]; \
} while (0)
#define QuatConj(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = -(a)[1]; \
(b)[2] = -(a)[2]; \
(b)[3] = -(a)[3]; \
} while (0)
#define QuatAdd(a,b,c) \
do { \
(c)[0] = (a)[0] + (b)[0]; \
(c)[1] = (a)[1] + (b)[1]; \
(c)[2] = (a)[2] + (b)[2]; \
(c)[3] = (a)[3] + (b)[3]; \
} while (0)
#define QuatCopy(a,b) \
do { \
(b)[0] = (a)[0]; \
(b)[1] = (a)[1]; \
(b)[2] = (a)[2]; \
(b)[3] = (a)[3]; \
} while (0)
#define QuatMultAdd(a,s,b,c) \
do { \
(c)[0] = (a)[0] + (s) * (b)[0]; \
(c)[1] = (a)[1] + (s) * (b)[1]; \
(c)[2] = (a)[2] + (s) * (b)[2]; \
(c)[3] = (a)[3] + (s) * (b)[3]; \
} while (0)
#define QuatMultSub(a,s,b,c) \
do { \
(c)[0] = (a)[0] - (s) * (b)[0]; \
(c)[1] = (a)[1] - (s) * (b)[1]; \
(c)[2] = (a)[2] - (s) * (b)[2]; \
(c)[3] = (a)[3] - (s) * (b)[3]; \
} while (0)
#define QuatLength(a) sqrt(QDotProduct(a, a))
#define QuatScale(a,b,c) \
do { \
(c)[0] = (a)[0] * (b); \
(c)[1] = (a)[1] * (b); \
(c)[2] = (a)[2] * (b); \
(c)[3] = (a)[3] * (b); \
} while (0)
#define QuatCompMult(a,b,c) \
do { \
(c)[0] = (a)[0] * (b)[0]; \
(c)[1] = (a)[1] * (b)[1]; \
(c)[2] = (a)[2] * (b)[2]; \
(c)[3] = (a)[3] * (b)[3]; \
} while (0)
#define QuatCompDiv(a,b,c) \
do { \
(c)[0] = (a)[0] / (b)[0]; \
(c)[1] = (a)[1] / (b)[1]; \
(c)[2] = (a)[2] / (b)[2]; \
(c)[3] = (a)[3] / (b)[3]; \
} while (0)
#define QuatCompCompare(x, op, y) \
(((x)[0] op (y)[0]) && ((x)[1] op (y)[1]) \
&& ((x)[2] op (y)[2]) && ((x)[3] op (y)[3]))
#define QuatCompare(x, y) QuatCompCompare (x, ==, y)
#define QuatCompMin(a, b, c) \
do { \
(c)[0] = min ((a)[0], (b)[0]); \
(c)[1] = min ((a)[1], (b)[1]); \
(c)[2] = min ((a)[2], (b)[2]); \
(c)[3] = min ((a)[3], (b)[3]); \
} while (0)
#define QuatCompMax(a, b, c) \
do { \
(c)[0] = max ((a)[0], (b)[0]); \
(c)[1] = max ((a)[1], (b)[1]); \
(c)[2] = max ((a)[2], (b)[2]); \
(c)[3] = max ((a)[3], (b)[3]); \
} while (0)
#define QuatCompBound(a, b, c, d) \
do { \
(d)[0] = bound ((a)[0], (b)[0], (c)[0]); \
(d)[1] = bound ((a)[1], (b)[1], (c)[1]); \
(d)[2] = bound ((a)[2], (b)[2], (c)[2]); \
(d)[3] = bound ((a)[3], (b)[3], (c)[3]); \
} while (0)
#define QuatIsZero(a) (!(a)[0] && !(a)[1] && !(a)[2] && !(a)[3])
#define QuatZero(a) ((a)[3] = (a)[2] = (a)[1] = (a)[0] = 0);
#define QuatSet(a,b,c,d,e) \
do { \
(e)[0] = a; \
(e)[1] = b; \
(e)[2] = c; \
(e)[3] = d; \
} while (0)
#define QuatBlend(q1,q2,b,q) \
do { \
(q)[0] = (q1)[0] * (1 - (b)) + (q2)[0] * (b); \
(q)[1] = (q1)[1] * (1 - (b)) + (q2)[1] * (b); \
(q)[2] = (q1)[2] * (1 - (b)) + (q2)[2] * (b); \
(q)[3] = (q1)[3] * (1 - (b)) + (q2)[3] * (b); \
} while (0)
//For printf etc
#define QuatExpand(q) (q)[0], (q)[1], (q)[2], (q)[3]
#define DualAdd(a,b,c) \
do { \
(c).r = (a).r + (b).r; \
(c).e = (a).e + (b).e; \
} while (0)
#define DualSubtract(a,b,c) \
do { \
(c).r = (a).r - (b).r; \
(c).e = (a).e - (b).e; \
} while (0)
#define DualNegate(a,b) \
do { \
(b).r = -(a).r; \
(b).e = -(a).e; \
} while (0)
#define DualConj(a,b) \
do { \
(b).r = (a).r; \
(b).e = -(a).e; \
} while (0)
#define DualMult(a,b,c) \
do { \
(c).e = (a).r * (b).e + (a).e * (b).r; \
(c).r = (a).r * (b).r; \
} while (0)
#define DualMultAdd(a,s,b,c) \
do { \
(c).r = (a).r + (s) * (b).r; \
(c).e = (a).e + (s) * (b).e; \
} while (0)
#define DualMultSub(a,s,b,c) \
do { \
(c).r = (a).r - (s) * (b).r; \
(c).e = (a).e - (s) * (b).e; \
} while (0)
#define DualNorm(a) ((a).r)
#define DualScale(a,b,c) \
do { \
(c).r = (a).r * (b); \
(c).e = (a).e * (b); \
} while (0)
#define DualCompCompare(x, op, y) ((x).r op (y).r) && ((x).e op (y).e)
#define DualCompare(x, y) DualCompCompare (x, ==, y)
#define DualIsZero(a) ((a).r == 0 && (a).e == 0)
#define DualIsUnit(a) (((a).r - 1) * ((a).r - 1) < 1e-6 && (a).e * (a).e < 1e-6)
#define DualSet(ar,ae,a) \
do { \
(a).ar = r; \
(a).er = r; \
} while (0)
#define DualZero(a) \
do { \
(a).e = (a).r = 0; \
} while (0)
#define DualBlend(d1,d2,b,d) \
do { \
(d).r = (d1).r * (1 - (b)) + (d2).r * (b); \
(d).e = (d1).e * (1 - (b)) + (d2).e * (b); \
} while (0)
#define DualExpand(d) (d).r, (d).e
#define DualQuatAdd(a,b,c) \
do { \
QuatAdd ((a).q0.q, (b).q0.q, (c).q0.q); \
QuatAdd ((a).qe.q, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatSubtract(a,b,c) \
do { \
QuatSub ((a).q0.q, (b).q0.q, (c).q0.q); \
QuatSub ((a).qe.q, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatNegate(a,b) \
do { \
QuatNegate ((a).q0.q, (b).q0.q); \
QuatNegate ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjQ(a,b) \
do { \
QuatConj ((a).q0.q, (b).q0.q); \
QuatConj ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjE(a,b) \
do { \
(b).q0 = (a).q0; \
QuatNegate ((a).qe.q, (b).qe.q); \
} while (0)
#define DualQuatConjQE(a,b) \
do { \
QuatConj ((a).q0.q, (b).q0.q); \
(b).qe.sv.s = -(a).qe.sv.s; \
VectorCopy ((a).qe.sv.v, (b).qe.sv.v); \
} while (0)
#define DualQuatMult(a,b,c) \
do { \
Quat_t t; \
QuatMult ((a).q0.q, (b).qe.q, t.q); \
QuatMult ((a).qe.q, (b).q0.q, (c).qe.q); \
QuatAdd (t.q, (c).qe.q, (c).qe.q); \
QuatMult ((a).q0.q, (b).q0.q, (c).q0.q); \
} while (0);
#define DualQuatMultAdd(a,s,b,c) \
do { \
QuatMultAdd ((a).q0.q, s, (b).q0.q, (c).q0.q); \
QuatMultAdd ((a).qe.q, s, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatMultSub(a,s,b,c) \
do { \
QuatMultSub ((a).q0.q, s, (b).q0.q, (c).q0.q); \
QuatMultSub ((a).qe.q, s, (b).qe.q, (c).qe.q); \
} while (0)
#define DualQuatNorm(a,b) \
do { \
(b).r = QuatLength ((a).q0.q); \
(b).e = 2 * QDotProduct ((a).q0.q, (a).qe.q); \
} while (0)
#define DualQuatScale(a,b,c) \
do { \
QuatSub ((a).q0.q, (b), (c).q0.q); \
QuatSub ((a).qe.q, (b), (c).qe.q); \
} while (0)
#define DualQuatCompCompare(x, op, y) \
(QuatCompCompare ((x).q0.q, op, (y).q0.q) \
&&QuatCompCompare ((x).qe.q, op, (y).qe.q))
#define DualQuatCompare(x, y) DualQuatCompCompare (x, ==, y)
#define DualQuatIsZero(a) (QuatIsZero ((a).q0.q) && QuatIsZero ((a).qe.q))
#define DualQuatSetVect(vec, a) \
do { \
(a).q0.sv.s = 1; \
VectorZero ((a).q0.sv.v); \
(a).qe.sv.s = 0; \
VectorCopy (vec, (a).qe.sv.v); \
} while (0)
#define DualQuatRotTrans(rot, trans, dq) \
do { \
QuatCopy (rot, (dq).q0.q); \
(dq).qe.sv.s = 0; \
VectorScale (trans, 0.5, (dq).qe.sv.v); \
QuatMult ((dq).qe.q, (dq).q0.q, (dq).qe.q); \
} while (0)
#define DualQuatZero(a) \
do { \
QuatZero ((a).q0.q); \
QuatZero ((a).qe.q); \
} while (0)
#define DualQuatBlend(dq1,dq2,b,dq) \
do { \
QuatBlend ((dq1).q0.q, (dq2).q0.q, (b), (dq).q0.q); \
QuatBlend ((dq1).qe.q, (dq2).qe.q, (b), (dq).qe.q); \
} while (0)
#define DualQuatExpand(dq) QuatExpand ((dq).q0.q), QuatExpand ((dq).qe.q)
#define Mat3Copy(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); \
VectorCopy ((a) + 3, (b) + 3); \
VectorCopy ((a) + 6, (b) + 6); \
} while (0)
#define Mat3Add(a, b, c) \
do { \
VectorAdd ((a) + 0, (b) + 0, (c) + 0); \
VectorAdd ((a) + 3, (b) + 3, (c) + 3); \
VectorAdd ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Subtract(a, b, c) \
do { \
VectorSubtract ((a) + 0, (b) + 0, (c) + 0); \
VectorSubtract ((a) + 3, (b) + 3, (c) + 3); \
VectorSubtract ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Scale(a, b, c) \
do { \
VectorScale ((a) + 0, (b), (c) + 0); \
VectorScale ((a) + 3, (b), (c) + 3); \
VectorScale ((a) + 6, (b), (c) + 6); \
} while (0)
#define Mat3CompMult(a, b, c) \
do { \
VectorCompMult ((a) + 0, (b) + 0, (c) + 0); \
VectorCompMult ((a) + 3, (b) + 3, (c) + 3); \
VectorCompMult ((a) + 6, (b) + 6, (c) + 6); \
} while (0)
#define Mat3Zero(a) \
memset ((a), 0, 9 * sizeof (a)[0])
#define Mat3Identity(a) \
do { \
Mat3Zero (a); \
(a)[8] = (a)[4] = (a)[0] = 1; \
} while (0)
#define Mat3Expand(a) \
VectorExpand ((a) + 0), \
VectorExpand ((a) + 3), \
VectorExpand ((a) + 6)
#define Mat3Blend(m1,m2,b,m) \
do { \
VectorBlend ((m1) + 0, (m2) + 0, (b), (m) + 0); \
VectorBlend ((m1) + 3, (m2) + 3, (b), (m) + 3); \
VectorBlend ((m1) + 6, (m2) + 6, (b), (m) + 6); \
} while (0)
#define Mat3MultAdd(a,s,b,c) \
do { \
VectorMultAdd ((a) + 0, s, (b) + 0, (c) + 0); \
VectorMultAdd ((a) + 3, s, (b) + 3, (c) + 3); \
VectorMultAdd ((a) + 6, s, (b) + 6, (c) + 6); \
} while (0)
#define Mat3toMat4(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); (b)[3] = 0; \
VectorCopy ((a) + 3, (b) + 4); (b)[7] = 0; \
VectorCopy ((a) + 6, (b) + 8); (b)[11] = 0; \
QuatZero ((b) + 12); \
} while (0)
#define Mat4toMat3(a, b) \
do { \
VectorCopy ((a) + 0, (b) + 0); \
VectorCopy ((a) + 4, (b) + 3); \
VectorCopy ((a) + 8, (b) + 6); \
} while (0)
#define Mat3Trace(a) ((a)[0] + (a)[4] + (a)[8])
#define Mat4Copy(a, b) \
do { \
QuatCopy ((a) + 0, (b) + 0); \
QuatCopy ((a) + 4, (b) + 4); \
QuatCopy ((a) + 8, (b) + 8); \
QuatCopy ((a) + 12, (b) + 12); \
} while (0)
#define Mat4Add(a, b, c) \
do { \
QuatAdd ((a) + 0, (b) + 0, (c) + 0); \
QuatAdd ((a) + 4, (b) + 4, (c) + 4); \
QuatAdd ((a) + 8, (b) + 8, (c) + 8); \
QuatAdd ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Subtract(a, b, c) \
do { \
QuatSubtract ((a) + 0, (b) + 0, (c) + 0); \
QuatSubtract ((a) + 4, (b) + 4, (c) + 4); \
QuatSubtract ((a) + 8, (b) + 8, (c) + 8); \
QuatSubtract ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Scale(a, b, c) \
do { \
QuatScale ((a) + 0, (b), (c) + 0); \
QuatScale ((a) + 4, (b), (c) + 4); \
QuatScale ((a) + 8, (b), (c) + 8); \
QuatScale ((a) + 12, (b), (c) + 12); \
} while (0)
#define Mat4CompMult(a, b, c) \
do { \
QuatCompMult ((a) + 0, (b) + 0, (c) + 0); \
QuatCompMult ((a) + 4, (b) + 4, (c) + 4); \
QuatCompMult ((a) + 8, (b) + 8, (c) + 8); \
QuatCompMult ((a) + 12, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Zero(a) \
memset ((a), 0, 16 * sizeof (a)[0])
#define Mat4Identity(a) \
do { \
Mat4Zero (a); \
(a)[15] = (a)[10] = (a)[5] = (a)[0] = 1; \
} while (0)
#define Mat4Expand(a) \
QuatExpand ((a) + 0), \
QuatExpand ((a) + 4), \
QuatExpand ((a) + 8), \
QuatExpand ((a) + 12)
#define Mat4Blend(m1,m2,b,m) \
do { \
QuatBlend ((m1) + 0, (m2) + 0, (b), (m) + 0); \
QuatBlend ((m1) + 4, (m2) + 4, (b), (m) + 4); \
QuatBlend ((m1) + 8, (m2) + 8, (b), (m) + 8); \
QuatBlend ((m1) + 12, (m2) + 12, (b), (m) + 12); \
} while (0)
#define Mat4MultAdd(a,s,b,c) \
do { \
QuatMultAdd ((a) + 0, s, (b) + 0, (c) + 0); \
QuatMultAdd ((a) + 4, s, (b) + 4, (c) + 4); \
QuatMultAdd ((a) + 8, s, (b) + 8, (c) + 8); \
QuatMultAdd ((a) + 12, s, (b) + 12, (c) + 12); \
} while (0)
#define Mat4Trace(a) ((a)[0] + (a)[5] + (a)[10] + (a)[15])
#include "QF/math/vector.h"
#include "QF/math/quaternion.h"
#include "QF/math/dual.h"
#include "QF/math/matrix3.h"
#include "QF/math/matrix4.h"
#include "QF/math/half.h"
#define qfrandom(MAX) ((float) MAX * (rand() * (1.0 / (RAND_MAX + 1.0))))
@ -607,17 +73,6 @@ extern const vec_t *const quat_origin;
// fall over
#define ROLL 2
vec_t _DotProduct (const vec3_t v1, const vec3_t v2);
void _VectorAdd (const vec3_t veca, const vec3_t vecb, vec3_t out);
void _VectorCopy (const vec3_t in, vec3_t out);
int _VectorCompare (const vec3_t v1, const vec3_t v2); // uses EQUAL_EPSILON
vec_t _VectorLength (const vec3_t v);
void _VectorMA (const vec3_t veca, float scale, const vec3_t vecb,
vec3_t vecc);
void _VectorScale (const vec3_t in, vec_t scale, vec3_t out);
void _VectorSubtract (const vec3_t veca, const vec3_t vecb, vec3_t out);
void CrossProduct (const vec3_t v1, const vec3_t v2, vec3_t cross);
vec_t _VectorNormalize (vec3_t v); // returns vector length
int Q_log2(int val);
void R_ConcatRotations (float in1[3][3], float in2[3][3], float out[3][3]);
@ -690,52 +145,6 @@ float anglemod (float a);
void RotatePointAroundVector (vec3_t dst, const vec3_t axis,
const vec3_t point, float degrees);
void QuatMult (const quat_t q1, const quat_t q2, quat_t out);
void QuatMultVec (const quat_t q, const vec3_t v, vec3_t out);
void QuatInverse (const quat_t in, quat_t out);
void QuatExp (const quat_t a, quat_t b);
void QuatToMatrix (const quat_t q, vec_t *m, int homogenous, int vertical);
void Mat3Init (const quat_t rot, const vec3_t scale, mat3_t mat);
void Mat3Transpose (const mat3_t a, mat3_t b);
vec_t Mat3Determinant (const mat3_t m);
int Mat3Inverse (const mat3_t a, mat3_t b);
void Mat3Mult (const mat3_t a, const mat3_t b, mat3_t c);
void Mat3MultVec (const mat3_t a, const vec3_t b, vec3_t c);
void Mat3SymEigen (const mat3_t m, vec3_t e);
/** Decompose a 3x3 column major matrix into its component transformations.
This gives the matrix's rotation as a quaternion, shear (XY, XZ, YZ),
and scale. Using the following sequence will give the same result as
multiplying \a v by \a mat.
QuatMultVec (rot, v, v);
VectorShear (shear, v, v);
VectorCompMult (scale, v, v);
*/
int Mat3Decompose (const mat4_t mat, quat_t rot, vec3_t shear, vec3_t scale);
void Mat4Init (const quat_t rot, const vec3_t scale, const vec3_t trans,
mat4_t mat);
void Mat4Transpose (const mat4_t a, mat4_t b);
int Mat4Inverse (const mat4_t a, mat4_t b);
void Mat4Mult (const mat4_t a, const mat4_t b, mat4_t c);
void Mat4MultVec (const mat4_t a, const vec3_t b, vec3_t c);
void Mat4as3MultVec (const mat4_t a, const vec3_t b, vec3_t c);
/** Decompose a 4x4 column major matrix into its component transformations.
This gives the matrix's rotation as a quaternion, shear (XY, XZ, YZ),
scale, and translation. Using the following sequence will give the
same result as multiplying \a v by \a mat.
QuatMultVec (rot, v, v);
VectorShear (shear, v, v);
VectorCompMult (scale, v, v);
VectorAdd (trans, v, v);
*/
int Mat4Decompose (const mat4_t mat, quat_t rot, vec3_t shear, vec3_t scale,
vec3_t trans);
#define BOX_ON_PLANE_SIDE(emins, emaxs, p) \
(((p)->type < 3)? \
( \
@ -764,12 +173,10 @@ int Mat4Decompose (const mat4_t mat, quat_t rot, vec3_t shear, vec3_t scale,
VectorNegate ((sp)->normal, (dp)->normal); \
} while (0)
int16_t FloatToHalf (float x);
float HalfToFloat (int16_t x);
extern plane_t * const frustum;
extern inline qboolean R_CullBox (const vec3_t mins, const vec3_t maxs);
extern inline qboolean R_CullSphere (const vec3_t origin, const float radius);
extern inline float VectorNormalize (vec3_t v); // returns vector length
#ifndef IMPLEMENT_R_Cull
extern inline
#else
@ -806,30 +213,6 @@ R_CullSphere (const vec3_t origin, const float radius)
return false;
}
#ifndef IMPLEMENT_VectorNormalize
extern inline
#else
VISIBLE
#endif
float
VectorNormalize (vec3_t v)
{
float length;
length = DotProduct (v, v);
if (length) {
float ilength;
length = sqrt (length);
ilength = 1.0 / length;
v[0] *= ilength;
v[1] *= ilength;
v[2] *= ilength;
}
return length;
}
//@}
#endif // __mathlib_h