1 | /* |
---|
2 | ----------------------------------------------------------------------------- |
---|
3 | This source file is part of OGRE |
---|
4 | (Object-oriented Graphics Rendering Engine) |
---|
5 | For the latest info, see http://www.ogre3d.org/ |
---|
6 | |
---|
7 | Copyright (c) 2000-2006 Torus Knot Software Ltd |
---|
8 | Also see acknowledgements in Readme.html |
---|
9 | |
---|
10 | This program is free software; you can redistribute it and/or modify it under |
---|
11 | the terms of the GNU Lesser General Public License as published by the Free Software |
---|
12 | Foundation; either version 2 of the License, or (at your option) any later |
---|
13 | version. |
---|
14 | |
---|
15 | This program is distributed in the hope that it will be useful, but WITHOUT |
---|
16 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS |
---|
17 | FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. |
---|
18 | |
---|
19 | You should have received a copy of the GNU Lesser General Public License along with |
---|
20 | this program; if not, write to the Free Software Foundation, Inc., 59 Temple |
---|
21 | Place - Suite 330, Boston, MA 02111-1307, USA, or go to |
---|
22 | http://www.gnu.org/copyleft/lesser.txt. |
---|
23 | |
---|
24 | You may alternatively use this source under the terms of a specific version of |
---|
25 | the OGRE Unrestricted License provided you have obtained such a license from |
---|
26 | Torus Knot Software Ltd. |
---|
27 | ----------------------------------------------------------------------------- |
---|
28 | */ |
---|
29 | #include "OgreStableHeaders.h" |
---|
30 | |
---|
31 | #include "OgreMath.h" |
---|
32 | #include "asm_math.h" |
---|
33 | #include "OgreVector2.h" |
---|
34 | #include "OgreVector3.h" |
---|
35 | #include "OgreVector4.h" |
---|
36 | #include "OgreRay.h" |
---|
37 | #include "OgreSphere.h" |
---|
38 | #include "OgreAxisAlignedBox.h" |
---|
39 | #include "OgrePlane.h" |
---|
40 | |
---|
41 | |
---|
42 | namespace Ogre |
---|
43 | { |
---|
44 | |
---|
45 | const Real Math::POS_INFINITY = std::numeric_limits<Real>::infinity(); |
---|
46 | const Real Math::NEG_INFINITY = -std::numeric_limits<Real>::infinity(); |
---|
47 | const Real Math::PI = Real( 4.0 * atan( 1.0 ) ); |
---|
48 | const Real Math::TWO_PI = Real( 2.0 * PI ); |
---|
49 | const Real Math::HALF_PI = Real( 0.5 * PI ); |
---|
50 | const Real Math::fDeg2Rad = PI / Real(180.0); |
---|
51 | const Real Math::fRad2Deg = Real(180.0) / PI; |
---|
52 | |
---|
53 | int Math::mTrigTableSize; |
---|
54 | Math::AngleUnit Math::msAngleUnit; |
---|
55 | |
---|
56 | Real Math::mTrigTableFactor; |
---|
57 | Real *Math::mSinTable = NULL; |
---|
58 | Real *Math::mTanTable = NULL; |
---|
59 | |
---|
60 | //----------------------------------------------------------------------- |
---|
61 | Math::Math( unsigned int trigTableSize ) |
---|
62 | { |
---|
63 | msAngleUnit = AU_DEGREE; |
---|
64 | |
---|
65 | mTrigTableSize = trigTableSize; |
---|
66 | mTrigTableFactor = mTrigTableSize / Math::TWO_PI; |
---|
67 | |
---|
68 | mSinTable = new Real[mTrigTableSize]; |
---|
69 | mTanTable = new Real[mTrigTableSize]; |
---|
70 | |
---|
71 | buildTrigTables(); |
---|
72 | } |
---|
73 | |
---|
74 | //----------------------------------------------------------------------- |
---|
75 | Math::~Math() |
---|
76 | { |
---|
77 | delete [] mSinTable; |
---|
78 | delete [] mTanTable; |
---|
79 | } |
---|
80 | |
---|
81 | //----------------------------------------------------------------------- |
---|
82 | void Math::buildTrigTables(void) |
---|
83 | { |
---|
84 | // Build trig lookup tables |
---|
85 | // Could get away with building only PI sized Sin table but simpler this |
---|
86 | // way. Who cares, it'll ony use an extra 8k of memory anyway and I like |
---|
87 | // simplicity. |
---|
88 | Real angle; |
---|
89 | for (int i = 0; i < mTrigTableSize; ++i) |
---|
90 | { |
---|
91 | angle = Math::TWO_PI * i / mTrigTableSize; |
---|
92 | mSinTable[i] = sin(angle); |
---|
93 | mTanTable[i] = tan(angle); |
---|
94 | } |
---|
95 | } |
---|
96 | //----------------------------------------------------------------------- |
---|
97 | Real Math::SinTable (Real fValue) |
---|
98 | { |
---|
99 | // Convert range to index values, wrap if required |
---|
100 | int idx; |
---|
101 | if (fValue >= 0) |
---|
102 | { |
---|
103 | idx = int(fValue * mTrigTableFactor) % mTrigTableSize; |
---|
104 | } |
---|
105 | else |
---|
106 | { |
---|
107 | idx = mTrigTableSize - (int(-fValue * mTrigTableFactor) % mTrigTableSize) - 1; |
---|
108 | } |
---|
109 | |
---|
110 | return mSinTable[idx]; |
---|
111 | } |
---|
112 | //----------------------------------------------------------------------- |
---|
113 | Real Math::TanTable (Real fValue) |
---|
114 | { |
---|
115 | // Convert range to index values, wrap if required |
---|
116 | int idx = int(fValue *= mTrigTableFactor) % mTrigTableSize; |
---|
117 | return mTanTable[idx]; |
---|
118 | } |
---|
119 | //----------------------------------------------------------------------- |
---|
120 | int Math::ISign (int iValue) |
---|
121 | { |
---|
122 | return ( iValue > 0 ? +1 : ( iValue < 0 ? -1 : 0 ) ); |
---|
123 | } |
---|
124 | //----------------------------------------------------------------------- |
---|
125 | Radian Math::ACos (Real fValue) |
---|
126 | { |
---|
127 | if ( -1.0 < fValue ) |
---|
128 | { |
---|
129 | if ( fValue < 1.0 ) |
---|
130 | return Radian(acos(fValue)); |
---|
131 | else |
---|
132 | return Radian(0.0); |
---|
133 | } |
---|
134 | else |
---|
135 | { |
---|
136 | return Radian(PI); |
---|
137 | } |
---|
138 | } |
---|
139 | //----------------------------------------------------------------------- |
---|
140 | Radian Math::ASin (Real fValue) |
---|
141 | { |
---|
142 | if ( -1.0 < fValue ) |
---|
143 | { |
---|
144 | if ( fValue < 1.0 ) |
---|
145 | return Radian(asin(fValue)); |
---|
146 | else |
---|
147 | return Radian(HALF_PI); |
---|
148 | } |
---|
149 | else |
---|
150 | { |
---|
151 | return Radian(-HALF_PI); |
---|
152 | } |
---|
153 | } |
---|
154 | //----------------------------------------------------------------------- |
---|
155 | Real Math::Sign (Real fValue) |
---|
156 | { |
---|
157 | if ( fValue > 0.0 ) |
---|
158 | return 1.0; |
---|
159 | |
---|
160 | if ( fValue < 0.0 ) |
---|
161 | return -1.0; |
---|
162 | |
---|
163 | return 0.0; |
---|
164 | } |
---|
165 | //----------------------------------------------------------------------- |
---|
166 | Real Math::InvSqrt(Real fValue) |
---|
167 | { |
---|
168 | return Real(asm_rsq(fValue)); |
---|
169 | } |
---|
170 | //----------------------------------------------------------------------- |
---|
171 | Real Math::UnitRandom () |
---|
172 | { |
---|
173 | return asm_rand() / asm_rand_max(); |
---|
174 | } |
---|
175 | |
---|
176 | //----------------------------------------------------------------------- |
---|
177 | Real Math::RangeRandom (Real fLow, Real fHigh) |
---|
178 | { |
---|
179 | return (fHigh-fLow)*UnitRandom() + fLow; |
---|
180 | } |
---|
181 | |
---|
182 | //----------------------------------------------------------------------- |
---|
183 | Real Math::SymmetricRandom () |
---|
184 | { |
---|
185 | return 2.0f * UnitRandom() - 1.0f; |
---|
186 | } |
---|
187 | |
---|
188 | //----------------------------------------------------------------------- |
---|
189 | void Math::setAngleUnit(Math::AngleUnit unit) |
---|
190 | { |
---|
191 | msAngleUnit = unit; |
---|
192 | } |
---|
193 | //----------------------------------------------------------------------- |
---|
194 | Math::AngleUnit Math::getAngleUnit(void) |
---|
195 | { |
---|
196 | return msAngleUnit; |
---|
197 | } |
---|
198 | //----------------------------------------------------------------------- |
---|
199 | Real Math::AngleUnitsToRadians(Real angleunits) |
---|
200 | { |
---|
201 | if (msAngleUnit == AU_DEGREE) |
---|
202 | return angleunits * fDeg2Rad; |
---|
203 | else |
---|
204 | return angleunits; |
---|
205 | } |
---|
206 | |
---|
207 | //----------------------------------------------------------------------- |
---|
208 | Real Math::RadiansToAngleUnits(Real radians) |
---|
209 | { |
---|
210 | if (msAngleUnit == AU_DEGREE) |
---|
211 | return radians * fRad2Deg; |
---|
212 | else |
---|
213 | return radians; |
---|
214 | } |
---|
215 | |
---|
216 | //----------------------------------------------------------------------- |
---|
217 | Real Math::AngleUnitsToDegrees(Real angleunits) |
---|
218 | { |
---|
219 | if (msAngleUnit == AU_RADIAN) |
---|
220 | return angleunits * fRad2Deg; |
---|
221 | else |
---|
222 | return angleunits; |
---|
223 | } |
---|
224 | |
---|
225 | //----------------------------------------------------------------------- |
---|
226 | Real Math::DegreesToAngleUnits(Real degrees) |
---|
227 | { |
---|
228 | if (msAngleUnit == AU_RADIAN) |
---|
229 | return degrees * fDeg2Rad; |
---|
230 | else |
---|
231 | return degrees; |
---|
232 | } |
---|
233 | |
---|
234 | //----------------------------------------------------------------------- |
---|
235 | bool Math::pointInTri2D(const Vector2& p, const Vector2& a, |
---|
236 | const Vector2& b, const Vector2& c) |
---|
237 | { |
---|
238 | // Winding must be consistent from all edges for point to be inside |
---|
239 | Vector2 v1, v2; |
---|
240 | Real dot[3]; |
---|
241 | bool zeroDot[3]; |
---|
242 | |
---|
243 | v1 = b - a; |
---|
244 | v2 = p - a; |
---|
245 | |
---|
246 | // Note we don't care about normalisation here since sign is all we need |
---|
247 | // It means we don't have to worry about magnitude of cross products either |
---|
248 | dot[0] = v1.crossProduct(v2); |
---|
249 | zeroDot[0] = Math::RealEqual(dot[0], 0.0f, 1e-3); |
---|
250 | |
---|
251 | |
---|
252 | v1 = c - b; |
---|
253 | v2 = p - b; |
---|
254 | |
---|
255 | dot[1] = v1.crossProduct(v2); |
---|
256 | zeroDot[1] = Math::RealEqual(dot[1], 0.0f, 1e-3); |
---|
257 | |
---|
258 | // Compare signs (ignore colinear / coincident points) |
---|
259 | if(!zeroDot[0] && !zeroDot[1] |
---|
260 | && Math::Sign(dot[0]) != Math::Sign(dot[1])) |
---|
261 | { |
---|
262 | return false; |
---|
263 | } |
---|
264 | |
---|
265 | v1 = a - c; |
---|
266 | v2 = p - c; |
---|
267 | |
---|
268 | dot[2] = v1.crossProduct(v2); |
---|
269 | zeroDot[2] = Math::RealEqual(dot[2], 0.0f, 1e-3); |
---|
270 | // Compare signs (ignore colinear / coincident points) |
---|
271 | if((!zeroDot[0] && !zeroDot[2] |
---|
272 | && Math::Sign(dot[0]) != Math::Sign(dot[2])) || |
---|
273 | (!zeroDot[1] && !zeroDot[2] |
---|
274 | && Math::Sign(dot[1]) != Math::Sign(dot[2]))) |
---|
275 | { |
---|
276 | return false; |
---|
277 | } |
---|
278 | |
---|
279 | |
---|
280 | return true; |
---|
281 | } |
---|
282 | //----------------------------------------------------------------------- |
---|
283 | bool Math::pointInTri3D(const Vector3& p, const Vector3& a, |
---|
284 | const Vector3& b, const Vector3& c, const Vector3& normal) |
---|
285 | { |
---|
286 | // Winding must be consistent from all edges for point to be inside |
---|
287 | Vector3 v1, v2; |
---|
288 | Real dot[3]; |
---|
289 | bool zeroDot[3]; |
---|
290 | |
---|
291 | v1 = b - a; |
---|
292 | v2 = p - a; |
---|
293 | |
---|
294 | // Note we don't care about normalisation here since sign is all we need |
---|
295 | // It means we don't have to worry about magnitude of cross products either |
---|
296 | dot[0] = v1.crossProduct(v2).dotProduct(normal); |
---|
297 | zeroDot[0] = Math::RealEqual(dot[0], 0.0f, 1e-3); |
---|
298 | |
---|
299 | |
---|
300 | v1 = c - b; |
---|
301 | v2 = p - b; |
---|
302 | |
---|
303 | dot[1] = v1.crossProduct(v2).dotProduct(normal); |
---|
304 | zeroDot[1] = Math::RealEqual(dot[1], 0.0f, 1e-3); |
---|
305 | |
---|
306 | // Compare signs (ignore colinear / coincident points) |
---|
307 | if(!zeroDot[0] && !zeroDot[1] |
---|
308 | && Math::Sign(dot[0]) != Math::Sign(dot[1])) |
---|
309 | { |
---|
310 | return false; |
---|
311 | } |
---|
312 | |
---|
313 | v1 = a - c; |
---|
314 | v2 = p - c; |
---|
315 | |
---|
316 | dot[2] = v1.crossProduct(v2).dotProduct(normal); |
---|
317 | zeroDot[2] = Math::RealEqual(dot[2], 0.0f, 1e-3); |
---|
318 | // Compare signs (ignore colinear / coincident points) |
---|
319 | if((!zeroDot[0] && !zeroDot[2] |
---|
320 | && Math::Sign(dot[0]) != Math::Sign(dot[2])) || |
---|
321 | (!zeroDot[1] && !zeroDot[2] |
---|
322 | && Math::Sign(dot[1]) != Math::Sign(dot[2]))) |
---|
323 | { |
---|
324 | return false; |
---|
325 | } |
---|
326 | |
---|
327 | |
---|
328 | return true; |
---|
329 | } |
---|
330 | //----------------------------------------------------------------------- |
---|
331 | bool Math::RealEqual( Real a, Real b, Real tolerance ) |
---|
332 | { |
---|
333 | if (fabs(b-a) <= tolerance) |
---|
334 | return true; |
---|
335 | else |
---|
336 | return false; |
---|
337 | } |
---|
338 | |
---|
339 | //----------------------------------------------------------------------- |
---|
340 | std::pair<bool, Real> Math::intersects(const Ray& ray, const Plane& plane) |
---|
341 | { |
---|
342 | |
---|
343 | Real denom = plane.normal.dotProduct(ray.getDirection()); |
---|
344 | if (Math::Abs(denom) < std::numeric_limits<Real>::epsilon()) |
---|
345 | { |
---|
346 | // Parallel |
---|
347 | return std::pair<bool, Real>(false, 0); |
---|
348 | } |
---|
349 | else |
---|
350 | { |
---|
351 | Real nom = plane.normal.dotProduct(ray.getOrigin()) + plane.d; |
---|
352 | Real t = -(nom/denom); |
---|
353 | return std::pair<bool, Real>(t >= 0, t); |
---|
354 | } |
---|
355 | |
---|
356 | } |
---|
357 | //----------------------------------------------------------------------- |
---|
358 | std::pair<bool, Real> Math::intersects(const Ray& ray, |
---|
359 | const std::vector<Plane>& planes, bool normalIsOutside) |
---|
360 | { |
---|
361 | std::vector<Plane>::const_iterator planeit, planeitend; |
---|
362 | planeitend = planes.end(); |
---|
363 | bool allInside = true; |
---|
364 | std::pair<bool, Real> ret; |
---|
365 | ret.first = false; |
---|
366 | ret.second = 0.0f; |
---|
367 | |
---|
368 | // derive side |
---|
369 | // NB we don't pass directly since that would require Plane::Side in |
---|
370 | // interface, which results in recursive includes since Math is so fundamental |
---|
371 | Plane::Side outside = normalIsOutside ? Plane::POSITIVE_SIDE : Plane::NEGATIVE_SIDE; |
---|
372 | |
---|
373 | for (planeit = planes.begin(); planeit != planeitend; ++planeit) |
---|
374 | { |
---|
375 | const Plane& plane = *planeit; |
---|
376 | // is origin outside? |
---|
377 | if (plane.getSide(ray.getOrigin()) == outside) |
---|
378 | { |
---|
379 | allInside = false; |
---|
380 | // Test single plane |
---|
381 | std::pair<bool, Real> planeRes = |
---|
382 | ray.intersects(plane); |
---|
383 | if (planeRes.first) |
---|
384 | { |
---|
385 | // Ok, we intersected |
---|
386 | ret.first = true; |
---|
387 | // Use the most distant result since convex volume |
---|
388 | ret.second = std::max(ret.second, planeRes.second); |
---|
389 | } |
---|
390 | } |
---|
391 | } |
---|
392 | |
---|
393 | if (allInside) |
---|
394 | { |
---|
395 | // Intersecting at 0 distance since inside the volume! |
---|
396 | ret.first = true; |
---|
397 | ret.second = 0.0f; |
---|
398 | } |
---|
399 | |
---|
400 | return ret; |
---|
401 | } |
---|
402 | //----------------------------------------------------------------------- |
---|
403 | std::pair<bool, Real> Math::intersects(const Ray& ray, |
---|
404 | const std::list<Plane>& planes, bool normalIsOutside) |
---|
405 | { |
---|
406 | std::list<Plane>::const_iterator planeit, planeitend; |
---|
407 | planeitend = planes.end(); |
---|
408 | bool allInside = true; |
---|
409 | std::pair<bool, Real> ret; |
---|
410 | ret.first = false; |
---|
411 | ret.second = 0.0f; |
---|
412 | |
---|
413 | // derive side |
---|
414 | // NB we don't pass directly since that would require Plane::Side in |
---|
415 | // interface, which results in recursive includes since Math is so fundamental |
---|
416 | Plane::Side outside = normalIsOutside ? Plane::POSITIVE_SIDE : Plane::NEGATIVE_SIDE; |
---|
417 | |
---|
418 | for (planeit = planes.begin(); planeit != planeitend; ++planeit) |
---|
419 | { |
---|
420 | const Plane& plane = *planeit; |
---|
421 | // is origin outside? |
---|
422 | if (plane.getSide(ray.getOrigin()) == outside) |
---|
423 | { |
---|
424 | allInside = false; |
---|
425 | // Test single plane |
---|
426 | std::pair<bool, Real> planeRes = |
---|
427 | ray.intersects(plane); |
---|
428 | if (planeRes.first) |
---|
429 | { |
---|
430 | // Ok, we intersected |
---|
431 | ret.first = true; |
---|
432 | // Use the most distant result since convex volume |
---|
433 | ret.second = std::max(ret.second, planeRes.second); |
---|
434 | } |
---|
435 | } |
---|
436 | } |
---|
437 | |
---|
438 | if (allInside) |
---|
439 | { |
---|
440 | // Intersecting at 0 distance since inside the volume! |
---|
441 | ret.first = true; |
---|
442 | ret.second = 0.0f; |
---|
443 | } |
---|
444 | |
---|
445 | return ret; |
---|
446 | } |
---|
447 | //----------------------------------------------------------------------- |
---|
448 | std::pair<bool, Real> Math::intersects(const Ray& ray, const Sphere& sphere, |
---|
449 | bool discardInside) |
---|
450 | { |
---|
451 | const Vector3& raydir = ray.getDirection(); |
---|
452 | // Adjust ray origin relative to sphere center |
---|
453 | const Vector3& rayorig = ray.getOrigin() - sphere.getCenter(); |
---|
454 | Real radius = sphere.getRadius(); |
---|
455 | |
---|
456 | // Check origin inside first |
---|
457 | if (rayorig.squaredLength() <= radius*radius && discardInside) |
---|
458 | { |
---|
459 | return std::pair<bool, Real>(true, 0); |
---|
460 | } |
---|
461 | |
---|
462 | // Mmm, quadratics |
---|
463 | // Build coeffs which can be used with std quadratic solver |
---|
464 | // ie t = (-b +/- sqrt(b*b + 4ac)) / 2a |
---|
465 | Real a = raydir.dotProduct(raydir); |
---|
466 | Real b = 2 * rayorig.dotProduct(raydir); |
---|
467 | Real c = rayorig.dotProduct(rayorig) - radius*radius; |
---|
468 | |
---|
469 | // Calc determinant |
---|
470 | Real d = (b*b) - (4 * a * c); |
---|
471 | if (d < 0) |
---|
472 | { |
---|
473 | // No intersection |
---|
474 | return std::pair<bool, Real>(false, 0); |
---|
475 | } |
---|
476 | else |
---|
477 | { |
---|
478 | // BTW, if d=0 there is one intersection, if d > 0 there are 2 |
---|
479 | // But we only want the closest one, so that's ok, just use the |
---|
480 | // '-' version of the solver |
---|
481 | Real t = ( -b - Math::Sqrt(d) ) / (2 * a); |
---|
482 | if (t < 0) |
---|
483 | t = ( -b + Math::Sqrt(d) ) / (2 * a); |
---|
484 | return std::pair<bool, Real>(true, t); |
---|
485 | } |
---|
486 | |
---|
487 | |
---|
488 | } |
---|
489 | //----------------------------------------------------------------------- |
---|
490 | std::pair<bool, Real> Math::intersects(const Ray& ray, const AxisAlignedBox& box) |
---|
491 | { |
---|
492 | if (box.isNull()) return std::pair<bool, Real>(false, 0); |
---|
493 | if (box.isInfinite()) return std::pair<bool, Real>(true, 0); |
---|
494 | |
---|
495 | Real lowt = 0.0f; |
---|
496 | Real t; |
---|
497 | bool hit = false; |
---|
498 | Vector3 hitpoint; |
---|
499 | const Vector3& min = box.getMinimum(); |
---|
500 | const Vector3& max = box.getMaximum(); |
---|
501 | const Vector3& rayorig = ray.getOrigin(); |
---|
502 | const Vector3& raydir = ray.getDirection(); |
---|
503 | |
---|
504 | // Check origin inside first |
---|
505 | if ( rayorig > min && rayorig < max ) |
---|
506 | { |
---|
507 | return std::pair<bool, Real>(true, 0); |
---|
508 | } |
---|
509 | |
---|
510 | // Check each face in turn, only check closest 3 |
---|
511 | // Min x |
---|
512 | if (rayorig.x <= min.x && raydir.x > 0) |
---|
513 | { |
---|
514 | t = (min.x - rayorig.x) / raydir.x; |
---|
515 | if (t >= 0) |
---|
516 | { |
---|
517 | // Substitute t back into ray and check bounds and dist |
---|
518 | hitpoint = rayorig + raydir * t; |
---|
519 | if (hitpoint.y >= min.y && hitpoint.y <= max.y && |
---|
520 | hitpoint.z >= min.z && hitpoint.z <= max.z && |
---|
521 | (!hit || t < lowt)) |
---|
522 | { |
---|
523 | hit = true; |
---|
524 | lowt = t; |
---|
525 | } |
---|
526 | } |
---|
527 | } |
---|
528 | // Max x |
---|
529 | if (rayorig.x >= max.x && raydir.x < 0) |
---|
530 | { |
---|
531 | t = (max.x - rayorig.x) / raydir.x; |
---|
532 | if (t >= 0) |
---|
533 | { |
---|
534 | // Substitute t back into ray and check bounds and dist |
---|
535 | hitpoint = rayorig + raydir * t; |
---|
536 | if (hitpoint.y >= min.y && hitpoint.y <= max.y && |
---|
537 | hitpoint.z >= min.z && hitpoint.z <= max.z && |
---|
538 | (!hit || t < lowt)) |
---|
539 | { |
---|
540 | hit = true; |
---|
541 | lowt = t; |
---|
542 | } |
---|
543 | } |
---|
544 | } |
---|
545 | // Min y |
---|
546 | if (rayorig.y <= min.y && raydir.y > 0) |
---|
547 | { |
---|
548 | t = (min.y - rayorig.y) / raydir.y; |
---|
549 | if (t >= 0) |
---|
550 | { |
---|
551 | // Substitute t back into ray and check bounds and dist |
---|
552 | hitpoint = rayorig + raydir * t; |
---|
553 | if (hitpoint.x >= min.x && hitpoint.x <= max.x && |
---|
554 | hitpoint.z >= min.z && hitpoint.z <= max.z && |
---|
555 | (!hit || t < lowt)) |
---|
556 | { |
---|
557 | hit = true; |
---|
558 | lowt = t; |
---|
559 | } |
---|
560 | } |
---|
561 | } |
---|
562 | // Max y |
---|
563 | if (rayorig.y >= max.y && raydir.y < 0) |
---|
564 | { |
---|
565 | t = (max.y - rayorig.y) / raydir.y; |
---|
566 | if (t >= 0) |
---|
567 | { |
---|
568 | // Substitute t back into ray and check bounds and dist |
---|
569 | hitpoint = rayorig + raydir * t; |
---|
570 | if (hitpoint.x >= min.x && hitpoint.x <= max.x && |
---|
571 | hitpoint.z >= min.z && hitpoint.z <= max.z && |
---|
572 | (!hit || t < lowt)) |
---|
573 | { |
---|
574 | hit = true; |
---|
575 | lowt = t; |
---|
576 | } |
---|
577 | } |
---|
578 | } |
---|
579 | // Min z |
---|
580 | if (rayorig.z <= min.z && raydir.z > 0) |
---|
581 | { |
---|
582 | t = (min.z - rayorig.z) / raydir.z; |
---|
583 | if (t >= 0) |
---|
584 | { |
---|
585 | // Substitute t back into ray and check bounds and dist |
---|
586 | hitpoint = rayorig + raydir * t; |
---|
587 | if (hitpoint.x >= min.x && hitpoint.x <= max.x && |
---|
588 | hitpoint.y >= min.y && hitpoint.y <= max.y && |
---|
589 | (!hit || t < lowt)) |
---|
590 | { |
---|
591 | hit = true; |
---|
592 | lowt = t; |
---|
593 | } |
---|
594 | } |
---|
595 | } |
---|
596 | // Max z |
---|
597 | if (rayorig.z >= max.z && raydir.z < 0) |
---|
598 | { |
---|
599 | t = (max.z - rayorig.z) / raydir.z; |
---|
600 | if (t >= 0) |
---|
601 | { |
---|
602 | // Substitute t back into ray and check bounds and dist |
---|
603 | hitpoint = rayorig + raydir * t; |
---|
604 | if (hitpoint.x >= min.x && hitpoint.x <= max.x && |
---|
605 | hitpoint.y >= min.y && hitpoint.y <= max.y && |
---|
606 | (!hit || t < lowt)) |
---|
607 | { |
---|
608 | hit = true; |
---|
609 | lowt = t; |
---|
610 | } |
---|
611 | } |
---|
612 | } |
---|
613 | |
---|
614 | return std::pair<bool, Real>(hit, lowt); |
---|
615 | |
---|
616 | } |
---|
617 | //----------------------------------------------------------------------- |
---|
618 | bool Math::intersects(const Ray& ray, const AxisAlignedBox& box, |
---|
619 | Real* d1, Real* d2) |
---|
620 | { |
---|
621 | if (box.isNull()) |
---|
622 | return false; |
---|
623 | |
---|
624 | if (box.isInfinite()) |
---|
625 | { |
---|
626 | if (d1) *d1 = 0; |
---|
627 | if (d2) *d2 = Math::POS_INFINITY; |
---|
628 | return true; |
---|
629 | } |
---|
630 | |
---|
631 | const Vector3& min = box.getMinimum(); |
---|
632 | const Vector3& max = box.getMaximum(); |
---|
633 | const Vector3& rayorig = ray.getOrigin(); |
---|
634 | const Vector3& raydir = ray.getDirection(); |
---|
635 | |
---|
636 | Vector3 absDir; |
---|
637 | absDir[0] = Math::Abs(raydir[0]); |
---|
638 | absDir[1] = Math::Abs(raydir[1]); |
---|
639 | absDir[2] = Math::Abs(raydir[2]); |
---|
640 | |
---|
641 | // Sort the axis, ensure check minimise floating error axis first |
---|
642 | int imax = 0, imid = 1, imin = 2; |
---|
643 | if (absDir[0] < absDir[2]) |
---|
644 | { |
---|
645 | imax = 2; |
---|
646 | imin = 0; |
---|
647 | } |
---|
648 | if (absDir[1] < absDir[imin]) |
---|
649 | { |
---|
650 | imid = imin; |
---|
651 | imin = 1; |
---|
652 | } |
---|
653 | else if (absDir[1] > absDir[imax]) |
---|
654 | { |
---|
655 | imid = imax; |
---|
656 | imax = 1; |
---|
657 | } |
---|
658 | |
---|
659 | Real start = 0, end = Math::POS_INFINITY; |
---|
660 | |
---|
661 | #define _CALC_AXIS(i) \ |
---|
662 | do { \ |
---|
663 | Real denom = 1 / raydir[i]; \ |
---|
664 | Real newstart = (min[i] - rayorig[i]) * denom; \ |
---|
665 | Real newend = (max[i] - rayorig[i]) * denom; \ |
---|
666 | if (newstart > newend) std::swap(newstart, newend); \ |
---|
667 | if (newstart > end || newend < start) return false; \ |
---|
668 | if (newstart > start) start = newstart; \ |
---|
669 | if (newend < end) end = newend; \ |
---|
670 | } while(0) |
---|
671 | |
---|
672 | // Check each axis in turn |
---|
673 | |
---|
674 | _CALC_AXIS(imax); |
---|
675 | |
---|
676 | if (absDir[imid] < std::numeric_limits<Real>::epsilon()) |
---|
677 | { |
---|
678 | // Parallel with middle and minimise axis, check bounds only |
---|
679 | if (rayorig[imid] < min[imid] || rayorig[imid] > max[imid] || |
---|
680 | rayorig[imin] < min[imin] || rayorig[imin] > max[imin]) |
---|
681 | return false; |
---|
682 | } |
---|
683 | else |
---|
684 | { |
---|
685 | _CALC_AXIS(imid); |
---|
686 | |
---|
687 | if (absDir[imin] < std::numeric_limits<Real>::epsilon()) |
---|
688 | { |
---|
689 | // Parallel with minimise axis, check bounds only |
---|
690 | if (rayorig[imin] < min[imin] || rayorig[imin] > max[imin]) |
---|
691 | return false; |
---|
692 | } |
---|
693 | else |
---|
694 | { |
---|
695 | _CALC_AXIS(imin); |
---|
696 | } |
---|
697 | } |
---|
698 | #undef _CALC_AXIS |
---|
699 | |
---|
700 | if (d1) *d1 = start; |
---|
701 | if (d2) *d2 = end; |
---|
702 | |
---|
703 | return true; |
---|
704 | } |
---|
705 | //----------------------------------------------------------------------- |
---|
706 | std::pair<bool, Real> Math::intersects(const Ray& ray, const Vector3& a, |
---|
707 | const Vector3& b, const Vector3& c, const Vector3& normal, |
---|
708 | bool positiveSide, bool negativeSide) |
---|
709 | { |
---|
710 | // |
---|
711 | // Calculate intersection with plane. |
---|
712 | // |
---|
713 | Real t; |
---|
714 | { |
---|
715 | Real denom = normal.dotProduct(ray.getDirection()); |
---|
716 | |
---|
717 | // Check intersect side |
---|
718 | if (denom > + std::numeric_limits<Real>::epsilon()) |
---|
719 | { |
---|
720 | if (!negativeSide) |
---|
721 | return std::pair<bool, Real>(false, 0); |
---|
722 | } |
---|
723 | else if (denom < - std::numeric_limits<Real>::epsilon()) |
---|
724 | { |
---|
725 | if (!positiveSide) |
---|
726 | return std::pair<bool, Real>(false, 0); |
---|
727 | } |
---|
728 | else |
---|
729 | { |
---|
730 | // Parallel or triangle area is close to zero when |
---|
731 | // the plane normal not normalised. |
---|
732 | return std::pair<bool, Real>(false, 0); |
---|
733 | } |
---|
734 | |
---|
735 | t = normal.dotProduct(a - ray.getOrigin()) / denom; |
---|
736 | |
---|
737 | if (t < 0) |
---|
738 | { |
---|
739 | // Intersection is behind origin |
---|
740 | return std::pair<bool, Real>(false, 0); |
---|
741 | } |
---|
742 | } |
---|
743 | |
---|
744 | // |
---|
745 | // Calculate the largest area projection plane in X, Y or Z. |
---|
746 | // |
---|
747 | size_t i0, i1; |
---|
748 | { |
---|
749 | Real n0 = Math::Abs(normal[0]); |
---|
750 | Real n1 = Math::Abs(normal[1]); |
---|
751 | Real n2 = Math::Abs(normal[2]); |
---|
752 | |
---|
753 | i0 = 1; i1 = 2; |
---|
754 | if (n1 > n2) |
---|
755 | { |
---|
756 | if (n1 > n0) i0 = 0; |
---|
757 | } |
---|
758 | else |
---|
759 | { |
---|
760 | if (n2 > n0) i1 = 0; |
---|
761 | } |
---|
762 | } |
---|
763 | |
---|
764 | // |
---|
765 | // Check the intersection point is inside the triangle. |
---|
766 | // |
---|
767 | { |
---|
768 | Real u1 = b[i0] - a[i0]; |
---|
769 | Real v1 = b[i1] - a[i1]; |
---|
770 | Real u2 = c[i0] - a[i0]; |
---|
771 | Real v2 = c[i1] - a[i1]; |
---|
772 | Real u0 = t * ray.getDirection()[i0] + ray.getOrigin()[i0] - a[i0]; |
---|
773 | Real v0 = t * ray.getDirection()[i1] + ray.getOrigin()[i1] - a[i1]; |
---|
774 | |
---|
775 | Real alpha = u0 * v2 - u2 * v0; |
---|
776 | Real beta = u1 * v0 - u0 * v1; |
---|
777 | Real area = u1 * v2 - u2 * v1; |
---|
778 | |
---|
779 | // epsilon to avoid float precision error |
---|
780 | const Real EPSILON = 1e-3f; |
---|
781 | |
---|
782 | Real tolerance = - EPSILON * area; |
---|
783 | |
---|
784 | if (area > 0) |
---|
785 | { |
---|
786 | if (alpha < tolerance || beta < tolerance || alpha+beta > area-tolerance) |
---|
787 | return std::pair<bool, Real>(false, 0); |
---|
788 | } |
---|
789 | else |
---|
790 | { |
---|
791 | if (alpha > tolerance || beta > tolerance || alpha+beta < area-tolerance) |
---|
792 | return std::pair<bool, Real>(false, 0); |
---|
793 | } |
---|
794 | } |
---|
795 | |
---|
796 | return std::pair<bool, Real>(true, t); |
---|
797 | } |
---|
798 | //----------------------------------------------------------------------- |
---|
799 | std::pair<bool, Real> Math::intersects(const Ray& ray, const Vector3& a, |
---|
800 | const Vector3& b, const Vector3& c, |
---|
801 | bool positiveSide, bool negativeSide) |
---|
802 | { |
---|
803 | Vector3 normal = calculateBasicFaceNormalWithoutNormalize(a, b, c); |
---|
804 | return intersects(ray, a, b, c, normal, positiveSide, negativeSide); |
---|
805 | } |
---|
806 | //----------------------------------------------------------------------- |
---|
807 | bool Math::intersects(const Sphere& sphere, const AxisAlignedBox& box) |
---|
808 | { |
---|
809 | if (box.isNull()) return false; |
---|
810 | if (box.isInfinite()) return true; |
---|
811 | |
---|
812 | // Use splitting planes |
---|
813 | const Vector3& center = sphere.getCenter(); |
---|
814 | Real radius = sphere.getRadius(); |
---|
815 | const Vector3& min = box.getMinimum(); |
---|
816 | const Vector3& max = box.getMaximum(); |
---|
817 | |
---|
818 | // Arvo's algorithm |
---|
819 | Real s, d = 0; |
---|
820 | for (int i = 0; i < 3; ++i) |
---|
821 | { |
---|
822 | if (center.ptr()[i] < min.ptr()[i]) |
---|
823 | { |
---|
824 | s = center.ptr()[i] - min.ptr()[i]; |
---|
825 | d += s * s; |
---|
826 | } |
---|
827 | else if(center.ptr()[i] > max.ptr()[i]) |
---|
828 | { |
---|
829 | s = center.ptr()[i] - max.ptr()[i]; |
---|
830 | d += s * s; |
---|
831 | } |
---|
832 | } |
---|
833 | return d <= radius * radius; |
---|
834 | |
---|
835 | } |
---|
836 | //----------------------------------------------------------------------- |
---|
837 | bool Math::intersects(const Plane& plane, const AxisAlignedBox& box) |
---|
838 | { |
---|
839 | return (plane.getSide(box) == Plane::BOTH_SIDE); |
---|
840 | } |
---|
841 | //----------------------------------------------------------------------- |
---|
842 | bool Math::intersects(const Sphere& sphere, const Plane& plane) |
---|
843 | { |
---|
844 | return ( |
---|
845 | Math::Abs(plane.getDistance(sphere.getCenter())) |
---|
846 | <= sphere.getRadius() ); |
---|
847 | } |
---|
848 | //----------------------------------------------------------------------- |
---|
849 | Vector3 Math::calculateTangentSpaceVector( |
---|
850 | const Vector3& position1, const Vector3& position2, const Vector3& position3, |
---|
851 | Real u1, Real v1, Real u2, Real v2, Real u3, Real v3) |
---|
852 | { |
---|
853 | //side0 is the vector along one side of the triangle of vertices passed in, |
---|
854 | //and side1 is the vector along another side. Taking the cross product of these returns the normal. |
---|
855 | Vector3 side0 = position1 - position2; |
---|
856 | Vector3 side1 = position3 - position1; |
---|
857 | //Calculate face normal |
---|
858 | Vector3 normal = side1.crossProduct(side0); |
---|
859 | normal.normalise(); |
---|
860 | //Now we use a formula to calculate the tangent. |
---|
861 | Real deltaV0 = v1 - v2; |
---|
862 | Real deltaV1 = v3 - v1; |
---|
863 | Vector3 tangent = deltaV1 * side0 - deltaV0 * side1; |
---|
864 | tangent.normalise(); |
---|
865 | //Calculate binormal |
---|
866 | Real deltaU0 = u1 - u2; |
---|
867 | Real deltaU1 = u3 - u1; |
---|
868 | Vector3 binormal = deltaU1 * side0 - deltaU0 * side1; |
---|
869 | binormal.normalise(); |
---|
870 | //Now, we take the cross product of the tangents to get a vector which |
---|
871 | //should point in the same direction as our normal calculated above. |
---|
872 | //If it points in the opposite direction (the dot product between the normals is less than zero), |
---|
873 | //then we need to reverse the s and t tangents. |
---|
874 | //This is because the triangle has been mirrored when going from tangent space to object space. |
---|
875 | //reverse tangents if necessary |
---|
876 | Vector3 tangentCross = tangent.crossProduct(binormal); |
---|
877 | if (tangentCross.dotProduct(normal) < 0.0f) |
---|
878 | { |
---|
879 | tangent = -tangent; |
---|
880 | binormal = -binormal; |
---|
881 | } |
---|
882 | |
---|
883 | return tangent; |
---|
884 | |
---|
885 | } |
---|
886 | //----------------------------------------------------------------------- |
---|
887 | Matrix4 Math::buildReflectionMatrix(const Plane& p) |
---|
888 | { |
---|
889 | return Matrix4( |
---|
890 | -2 * p.normal.x * p.normal.x + 1, -2 * p.normal.x * p.normal.y, -2 * p.normal.x * p.normal.z, -2 * p.normal.x * p.d, |
---|
891 | -2 * p.normal.y * p.normal.x, -2 * p.normal.y * p.normal.y + 1, -2 * p.normal.y * p.normal.z, -2 * p.normal.y * p.d, |
---|
892 | -2 * p.normal.z * p.normal.x, -2 * p.normal.z * p.normal.y, -2 * p.normal.z * p.normal.z + 1, -2 * p.normal.z * p.d, |
---|
893 | 0, 0, 0, 1); |
---|
894 | } |
---|
895 | //----------------------------------------------------------------------- |
---|
896 | Vector4 Math::calculateFaceNormal(const Vector3& v1, const Vector3& v2, const Vector3& v3) |
---|
897 | { |
---|
898 | Vector3 normal = calculateBasicFaceNormal(v1, v2, v3); |
---|
899 | // Now set up the w (distance of tri from origin |
---|
900 | return Vector4(normal.x, normal.y, normal.z, -(normal.dotProduct(v1))); |
---|
901 | } |
---|
902 | //----------------------------------------------------------------------- |
---|
903 | Vector3 Math::calculateBasicFaceNormal(const Vector3& v1, const Vector3& v2, const Vector3& v3) |
---|
904 | { |
---|
905 | Vector3 normal = (v2 - v1).crossProduct(v3 - v1); |
---|
906 | normal.normalise(); |
---|
907 | return normal; |
---|
908 | } |
---|
909 | //----------------------------------------------------------------------- |
---|
910 | Vector4 Math::calculateFaceNormalWithoutNormalize(const Vector3& v1, const Vector3& v2, const Vector3& v3) |
---|
911 | { |
---|
912 | Vector3 normal = calculateBasicFaceNormalWithoutNormalize(v1, v2, v3); |
---|
913 | // Now set up the w (distance of tri from origin) |
---|
914 | return Vector4(normal.x, normal.y, normal.z, -(normal.dotProduct(v1))); |
---|
915 | } |
---|
916 | //----------------------------------------------------------------------- |
---|
917 | Vector3 Math::calculateBasicFaceNormalWithoutNormalize(const Vector3& v1, const Vector3& v2, const Vector3& v3) |
---|
918 | { |
---|
919 | Vector3 normal = (v2 - v1).crossProduct(v3 - v1); |
---|
920 | return normal; |
---|
921 | } |
---|
922 | //----------------------------------------------------------------------- |
---|
923 | Real Math::gaussianDistribution(Real x, Real offset, Real scale) |
---|
924 | { |
---|
925 | Real nom = Math::Exp( |
---|
926 | -Math::Sqr(x - offset) / (2 * Math::Sqr(scale))); |
---|
927 | Real denom = scale * Math::Sqrt(2 * Math::PI); |
---|
928 | |
---|
929 | return nom / denom; |
---|
930 | |
---|
931 | } |
---|
932 | } |
---|