1 /* 2 Bullet Continuous Collision Detection and Physics Library 3 Copyright (c) 2003-2006 Erwin Coumans http://continuousphysics.com/Bullet/ 4 5 This software is provided 'as-is', without any express or implied warranty. 6 In no event will the authors be held liable for any damages arising from the use of this software. 7 Permission is granted to anyone to use this software for any purpose, 8 including commercial applications, and to alter it and redistribute it freely, 9 subject to the following restrictions: 10 11 1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required. 12 2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software. 13 3. This notice may not be removed or altered from any source distribution. 14 */ 15 16 #ifndef BT_QUANTIZED_BVH_H 17 #define BT_QUANTIZED_BVH_H 18 19 class btSerializer; 20 21 //#define DEBUG_CHECK_DEQUANTIZATION 1 22 #ifdef DEBUG_CHECK_DEQUANTIZATION 23 #ifdef __SPU__ 24 #define printf spu_printf 25 #endif //__SPU__ 26 27 #include <stdio.h> 28 #include <stdlib.h> 29 #endif //DEBUG_CHECK_DEQUANTIZATION 30 31 #include "LinearMath/btVector3.h" 32 #include "LinearMath/btAlignedAllocator.h" 33 34 #ifdef BT_USE_DOUBLE_PRECISION 35 #define btQuantizedBvhData btQuantizedBvhDoubleData 36 #define btOptimizedBvhNodeData btOptimizedBvhNodeDoubleData 37 #define btQuantizedBvhDataName "btQuantizedBvhDoubleData" 38 #else 39 #define btQuantizedBvhData btQuantizedBvhFloatData 40 #define btOptimizedBvhNodeData btOptimizedBvhNodeFloatData 41 #define btQuantizedBvhDataName "btQuantizedBvhFloatData" 42 #endif 43 44 45 46 //http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclang/html/vclrf__m128.asp 47 48 49 //Note: currently we have 16 bytes per quantized node 50 #define MAX_SUBTREE_SIZE_IN_BYTES 2048 51 52 // 10 gives the potential for 1024 parts, with at most 2^21 (2097152) (minus one 53 // actually) triangles each (since the sign bit is reserved 54 #define MAX_NUM_PARTS_IN_BITS 10 55 56 ///btQuantizedBvhNode is a compressed aabb node, 16 bytes. 57 ///Node can be used for leafnode or internal node. Leafnodes can point to 32-bit triangle index (non-negative range). 58 ATTRIBUTE_ALIGNED16 (struct) btQuantizedBvhNode 59 { 60 BT_DECLARE_ALIGNED_ALLOCATOR(); 61 62 //12 bytes 63 unsigned short int m_quantizedAabbMin[3]; 64 unsigned short int m_quantizedAabbMax[3]; 65 //4 bytes 66 int m_escapeIndexOrTriangleIndex; 67 68 bool isLeafNode() const 69 { 70 //skipindex is negative (internal node), triangleindex >=0 (leafnode) 71 return (m_escapeIndexOrTriangleIndex >= 0); 72 } 73 int getEscapeIndex() const 74 { 75 btAssert(!isLeafNode()); 76 return -m_escapeIndexOrTriangleIndex; 77 } 78 int getTriangleIndex() const 79 { 80 btAssert(isLeafNode()); 81 unsigned int x=0; 82 unsigned int y = (~(x&0))<<(31-MAX_NUM_PARTS_IN_BITS); 83 // Get only the lower bits where the triangle index is stored 84 return (m_escapeIndexOrTriangleIndex&~(y)); 85 } 86 int getPartId() const 87 { 88 btAssert(isLeafNode()); 89 // Get only the highest bits where the part index is stored 90 return (m_escapeIndexOrTriangleIndex>>(31-MAX_NUM_PARTS_IN_BITS)); 91 } 92 } 93 ; 94 95 /// btOptimizedBvhNode contains both internal and leaf node information. 96 /// Total node size is 44 bytes / node. You can use the compressed version of 16 bytes. 97 ATTRIBUTE_ALIGNED16 (struct) btOptimizedBvhNode 98 { 99 BT_DECLARE_ALIGNED_ALLOCATOR(); 100 101 //32 bytes 102 btVector3 m_aabbMinOrg; 103 btVector3 m_aabbMaxOrg; 104 105 //4 106 int m_escapeIndex; 107 108 //8 109 //for child nodes 110 int m_subPart; 111 int m_triangleIndex; 112 113 //pad the size to 64 bytes 114 char m_padding[20]; 115 }; 116 117 118 ///btBvhSubtreeInfo provides info to gather a subtree of limited size 119 ATTRIBUTE_ALIGNED16(class) btBvhSubtreeInfo 120 { 121 public: 122 BT_DECLARE_ALIGNED_ALLOCATOR(); 123 124 //12 bytes 125 unsigned short int m_quantizedAabbMin[3]; 126 unsigned short int m_quantizedAabbMax[3]; 127 //4 bytes, points to the root of the subtree 128 int m_rootNodeIndex; 129 //4 bytes 130 int m_subtreeSize; 131 int m_padding[3]; 132 133 btBvhSubtreeInfo() 134 { 135 //memset(&m_padding[0], 0, sizeof(m_padding)); 136 } 137 138 139 void setAabbFromQuantizeNode(const btQuantizedBvhNode& quantizedNode) 140 { 141 m_quantizedAabbMin[0] = quantizedNode.m_quantizedAabbMin[0]; 142 m_quantizedAabbMin[1] = quantizedNode.m_quantizedAabbMin[1]; 143 m_quantizedAabbMin[2] = quantizedNode.m_quantizedAabbMin[2]; 144 m_quantizedAabbMax[0] = quantizedNode.m_quantizedAabbMax[0]; 145 m_quantizedAabbMax[1] = quantizedNode.m_quantizedAabbMax[1]; 146 m_quantizedAabbMax[2] = quantizedNode.m_quantizedAabbMax[2]; 147 } 148 } 149 ; 150 151 152 class btNodeOverlapCallback 153 { 154 public: 155 virtual ~btNodeOverlapCallback() {}; 156 157 virtual void processNode(int subPart, int triangleIndex) = 0; 158 }; 159 160 #include "LinearMath/btAlignedAllocator.h" 161 #include "LinearMath/btAlignedObjectArray.h" 162 163 164 165 ///for code readability: 166 typedef btAlignedObjectArray<btOptimizedBvhNode> NodeArray; 167 typedef btAlignedObjectArray<btQuantizedBvhNode> QuantizedNodeArray; 168 typedef btAlignedObjectArray<btBvhSubtreeInfo> BvhSubtreeInfoArray; 169 170 171 ///The btQuantizedBvh class stores an AABB tree that can be quickly traversed on CPU and Cell SPU. 172 ///It is used by the btBvhTriangleMeshShape as midphase, and by the btMultiSapBroadphase. 173 ///It is recommended to use quantization for better performance and lower memory requirements. 174 ATTRIBUTE_ALIGNED16(class) btQuantizedBvh 175 { 176 public: 177 enum btTraversalMode 178 { 179 TRAVERSAL_STACKLESS = 0, 180 TRAVERSAL_STACKLESS_CACHE_FRIENDLY, 181 TRAVERSAL_RECURSIVE 182 }; 183 184 protected: 185 186 187 btVector3 m_bvhAabbMin; 188 btVector3 m_bvhAabbMax; 189 btVector3 m_bvhQuantization; 190 191 int m_bulletVersion; //for serialization versioning. It could also be used to detect endianess. 192 193 int m_curNodeIndex; 194 //quantization data 195 bool m_useQuantization; 196 197 198 199 NodeArray m_leafNodes; 200 NodeArray m_contiguousNodes; 201 QuantizedNodeArray m_quantizedLeafNodes; 202 QuantizedNodeArray m_quantizedContiguousNodes; 203 204 btTraversalMode m_traversalMode; 205 BvhSubtreeInfoArray m_SubtreeHeaders; 206 207 //This is only used for serialization so we don't have to add serialization directly to btAlignedObjectArray 208 mutable int m_subtreeHeaderCount; 209 210 211 212 213 214 ///two versions, one for quantized and normal nodes. This allows code-reuse while maintaining readability (no template/macro!) 215 ///this might be refactored into a virtual, it is usually not calculated at run-time 216 void setInternalNodeAabbMin(int nodeIndex, const btVector3& aabbMin) 217 { 218 if (m_useQuantization) 219 { 220 quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[0] ,aabbMin,0); 221 } else 222 { 223 m_contiguousNodes[nodeIndex].m_aabbMinOrg = aabbMin; 224 225 } 226 } 227 void setInternalNodeAabbMax(int nodeIndex,const btVector3& aabbMax) 228 { 229 if (m_useQuantization) 230 { 231 quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[0],aabbMax,1); 232 } else 233 { 234 m_contiguousNodes[nodeIndex].m_aabbMaxOrg = aabbMax; 235 } 236 } 237 238 btVector3 getAabbMin(int nodeIndex) const 239 { 240 if (m_useQuantization) 241 { 242 return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMin[0]); 243 } 244 //non-quantized 245 return m_leafNodes[nodeIndex].m_aabbMinOrg; 246 247 } 248 btVector3 getAabbMax(int nodeIndex) const 249 { 250 if (m_useQuantization) 251 { 252 return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMax[0]); 253 } 254 //non-quantized 255 return m_leafNodes[nodeIndex].m_aabbMaxOrg; 256 257 } 258 259 260 void setInternalNodeEscapeIndex(int nodeIndex, int escapeIndex) 261 { 262 if (m_useQuantization) 263 { 264 m_quantizedContiguousNodes[nodeIndex].m_escapeIndexOrTriangleIndex = -escapeIndex; 265 } 266 else 267 { 268 m_contiguousNodes[nodeIndex].m_escapeIndex = escapeIndex; 269 } 270 271 } 272 273 void mergeInternalNodeAabb(int nodeIndex,const btVector3& newAabbMin,const btVector3& newAabbMax) 274 { 275 if (m_useQuantization) 276 { 277 unsigned short int quantizedAabbMin[3]; 278 unsigned short int quantizedAabbMax[3]; 279 quantize(quantizedAabbMin,newAabbMin,0); 280 quantize(quantizedAabbMax,newAabbMax,1); 281 for (int i=0;i<3;i++) 282 { 283 if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] > quantizedAabbMin[i]) 284 m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] = quantizedAabbMin[i]; 285 286 if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] < quantizedAabbMax[i]) 287 m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] = quantizedAabbMax[i]; 288 289 } 290 } else 291 { 292 //non-quantized 293 m_contiguousNodes[nodeIndex].m_aabbMinOrg.setMin(newAabbMin); 294 m_contiguousNodes[nodeIndex].m_aabbMaxOrg.setMax(newAabbMax); 295 } 296 } 297 298 void swapLeafNodes(int firstIndex,int secondIndex); 299 300 void assignInternalNodeFromLeafNode(int internalNode,int leafNodeIndex); 301 302 protected: 303 304 305 306 void buildTree (int startIndex,int endIndex); 307 308 int calcSplittingAxis(int startIndex,int endIndex); 309 310 int sortAndCalcSplittingIndex(int startIndex,int endIndex,int splitAxis); 311 312 void walkStacklessTree(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const; 313 314 void walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax, int startNodeIndex,int endNodeIndex) const; 315 void walkStacklessQuantizedTree(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax,int startNodeIndex,int endNodeIndex) const; 316 void walkStacklessTreeAgainstRay(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax, int startNodeIndex,int endNodeIndex) const; 317 318 ///tree traversal designed for small-memory processors like PS3 SPU 319 void walkStacklessQuantizedTreeCacheFriendly(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const; 320 321 ///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal 322 void walkRecursiveQuantizedTreeAgainstQueryAabb(const btQuantizedBvhNode* currentNode,btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const; 323 324 ///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal 325 void walkRecursiveQuantizedTreeAgainstQuantizedTree(const btQuantizedBvhNode* treeNodeA,const btQuantizedBvhNode* treeNodeB,btNodeOverlapCallback* nodeCallback) const; 326 327 328 329 330 void updateSubtreeHeaders(int leftChildNodexIndex,int rightChildNodexIndex); 331 332 public: 333 334 BT_DECLARE_ALIGNED_ALLOCATOR(); 335 336 btQuantizedBvh(); 337 338 virtual ~btQuantizedBvh(); 339 340 341 ///***************************************** expert/internal use only ************************* 342 void setQuantizationValues(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax,btScalar quantizationMargin=btScalar(1.0)); 343 QuantizedNodeArray& getLeafNodeArray() { return m_quantizedLeafNodes; } 344 ///buildInternal is expert use only: assumes that setQuantizationValues and LeafNodeArray are initialized 345 void buildInternal(); 346 ///***************************************** expert/internal use only ************************* 347 348 void reportAabbOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const; 349 void reportRayOverlappingNodex (btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget) const; 350 void reportBoxCastOverlappingNodex(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin,const btVector3& aabbMax) const; 351 352 SIMD_FORCE_INLINE void quantize(unsigned short* out, const btVector3& point,int isMax) const 353 { 354 355 btAssert(m_useQuantization); 356 357 btAssert(point.getX() <= m_bvhAabbMax.getX()); 358 btAssert(point.getY() <= m_bvhAabbMax.getY()); 359 btAssert(point.getZ() <= m_bvhAabbMax.getZ()); 360 361 btAssert(point.getX() >= m_bvhAabbMin.getX()); 362 btAssert(point.getY() >= m_bvhAabbMin.getY()); 363 btAssert(point.getZ() >= m_bvhAabbMin.getZ()); 364 365 btVector3 v = (point - m_bvhAabbMin) * m_bvhQuantization; 366 ///Make sure rounding is done in a way that unQuantize(quantizeWithClamp(...)) is conservative 367 ///end-points always set the first bit, so that they are sorted properly (so that neighbouring AABBs overlap properly) 368 ///@todo: double-check this 369 if (isMax) 370 { 371 out[0] = (unsigned short) (((unsigned short)(v.getX()+btScalar(1.)) | 1)); 372 out[1] = (unsigned short) (((unsigned short)(v.getY()+btScalar(1.)) | 1)); 373 out[2] = (unsigned short) (((unsigned short)(v.getZ()+btScalar(1.)) | 1)); 374 } else 375 { 376 out[0] = (unsigned short) (((unsigned short)(v.getX()) & 0xfffe)); 377 out[1] = (unsigned short) (((unsigned short)(v.getY()) & 0xfffe)); 378 out[2] = (unsigned short) (((unsigned short)(v.getZ()) & 0xfffe)); 379 } 380 381 382 #ifdef DEBUG_CHECK_DEQUANTIZATION 383 btVector3 newPoint = unQuantize(out); 384 if (isMax) 385 { 386 if (newPoint.getX() < point.getX()) 387 { 388 printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX()); 389 } 390 if (newPoint.getY() < point.getY()) 391 { 392 printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY()); 393 } 394 if (newPoint.getZ() < point.getZ()) 395 { 396 397 printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ()); 398 } 399 } else 400 { 401 if (newPoint.getX() > point.getX()) 402 { 403 printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX()); 404 } 405 if (newPoint.getY() > point.getY()) 406 { 407 printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY()); 408 } 409 if (newPoint.getZ() > point.getZ()) 410 { 411 printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ()); 412 } 413 } 414 #endif //DEBUG_CHECK_DEQUANTIZATION 415 416 } 417 418 419 SIMD_FORCE_INLINE void quantizeWithClamp(unsigned short* out, const btVector3& point2,int isMax) const 420 { 421 422 btAssert(m_useQuantization); 423 424 btVector3 clampedPoint(point2); 425 clampedPoint.setMax(m_bvhAabbMin); 426 clampedPoint.setMin(m_bvhAabbMax); 427 428 quantize(out,clampedPoint,isMax); 429 430 } 431 432 SIMD_FORCE_INLINE btVector3 unQuantize(const unsigned short* vecIn) const 433 { 434 btVector3 vecOut; 435 vecOut.setValue( 436 (btScalar)(vecIn[0]) / (m_bvhQuantization.getX()), 437 (btScalar)(vecIn[1]) / (m_bvhQuantization.getY()), 438 (btScalar)(vecIn[2]) / (m_bvhQuantization.getZ())); 439 vecOut += m_bvhAabbMin; 440 return vecOut; 441 } 442 443 ///setTraversalMode let's you choose between stackless, recursive or stackless cache friendly tree traversal. Note this is only implemented for quantized trees. 444 void setTraversalMode(btTraversalMode traversalMode) 445 { 446 m_traversalMode = traversalMode; 447 } 448 449 450 SIMD_FORCE_INLINE QuantizedNodeArray& getQuantizedNodeArray() 451 { 452 return m_quantizedContiguousNodes; 453 } 454 455 456 SIMD_FORCE_INLINE BvhSubtreeInfoArray& getSubtreeInfoArray() 457 { 458 return m_SubtreeHeaders; 459 } 460 461 //////////////////////////////////////////////////////////////////// 462 463 /////Calculate space needed to store BVH for serialization 464 unsigned calculateSerializeBufferSize() const; 465 466 /// Data buffer MUST be 16 byte aligned 467 virtual bool serialize(void *o_alignedDataBuffer, unsigned i_dataBufferSize, bool i_swapEndian) const; 468 469 ///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place' 470 static btQuantizedBvh *deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian); 471 472 static unsigned int getAlignmentSerializationPadding(); 473 ////////////////////////////////////////////////////////////////////// 474 475 476 virtual int calculateSerializeBufferSizeNew() const; 477 478 ///fills the dataBuffer and returns the struct name (and 0 on failure) 479 virtual const char* serialize(void* dataBuffer, btSerializer* serializer) const; 480 481 virtual void deSerializeFloat(struct btQuantizedBvhFloatData& quantizedBvhFloatData); 482 483 virtual void deSerializeDouble(struct btQuantizedBvhDoubleData& quantizedBvhDoubleData); 484 485 486 //////////////////////////////////////////////////////////////////// 487 488 SIMD_FORCE_INLINE bool isQuantized() 489 { 490 return m_useQuantization; 491 } 492 493 private: 494 // Special "copy" constructor that allows for in-place deserialization 495 // Prevents btVector3's default constructor from being called, but doesn't inialize much else 496 // ownsMemory should most likely be false if deserializing, and if you are not, don't call this (it also changes the function signature, which we need) 497 btQuantizedBvh(btQuantizedBvh &other, bool ownsMemory); 498 499 } 500 ; 501 502 503 struct btBvhSubtreeInfoData 504 { 505 int m_rootNodeIndex; 506 int m_subtreeSize; 507 unsigned short m_quantizedAabbMin[3]; 508 unsigned short m_quantizedAabbMax[3]; 509 }; 510 511 struct btOptimizedBvhNodeFloatData 512 { 513 btVector3FloatData m_aabbMinOrg; 514 btVector3FloatData m_aabbMaxOrg; 515 int m_escapeIndex; 516 int m_subPart; 517 int m_triangleIndex; 518 char m_pad[4]; 519 }; 520 521 struct btOptimizedBvhNodeDoubleData 522 { 523 btVector3DoubleData m_aabbMinOrg; 524 btVector3DoubleData m_aabbMaxOrg; 525 int m_escapeIndex; 526 int m_subPart; 527 int m_triangleIndex; 528 char m_pad[4]; 529 }; 530 531 532 struct btQuantizedBvhNodeData 533 { 534 unsigned short m_quantizedAabbMin[3]; 535 unsigned short m_quantizedAabbMax[3]; 536 int m_escapeIndexOrTriangleIndex; 537 }; 538 539 struct btQuantizedBvhFloatData 540 { 541 btVector3FloatData m_bvhAabbMin; 542 btVector3FloatData m_bvhAabbMax; 543 btVector3FloatData m_bvhQuantization; 544 int m_curNodeIndex; 545 int m_useQuantization; 546 int m_numContiguousLeafNodes; 547 int m_numQuantizedContiguousNodes; 548 btOptimizedBvhNodeFloatData *m_contiguousNodesPtr; 549 btQuantizedBvhNodeData *m_quantizedContiguousNodesPtr; 550 btBvhSubtreeInfoData *m_subTreeInfoPtr; 551 int m_traversalMode; 552 int m_numSubtreeHeaders; 553 554 }; 555 556 struct btQuantizedBvhDoubleData 557 { 558 btVector3DoubleData m_bvhAabbMin; 559 btVector3DoubleData m_bvhAabbMax; 560 btVector3DoubleData m_bvhQuantization; 561 int m_curNodeIndex; 562 int m_useQuantization; 563 int m_numContiguousLeafNodes; 564 int m_numQuantizedContiguousNodes; 565 btOptimizedBvhNodeDoubleData *m_contiguousNodesPtr; 566 btQuantizedBvhNodeData *m_quantizedContiguousNodesPtr; 567 568 int m_traversalMode; 569 int m_numSubtreeHeaders; 570 btBvhSubtreeInfoData *m_subTreeInfoPtr; 571 }; 572 573 574 SIMD_FORCE_INLINE int btQuantizedBvh::calculateSerializeBufferSizeNew() const 575 { 576 return sizeof(btQuantizedBvhData); 577 } 578 579 580 581 #endif //BT_QUANTIZED_BVH_H 582