Finally upgraded to latest Bullet subversion, about to release 2.71. Some recent changes in extern/bullet2 need to be re-applied, will check with Benoit. Ray tests in 0_FPS_Template.blend is broken, didn't figure out why yet.

HELP BUILD SYSTEM MAINTAINERS: Please help with updating all build systems: the newly added files need to be added. Note that the src/SoftBody has been added for future extension of real-time soft bodies.

extern/bullet2/src/Bullet-C-Api.h

@@ -23,11 +23,145 @@ subject to the following restrictions:
 #ifndef BULLET_C_API_H
 #define BULLET_C_API_H
 
+#define PL_DECLARE_HANDLE(name) typedef struct name##__ { int unused; } *name
+
+#ifdef BT_USE_DOUBLE_PRECISION
+typedef double	plReal;
+#else
+typedef float	plReal;
+#endif
+
+typedef plReal	plVector3[3];
+typedef plReal	plQuaternion[4];
+
 #ifdef __cplusplus
 extern "C" { 
 #endif
 
-double plNearestPoints(float p1[3], float p2[3], float p3[3], float q1[3], float q2[3], float q3[3], float *pa, float *pb, float normal[3]);
+/*	Particular physics SDK */
+	PL_DECLARE_HANDLE(plPhysicsSdkHandle);
+
+/* 	Dynamics world, belonging to some physics SDK */
+	PL_DECLARE_HANDLE(plDynamicsWorldHandle);
+
+/* Rigid Body that can be part of a Dynamics World */	
+	PL_DECLARE_HANDLE(plRigidBodyHandle);
+
+/* 	Collision Shape/Geometry, property of a Rigid Body */
+	PL_DECLARE_HANDLE(plCollisionShapeHandle);
+
+/* Constraint for Rigid Bodies */
+	PL_DECLARE_HANDLE(plConstraintHandle);
+
+/* Triangle Mesh interface */
+	PL_DECLARE_HANDLE(plMeshInterfaceHandle);
+
+/* Broadphase Scene/Proxy Handles */
+	PL_DECLARE_HANDLE(plCollisionBroadphaseHandle);
+	PL_DECLARE_HANDLE(plBroadphaseProxyHandle);
+	PL_DECLARE_HANDLE(plCollisionWorldHandle);
+
+/*
+	Create and Delete a Physics SDK	
+*/
+
+	extern	plPhysicsSdkHandle	plNewBulletSdk(); //this could be also another sdk, like ODE, PhysX etc.
+	extern	void		plDeletePhysicsSdk(plPhysicsSdkHandle	physicsSdk);
+
+/* Collision World, not strictly necessary, you can also just create a Dynamics World with Rigid Bodies which internally manages the Collision World with Collision Objects */
+
+	typedef void(*btBroadphaseCallback)(void* clientData, void* object1,void* object2);
+
+	extern plCollisionBroadphaseHandle	plCreateSapBroadphase(btBroadphaseCallback beginCallback,btBroadphaseCallback endCallback);
+
+	extern void	plDestroyBroadphase(plCollisionBroadphaseHandle bp);
+
+	extern 	plBroadphaseProxyHandle plCreateProxy(plCollisionBroadphaseHandle bp, void* clientData, plReal minX,plReal minY,plReal minZ, plReal maxX,plReal maxY, plReal maxZ);
+
+	extern void plDestroyProxy(plCollisionBroadphaseHandle bp, plBroadphaseProxyHandle proxyHandle);
+
+	extern void plSetBoundingBox(plBroadphaseProxyHandle proxyHandle, plReal minX,plReal minY,plReal minZ, plReal maxX,plReal maxY, plReal maxZ);
+
+/* todo: add pair cache support with queries like add/remove/find pair */
+	
+	extern plCollisionWorldHandle plCreateCollisionWorld(plPhysicsSdkHandle physicsSdk);
+
+/* todo: add/remove objects */
+	
+
+/* Dynamics World */
+
+	extern  plDynamicsWorldHandle plCreateDynamicsWorld(plPhysicsSdkHandle physicsSdk);
+
+	extern  void           plDeleteDynamicsWorld(plDynamicsWorldHandle world);
+
+	extern	void	plStepSimulation(plDynamicsWorldHandle,	plReal	timeStep);
+
+	extern  void plAddRigidBody(plDynamicsWorldHandle world, plRigidBodyHandle object);
+
+	extern  void plRemoveRigidBody(plDynamicsWorldHandle world, plRigidBodyHandle object);
+
+
+/* Rigid Body  */
+
+	extern  plRigidBodyHandle plCreateRigidBody(	void* user_data,  float mass, plCollisionShapeHandle cshape );
+
+	extern  void plDeleteRigidBody(plRigidBodyHandle body);
+
+
+/* Collision Shape definition */
+
+	extern  plCollisionShapeHandle plNewSphereShape(plReal radius);
+	extern  plCollisionShapeHandle plNewBoxShape(plReal x, plReal y, plReal z);
+	extern  plCollisionShapeHandle plNewCapsuleShape(plReal radius, plReal height);	
+	extern  plCollisionShapeHandle plNewConeShape(plReal radius, plReal height);
+	extern  plCollisionShapeHandle plNewCylinderShape(plReal radius, plReal height);
+	extern	plCollisionShapeHandle plNewCompoundShape();
+	extern	void	plAddChildShape(plCollisionShapeHandle compoundShape,plCollisionShapeHandle childShape, plVector3 childPos,plQuaternion childOrn);
+
+	extern  void plDeleteShape(plCollisionShapeHandle shape);
+
+	/* Convex Meshes */
+	extern  plCollisionShapeHandle plNewConvexHullShape();
+	extern  void		plAddVertex(plCollisionShapeHandle convexHull, plReal x,plReal y,plReal z);
+/* Concave static triangle meshes */
+	extern  plMeshInterfaceHandle		   plNewMeshInterface();
+	extern  void		plAddTriangle(plMeshInterfaceHandle meshHandle, plVector3 v0,plVector3 v1,plVector3 v2);
+	extern  plCollisionShapeHandle plNewStaticTriangleMeshShape(plMeshInterfaceHandle);
+
+	extern  void plSetScaling(plCollisionShapeHandle shape, plVector3 scaling);
+
+/* SOLID has Response Callback/Table/Management */
+/* PhysX has Triggers, User Callbacks and filtering */
+/* ODE has the typedef void dNearCallback (void *data, dGeomID o1, dGeomID o2); */
+
+/*	typedef void plUpdatedPositionCallback(void* userData, plRigidBodyHandle	rbHandle, plVector3 pos); */
+/*	typedef void plUpdatedOrientationCallback(void* userData, plRigidBodyHandle	rbHandle, plQuaternion orientation); */
+
+	/* get world transform */
+	extern void	plGetOpenGLMatrix(plRigidBodyHandle object, plReal* matrix);
+	extern void	plGetPosition(plRigidBodyHandle object,plVector3 position);
+	extern void plGetOrientation(plRigidBodyHandle object,plQuaternion orientation);
+
+	/* set world transform (position/orientation) */
+	extern  void plSetPosition(plRigidBodyHandle object, const plVector3 position);
+	extern  void plSetOrientation(plRigidBodyHandle object, const plQuaternion orientation);
+	extern	void plSetEuler(plReal yaw,plReal pitch,plReal roll, plQuaternion orient);
+
+	typedef struct plRayCastResult {
+		plRigidBodyHandle		m_body;  
+		plCollisionShapeHandle	m_shape; 		
+		plVector3				m_positionWorld; 		
+		plVector3				m_normalWorld;
+	} plRayCastResult;
+
+	extern  int plRayCast(plDynamicsWorldHandle world, const plVector3 rayStart, const plVector3 rayEnd, plRayCastResult res);
+
+	/* Sweep API */
+
+	/* extern  plRigidBodyHandle plObjectCast(plDynamicsWorldHandle world, const plVector3 rayStart, const plVector3 rayEnd, plVector3 hitpoint, plVector3 normal); */
+
+	/* Continuous Collision Detection API */
 
 #ifdef __cplusplus
 }

extern/bullet2/src/BulletCollision/BroadphaseCollision/btAxisSweep3.cpp

@@ -21,640 +21,18 @@
 
 #include <assert.h>
 
-#ifdef DEBUG_BROADPHASE
+btAxisSweep3::btAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, unsigned short int maxHandles, btOverlappingPairCache* pairCache)
-#include <stdio.h>
+:btAxisSweep3Internal<unsigned short int>(worldAabbMin,worldAabbMax,0xfffe,0xffff,maxHandles,pairCache)
-void btAxisSweep3::debugPrintAxis(int axis, bool checkCardinality)
 {
-	int numEdges = m_pHandles[0].m_maxEdges[axis];
-	printf("SAP Axis %d, numEdges=%d\n",axis,numEdges);
-
-	int i;
-	for (i=0;i<numEdges+1;i++)
-	{
-		Edge* pEdge = m_pEdges[axis] + i;
-		Handle* pHandlePrev = getHandle(pEdge->m_handle);
-		int handleIndex = pEdge->IsMax()? pHandlePrev->m_maxEdges[axis] : pHandlePrev->m_minEdges[axis];
-		char beginOrEnd;
-		beginOrEnd=pEdge->IsMax()?'E':'B';
-		printf("	[%c,h=%d,p=%x,i=%d]\n",beginOrEnd,pEdge->m_handle,pEdge->m_pos,handleIndex);
-	}
-
-	if (checkCardinality)
-		assert(numEdges == m_numHandles*2+1);
-}
-#endif //DEBUG_BROADPHASE
-
-
-btBroadphaseProxy*	btAxisSweep3::createProxy(  const btVector3& min,  const btVector3& max,int shapeType,void* userPtr,short int collisionFilterGroup,short int collisionFilterMask)
-{
-		(void)shapeType;
-		BP_FP_INT_TYPE handleId = addHandle(min,max, userPtr,collisionFilterGroup,collisionFilterMask);
-		
-		Handle* handle = getHandle(handleId);
-				
-		return handle;
-}
-
-void	btAxisSweep3::destroyProxy(btBroadphaseProxy* proxy)
-{
-	Handle* handle = static_cast<Handle*>(proxy);
-	removeHandle(handle->m_handleId);
-}
-
-void	btAxisSweep3::setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax)
-{
-	Handle* handle = static_cast<Handle*>(proxy);
-	updateHandle(handle->m_handleId,aabbMin,aabbMax);
-
-}
-
-
-
-
-
-
-btAxisSweep3::btAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, int maxHandles)
-:btOverlappingPairCache()
-{
-	m_invalidPair = 0;
-	//assert(bounds.HasVolume());
-
 	// 1 handle is reserved as sentinel
-	btAssert(maxHandles > 1 && maxHandles < BP_MAX_HANDLES);
+	btAssert(maxHandles > 1 && maxHandles < 32767);
-
-	// init bounds
-	m_worldAabbMin = worldAabbMin;
-	m_worldAabbMax = worldAabbMax;
-
-	btVector3 aabbSize = m_worldAabbMax - m_worldAabbMin;
-
-	BP_FP_INT_TYPE	maxInt = BP_HANDLE_SENTINEL;
-
-	m_quantize = btVector3(btScalar(maxInt),btScalar(maxInt),btScalar(maxInt)) / aabbSize;
-
-	// allocate handles buffer and put all handles on free list
-	m_pHandles = new Handle[maxHandles];
-	m_maxHandles = maxHandles;
-	m_numHandles = 0;
-
-	// handle 0 is reserved as the null index, and is also used as the sentinel
-	m_firstFreeHandle = 1;
-	{
-		for (BP_FP_INT_TYPE i = m_firstFreeHandle; i < maxHandles; i++)
-			m_pHandles[i].SetNextFree(i + 1);
-		m_pHandles[maxHandles - 1].SetNextFree(0);
-	}
-
-	{
-	// allocate edge buffers
-	for (int i = 0; i < 3; i++)
-		m_pEdges[i] = new Edge[maxHandles * 2];
-	}
-	//removed overlap management
-
-	// make boundary sentinels
-	
-	m_pHandles[0].m_clientObject = 0;
-
-	for (int axis = 0; axis < 3; axis++)
-	{
-		m_pHandles[0].m_minEdges[axis] = 0;
-		m_pHandles[0].m_maxEdges[axis] = 1;
-
-		m_pEdges[axis][0].m_pos = 0;
-		m_pEdges[axis][0].m_handle = 0;
-		m_pEdges[axis][1].m_pos = BP_HANDLE_SENTINEL;
-		m_pEdges[axis][1].m_handle = 0;
-#ifdef DEBUG_BROADPHASE
-		debugPrintAxis(axis);
-#endif //DEBUG_BROADPHASE
-
-	}
 
 }
 
-btAxisSweep3::~btAxisSweep3()
+
+bt32BitAxisSweep3::bt32BitAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, unsigned int maxHandles , btOverlappingPairCache* pairCache )
+:btAxisSweep3Internal<unsigned int>(worldAabbMin,worldAabbMax,0xfffffffe,0x7fffffff,maxHandles,pairCache)
 {
-	
+	// 1 handle is reserved as sentinel
-	for (int i = 2; i >= 0; i--)
+	btAssert(maxHandles > 1 && maxHandles < 2147483647);
-		delete[] m_pEdges[i];
-	delete[] m_pHandles;
-}
-
-void btAxisSweep3::quantize(BP_FP_INT_TYPE* out, const btPoint3& point, int isMax) const
-{
-	btPoint3 clampedPoint(point);
-	
-
-
-	clampedPoint.setMax(m_worldAabbMin);
-	clampedPoint.setMin(m_worldAabbMax);
-
-	btVector3 v = (clampedPoint - m_worldAabbMin) * m_quantize;
-	out[0] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getX() & BP_HANDLE_MASK) | isMax);
-	out[1] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getY() & BP_HANDLE_MASK) | isMax);
-	out[2] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getZ() & BP_HANDLE_MASK) | isMax);
-	
-}
-
-
-
-BP_FP_INT_TYPE btAxisSweep3::allocHandle()
-{
-	assert(m_firstFreeHandle);
-
-	BP_FP_INT_TYPE handle = m_firstFreeHandle;
-	m_firstFreeHandle = getHandle(handle)->GetNextFree();
-	m_numHandles++;
-
-	return handle;
-}
-
-void btAxisSweep3::freeHandle(BP_FP_INT_TYPE handle)
-{
-	assert(handle > 0 && handle < m_maxHandles);
-
-	getHandle(handle)->SetNextFree(m_firstFreeHandle);
-	m_firstFreeHandle = handle;
-
-	m_numHandles--;
-}
-
-
-
-BP_FP_INT_TYPE btAxisSweep3::addHandle(const btPoint3& aabbMin,const btPoint3& aabbMax, void* pOwner,short int collisionFilterGroup,short int collisionFilterMask)
-{
-	// quantize the bounds
-	BP_FP_INT_TYPE min[3], max[3];
-	quantize(min, aabbMin, 0);
-	quantize(max, aabbMax, 1);
-
-	// allocate a handle
-	BP_FP_INT_TYPE handle = allocHandle();
-	assert(handle!= 0xcdcd);
-
-	Handle* pHandle = getHandle(handle);
-	
-	pHandle->m_handleId = handle;
-	//pHandle->m_pOverlaps = 0;
-	pHandle->m_clientObject = pOwner;
-	pHandle->m_collisionFilterGroup = collisionFilterGroup;
-	pHandle->m_collisionFilterMask = collisionFilterMask;
-
-	// compute current limit of edge arrays
-	BP_FP_INT_TYPE limit = m_numHandles * 2;
-
-	
-	// insert new edges just inside the max boundary edge
-	for (BP_FP_INT_TYPE axis = 0; axis < 3; axis++)
-	{
-
-		m_pHandles[0].m_maxEdges[axis] += 2;
-
-		m_pEdges[axis][limit + 1] = m_pEdges[axis][limit - 1];
-
-		m_pEdges[axis][limit - 1].m_pos = min[axis];
-		m_pEdges[axis][limit - 1].m_handle = handle;
-
-		m_pEdges[axis][limit].m_pos = max[axis];
-		m_pEdges[axis][limit].m_handle = handle;
-
-		pHandle->m_minEdges[axis] = limit - 1;
-		pHandle->m_maxEdges[axis] = limit;
-	}
-
-	// now sort the new edges to their correct position
-	sortMinDown(0, pHandle->m_minEdges[0], false);
-	sortMaxDown(0, pHandle->m_maxEdges[0], false);
-	sortMinDown(1, pHandle->m_minEdges[1], false);
-	sortMaxDown(1, pHandle->m_maxEdges[1], false);
-	sortMinDown(2, pHandle->m_minEdges[2], true);
-	sortMaxDown(2, pHandle->m_maxEdges[2], true);
-
-
-	return handle;
-}
-
-
-void btAxisSweep3::removeHandle(BP_FP_INT_TYPE handle)
-{
-	
-	Handle* pHandle = getHandle(handle);
-
-	//explicitly remove the pairs containing the proxy
-	//we could do it also in the sortMinUp (passing true)
-	//todo: compare performance
-	removeOverlappingPairsContainingProxy(pHandle);
-
-
-	// compute current limit of edge arrays
-	int limit = m_numHandles * 2;
-	
-	int axis;
-
-	for (axis = 0;axis<3;axis++)
-	{
-		m_pHandles[0].m_maxEdges[axis] -= 2;
-	}
-
-	// remove the edges by sorting them up to the end of the list
-	for ( axis = 0; axis < 3; axis++)
-	{
-		Edge* pEdges = m_pEdges[axis];
-		BP_FP_INT_TYPE max = pHandle->m_maxEdges[axis];
-		pEdges[max].m_pos = BP_HANDLE_SENTINEL;
-
-		sortMaxUp(axis,max,false);
-
-
-		BP_FP_INT_TYPE i = pHandle->m_minEdges[axis];
-		pEdges[i].m_pos = BP_HANDLE_SENTINEL;
-
-
-		sortMinUp(axis,i,false);
-
-		pEdges[limit-1].m_handle = 0;
-		pEdges[limit-1].m_pos = BP_HANDLE_SENTINEL;
-		
-#ifdef DEBUG_BROADPHASE
-			debugPrintAxis(axis,false);
-#endif //DEBUG_BROADPHASE
-
-
-	}
-
-
-	// free the handle
-	freeHandle(handle);
-
-	
-}
-
-extern int gOverlappingPairs;
-
-
-void	btAxisSweep3::refreshOverlappingPairs()
-{
-
-}
-void	btAxisSweep3::processAllOverlappingPairs(btOverlapCallback* callback)
-{
-
-	//perform a sort, to find duplicates and to sort 'invalid' pairs to the end
-	m_overlappingPairArray.heapSort(btBroadphasePairSortPredicate());
-
-	//remove the 'invalid' ones
-#ifdef USE_POPBACK_REMOVAL
-	while (m_invalidPair>0)
-	{
-		m_invalidPair--;
-		m_overlappingPairArray.pop_back();
-	}
-#else	
-	m_overlappingPairArray.resize(m_overlappingPairArray.size() - m_invalidPair);
-	m_invalidPair = 0;
-#endif
-
-	
-	int i;
-
-	btBroadphasePair previousPair;
-	previousPair.m_pProxy0 = 0;
-	previousPair.m_pProxy1 = 0;
-	previousPair.m_algorithm = 0;
-	
-	
-	for (i=0;i<m_overlappingPairArray.size();i++)
-	{
-	
-		btBroadphasePair& pair = m_overlappingPairArray[i];
-
-		bool isDuplicate = (pair == previousPair);
-
-		previousPair = pair;
-
-		bool needsRemoval = false;
-
-		if (!isDuplicate)
-		{
-			bool hasOverlap = testOverlap(pair.m_pProxy0,pair.m_pProxy1);
-
-			if (hasOverlap)
-			{
-				needsRemoval = callback->processOverlap(pair);
-			} else
-			{
-				needsRemoval = true;
-			}
-		} else
-		{
-			//remove duplicate
-			needsRemoval = true;
-			//should have no algorithm
-			btAssert(!pair.m_algorithm);
-		}
-		
-		if (needsRemoval)
-		{
-			cleanOverlappingPair(pair);
-
-	//		m_overlappingPairArray.swap(i,m_overlappingPairArray.size()-1);
-	//		m_overlappingPairArray.pop_back();
-			pair.m_pProxy0 = 0;
-			pair.m_pProxy1 = 0;
-			m_invalidPair++;
-			gOverlappingPairs--;
-		} 
-		
-	}
-}
-
-
-bool btAxisSweep3::testOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
-{
-	const Handle* pHandleA = static_cast<Handle*>(proxy0);
-	const Handle* pHandleB = static_cast<Handle*>(proxy1);
-	
-	//optimization 1: check the array index (memory address), instead of the m_pos
-
-	for (int axis = 0; axis < 3; axis++)
-	{ 
-		if (pHandleA->m_maxEdges[axis] < pHandleB->m_minEdges[axis] || 
-			pHandleB->m_maxEdges[axis] < pHandleA->m_minEdges[axis]) 
-		{ 
-			return false; 
-		} 
-	} 
-	return true;
-}
-
-bool btAxisSweep3::testOverlap(int ignoreAxis,const Handle* pHandleA, const Handle* pHandleB)
-{
-	//optimization 1: check the array index (memory address), instead of the m_pos
-
-	for (int axis = 0; axis < 3; axis++)
-	{ 
-		if (axis != ignoreAxis)
-		{
-			if (pHandleA->m_maxEdges[axis] < pHandleB->m_minEdges[axis] || 
-				pHandleB->m_maxEdges[axis] < pHandleA->m_minEdges[axis]) 
-			{ 
-				return false; 
-			} 
-		}
-	} 
-
-	//optimization 2: only 2 axis need to be tested (conflicts with 'delayed removal' optimization)
-
-	/*for (int axis = 0; axis < 3; axis++)
-	{
-		if (m_pEdges[axis][pHandleA->m_maxEdges[axis]].m_pos < m_pEdges[axis][pHandleB->m_minEdges[axis]].m_pos ||
-			m_pEdges[axis][pHandleB->m_maxEdges[axis]].m_pos < m_pEdges[axis][pHandleA->m_minEdges[axis]].m_pos)
-		{
-			return false;
-		}
-	}
-	*/
-
-	return true;
-}
-
-void btAxisSweep3::updateHandle(BP_FP_INT_TYPE handle, const btPoint3& aabbMin,const btPoint3& aabbMax)
-{
-//	assert(bounds.IsFinite());
-	//assert(bounds.HasVolume());
-
-	Handle* pHandle = getHandle(handle);
-
-	// quantize the new bounds
-	BP_FP_INT_TYPE min[3], max[3];
-	quantize(min, aabbMin, 0);
-	quantize(max, aabbMax, 1);
-
-	// update changed edges
-	for (int axis = 0; axis < 3; axis++)
-	{
-		BP_FP_INT_TYPE emin = pHandle->m_minEdges[axis];
-		BP_FP_INT_TYPE emax = pHandle->m_maxEdges[axis];
-
-		int dmin = (int)min[axis] - (int)m_pEdges[axis][emin].m_pos;
-		int dmax = (int)max[axis] - (int)m_pEdges[axis][emax].m_pos;
-
-		m_pEdges[axis][emin].m_pos = min[axis];
-		m_pEdges[axis][emax].m_pos = max[axis];
-
-		// expand (only adds overlaps)
-		if (dmin < 0)
-			sortMinDown(axis, emin);
-
-		if (dmax > 0)
-			sortMaxUp(axis, emax);
-
-		// shrink (only removes overlaps)
-		if (dmin > 0)
-			sortMinUp(axis, emin);
-
-		if (dmax < 0)
-			sortMaxDown(axis, emax);
-
-#ifdef DEBUG_BROADPHASE
-	debugPrintAxis(axis);
-#endif //DEBUG_BROADPHASE
-	}
-
-	
-}
-
-
-
-
-// sorting a min edge downwards can only ever *add* overlaps
-void btAxisSweep3::sortMinDown(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps)
-{
-
-	Edge* pEdge = m_pEdges[axis] + edge;
-	Edge* pPrev = pEdge - 1;
-	Handle* pHandleEdge = getHandle(pEdge->m_handle);
-
-	while (pEdge->m_pos < pPrev->m_pos)
-	{
-		Handle* pHandlePrev = getHandle(pPrev->m_handle);
-
-		if (pPrev->IsMax())
-		{
-			// if previous edge is a maximum check the bounds and add an overlap if necessary
-			if (updateOverlaps && testOverlap(axis,pHandleEdge, pHandlePrev))
-			{
-				addOverlappingPair(pHandleEdge,pHandlePrev);
-
-				//AddOverlap(pEdge->m_handle, pPrev->m_handle);
-
-			}
-
-			// update edge reference in other handle
-			pHandlePrev->m_maxEdges[axis]++;
-		}
-		else
-			pHandlePrev->m_minEdges[axis]++;
-
-		pHandleEdge->m_minEdges[axis]--;
-
-		// swap the edges
-		Edge swap = *pEdge;
-		*pEdge = *pPrev;
-		*pPrev = swap;
-
-		// decrement
-		pEdge--;
-		pPrev--;
-	}
-
-#ifdef DEBUG_BROADPHASE
-	debugPrintAxis(axis);
-#endif //DEBUG_BROADPHASE
-
-}
-
-// sorting a min edge upwards can only ever *remove* overlaps
-void btAxisSweep3::sortMinUp(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps)
-{
-	Edge* pEdge = m_pEdges[axis] + edge;
-	Edge* pNext = pEdge + 1;
-	Handle* pHandleEdge = getHandle(pEdge->m_handle);
-
-	while (pNext->m_handle && (pEdge->m_pos >= pNext->m_pos))
-	{
-		Handle* pHandleNext = getHandle(pNext->m_handle);
-
-		if (pNext->IsMax())
-		{
-			// if next edge is maximum remove any overlap between the two handles
-			if (updateOverlaps)
-			{
-				/*
-				Handle* handle0 = getHandle(pEdge->m_handle);
-				Handle* handle1 = getHandle(pNext->m_handle);
-				btBroadphasePair tmpPair(*handle0,*handle1);
-				removeOverlappingPair(tmpPair);
-				*/
-
-			}
-
-			// update edge reference in other handle
-			pHandleNext->m_maxEdges[axis]--;
-		}
-		else
-			pHandleNext->m_minEdges[axis]--;
-
-		pHandleEdge->m_minEdges[axis]++;
-
-		// swap the edges
-		Edge swap = *pEdge;
-		*pEdge = *pNext;
-		*pNext = swap;
-
-		// increment
-		pEdge++;
-		pNext++;
-	}
-
-
-}
-
-// sorting a max edge downwards can only ever *remove* overlaps
-void btAxisSweep3::sortMaxDown(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps)
-{
-
-	Edge* pEdge = m_pEdges[axis] + edge;
-	Edge* pPrev = pEdge - 1;
-	Handle* pHandleEdge = getHandle(pEdge->m_handle);
-
-	while (pEdge->m_pos < pPrev->m_pos)
-	{
-		Handle* pHandlePrev = getHandle(pPrev->m_handle);
-
-		if (!pPrev->IsMax())
-		{
-			// if previous edge was a minimum remove any overlap between the two handles
-			if (updateOverlaps)
-			{
-				//this is done during the overlappingpairarray iteration/narrowphase collision
-				/*
-				Handle* handle0 = getHandle(pEdge->m_handle);
-				Handle* handle1 = getHandle(pPrev->m_handle);
-				btBroadphasePair* pair = findPair(handle0,handle1);
-				//assert(pair);
-
-				if (pair)
-				{
-					removeOverlappingPair(*pair);
-				}
-				*/
-
-			}
-
-			// update edge reference in other handle
-			pHandlePrev->m_minEdges[axis]++;;
-		}
-		else
-			pHandlePrev->m_maxEdges[axis]++;
-
-		pHandleEdge->m_maxEdges[axis]--;
-
-		// swap the edges
-		Edge swap = *pEdge;
-		*pEdge = *pPrev;
-		*pPrev = swap;
-
-		// decrement
-		pEdge--;
-		pPrev--;
-	}
-
-	
-#ifdef DEBUG_BROADPHASE
-	debugPrintAxis(axis);
-#endif //DEBUG_BROADPHASE
-
-}
-
-// sorting a max edge upwards can only ever *add* overlaps
-void btAxisSweep3::sortMaxUp(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps)
-{
-	Edge* pEdge = m_pEdges[axis] + edge;
-	Edge* pNext = pEdge + 1;
-	Handle* pHandleEdge = getHandle(pEdge->m_handle);
-
-	while (pNext->m_handle && (pEdge->m_pos >= pNext->m_pos))
-	{
-		Handle* pHandleNext = getHandle(pNext->m_handle);
-
-		if (!pNext->IsMax())
-		{
-			// if next edge is a minimum check the bounds and add an overlap if necessary
-			if (updateOverlaps && testOverlap(axis, pHandleEdge, pHandleNext))
-			{
-				Handle* handle0 = getHandle(pEdge->m_handle);
-				Handle* handle1 = getHandle(pNext->m_handle);
-				addOverlappingPair(handle0,handle1);
-			}
-
-			// update edge reference in other handle
-			pHandleNext->m_minEdges[axis]--;
-		}
-		else
-			pHandleNext->m_maxEdges[axis]--;
-
-		pHandleEdge->m_maxEdges[axis]++;
-
-		// swap the edges
-		Edge swap = *pEdge;
-		*pEdge = *pNext;
-		*pNext = swap;
-
-		// increment
-		pEdge++;
-		pNext++;
-	}
-	
 }

extern/bullet2/src/BulletCollision/BroadphaseCollision/btAxisSweep3.h

@@ -19,34 +19,25 @@
 #ifndef AXIS_SWEEP_3_H
 #define AXIS_SWEEP_3_H
 
-#include "../../LinearMath/btPoint3.h"
+#include "LinearMath/btPoint3.h"
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
 #include "btOverlappingPairCache.h"
+#include "btBroadphaseInterface.h"
 #include "btBroadphaseProxy.h"
-
+#include "btOverlappingPairCallback.h"
-
-//Enable BP_USE_FIXEDPOINT_INT_32 if you need more then 32767 objects
-//#define BP_USE_FIXEDPOINT_INT_32 1
-
-#ifdef BP_USE_FIXEDPOINT_INT_32
-	#define BP_FP_INT_TYPE unsigned int
-	#define BP_MAX_HANDLES 1500000 //arbitrary maximum number of handles
-	#define BP_HANDLE_SENTINEL 0x7fffffff
-	#define BP_HANDLE_MASK	0xfffffffe
-#else
-	#define BP_FP_INT_TYPE unsigned short int
-	#define BP_MAX_HANDLES 32767
-	#define BP_HANDLE_SENTINEL 0xffff
-	#define BP_HANDLE_MASK	0xfffe
-#endif //BP_USE_FIXEDPOINT_INT_32
 
 //#define DEBUG_BROADPHASE 1
 
-/// btAxisSweep3 is an efficient implementation of the 3d axis sweep and prune broadphase.
+/// The internal templace class btAxisSweep3Internal implements the sweep and prune broadphase.
-/// It uses arrays rather then lists for storage of the 3 axis. Also it operates using integer coordinates instead of floats.
+/// It uses quantized integers to represent the begin and end points for each of the 3 axis.
-/// The testOverlap check is optimized to check the array index, rather then the actual AABB coordinates/pos
+/// Dont use this class directly, use btAxisSweep3 or bt32BitAxisSweep3 instead.
-class btAxisSweep3 : public btOverlappingPairCache
+template <typename BP_FP_INT_TYPE>
+class btAxisSweep3Internal : public btBroadphaseInterface
 {
+protected:
+
+	BP_FP_INT_TYPE	m_bpHandleMask;
+	BP_FP_INT_TYPE	m_handleSentinel;
 
 public:
 	
@@ -57,40 +48,52 @@ public:
 		BP_FP_INT_TYPE m_pos;			// low bit is min/max
 		BP_FP_INT_TYPE m_handle;
 
-		BP_FP_INT_TYPE IsMax() const {return m_pos & 1;}
+		BP_FP_INT_TYPE IsMax() const {return static_cast<BP_FP_INT_TYPE>(m_pos & 1);}
 	};
 
 public:
+	//This breaks the Intel compiler, see http://softwarecommunity.intel.com/isn/Community/en-US/forums/thread/30253577.aspx
 	class Handle : public btBroadphaseProxy
+	//ATTRIBUTE_ALIGNED16(class) Handle : public btBroadphaseProxy
 	{
 	public:
-		
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+	
 		// indexes into the edge arrays
 		BP_FP_INT_TYPE m_minEdges[3], m_maxEdges[3];		// 6 * 2 = 12
-		BP_FP_INT_TYPE m_handleId;
+//		BP_FP_INT_TYPE m_uniqueId;
 		BP_FP_INT_TYPE m_pad;
 		
 		//void* m_pOwner; this is now in btBroadphaseProxy.m_clientObject
 	
-		inline void SetNextFree(BP_FP_INT_TYPE next) {m_minEdges[0] = next;}
+		SIMD_FORCE_INLINE void SetNextFree(BP_FP_INT_TYPE next) {m_minEdges[0] = next;}
-		inline BP_FP_INT_TYPE GetNextFree() const {return m_minEdges[0];}
+		SIMD_FORCE_INLINE BP_FP_INT_TYPE GetNextFree() const {return m_minEdges[0];}
 	};		// 24 bytes + 24 for Edge structures = 44 bytes total per entry
 
 	
-private:
+protected:
 	btPoint3 m_worldAabbMin;						// overall system bounds
 	btPoint3 m_worldAabbMax;						// overall system bounds
 
 	btVector3 m_quantize;						// scaling factor for quantization
 
 	BP_FP_INT_TYPE m_numHandles;						// number of active handles
-	int m_maxHandles;						// max number of handles
+	BP_FP_INT_TYPE m_maxHandles;						// max number of handles
 	Handle* m_pHandles;						// handles pool
+	void* m_pHandlesRawPtr;
 	BP_FP_INT_TYPE m_firstFreeHandle;		// free handles list
 
 	Edge* m_pEdges[3];						// edge arrays for the 3 axes (each array has m_maxHandles * 2 + 2 sentinel entries)
+	void* m_pEdgesRawPtr[3];
 
-	int m_invalidPair;
+	btOverlappingPairCache* m_pairCache;
+
+	///btOverlappingPairCallback is an additional optional user callback for adding/removing overlapping pairs, similar interface to btOverlappingPairCache.
+	btOverlappingPairCallback* m_userPairCallback;
+	
+	bool	m_ownsPairCache;
+
+	int	m_invalidPair;
 
 	// allocation/deallocation
 	BP_FP_INT_TYPE allocHandle();
@@ -108,29 +111,801 @@ private:
 
 	void quantize(BP_FP_INT_TYPE* out, const btPoint3& point, int isMax) const;
 
-	void sortMinDown(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps = true);
+	void sortMinDown(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps );
-	void sortMinUp(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps = true);
+	void sortMinUp(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps );
-	void sortMaxDown(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps = true);
+	void sortMaxDown(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps );
-	void sortMaxUp(int axis, BP_FP_INT_TYPE edge, bool updateOverlaps = true);
+	void sortMaxUp(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps );
 
 public:
-	btAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, int maxHandles = 16384);
-	virtual	~btAxisSweep3();
 
-	virtual void	refreshOverlappingPairs();
+	btAxisSweep3Internal(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, BP_FP_INT_TYPE handleMask, BP_FP_INT_TYPE handleSentinel, BP_FP_INT_TYPE maxHandles = 16384, btOverlappingPairCache* pairCache=0);
+
+	virtual	~btAxisSweep3Internal();
+
+	BP_FP_INT_TYPE getNumHandles() const
+	{
+		return m_numHandles;
+	}
+
+	virtual void	calculateOverlappingPairs(btDispatcher* dispatcher);
 	
-	BP_FP_INT_TYPE addHandle(const btPoint3& aabbMin,const btPoint3& aabbMax, void* pOwner,short int collisionFilterGroup,short int collisionFilterMask);
+	BP_FP_INT_TYPE addHandle(const btPoint3& aabbMin,const btPoint3& aabbMax, void* pOwner,short int collisionFilterGroup,short int collisionFilterMask,btDispatcher* dispatcher,void* multiSapProxy);
-	void removeHandle(BP_FP_INT_TYPE handle);
+	void removeHandle(BP_FP_INT_TYPE handle,btDispatcher* dispatcher);
-	void updateHandle(BP_FP_INT_TYPE handle, const btPoint3& aabbMin,const btPoint3& aabbMax);
+	void updateHandle(BP_FP_INT_TYPE handle, const btPoint3& aabbMin,const btPoint3& aabbMax,btDispatcher* dispatcher);
-	inline Handle* getHandle(BP_FP_INT_TYPE index) const {return m_pHandles + index;}
+	SIMD_FORCE_INLINE Handle* getHandle(BP_FP_INT_TYPE index) const {return m_pHandles + index;}
 
 	void	processAllOverlappingPairs(btOverlapCallback* callback);
 
 	//Broadphase Interface
-	virtual btBroadphaseProxy*	createProxy(  const btVector3& min,  const btVector3& max,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
+	virtual btBroadphaseProxy*	createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask,btDispatcher* dispatcher,void* multiSapProxy);
-	virtual void	destroyProxy(btBroadphaseProxy* proxy);
+	virtual void	destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
-	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax);
+	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax,btDispatcher* dispatcher);
-	bool testOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
+	
+	bool	testAabbOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
+
+	btOverlappingPairCache*	getOverlappingPairCache()
+	{
+		return m_pairCache;
+	}
+	const btOverlappingPairCache*	getOverlappingPairCache() const
+	{
+		return m_pairCache;
+	}
+
+	void	setOverlappingPairUserCallback(btOverlappingPairCallback* pairCallback)
+	{
+		m_userPairCallback = pairCallback;
+	}
+	const btOverlappingPairCallback*	getOverlappingPairUserCallback() const
+	{
+		return m_userPairCallback;
+	}
+
+	///getAabb returns the axis aligned bounding box in the 'global' coordinate frame
+	///will add some transform later
+	virtual void getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const
+	{
+		aabbMin = m_worldAabbMin;
+		aabbMax = m_worldAabbMax;
+	}
+
+	virtual void	printStats()
+	{
+/*		printf("btAxisSweep3.h\n");
+		printf("numHandles = %d, maxHandles = %d\n",m_numHandles,m_maxHandles);
+		printf("aabbMin=%f,%f,%f,aabbMax=%f,%f,%f\n",m_worldAabbMin.getX(),m_worldAabbMin.getY(),m_worldAabbMin.getZ(),
+			m_worldAabbMax.getX(),m_worldAabbMax.getY(),m_worldAabbMax.getZ());
+			*/
+
+	}
+
+};
+
+////////////////////////////////////////////////////////////////////
+
+
+
+
+#ifdef DEBUG_BROADPHASE
+#include <stdio.h>
+
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3<BP_FP_INT_TYPE>::debugPrintAxis(int axis, bool checkCardinality)
+{
+	int numEdges = m_pHandles[0].m_maxEdges[axis];
+	printf("SAP Axis %d, numEdges=%d\n",axis,numEdges);
+
+	int i;
+	for (i=0;i<numEdges+1;i++)
+	{
+		Edge* pEdge = m_pEdges[axis] + i;
+		Handle* pHandlePrev = getHandle(pEdge->m_handle);
+		int handleIndex = pEdge->IsMax()? pHandlePrev->m_maxEdges[axis] : pHandlePrev->m_minEdges[axis];
+		char beginOrEnd;
+		beginOrEnd=pEdge->IsMax()?'E':'B';
+		printf("	[%c,h=%d,p=%x,i=%d]\n",beginOrEnd,pEdge->m_handle,pEdge->m_pos,handleIndex);
+	}
+
+	if (checkCardinality)
+		assert(numEdges == m_numHandles*2+1);
+}
+#endif //DEBUG_BROADPHASE
+
+template <typename BP_FP_INT_TYPE>
+btBroadphaseProxy*	btAxisSweep3Internal<BP_FP_INT_TYPE>::createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr,short int collisionFilterGroup,short int collisionFilterMask,btDispatcher* dispatcher,void* multiSapProxy)
+{
+		(void)shapeType;
+		BP_FP_INT_TYPE handleId = addHandle(aabbMin,aabbMax, userPtr,collisionFilterGroup,collisionFilterMask,dispatcher,multiSapProxy);
+		
+		Handle* handle = getHandle(handleId);
+				
+		return handle;
+}
+
+
+
+template <typename BP_FP_INT_TYPE>
+void	btAxisSweep3Internal<BP_FP_INT_TYPE>::destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher)
+{
+	Handle* handle = static_cast<Handle*>(proxy);
+	removeHandle(static_cast<BP_FP_INT_TYPE>(handle->m_uniqueId), dispatcher);
+}
+
+template <typename BP_FP_INT_TYPE>
+void	btAxisSweep3Internal<BP_FP_INT_TYPE>::setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax,btDispatcher* dispatcher)
+{
+	Handle* handle = static_cast<Handle*>(proxy);
+	updateHandle(static_cast<BP_FP_INT_TYPE>(handle->m_uniqueId), aabbMin, aabbMax,dispatcher);
+
+}
+
+
+
+
+
+template <typename BP_FP_INT_TYPE>
+btAxisSweep3Internal<BP_FP_INT_TYPE>::btAxisSweep3Internal(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, BP_FP_INT_TYPE handleMask, BP_FP_INT_TYPE handleSentinel,BP_FP_INT_TYPE userMaxHandles, btOverlappingPairCache* pairCache )
+:m_bpHandleMask(handleMask),
+m_handleSentinel(handleSentinel),
+m_pairCache(pairCache),
+m_userPairCallback(0),
+m_ownsPairCache(false),
+m_invalidPair(0)
+{
+	BP_FP_INT_TYPE maxHandles = static_cast<BP_FP_INT_TYPE>(userMaxHandles+1);//need to add one sentinel handle
+
+	if (!m_pairCache)
+	{
+		void* ptr = btAlignedAlloc(sizeof(btHashedOverlappingPairCache),16);
+		m_pairCache = new(ptr) btHashedOverlappingPairCache();
+		m_ownsPairCache = true;
+	}
+
+	//assert(bounds.HasVolume());
+
+	// init bounds
+	m_worldAabbMin = worldAabbMin;
+	m_worldAabbMax = worldAabbMax;
+
+	btVector3 aabbSize = m_worldAabbMax - m_worldAabbMin;
+
+	BP_FP_INT_TYPE	maxInt = m_handleSentinel;
+
+	m_quantize = btVector3(btScalar(maxInt),btScalar(maxInt),btScalar(maxInt)) / aabbSize;
+
+	// allocate handles buffer and put all handles on free list
+	m_pHandlesRawPtr = btAlignedAlloc(sizeof(Handle)*maxHandles,16);
+	m_pHandles = new(m_pHandlesRawPtr) Handle[maxHandles];
+
+	m_maxHandles = maxHandles;
+	m_numHandles = 0;
+
+	// handle 0 is reserved as the null index, and is also used as the sentinel
+	m_firstFreeHandle = 1;
+	{
+		for (BP_FP_INT_TYPE i = m_firstFreeHandle; i < maxHandles; i++)
+			m_pHandles[i].SetNextFree(static_cast<BP_FP_INT_TYPE>(i + 1));
+		m_pHandles[maxHandles - 1].SetNextFree(0);
+	}
+
+	{
+		// allocate edge buffers
+		for (int i = 0; i < 3; i++)
+		{
+			m_pEdgesRawPtr[i] = btAlignedAlloc(sizeof(Edge)*maxHandles*2,16);
+			m_pEdges[i] = new(m_pEdgesRawPtr[i]) Edge[maxHandles * 2];
+		}
+	}
+	//removed overlap management
+
+	// make boundary sentinels
+	
+	m_pHandles[0].m_clientObject = 0;
+
+	for (int axis = 0; axis < 3; axis++)
+	{
+		m_pHandles[0].m_minEdges[axis] = 0;
+		m_pHandles[0].m_maxEdges[axis] = 1;
+
+		m_pEdges[axis][0].m_pos = 0;
+		m_pEdges[axis][0].m_handle = 0;
+		m_pEdges[axis][1].m_pos = m_handleSentinel;
+		m_pEdges[axis][1].m_handle = 0;
+#ifdef DEBUG_BROADPHASE
+		debugPrintAxis(axis);
+#endif //DEBUG_BROADPHASE
+
+	}
+
+}
+
+template <typename BP_FP_INT_TYPE>
+btAxisSweep3Internal<BP_FP_INT_TYPE>::~btAxisSweep3Internal()
+{
+	
+	for (int i = 2; i >= 0; i--)
+	{
+		btAlignedFree(m_pEdgesRawPtr[i]);
+	}
+	btAlignedFree(m_pHandlesRawPtr);
+
+	if (m_ownsPairCache)
+	{
+		m_pairCache->~btOverlappingPairCache();
+		btAlignedFree(m_pairCache);
+	}
+}
+
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::quantize(BP_FP_INT_TYPE* out, const btPoint3& point, int isMax) const
+{
+	btPoint3 clampedPoint(point);
+	
+
+
+	clampedPoint.setMax(m_worldAabbMin);
+	clampedPoint.setMin(m_worldAabbMax);
+
+	btVector3 v = (clampedPoint - m_worldAabbMin) * m_quantize;
+	out[0] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getX() & m_bpHandleMask) | isMax);
+	out[1] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getY() & m_bpHandleMask) | isMax);
+	out[2] = (BP_FP_INT_TYPE)(((BP_FP_INT_TYPE)v.getZ() & m_bpHandleMask) | isMax);
+	
+}
+
+
+template <typename BP_FP_INT_TYPE>
+BP_FP_INT_TYPE btAxisSweep3Internal<BP_FP_INT_TYPE>::allocHandle()
+{
+	assert(m_firstFreeHandle);
+
+	BP_FP_INT_TYPE handle = m_firstFreeHandle;
+	m_firstFreeHandle = getHandle(handle)->GetNextFree();
+	m_numHandles++;
+
+	return handle;
+}
+
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::freeHandle(BP_FP_INT_TYPE handle)
+{
+	assert(handle > 0 && handle < m_maxHandles);
+
+	getHandle(handle)->SetNextFree(m_firstFreeHandle);
+	m_firstFreeHandle = handle;
+
+	m_numHandles--;
+}
+
+
+template <typename BP_FP_INT_TYPE>
+BP_FP_INT_TYPE btAxisSweep3Internal<BP_FP_INT_TYPE>::addHandle(const btPoint3& aabbMin,const btPoint3& aabbMax, void* pOwner,short int collisionFilterGroup,short int collisionFilterMask,btDispatcher* dispatcher,void* multiSapProxy)
+{
+	// quantize the bounds
+	BP_FP_INT_TYPE min[3], max[3];
+	quantize(min, aabbMin, 0);
+	quantize(max, aabbMax, 1);
+
+	// allocate a handle
+	BP_FP_INT_TYPE handle = allocHandle();
+	
+
+	Handle* pHandle = getHandle(handle);
+	
+	pHandle->m_uniqueId = static_cast<int>(handle);
+	//pHandle->m_pOverlaps = 0;
+	pHandle->m_clientObject = pOwner;
+	pHandle->m_collisionFilterGroup = collisionFilterGroup;
+	pHandle->m_collisionFilterMask = collisionFilterMask;
+	pHandle->m_multiSapParentProxy = multiSapProxy;
+
+	// compute current limit of edge arrays
+	BP_FP_INT_TYPE limit = static_cast<BP_FP_INT_TYPE>(m_numHandles * 2);
+
+	
+	// insert new edges just inside the max boundary edge
+	for (BP_FP_INT_TYPE axis = 0; axis < 3; axis++)
+	{
+
+		m_pHandles[0].m_maxEdges[axis] += 2;
+
+		m_pEdges[axis][limit + 1] = m_pEdges[axis][limit - 1];
+
+		m_pEdges[axis][limit - 1].m_pos = min[axis];
+		m_pEdges[axis][limit - 1].m_handle = handle;
+
+		m_pEdges[axis][limit].m_pos = max[axis];
+		m_pEdges[axis][limit].m_handle = handle;
+
+		pHandle->m_minEdges[axis] = static_cast<BP_FP_INT_TYPE>(limit - 1);
+		pHandle->m_maxEdges[axis] = limit;
+	}
+
+	// now sort the new edges to their correct position
+	sortMinDown(0, pHandle->m_minEdges[0], dispatcher,false);
+	sortMaxDown(0, pHandle->m_maxEdges[0], dispatcher,false);
+	sortMinDown(1, pHandle->m_minEdges[1], dispatcher,false);
+	sortMaxDown(1, pHandle->m_maxEdges[1], dispatcher,false);
+	sortMinDown(2, pHandle->m_minEdges[2], dispatcher,true);
+	sortMaxDown(2, pHandle->m_maxEdges[2], dispatcher,true);
+
+
+	return handle;
+}
+
+
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::removeHandle(BP_FP_INT_TYPE handle,btDispatcher* dispatcher)
+{
+
+	Handle* pHandle = getHandle(handle);
+
+	//explicitly remove the pairs containing the proxy
+	//we could do it also in the sortMinUp (passing true)
+	//todo: compare performance
+	if (!m_pairCache->hasDeferredRemoval())
+	{
+		m_pairCache->removeOverlappingPairsContainingProxy(pHandle,dispatcher);
+	}
+
+	// compute current limit of edge arrays
+	int limit = static_cast<int>(m_numHandles * 2);
+	
+	int axis;
+
+	for (axis = 0;axis<3;axis++)
+	{
+		m_pHandles[0].m_maxEdges[axis] -= 2;
+	}
+
+	// remove the edges by sorting them up to the end of the list
+	for ( axis = 0; axis < 3; axis++)
+	{
+		Edge* pEdges = m_pEdges[axis];
+		BP_FP_INT_TYPE max = pHandle->m_maxEdges[axis];
+		pEdges[max].m_pos = m_handleSentinel;
+
+		sortMaxUp(axis,max,dispatcher,false);
+
+
+		BP_FP_INT_TYPE i = pHandle->m_minEdges[axis];
+		pEdges[i].m_pos = m_handleSentinel;
+
+
+		sortMinUp(axis,i,dispatcher,false);
+
+		pEdges[limit-1].m_handle = 0;
+		pEdges[limit-1].m_pos = m_handleSentinel;
+		
+#ifdef DEBUG_BROADPHASE
+			debugPrintAxis(axis,false);
+#endif //DEBUG_BROADPHASE
+
+
+	}
+
+
+	// free the handle
+	freeHandle(handle);
+
+	
+}
+
+extern int gOverlappingPairs;
+//#include <stdio.h>
+
+template <typename BP_FP_INT_TYPE>
+void	btAxisSweep3Internal<BP_FP_INT_TYPE>::calculateOverlappingPairs(btDispatcher* dispatcher)
+{
+
+	if (m_pairCache->hasDeferredRemoval())
+	{
+	
+		btBroadphasePairArray&	overlappingPairArray = m_pairCache->getOverlappingPairArray();
+
+		//perform a sort, to find duplicates and to sort 'invalid' pairs to the end
+		overlappingPairArray.quickSort(btBroadphasePairSortPredicate());
+
+		overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+		m_invalidPair = 0;
+
+		
+		int i;
+
+		btBroadphasePair previousPair;
+		previousPair.m_pProxy0 = 0;
+		previousPair.m_pProxy1 = 0;
+		previousPair.m_algorithm = 0;
+		
+		
+		for (i=0;i<overlappingPairArray.size();i++)
+		{
+		
+			btBroadphasePair& pair = overlappingPairArray[i];
+
+			bool isDuplicate = (pair == previousPair);
+
+			previousPair = pair;
+
+			bool needsRemoval = false;
+
+			if (!isDuplicate)
+			{
+				bool hasOverlap = testAabbOverlap(pair.m_pProxy0,pair.m_pProxy1);
+
+				if (hasOverlap)
+				{
+					needsRemoval = false;//callback->processOverlap(pair);
+				} else
+				{
+					needsRemoval = true;
+				}
+			} else
+			{
+				//remove duplicate
+				needsRemoval = true;
+				//should have no algorithm
+				btAssert(!pair.m_algorithm);
+			}
+			
+			if (needsRemoval)
+			{
+				m_pairCache->cleanOverlappingPair(pair,dispatcher);
+
+		//		m_overlappingPairArray.swap(i,m_overlappingPairArray.size()-1);
+		//		m_overlappingPairArray.pop_back();
+				pair.m_pProxy0 = 0;
+				pair.m_pProxy1 = 0;
+				m_invalidPair++;
+				gOverlappingPairs--;
+			} 
+			
+		}
+
+	///if you don't like to skip the invalid pairs in the array, execute following code:
+	#define CLEAN_INVALID_PAIRS 1
+	#ifdef CLEAN_INVALID_PAIRS
+
+		//perform a sort, to sort 'invalid' pairs to the end
+		overlappingPairArray.quickSort(btBroadphasePairSortPredicate());
+
+		overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+		m_invalidPair = 0;
+	#endif//CLEAN_INVALID_PAIRS
+		
+		//printf("overlappingPairArray.size()=%d\n",overlappingPairArray.size());
+	}
+
+
+
+	
+
+}
+
+
+template <typename BP_FP_INT_TYPE>
+bool btAxisSweep3Internal<BP_FP_INT_TYPE>::testAabbOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+{
+	const Handle* pHandleA = static_cast<Handle*>(proxy0);
+	const Handle* pHandleB = static_cast<Handle*>(proxy1);
+	
+	//optimization 1: check the array index (memory address), instead of the m_pos
+
+	for (int axis = 0; axis < 3; axis++)
+	{ 
+		if (pHandleA->m_maxEdges[axis] < pHandleB->m_minEdges[axis] || 
+			pHandleB->m_maxEdges[axis] < pHandleA->m_minEdges[axis]) 
+		{ 
+			return false; 
+		} 
+	} 
+	return true;
+}
+
+template <typename BP_FP_INT_TYPE>
+bool btAxisSweep3Internal<BP_FP_INT_TYPE>::testOverlap(int ignoreAxis,const Handle* pHandleA, const Handle* pHandleB)
+{
+	//optimization 1: check the array index (memory address), instead of the m_pos
+
+	for (int axis = 0; axis < 3; axis++)
+	{ 
+		if (axis != ignoreAxis)
+		{
+			if (pHandleA->m_maxEdges[axis] < pHandleB->m_minEdges[axis] || 
+				pHandleB->m_maxEdges[axis] < pHandleA->m_minEdges[axis]) 
+			{ 
+				return false; 
+			} 
+		}
+	} 
+
+	//optimization 2: only 2 axis need to be tested (conflicts with 'delayed removal' optimization)
+
+	/*for (int axis = 0; axis < 3; axis++)
+	{
+		if (m_pEdges[axis][pHandleA->m_maxEdges[axis]].m_pos < m_pEdges[axis][pHandleB->m_minEdges[axis]].m_pos ||
+			m_pEdges[axis][pHandleB->m_maxEdges[axis]].m_pos < m_pEdges[axis][pHandleA->m_minEdges[axis]].m_pos)
+		{
+			return false;
+		}
+	}
+	*/
+
+	return true;
+}
+
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::updateHandle(BP_FP_INT_TYPE handle, const btPoint3& aabbMin,const btPoint3& aabbMax,btDispatcher* dispatcher)
+{
+//	assert(bounds.IsFinite());
+	//assert(bounds.HasVolume());
+
+	Handle* pHandle = getHandle(handle);
+
+	// quantize the new bounds
+	BP_FP_INT_TYPE min[3], max[3];
+	quantize(min, aabbMin, 0);
+	quantize(max, aabbMax, 1);
+
+	// update changed edges
+	for (int axis = 0; axis < 3; axis++)
+	{
+		BP_FP_INT_TYPE emin = pHandle->m_minEdges[axis];
+		BP_FP_INT_TYPE emax = pHandle->m_maxEdges[axis];
+
+		int dmin = (int)min[axis] - (int)m_pEdges[axis][emin].m_pos;
+		int dmax = (int)max[axis] - (int)m_pEdges[axis][emax].m_pos;
+
+		m_pEdges[axis][emin].m_pos = min[axis];
+		m_pEdges[axis][emax].m_pos = max[axis];
+
+		// expand (only adds overlaps)
+		if (dmin < 0)
+			sortMinDown(axis, emin,dispatcher,true);
+
+		if (dmax > 0)
+			sortMaxUp(axis, emax,dispatcher,true);
+
+		// shrink (only removes overlaps)
+		if (dmin > 0)
+			sortMinUp(axis, emin,dispatcher,true);
+
+		if (dmax < 0)
+			sortMaxDown(axis, emax,dispatcher,true);
+
+#ifdef DEBUG_BROADPHASE
+	debugPrintAxis(axis);
+#endif //DEBUG_BROADPHASE
+	}
+
+	
+}
+
+
+
+
+// sorting a min edge downwards can only ever *add* overlaps
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::sortMinDown(int axis, BP_FP_INT_TYPE edge, btDispatcher* /* dispatcher */, bool updateOverlaps)
+{
+
+	Edge* pEdge = m_pEdges[axis] + edge;
+	Edge* pPrev = pEdge - 1;
+	Handle* pHandleEdge = getHandle(pEdge->m_handle);
+
+	while (pEdge->m_pos < pPrev->m_pos)
+	{
+		Handle* pHandlePrev = getHandle(pPrev->m_handle);
+
+		if (pPrev->IsMax())
+		{
+			// if previous edge is a maximum check the bounds and add an overlap if necessary
+			if (updateOverlaps && testOverlap(axis,pHandleEdge, pHandlePrev))
+			{
+				m_pairCache->addOverlappingPair(pHandleEdge,pHandlePrev);
+				if (m_userPairCallback)
+					m_userPairCallback->addOverlappingPair(pHandleEdge,pHandlePrev);
+
+				//AddOverlap(pEdge->m_handle, pPrev->m_handle);
+
+			}
+
+			// update edge reference in other handle
+			pHandlePrev->m_maxEdges[axis]++;
+		}
+		else
+			pHandlePrev->m_minEdges[axis]++;
+
+		pHandleEdge->m_minEdges[axis]--;
+
+		// swap the edges
+		Edge swap = *pEdge;
+		*pEdge = *pPrev;
+		*pPrev = swap;
+
+		// decrement
+		pEdge--;
+		pPrev--;
+	}
+
+#ifdef DEBUG_BROADPHASE
+	debugPrintAxis(axis);
+#endif //DEBUG_BROADPHASE
+
+}
+
+// sorting a min edge upwards can only ever *remove* overlaps
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::sortMinUp(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps)
+{
+	Edge* pEdge = m_pEdges[axis] + edge;
+	Edge* pNext = pEdge + 1;
+	Handle* pHandleEdge = getHandle(pEdge->m_handle);
+
+	while (pNext->m_handle && (pEdge->m_pos >= pNext->m_pos))
+	{
+		Handle* pHandleNext = getHandle(pNext->m_handle);
+
+		if (pNext->IsMax())
+		{
+
+			// if next edge is maximum remove any overlap between the two handles
+			if (updateOverlaps)
+			{
+				Handle* handle0 = getHandle(pEdge->m_handle);
+				Handle* handle1 = getHandle(pNext->m_handle);
+
+				m_pairCache->removeOverlappingPair(handle0,handle1,dispatcher);	
+				if (m_userPairCallback)
+					m_userPairCallback->removeOverlappingPair(handle0,handle1,dispatcher);
+				
+			}
+
+
+			// update edge reference in other handle
+			pHandleNext->m_maxEdges[axis]--;
+		}
+		else
+			pHandleNext->m_minEdges[axis]--;
+
+		pHandleEdge->m_minEdges[axis]++;
+
+		// swap the edges
+		Edge swap = *pEdge;
+		*pEdge = *pNext;
+		*pNext = swap;
+
+		// increment
+		pEdge++;
+		pNext++;
+	}
+
+
+}
+
+// sorting a max edge downwards can only ever *remove* overlaps
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::sortMaxDown(int axis, BP_FP_INT_TYPE edge, btDispatcher* dispatcher, bool updateOverlaps)
+{
+
+	Edge* pEdge = m_pEdges[axis] + edge;
+	Edge* pPrev = pEdge - 1;
+	Handle* pHandleEdge = getHandle(pEdge->m_handle);
+
+	while (pEdge->m_pos < pPrev->m_pos)
+	{
+		Handle* pHandlePrev = getHandle(pPrev->m_handle);
+
+		if (!pPrev->IsMax())
+		{
+			// if previous edge was a minimum remove any overlap between the two handles
+			if (updateOverlaps)
+			{
+				//this is done during the overlappingpairarray iteration/narrowphase collision
+
+				Handle* handle0 = getHandle(pEdge->m_handle);
+				Handle* handle1 = getHandle(pPrev->m_handle);
+				m_pairCache->removeOverlappingPair(handle0,handle1,dispatcher);
+				if (m_userPairCallback)
+					m_userPairCallback->removeOverlappingPair(handle0,handle1,dispatcher);
+			
+
+
+			}
+
+			// update edge reference in other handle
+			pHandlePrev->m_minEdges[axis]++;;
+		}
+		else
+			pHandlePrev->m_maxEdges[axis]++;
+
+		pHandleEdge->m_maxEdges[axis]--;
+
+		// swap the edges
+		Edge swap = *pEdge;
+		*pEdge = *pPrev;
+		*pPrev = swap;
+
+		// decrement
+		pEdge--;
+		pPrev--;
+	}
+
+	
+#ifdef DEBUG_BROADPHASE
+	debugPrintAxis(axis);
+#endif //DEBUG_BROADPHASE
+
+}
+
+// sorting a max edge upwards can only ever *add* overlaps
+template <typename BP_FP_INT_TYPE>
+void btAxisSweep3Internal<BP_FP_INT_TYPE>::sortMaxUp(int axis, BP_FP_INT_TYPE edge, btDispatcher* /* dispatcher */, bool updateOverlaps)
+{
+	Edge* pEdge = m_pEdges[axis] + edge;
+	Edge* pNext = pEdge + 1;
+	Handle* pHandleEdge = getHandle(pEdge->m_handle);
+
+	while (pNext->m_handle && (pEdge->m_pos >= pNext->m_pos))
+	{
+		Handle* pHandleNext = getHandle(pNext->m_handle);
+
+		if (!pNext->IsMax())
+		{
+			// if next edge is a minimum check the bounds and add an overlap if necessary
+			if (updateOverlaps && testOverlap(axis, pHandleEdge, pHandleNext))
+			{
+				Handle* handle0 = getHandle(pEdge->m_handle);
+				Handle* handle1 = getHandle(pNext->m_handle);
+				m_pairCache->addOverlappingPair(handle0,handle1);
+				if (m_userPairCallback)
+					m_userPairCallback->addOverlappingPair(handle0,handle1);
+			}
+
+			// update edge reference in other handle
+			pHandleNext->m_minEdges[axis]--;
+		}
+		else
+			pHandleNext->m_maxEdges[axis]--;
+
+		pHandleEdge->m_maxEdges[axis]++;
+
+		// swap the edges
+		Edge swap = *pEdge;
+		*pEdge = *pNext;
+		*pNext = swap;
+
+		// increment
+		pEdge++;
+		pNext++;
+	}
+	
+}
+
+
+
+////////////////////////////////////////////////////////////////////
+
+
+/// The btAxisSweep3 is an efficient implementation of the 3d axis sweep and prune broadphase.
+/// It uses arrays rather then lists for storage of the 3 axis. Also it operates using 16 bit integer coordinates instead of floats.
+/// For large worlds and many objects, use bt32BitAxisSweep3 or btDbvtBroadphase instead. bt32BitAxisSweep3 has higher precision and allows more then 16384 objects at the cost of more memory and bit of performance.
+class btAxisSweep3 : public btAxisSweep3Internal<unsigned short int>
+{
+public:
+
+	btAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, unsigned short int maxHandles = 16384, btOverlappingPairCache* pairCache = 0);
+
+};
+
+/// The bt32BitAxisSweep3 allows higher precision quantization and more objects compared to the btAxisSweep3 sweep and prune.
+/// This comes at the cost of more memory per handle, and a bit slower performance.
+/// It uses arrays rather then lists for storage of the 3 axis.
+class bt32BitAxisSweep3 : public btAxisSweep3Internal<unsigned int>
+{
+public:
+
+	bt32BitAxisSweep3(const btPoint3& worldAabbMin,const btPoint3& worldAabbMax, unsigned int maxHandles = 1500000, btOverlappingPairCache* pairCache = 0);
 
 };
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btBroadphaseInterface.h

@@ -20,20 +20,34 @@ subject to the following restrictions:
 
 struct btDispatcherInfo;
 class btDispatcher;
-struct btBroadphaseProxy;
+#include "btBroadphaseProxy.h"
-#include "../../LinearMath/btVector3.h"
+class btOverlappingPairCache;
 
-///BroadphaseInterface for aabb-overlapping object pairs
+#include "LinearMath/btVector3.h"
+
+///The btBroadphaseInterface class provides an interface to detect aabb-overlapping object pairs.
+///Some implementations for this broadphase interface include btAxisSweep3, bt32BitAxisSweep3 and btDbvtBroadphase.
+///The actual overlapping pair management, storage, adding and removing of pairs is dealt by the btOverlappingPairCache class.
 class btBroadphaseInterface
 {
 public:
 	virtual ~btBroadphaseInterface() {}
 
-	virtual btBroadphaseProxy*	createProxy(  const btVector3& min,  const btVector3& max,int shapeType,void* userPtr, short int collisionFilterGroup,short int collisionFilterMask) =0;
+	virtual btBroadphaseProxy*	createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr, short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* multiSapProxy) =0;
-	virtual void	destroyProxy(btBroadphaseProxy* proxy)=0;
+	virtual void	destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher)=0;
-	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax)=0;
+	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax, btDispatcher* dispatcher)=0;
-	virtual void	cleanProxyFromPairs(btBroadphaseProxy* proxy)=0;
 	
+	///calculateOverlappingPairs is optional: incremental algorithms (sweep and prune) might do it during the set aabb
+	virtual void	calculateOverlappingPairs(btDispatcher* dispatcher)=0;
+
+	virtual	btOverlappingPairCache*	getOverlappingPairCache()=0;
+	virtual	const btOverlappingPairCache*	getOverlappingPairCache() const =0;
+
+	///getAabb returns the axis aligned bounding box in the 'global' coordinate frame
+	///will add some transform later
+	virtual void getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const =0;
+
+	virtual void	printStats() = 0;
 
 };
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btBroadphaseProxy.h

@@ -16,7 +16,8 @@ subject to the following restrictions:
 #ifndef BROADPHASE_PROXY_H
 #define BROADPHASE_PROXY_H
 
-#include "../../LinearMath/btScalar.h" //for SIMD_FORCE_INLINE
+#include "LinearMath/btScalar.h" //for SIMD_FORCE_INLINE
+#include "LinearMath/btAlignedAllocator.h"
 
 
 /// btDispatcher uses these types
@@ -38,6 +39,7 @@ IMPLICIT_CONVEX_SHAPES_START_HERE,
 	CONE_SHAPE_PROXYTYPE,
 	CONVEX_SHAPE_PROXYTYPE,
 	CYLINDER_SHAPE_PROXYTYPE,
+	UNIFORM_SCALING_SHAPE_PROXYTYPE,
 	MINKOWSKI_SUM_SHAPE_PROXYTYPE,
 	MINKOWSKI_DIFFERENCE_SHAPE_PROXYTYPE,
 //concave shapes
@@ -50,6 +52,8 @@ CONCAVE_SHAPES_START_HERE,
 	TERRAIN_SHAPE_PROXYTYPE,
 ///Used for GIMPACT Trimesh integration
 	GIMPACT_SHAPE_PROXYTYPE,
+///Multimaterial mesh
+    MULTIMATERIAL_TRIANGLE_MESH_PROXYTYPE,
 	
 	EMPTY_SHAPE_PROXYTYPE,
 	STATIC_PLANE_PROXYTYPE,
@@ -57,13 +61,18 @@ CONCAVE_SHAPES_END_HERE,
 
 	COMPOUND_SHAPE_PROXYTYPE,
 
+	SOFTBODY_SHAPE_PROXYTYPE,
+
 	MAX_BROADPHASE_COLLISION_TYPES
 };
 
 
-///btBroadphaseProxy
+///The btBroadphaseProxy is the main class that can be used with the Bullet broadphases. 
-struct btBroadphaseProxy
+///It stores collision shape type information, collision filter information and a client object, typically a btCollisionObject or btRigidBody.
+ATTRIBUTE_ALIGNED16(struct) btBroadphaseProxy
 {
+
+BT_DECLARE_ALIGNED_ALLOCATOR();
 	
 	///optional filtering to cull potential collisions
 	enum CollisionFilterGroups
@@ -73,44 +82,60 @@ struct btBroadphaseProxy
 	        KinematicFilter = 4,
 	        DebrisFilter = 8,
 			SensorTrigger = 16,
-	        AllFilter = DefaultFilter | StaticFilter | KinematicFilter | DebrisFilter | SensorTrigger
+	        AllFilter = -1 //all bits sets: DefaultFilter | StaticFilter | KinematicFilter | DebrisFilter | SensorTrigger
 	};
 
 	//Usually the client btCollisionObject or Rigidbody class
 	void*	m_clientObject;
+
 	short int m_collisionFilterGroup;
 	short int m_collisionFilterMask;
 
-	//used for memory pools
+	void*	m_multiSapParentProxy;		
-	btBroadphaseProxy() :m_clientObject(0){}
 
-	btBroadphaseProxy(void* userPtr,short int collisionFilterGroup, short int collisionFilterMask)
+
+	int			m_uniqueId;//m_uniqueId is introduced for paircache. could get rid of this, by calculating the address offset etc.
+
+	SIMD_FORCE_INLINE int getUid() const
+	{
+		return m_uniqueId;
+	}
+
+	//used for memory pools
+	btBroadphaseProxy() :m_clientObject(0),m_multiSapParentProxy(0)
+	{
+	}
+
+	btBroadphaseProxy(void* userPtr,short int collisionFilterGroup, short int collisionFilterMask,void* multiSapParentProxy=0)
 		:m_clientObject(userPtr),
 		m_collisionFilterGroup(collisionFilterGroup),
 		m_collisionFilterMask(collisionFilterMask)
 	{
+		m_multiSapParentProxy = multiSapParentProxy;
 	}
 
-	static inline bool isPolyhedral(int proxyType)
+	
+
+	static SIMD_FORCE_INLINE bool isPolyhedral(int proxyType)
 	{
 		return (proxyType  < IMPLICIT_CONVEX_SHAPES_START_HERE);
 	}
 
-	static inline bool	isConvex(int proxyType)
+	static SIMD_FORCE_INLINE bool	isConvex(int proxyType)
 	{
 		return (proxyType < CONCAVE_SHAPES_START_HERE);
 	}
 
-	static inline bool	isConcave(int proxyType)
+	static SIMD_FORCE_INLINE bool	isConcave(int proxyType)
 	{
 		return ((proxyType > CONCAVE_SHAPES_START_HERE) &&
 			(proxyType < CONCAVE_SHAPES_END_HERE));
 	}
-	static inline bool	isCompound(int proxyType)
+	static SIMD_FORCE_INLINE bool	isCompound(int proxyType)
 	{
 		return (proxyType == COMPOUND_SHAPE_PROXYTYPE);
 	}
-	static inline bool isInfinite(int proxyType)
+	static SIMD_FORCE_INLINE bool isInfinite(int proxyType)
 	{
 		return (proxyType == STATIC_PLANE_PROXYTYPE);
 	}
@@ -124,8 +149,9 @@ struct btBroadphaseProxy;
 
 
 
-/// contains a pair of aabb-overlapping objects
+///The btBroadphasePair class contains a pair of aabb-overlapping objects.
-struct btBroadphasePair
+///A btDispatcher can search a btCollisionAlgorithm that performs exact/narrowphase collision detection on the actual collision shapes.
+ATTRIBUTE_ALIGNED16(struct) btBroadphasePair
 {
 	btBroadphasePair ()
 		:
@@ -136,6 +162,8 @@ struct btBroadphasePair
 	{
 	}
 
+BT_DECLARE_ALIGNED_ALLOCATOR();
+
 	btBroadphasePair(const btBroadphasePair& other)
 		:		m_pProxy0(other.m_pProxy0),
 				m_pProxy1(other.m_pProxy1),
@@ -181,6 +209,7 @@ SIMD_FORCE_INLINE bool operator<(const btBroadphasePair& a, const btBroadphasePa
 */
 
 
+
 class btBroadphasePairSortPredicate
 {
 	public:

extern/bullet2/src/BulletCollision/BroadphaseCollision/btCollisionAlgorithm.cpp

@@ -18,6 +18,6 @@ subject to the following restrictions:
 
 btCollisionAlgorithm::btCollisionAlgorithm(const btCollisionAlgorithmConstructionInfo& ci)
 {
-	m_dispatcher = ci.m_dispatcher;
+	m_dispatcher = ci.m_dispatcher1;
 }
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h

@@ -16,7 +16,8 @@ subject to the following restrictions:
 #ifndef COLLISION_ALGORITHM_H
 #define COLLISION_ALGORITHM_H
 
-#include "../../LinearMath/btScalar.h"
+#include "LinearMath/btScalar.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
 struct btBroadphaseProxy;
 class btDispatcher;
@@ -25,21 +26,22 @@ class btCollisionObject;
 struct btDispatcherInfo;
 class	btPersistentManifold;
 
+typedef btAlignedObjectArray<btPersistentManifold*>	btManifoldArray;
 
 struct btCollisionAlgorithmConstructionInfo
 {
 	btCollisionAlgorithmConstructionInfo()
-		:m_dispatcher(0),
+		:m_dispatcher1(0),
 		m_manifold(0)
 	{
 	}
 	btCollisionAlgorithmConstructionInfo(btDispatcher* dispatcher,int temp)
-		:m_dispatcher(dispatcher)
+		:m_dispatcher1(dispatcher)
 	{
 		(void)temp;
 	}
 
-	btDispatcher*	m_dispatcher;
+	btDispatcher*	m_dispatcher1;
 	btPersistentManifold*	m_manifold;
 
 	int	getDispatcherId();
@@ -71,6 +73,7 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut) = 0;
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray) = 0;
 };
 
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btDbvtBroadphase.cpp

@@ -0,0 +1,344 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2007 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+///btDbvtBroadphase implementation by Nathanael Presson
+
+#include "btDbvtBroadphase.h"
+
+//
+// Profiling
+//
+
+#if DBVT_BP_PROFILE
+#include <stdio.h>
+struct	ProfileScope
+	{
+	ProfileScope(btClock& clock,unsigned long& value)
+		{
+		m_clock=&clock;
+		m_value=&value;
+		m_base=clock.getTimeMicroseconds();
+		}
+	~ProfileScope()
+		{
+		(*m_value)+=m_clock->getTimeMicroseconds()-m_base;
+		}
+	btClock*		m_clock;
+	unsigned long*	m_value;
+	unsigned long	m_base;
+	};
+#define	SPC(_value_)	ProfileScope	spc_scope(m_clock,_value_)
+#else
+#define	SPC(_value_)
+#endif
+
+//
+// Helpers
+//
+
+//
+template <typename T>
+static inline void	listappend(T* item,T*& list)
+{
+item->links[0]=0;
+item->links[1]=list;
+if(list) list->links[0]=item;
+list=item;
+}
+
+//
+template <typename T>
+static inline void	listremove(T* item,T*& list)
+{
+if(item->links[0]) item->links[0]->links[1]=item->links[1]; else list=item->links[1];
+if(item->links[1]) item->links[1]->links[0]=item->links[0];
+}
+
+//
+template <typename T>
+static inline int	listcount(T* root)
+{
+int	n=0;
+while(root) { ++n;root=root->links[1]; }
+return(n);
+}
+
+//
+template <typename T>
+static inline void	clear(T& value)
+{
+static const struct ZeroDummy : T {} zerodummy;
+value=zerodummy;
+}
+
+//
+// Colliders
+//
+
+/* Tree collider	*/ 
+struct	btDbvtTreeCollider : btDbvt::ICollide
+{
+btDbvtBroadphase*	pbp;
+		btDbvtTreeCollider(btDbvtBroadphase* p) : pbp(p) {}
+void	Process(const btDbvtNode* na,const btDbvtNode* nb)
+	{
+	btDbvtProxy*	pa=(btDbvtProxy*)na->data;
+	btDbvtProxy*	pb=(btDbvtProxy*)nb->data;
+	#if DBVT_BP_DISCRETPAIRS
+	if(Intersect(pa->aabb,pb->aabb))
+	#endif
+		{
+		if(pa>pb) btSwap(pa,pb);
+		pbp->m_paircache->addOverlappingPair(pa,pb);
+		}
+	}
+};
+
+//
+// btDbvtBroadphase
+//
+
+//
+btDbvtBroadphase::btDbvtBroadphase(btOverlappingPairCache* paircache)
+{
+m_releasepaircache	=	(paircache!=0)?false:true;
+m_predictedframes	=	2;
+m_stageCurrent		=	0;
+m_fupdates			=	1;
+m_dupdates			=	1;
+m_paircache			=	paircache?
+							paircache	:
+							new(btAlignedAlloc(sizeof(btHashedOverlappingPairCache),16)) btHashedOverlappingPairCache();
+m_gid				=	0;
+m_pid				=	0;
+for(int i=0;i<=STAGECOUNT;++i)
+	{
+	m_stageRoots[i]=0;
+	}
+#if DBVT_BP_PROFILE
+clear(m_profiling);
+#endif
+}
+
+//
+btDbvtBroadphase::~btDbvtBroadphase()
+{
+if(m_releasepaircache) 
+{
+	m_paircache->~btOverlappingPairCache();
+	btAlignedFree(m_paircache);
+}
+}
+
+//
+btBroadphaseProxy*				btDbvtBroadphase::createProxy(	const btVector3& aabbMin,
+																const btVector3& aabbMax,
+																int /*shapeType*/,
+																void* userPtr,
+																short int collisionFilterGroup,
+																short int collisionFilterMask,
+																btDispatcher* /*dispatcher*/,
+																void* /*multiSapProxy*/)
+{
+btDbvtProxy*	proxy=new(btAlignedAlloc(sizeof(btDbvtProxy),16)) btDbvtProxy(	userPtr,
+																				collisionFilterGroup,
+																				collisionFilterMask);
+proxy->aabb			=	btDbvtVolume::FromMM(aabbMin,aabbMax);
+proxy->leaf			=	m_sets[0].insert(proxy->aabb,proxy);
+proxy->stage		=	m_stageCurrent;
+proxy->m_uniqueId	=	++m_gid;
+listappend(proxy,m_stageRoots[m_stageCurrent]);
+return(proxy);
+}
+
+//
+void							btDbvtBroadphase::destroyProxy(	btBroadphaseProxy* absproxy,
+																btDispatcher* dispatcher)
+{
+btDbvtProxy*	proxy=(btDbvtProxy*)absproxy;
+if(proxy->stage==STAGECOUNT)
+	m_sets[1].remove(proxy->leaf);
+	else
+	m_sets[0].remove(proxy->leaf);
+listremove(proxy,m_stageRoots[proxy->stage]);
+m_paircache->removeOverlappingPairsContainingProxy(proxy,dispatcher);
+btAlignedFree(proxy);
+}
+
+//
+void							btDbvtBroadphase::setAabb(		btBroadphaseProxy* absproxy,
+																const btVector3& aabbMin,
+																const btVector3& aabbMax,
+																btDispatcher* /*dispatcher*/)
+{
+btDbvtProxy*	proxy=(btDbvtProxy*)absproxy;
+btDbvtVolume	aabb=btDbvtVolume::FromMM(aabbMin,aabbMax);
+if(NotEqual(aabb,proxy->leaf->volume))
+	{
+	if(proxy->stage==STAGECOUNT)
+		{/* fixed -> dynamic set	*/ 
+		m_sets[1].remove(proxy->leaf);
+		proxy->leaf=m_sets[0].insert(aabb,proxy);
+		}
+		else
+		{/* dynamic set				*/ 
+		if(Intersect(proxy->leaf->volume,aabb))
+			{/* Moving				*/ 
+			const btVector3	delta=(aabbMin+aabbMax)/2-proxy->aabb.Center();
+			#ifdef DBVT_BP_MARGIN
+			m_sets[0].update(proxy->leaf,aabb,delta*m_predictedframes,DBVT_BP_MARGIN);
+			#else
+			m_sets[0].update(proxy->leaf,aabb,delta*m_predictedframes);
+			#endif
+			}
+			else
+			{/* Teleporting			*/ 
+			m_sets[0].update(proxy->leaf,aabb);		
+			}	
+		}
+	listremove(proxy,m_stageRoots[proxy->stage]);
+	proxy->aabb		=	aabb;
+	proxy->stage	=	m_stageCurrent;
+	listappend(proxy,m_stageRoots[m_stageCurrent]);
+	}
+}
+
+//
+void							btDbvtBroadphase::calculateOverlappingPairs(btDispatcher* dispatcher)
+{
+collide(dispatcher);
+#if DBVT_BP_PROFILE
+if(0==(m_pid%DBVT_BP_PROFILING_RATE))
+	{	
+	printf("fixed(%u) dynamics(%u) pairs(%u)\r\n",m_sets[1].m_leaves,m_sets[0].m_leaves,m_paircache->getNumOverlappingPairs());
+	unsigned int	total=m_profiling.m_total;
+	if(total<=0) total=1;
+	printf("ddcollide: %u%% (%uus)\r\n",(50+m_profiling.m_ddcollide*100)/total,m_profiling.m_ddcollide/DBVT_BP_PROFILING_RATE);
+	printf("fdcollide: %u%% (%uus)\r\n",(50+m_profiling.m_fdcollide*100)/total,m_profiling.m_fdcollide/DBVT_BP_PROFILING_RATE);
+	printf("cleanup:   %u%% (%uus)\r\n",(50+m_profiling.m_cleanup*100)/total,m_profiling.m_cleanup/DBVT_BP_PROFILING_RATE);
+	printf("total:     %uus\r\n",total/DBVT_BP_PROFILING_RATE);
+	const unsigned long	sum=m_profiling.m_ddcollide+
+							m_profiling.m_fdcollide+
+							m_profiling.m_cleanup;
+	printf("leaked: %u%% (%uus)\r\n",100-((50+sum*100)/total),(total-sum)/DBVT_BP_PROFILING_RATE);
+	printf("job counts: %u%%\r\n",(m_profiling.m_jobcount*100)/((m_sets[0].m_leaves+m_sets[1].m_leaves)*DBVT_BP_PROFILING_RATE));
+	clear(m_profiling);
+	m_clock.reset();
+	}
+#endif
+}
+
+//
+void							btDbvtBroadphase::collide(btDispatcher* dispatcher)
+{
+SPC(m_profiling.m_total);
+/* optimize				*/ 
+m_sets[0].optimizeIncremental(1+(m_sets[0].m_leaves*m_dupdates)/100);
+m_sets[1].optimizeIncremental(1+(m_sets[1].m_leaves*m_fupdates)/100);
+/* dynamic -> fixed set	*/ 
+m_stageCurrent=(m_stageCurrent+1)%STAGECOUNT;
+btDbvtProxy*	current=m_stageRoots[m_stageCurrent];
+if(current)
+	{
+	btDbvtTreeCollider	collider(this);
+	do	{
+		btDbvtProxy*	next=current->links[1];
+		listremove(current,m_stageRoots[current->stage]);
+		listappend(current,m_stageRoots[STAGECOUNT]);
+		btDbvt::collideTT(m_sets[1].m_root,current->leaf,collider);
+		m_sets[0].remove(current->leaf);
+		current->leaf	=	m_sets[1].insert(current->aabb,current);
+		current->stage	=	STAGECOUNT;	
+		current			=	next;
+		} while(current);
+	}
+/* collide dynamics		*/ 
+	{
+	btDbvtTreeCollider	collider(this);
+		{
+		SPC(m_profiling.m_fdcollide);
+		btDbvt::collideTT(m_sets[0].m_root,m_sets[1].m_root,collider);
+		}
+		{
+		SPC(m_profiling.m_ddcollide);
+		btDbvt::collideTT(m_sets[0].m_root,m_sets[0].m_root,collider);
+		}
+	}
+/* clean up				*/ 
+	{
+	SPC(m_profiling.m_cleanup);
+	btBroadphasePairArray&	pairs=m_paircache->getOverlappingPairArray();
+	if(pairs.size()>0)
+		{
+		for(int i=0,ni=pairs.size();i<ni;++i)
+			{
+			btBroadphasePair&	p=pairs[i];
+			btDbvtProxy*	pa=(btDbvtProxy*)p.m_pProxy0;
+			btDbvtProxy*	pb=(btDbvtProxy*)p.m_pProxy1;
+			if(!Intersect(pa->aabb,pb->aabb))
+				{
+				if(pa>pb) btSwap(pa,pb);
+				m_paircache->removeOverlappingPair(pa,pb,dispatcher);
+				--ni;--i;
+				}
+			}
+		}
+	}
+++m_pid;
+}
+
+//
+void							btDbvtBroadphase::optimize()
+{
+m_sets[0].optimizeTopDown();
+m_sets[1].optimizeTopDown();
+}
+
+//
+btOverlappingPairCache*			btDbvtBroadphase::getOverlappingPairCache()
+{
+return(m_paircache);
+}
+
+//
+const btOverlappingPairCache*	btDbvtBroadphase::getOverlappingPairCache() const
+{
+return(m_paircache);
+}
+
+//
+void							btDbvtBroadphase::getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const
+{
+
+	ATTRIBUTE_ALIGNED16(btDbvtVolume)	bounds;
+
+if(!m_sets[0].empty())
+	if(!m_sets[1].empty())	Merge(	m_sets[0].m_root->volume,
+									m_sets[1].m_root->volume,bounds);
+							else
+							bounds=m_sets[0].m_root->volume;
+else if(!m_sets[1].empty())	bounds=m_sets[1].m_root->volume;
+							else
+							bounds=btDbvtVolume::FromCR(btVector3(0,0,0),0);
+aabbMin=bounds.Mins();
+aabbMax=bounds.Maxs();
+}
+
+//
+void							btDbvtBroadphase::printStats()
+{}
+
+#if DBVT_BP_PROFILE
+#undef	SPC
+#endif

extern/bullet2/src/BulletCollision/BroadphaseCollision/btDbvtBroadphase.h

@@ -0,0 +1,103 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2007 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+///btDbvtBroadphase implementation by Nathanael Presson
+#ifndef BT_DBVT_BROADPHASE_H
+#define BT_DBVT_BROADPHASE_H
+
+#include "BulletCollision/BroadphaseCollision/btDbvt.h"
+#include "BulletCollision/BroadphaseCollision/btOverlappingPairCache.h"
+
+//
+// Compile time config
+//
+
+#define	DBVT_BP_PROFILE			0
+#define DBVT_BP_DISCRETPAIRS	1
+#define DBVT_BP_MARGIN			(btScalar)0.05
+
+#if DBVT_BP_PROFILE
+	#define	DBVT_BP_PROFILING_RATE	256
+	#include "LinearMath/btQuickprof.h"
+#endif
+
+//
+// btDbvtProxy
+//
+struct btDbvtProxy : btBroadphaseProxy
+{
+/* Fields		*/ 
+btDbvtAabbMm		aabb;
+btDbvtNode*		leaf;
+btDbvtProxy*		links[2];
+int					stage;
+/* ctor			*/ 
+btDbvtProxy(void* userPtr,short int collisionFilterGroup, short int collisionFilterMask) :
+	btBroadphaseProxy(userPtr,collisionFilterGroup,collisionFilterMask)
+	{
+	links[0]=links[1]=0;
+	}
+};
+
+typedef btAlignedObjectArray<btDbvtProxy*>	btDbvtProxyArray;
+
+///The btDbvtBroadphase implements a broadphase using two dynamic AABB bounding volume hierarchies/trees (see btDbvt).
+///One tree is used for static/non-moving objects, and another tree is used for dynamic objects. Objects can move from one tree to the other.
+///This is a very fast broadphase, especially for very dynamic worlds where many objects are moving. Its insert/add and remove of objects is generally faster than the sweep and prune broadphases btAxisSweep3 and bt32BitAxisSweep3.
+struct	btDbvtBroadphase : btBroadphaseInterface
+{
+/* Config		*/ 
+enum	{
+		DYNAMIC_SET			=	0,	/* Dynamic set index	*/ 
+		FIXED_SET			=	1,	/* Fixed set index		*/ 
+		STAGECOUNT			=	2	/* Number of stages		*/ 
+		};
+/* Fields		*/ 
+btDbvt					m_sets[2];					// Dbvt sets
+btDbvtProxy*			m_stageRoots[STAGECOUNT+1];	// Stages list
+btOverlappingPairCache*	m_paircache;				// Pair cache
+btScalar				m_predictedframes;			// Frames predicted
+int						m_stageCurrent;				// Current stage
+int						m_fupdates;					// % of fixed updates per frame
+int						m_dupdates;					// % of dynamic updates per frame
+int						m_pid;						// Parse id
+int						m_gid;						// Gen id
+bool					m_releasepaircache;			// Release pair cache on delete
+#if DBVT_BP_PROFILE
+btClock					m_clock;
+struct	{
+		unsigned long		m_total;
+		unsigned long		m_ddcollide;
+		unsigned long		m_fdcollide;
+		unsigned long		m_cleanup;
+		unsigned long		m_jobcount;
+		}				m_profiling;
+#endif
+/* Methods		*/ 
+btDbvtBroadphase(btOverlappingPairCache* paircache=0);
+~btDbvtBroadphase();
+void							collide(btDispatcher* dispatcher);
+void							optimize();
+/* btBroadphaseInterface Implementation	*/ 
+btBroadphaseProxy*				createProxy(const btVector3& aabbMin,const btVector3& aabbMax,int shapeType,void* userPtr,short int collisionFilterGroup,short int collisionFilterMask,btDispatcher* dispatcher,void* multiSapProxy);
+void							destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
+void							setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax,btDispatcher* dispatcher);
+void							calculateOverlappingPairs(btDispatcher* dispatcher);
+btOverlappingPairCache*			getOverlappingPairCache();
+const btOverlappingPairCache*	getOverlappingPairCache() const;
+void							getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const;
+void							printStats();
+};
+
+#endif

extern/bullet2/src/BulletCollision/BroadphaseCollision/btDispatcher.h

@@ -16,7 +16,7 @@ subject to the following restrictions:
 #ifndef _DISPATCHER_H
 #define _DISPATCHER_H
 
-#include "../../LinearMath/btScalar.h"
+#include "LinearMath/btScalar.h"
 
 class btCollisionAlgorithm;
 struct btBroadphaseProxy;
@@ -43,7 +43,9 @@ struct btDispatcherInfo
 		m_useContinuous(false),
 		m_debugDraw(0),
 		m_enableSatConvex(false),
-		m_enableSPU(false),
+		m_enableSPU(true),
+		m_useEpa(true),
+		m_allowedCcdPenetration(btScalar(0.04)),
 		m_stackAllocator(0)
 	{
 
@@ -51,17 +53,19 @@ struct btDispatcherInfo
 	btScalar	m_timeStep;
 	int		m_stepCount;
 	int		m_dispatchFunc;
-	btScalar	m_timeOfImpact;
+	mutable btScalar	m_timeOfImpact;
 	bool	m_useContinuous;
 	class btIDebugDraw*	m_debugDraw;
 	bool	m_enableSatConvex;
 	bool	m_enableSPU;
+	bool	m_useEpa;
+	btScalar	m_allowedCcdPenetration;
 	btStackAlloc*	m_stackAllocator;
 	
 };
 
-/// btDispatcher can be used in combination with broadphase to dispatch overlapping pairs.
+///The btDispatcher interface class can be used in combination with broadphase to dispatch calculations for overlapping pairs.
-/// For example for pairwise collision detection or user callbacks (game logic).
+///For example for pairwise collision detection, calculating contact points stored in btPersistentManifold or user callbacks (game logic).
 class btDispatcher
 {
 
@@ -81,12 +85,18 @@ public:
 
 	virtual bool	needsResponse(btCollisionObject* body0,btCollisionObject* body1)=0;
 
-	virtual void	dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,btDispatcherInfo& dispatchInfo)=0;
+	virtual void	dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,const btDispatcherInfo& dispatchInfo,btDispatcher* dispatcher)  =0;
 
 	virtual int getNumManifolds() const = 0;
 
 	virtual btPersistentManifold* getManifoldByIndexInternal(int index) = 0;
 
+	virtual	btPersistentManifold**	getInternalManifoldPointer() = 0;
+
+	virtual	void* allocateCollisionAlgorithm(int size)  = 0;
+
+	virtual	void freeCollisionAlgorithm(void* ptr) = 0;
+
 };
 
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btMultiSapBroadphase.cpp

@@ -0,0 +1,466 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "btMultiSapBroadphase.h"
+
+#include "btSimpleBroadphase.h"
+#include "LinearMath/btAabbUtil2.h"
+#include "btQuantizedBvh.h"
+
+///	btSapBroadphaseArray	m_sapBroadphases;
+
+///	btOverlappingPairCache*	m_overlappingPairs;
+extern int gOverlappingPairs;
+
+/*
+class btMultiSapSortedOverlappingPairCache : public btSortedOverlappingPairCache
+{
+public:
+
+	virtual btBroadphasePair*	addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+	{
+		return btSortedOverlappingPairCache::addOverlappingPair((btBroadphaseProxy*)proxy0->m_multiSapParentProxy,(btBroadphaseProxy*)proxy1->m_multiSapParentProxy);
+	}
+};
+
+*/
+
+btMultiSapBroadphase::btMultiSapBroadphase(int /*maxProxies*/,btOverlappingPairCache* pairCache)
+:m_overlappingPairs(pairCache),
+m_optimizedAabbTree(0),
+m_ownsPairCache(false),
+m_invalidPair(0)
+{
+	if (!m_overlappingPairs)
+	{
+		m_ownsPairCache = true;
+		void* mem = btAlignedAlloc(sizeof(btSortedOverlappingPairCache),16);
+		m_overlappingPairs = new (mem)btSortedOverlappingPairCache();
+	}
+
+	struct btMultiSapOverlapFilterCallback : public btOverlapFilterCallback
+	{
+		virtual ~btMultiSapOverlapFilterCallback()
+		{}
+		// return true when pairs need collision
+		virtual bool	needBroadphaseCollision(btBroadphaseProxy* childProxy0,btBroadphaseProxy* childProxy1) const
+		{
+			btBroadphaseProxy* multiProxy0 = (btBroadphaseProxy*)childProxy0->m_multiSapParentProxy;
+			btBroadphaseProxy* multiProxy1 = (btBroadphaseProxy*)childProxy1->m_multiSapParentProxy;
+			
+			bool collides = (multiProxy0->m_collisionFilterGroup & multiProxy1->m_collisionFilterMask) != 0;
+			collides = collides && (multiProxy1->m_collisionFilterGroup & multiProxy0->m_collisionFilterMask);
+	
+			return collides;
+		}
+	};
+
+	void* mem = btAlignedAlloc(sizeof(btMultiSapOverlapFilterCallback),16);
+	m_filterCallback = new (mem)btMultiSapOverlapFilterCallback();
+
+	m_overlappingPairs->setOverlapFilterCallback(m_filterCallback);
+//	mem = btAlignedAlloc(sizeof(btSimpleBroadphase),16);
+//	m_simpleBroadphase = new (mem) btSimpleBroadphase(maxProxies,m_overlappingPairs);
+}
+
+btMultiSapBroadphase::~btMultiSapBroadphase()
+{
+	if (m_ownsPairCache)
+	{
+		m_overlappingPairs->~btOverlappingPairCache();
+		btAlignedFree(m_overlappingPairs);
+	}
+}
+
+
+void	btMultiSapBroadphase::buildTree(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax)
+{
+	m_optimizedAabbTree = new btQuantizedBvh();
+	m_optimizedAabbTree->setQuantizationValues(bvhAabbMin,bvhAabbMax);
+	QuantizedNodeArray&	nodes = m_optimizedAabbTree->getLeafNodeArray();
+	for (int i=0;i<m_sapBroadphases.size();i++)
+	{
+		btQuantizedBvhNode node;
+		btVector3 aabbMin,aabbMax;
+		m_sapBroadphases[i]->getBroadphaseAabb(aabbMin,aabbMax);
+		m_optimizedAabbTree->quantize(&node.m_quantizedAabbMin[0],aabbMin,0);
+		m_optimizedAabbTree->quantize(&node.m_quantizedAabbMax[0],aabbMax,1);
+		int partId = 0;
+		node.m_escapeIndexOrTriangleIndex = (partId<<(31-MAX_NUM_PARTS_IN_BITS)) | i;
+		nodes.push_back(node);
+	}
+	m_optimizedAabbTree->buildInternal();
+}
+
+btBroadphaseProxy*	btMultiSapBroadphase::createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr, short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* /*ignoreMe*/)
+{
+	//void* ignoreMe -> we could think of recursive multi-sap, if someone is interested
+
+	void* mem = btAlignedAlloc(sizeof(btMultiSapProxy),16);
+	btMultiSapProxy* proxy = new (mem)btMultiSapProxy(aabbMin,  aabbMax,shapeType,userPtr, collisionFilterGroup,collisionFilterMask);
+	m_multiSapProxies.push_back(proxy);
+
+	///this should deal with inserting/removal into child broadphases
+	setAabb(proxy,aabbMin,aabbMax,dispatcher);
+	return proxy;
+}
+
+void	btMultiSapBroadphase::destroyProxy(btBroadphaseProxy* /*proxy*/,btDispatcher* /*dispatcher*/)
+{
+	///not yet
+	btAssert(0);
+
+}
+
+
+void	btMultiSapBroadphase::addToChildBroadphase(btMultiSapProxy* parentMultiSapProxy, btBroadphaseProxy* childProxy, btBroadphaseInterface*	childBroadphase)
+{
+	void* mem = btAlignedAlloc(sizeof(btBridgeProxy),16);
+	btBridgeProxy* bridgeProxyRef = new(mem) btBridgeProxy;
+	bridgeProxyRef->m_childProxy = childProxy;
+	bridgeProxyRef->m_childBroadphase = childBroadphase;
+	parentMultiSapProxy->m_bridgeProxies.push_back(bridgeProxyRef);
+}
+
+
+bool boxIsContainedWithinBox(const btVector3& amin,const btVector3& amax,const btVector3& bmin,const btVector3& bmax);
+bool boxIsContainedWithinBox(const btVector3& amin,const btVector3& amax,const btVector3& bmin,const btVector3& bmax)
+{
+return
+amin.getX() >= bmin.getX() && amax.getX() <= bmax.getX() &&
+amin.getY() >= bmin.getY() && amax.getY() <= bmax.getY() &&
+amin.getZ() >= bmin.getZ() && amax.getZ() <= bmax.getZ();
+}
+
+
+
+
+
+
+//#include <stdio.h>
+
+void	btMultiSapBroadphase::setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax, btDispatcher* dispatcher)
+{
+	btMultiSapProxy* multiProxy = static_cast<btMultiSapProxy*>(proxy);
+	multiProxy->m_aabbMin = aabbMin;
+	multiProxy->m_aabbMax = aabbMax;
+	
+	
+//	bool fullyContained = false;
+//	bool alreadyInSimple = false;
+	
+
+
+	
+	struct MyNodeOverlapCallback : public btNodeOverlapCallback
+	{
+		btMultiSapBroadphase*	m_multiSap;
+		btMultiSapProxy*		m_multiProxy;
+		btDispatcher*			m_dispatcher;
+
+		MyNodeOverlapCallback(btMultiSapBroadphase* multiSap,btMultiSapProxy* multiProxy,btDispatcher* dispatcher)
+			:m_multiSap(multiSap),
+			m_multiProxy(multiProxy),
+			m_dispatcher(dispatcher)
+		{
+
+		}
+
+		virtual void processNode(int /*nodeSubPart*/, int broadphaseIndex)
+		{
+			btBroadphaseInterface* childBroadphase = m_multiSap->getBroadphaseArray()[broadphaseIndex];
+
+			int containingBroadphaseIndex = -1;
+			//already found?
+			for (int i=0;i<m_multiProxy->m_bridgeProxies.size();i++)
+			{
+
+				if (m_multiProxy->m_bridgeProxies[i]->m_childBroadphase == childBroadphase)
+				{
+					containingBroadphaseIndex = i;
+					break;
+				}
+			}
+			if (containingBroadphaseIndex<0)
+			{
+				//add it
+				btBroadphaseProxy* childProxy = childBroadphase->createProxy(m_multiProxy->m_aabbMin,m_multiProxy->m_aabbMax,m_multiProxy->m_shapeType,m_multiProxy->m_clientObject,m_multiProxy->m_collisionFilterGroup,m_multiProxy->m_collisionFilterMask, m_dispatcher,m_multiProxy);
+				m_multiSap->addToChildBroadphase(m_multiProxy,childProxy,childBroadphase);
+
+			}
+		}
+	};
+
+	MyNodeOverlapCallback	myNodeCallback(this,multiProxy,dispatcher);
+
+
+
+	
+	m_optimizedAabbTree->reportAabbOverlappingNodex(&myNodeCallback,aabbMin,aabbMax);
+	int i;
+
+	for ( i=0;i<multiProxy->m_bridgeProxies.size();i++)
+	{
+		btVector3 worldAabbMin,worldAabbMax;
+		multiProxy->m_bridgeProxies[i]->m_childBroadphase->getBroadphaseAabb(worldAabbMin,worldAabbMax);
+		bool overlapsBroadphase = TestAabbAgainstAabb2(worldAabbMin,worldAabbMax,multiProxy->m_aabbMin,multiProxy->m_aabbMax);
+		if (!overlapsBroadphase)
+		{
+			//remove it now
+			btBridgeProxy* bridgeProxy = multiProxy->m_bridgeProxies[i];
+
+			btBroadphaseProxy* childProxy = bridgeProxy->m_childProxy;
+			bridgeProxy->m_childBroadphase->destroyProxy(childProxy,dispatcher);
+			
+			multiProxy->m_bridgeProxies.swap( i,multiProxy->m_bridgeProxies.size()-1);
+			multiProxy->m_bridgeProxies.pop_back();
+
+		}
+	}
+
+
+	/*
+
+	if (1)
+	{
+
+		//find broadphase that contain this multiProxy
+		int numChildBroadphases = getBroadphaseArray().size();
+		for (int i=0;i<numChildBroadphases;i++)
+		{
+			btBroadphaseInterface* childBroadphase = getBroadphaseArray()[i];
+			btVector3 worldAabbMin,worldAabbMax;
+			childBroadphase->getBroadphaseAabb(worldAabbMin,worldAabbMax);
+			bool overlapsBroadphase = TestAabbAgainstAabb2(worldAabbMin,worldAabbMax,multiProxy->m_aabbMin,multiProxy->m_aabbMax);
+			
+		//	fullyContained = fullyContained || boxIsContainedWithinBox(worldAabbMin,worldAabbMax,multiProxy->m_aabbMin,multiProxy->m_aabbMax);
+			int containingBroadphaseIndex = -1;
+			
+			//if already contains this
+			
+			for (int i=0;i<multiProxy->m_bridgeProxies.size();i++)
+			{
+				if (multiProxy->m_bridgeProxies[i]->m_childBroadphase == childBroadphase)
+				{
+					containingBroadphaseIndex = i;
+				}
+				alreadyInSimple = alreadyInSimple || (multiProxy->m_bridgeProxies[i]->m_childBroadphase == m_simpleBroadphase);
+			}
+
+			if (overlapsBroadphase)
+			{
+				if (containingBroadphaseIndex<0)
+				{
+					btBroadphaseProxy* childProxy = childBroadphase->createProxy(aabbMin,aabbMax,multiProxy->m_shapeType,multiProxy->m_clientObject,multiProxy->m_collisionFilterGroup,multiProxy->m_collisionFilterMask, dispatcher);
+					childProxy->m_multiSapParentProxy = multiProxy;
+					addToChildBroadphase(multiProxy,childProxy,childBroadphase);
+				}
+			} else
+			{
+				if (containingBroadphaseIndex>=0)
+				{
+					//remove
+					btBridgeProxy* bridgeProxy = multiProxy->m_bridgeProxies[containingBroadphaseIndex];
+
+					btBroadphaseProxy* childProxy = bridgeProxy->m_childProxy;
+					bridgeProxy->m_childBroadphase->destroyProxy(childProxy,dispatcher);
+					
+					multiProxy->m_bridgeProxies.swap( containingBroadphaseIndex,multiProxy->m_bridgeProxies.size()-1);
+					multiProxy->m_bridgeProxies.pop_back();
+				}
+			}
+		}
+
+
+		///If we are in no other child broadphase, stick the proxy in the global 'simple' broadphase (brute force)
+		///hopefully we don't end up with many entries here (can assert/provide feedback on stats)
+		if (0)//!multiProxy->m_bridgeProxies.size())
+		{
+			///we don't pass the userPtr but our multisap proxy. We need to patch this, before processing an actual collision
+			///this is needed to be able to calculate the aabb overlap
+			btBroadphaseProxy* childProxy = m_simpleBroadphase->createProxy(aabbMin,aabbMax,multiProxy->m_shapeType,multiProxy->m_clientObject,multiProxy->m_collisionFilterGroup,multiProxy->m_collisionFilterMask, dispatcher);
+			childProxy->m_multiSapParentProxy = multiProxy;
+			addToChildBroadphase(multiProxy,childProxy,m_simpleBroadphase);
+		}
+	}
+
+	if (!multiProxy->m_bridgeProxies.size())
+	{
+		///we don't pass the userPtr but our multisap proxy. We need to patch this, before processing an actual collision
+		///this is needed to be able to calculate the aabb overlap
+		btBroadphaseProxy* childProxy = m_simpleBroadphase->createProxy(aabbMin,aabbMax,multiProxy->m_shapeType,multiProxy->m_clientObject,multiProxy->m_collisionFilterGroup,multiProxy->m_collisionFilterMask, dispatcher);
+		childProxy->m_multiSapParentProxy = multiProxy;
+		addToChildBroadphase(multiProxy,childProxy,m_simpleBroadphase);
+	}
+*/
+
+
+	//update
+	for ( i=0;i<multiProxy->m_bridgeProxies.size();i++)
+	{
+		btBridgeProxy* bridgeProxyRef = multiProxy->m_bridgeProxies[i];
+		bridgeProxyRef->m_childBroadphase->setAabb(bridgeProxyRef->m_childProxy,aabbMin,aabbMax,dispatcher);
+	}
+
+}
+bool stopUpdating=false;
+
+
+
+class btMultiSapBroadphasePairSortPredicate
+{
+	public:
+
+		bool operator() ( const btBroadphasePair& a1, const btBroadphasePair& b1 )
+		{
+				btMultiSapBroadphase::btMultiSapProxy* aProxy0 = a1.m_pProxy0 ? (btMultiSapBroadphase::btMultiSapProxy*)a1.m_pProxy0->m_multiSapParentProxy : 0;
+				btMultiSapBroadphase::btMultiSapProxy* aProxy1 = a1.m_pProxy1 ? (btMultiSapBroadphase::btMultiSapProxy*)a1.m_pProxy1->m_multiSapParentProxy : 0;
+				btMultiSapBroadphase::btMultiSapProxy* bProxy0 = b1.m_pProxy0 ? (btMultiSapBroadphase::btMultiSapProxy*)b1.m_pProxy0->m_multiSapParentProxy : 0;
+				btMultiSapBroadphase::btMultiSapProxy* bProxy1 = b1.m_pProxy1 ? (btMultiSapBroadphase::btMultiSapProxy*)b1.m_pProxy1->m_multiSapParentProxy : 0;
+
+				 return aProxy0 > bProxy0 || 
+					(aProxy0 == bProxy0 && aProxy1 > bProxy1) ||
+					(aProxy0 == bProxy0 && aProxy1 == bProxy1 && a1.m_algorithm > b1.m_algorithm); 
+		}
+};
+
+
+        ///calculateOverlappingPairs is optional: incremental algorithms (sweep and prune) might do it during the set aabb
+void    btMultiSapBroadphase::calculateOverlappingPairs(btDispatcher* dispatcher)
+{
+
+//	m_simpleBroadphase->calculateOverlappingPairs(dispatcher);
+
+	if (!stopUpdating && getOverlappingPairCache()->hasDeferredRemoval())
+	{
+	
+		btBroadphasePairArray&	overlappingPairArray = getOverlappingPairCache()->getOverlappingPairArray();
+
+	//	quicksort(overlappingPairArray,0,overlappingPairArray.size());
+
+		overlappingPairArray.quickSort(btMultiSapBroadphasePairSortPredicate());
+
+		//perform a sort, to find duplicates and to sort 'invalid' pairs to the end
+	//	overlappingPairArray.heapSort(btMultiSapBroadphasePairSortPredicate());
+
+		overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+		m_invalidPair = 0;
+
+		
+		int i;
+
+		btBroadphasePair previousPair;
+		previousPair.m_pProxy0 = 0;
+		previousPair.m_pProxy1 = 0;
+		previousPair.m_algorithm = 0;
+		
+		
+		for (i=0;i<overlappingPairArray.size();i++)
+		{
+		
+			btBroadphasePair& pair = overlappingPairArray[i];
+
+			btMultiSapProxy* aProxy0 = pair.m_pProxy0 ? (btMultiSapProxy*)pair.m_pProxy0->m_multiSapParentProxy : 0;
+			btMultiSapProxy* aProxy1 = pair.m_pProxy1 ? (btMultiSapProxy*)pair.m_pProxy1->m_multiSapParentProxy : 0;
+			btMultiSapProxy* bProxy0 = previousPair.m_pProxy0 ? (btMultiSapProxy*)previousPair.m_pProxy0->m_multiSapParentProxy : 0;
+			btMultiSapProxy* bProxy1 = previousPair.m_pProxy1 ? (btMultiSapProxy*)previousPair.m_pProxy1->m_multiSapParentProxy : 0;
+
+			bool isDuplicate = (aProxy0 == bProxy0) && (aProxy1 == bProxy1);
+			
+			previousPair = pair;
+
+			bool needsRemoval = false;
+
+			if (!isDuplicate)
+			{
+				bool hasOverlap = testAabbOverlap(pair.m_pProxy0,pair.m_pProxy1);
+
+				if (hasOverlap)
+				{
+					needsRemoval = false;//callback->processOverlap(pair);
+				} else
+				{
+					needsRemoval = true;
+				}
+			} else
+			{
+				//remove duplicate
+				needsRemoval = true;
+				//should have no algorithm
+				btAssert(!pair.m_algorithm);
+			}
+			
+			if (needsRemoval)
+			{
+				getOverlappingPairCache()->cleanOverlappingPair(pair,dispatcher);
+
+		//		m_overlappingPairArray.swap(i,m_overlappingPairArray.size()-1);
+		//		m_overlappingPairArray.pop_back();
+				pair.m_pProxy0 = 0;
+				pair.m_pProxy1 = 0;
+				m_invalidPair++;
+				gOverlappingPairs--;
+			} 
+			
+		}
+
+	///if you don't like to skip the invalid pairs in the array, execute following code:
+	#define CLEAN_INVALID_PAIRS 1
+	#ifdef CLEAN_INVALID_PAIRS
+
+		//perform a sort, to sort 'invalid' pairs to the end
+		//overlappingPairArray.heapSort(btMultiSapBroadphasePairSortPredicate());
+		overlappingPairArray.quickSort(btMultiSapBroadphasePairSortPredicate());
+
+		overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+		m_invalidPair = 0;
+	#endif//CLEAN_INVALID_PAIRS
+		
+		//printf("overlappingPairArray.size()=%d\n",overlappingPairArray.size());
+	}
+
+
+}
+
+
+bool	btMultiSapBroadphase::testAabbOverlap(btBroadphaseProxy* childProxy0,btBroadphaseProxy* childProxy1)
+{
+	btMultiSapProxy* multiSapProxy0 = (btMultiSapProxy*)childProxy0->m_multiSapParentProxy;
+		btMultiSapProxy* multiSapProxy1 = (btMultiSapProxy*)childProxy1->m_multiSapParentProxy;
+
+		return	TestAabbAgainstAabb2(multiSapProxy0->m_aabbMin,multiSapProxy0->m_aabbMax,
+			multiSapProxy1->m_aabbMin,multiSapProxy1->m_aabbMax);
+		
+}
+
+
+void	btMultiSapBroadphase::printStats()
+{
+/*	printf("---------------------------------\n");
+	
+		printf("btMultiSapBroadphase.h\n");
+		printf("numHandles = %d\n",m_multiSapProxies.size());
+			//find broadphase that contain this multiProxy
+		int numChildBroadphases = getBroadphaseArray().size();
+		for (int i=0;i<numChildBroadphases;i++)
+		{
+
+			btBroadphaseInterface* childBroadphase = getBroadphaseArray()[i];
+			childBroadphase->printStats();
+
+		}
+		*/
+
+}

extern/bullet2/src/BulletCollision/BroadphaseCollision/btMultiSapBroadphase.h

@@ -0,0 +1,144 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+#ifndef BT_MULTI_SAP_BROADPHASE
+#define BT_MULTI_SAP_BROADPHASE
+
+#include "btBroadphaseInterface.h"
+#include "LinearMath/btAlignedObjectArray.h"
+#include "btOverlappingPairCache.h"
+
+
+class btBroadphaseInterface;
+class btSimpleBroadphase;
+
+
+typedef btAlignedObjectArray<btBroadphaseInterface*> btSapBroadphaseArray;
+
+///The btMultiSapBroadphase is a broadphase that contains multiple SAP broadphases.
+///The user can add SAP broadphases that cover the world. A btBroadphaseProxy can be in multiple child broadphases at the same time.
+///A btQuantizedBvh acceleration structures finds overlapping SAPs for each btBroadphaseProxy.
+///See http://www.continuousphysics.com/Bullet/phpBB2/viewtopic.php?t=328
+///and http://www.continuousphysics.com/Bullet/phpBB2/viewtopic.php?t=1329
+class btMultiSapBroadphase :public btBroadphaseInterface
+{
+	btSapBroadphaseArray	m_sapBroadphases;
+	
+	btSimpleBroadphase*		m_simpleBroadphase;
+
+	btOverlappingPairCache*	m_overlappingPairs;
+
+	class btQuantizedBvh*			m_optimizedAabbTree;
+
+
+	bool					m_ownsPairCache;
+	
+	btOverlapFilterCallback*	m_filterCallback;
+
+	int			m_invalidPair;
+
+	struct	btBridgeProxy
+	{
+		btBroadphaseProxy*		m_childProxy;
+		btBroadphaseInterface*	m_childBroadphase;
+	};
+
+
+public:
+
+	struct	btMultiSapProxy	: public btBroadphaseProxy
+	{
+
+		///array with all the entries that this proxy belongs to
+		btAlignedObjectArray<btBridgeProxy*> m_bridgeProxies;
+		btVector3	m_aabbMin;
+		btVector3	m_aabbMax;
+
+		int	m_shapeType;
+
+/*		void*	m_userPtr;
+		short int	m_collisionFilterGroup;
+		short int	m_collisionFilterMask;
+*/
+		btMultiSapProxy(const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr, short int collisionFilterGroup,short int collisionFilterMask)
+			:btBroadphaseProxy(userPtr,collisionFilterGroup,collisionFilterMask),
+			m_aabbMin(aabbMin),
+			m_aabbMax(aabbMax),
+			m_shapeType(shapeType)
+		{
+			m_multiSapParentProxy =this;
+		}
+
+		
+	};
+
+protected:
+
+
+	btAlignedObjectArray<btMultiSapProxy*> m_multiSapProxies;
+
+public:
+
+	btMultiSapBroadphase(int maxProxies = 16384,btOverlappingPairCache* pairCache=0);
+
+
+	btSapBroadphaseArray&	getBroadphaseArray()
+	{
+		return m_sapBroadphases;
+	}
+
+	const btSapBroadphaseArray&	getBroadphaseArray() const
+	{
+		return m_sapBroadphases;
+	}
+
+	virtual ~btMultiSapBroadphase();
+
+	virtual btBroadphaseProxy*	createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr, short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* multiSapProxy);
+	virtual void	destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
+	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax, btDispatcher* dispatcher);
+
+	void	addToChildBroadphase(btMultiSapProxy* parentMultiSapProxy, btBroadphaseProxy* childProxy, btBroadphaseInterface*	childBroadphase);
+
+	///calculateOverlappingPairs is optional: incremental algorithms (sweep and prune) might do it during the set aabb
+	virtual void	calculateOverlappingPairs(btDispatcher* dispatcher);
+
+	bool	testAabbOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
+
+	virtual	btOverlappingPairCache*	getOverlappingPairCache()
+	{
+		return m_overlappingPairs;
+	}
+	virtual	const btOverlappingPairCache*	getOverlappingPairCache() const
+	{
+		return m_overlappingPairs;
+	}
+
+	///getAabb returns the axis aligned bounding box in the 'global' coordinate frame
+	///will add some transform later
+	virtual void getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const
+	{
+		aabbMin.setValue(-1e30f,-1e30f,-1e30f);
+		aabbMax.setValue(1e30f,1e30f,1e30f);
+	}
+
+	void	buildTree(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax);
+
+	virtual void	printStats();
+
+	void quicksort (btBroadphasePairArray& a, int lo, int hi);
+
+};
+
+#endif //BT_MULTI_SAP_BROADPHASE

extern/bullet2/src/BulletCollision/BroadphaseCollision/btOverlappingPairCache.cpp

@@ -1,4 +1,3 @@
-
 /*
 Bullet Continuous Collision Detection and Physics Library
 Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
@@ -21,44 +20,43 @@ subject to the following restrictions:
 #include "btDispatcher.h"
 #include "btCollisionAlgorithm.h"
 
+#include <stdio.h>
+
 int	gOverlappingPairs = 0;
 
-btOverlappingPairCache::btOverlappingPairCache():
+int gRemovePairs =0;
-m_blockedForChanges(false),
+int gAddedPairs =0;
-m_overlapFilterCallback(0)
+int gFindPairs =0;
-//m_NumOverlapBroadphasePair(0)
+
+
+
+
+btHashedOverlappingPairCache::btHashedOverlappingPairCache():
+	m_overlapFilterCallback(0),
+	m_blockedForChanges(false)
 {
+	int initialAllocatedSize= 2;
+	m_overlappingPairArray.reserve(initialAllocatedSize);
+	growTables();
 }
 
 
-btOverlappingPairCache::~btOverlappingPairCache()
+
+
+btHashedOverlappingPairCache::~btHashedOverlappingPairCache()
 {
 	//todo/test: show we erase/delete data, or is it automatic
 }
 
 
-void	btOverlappingPairCache::removeOverlappingPair(btBroadphasePair& findPair)
-{
-	
-	int findIndex = m_overlappingPairArray.findLinearSearch(findPair);
-	if (findIndex < m_overlappingPairArray.size())
-	{
-		gOverlappingPairs--;
-		btBroadphasePair& pair = m_overlappingPairArray[findIndex];
-		cleanOverlappingPair(pair);
-		
-		m_overlappingPairArray.swap(findIndex,m_overlappingPairArray.size()-1);
-		m_overlappingPairArray.pop_back();
-	}
-}
 
-
+void	btHashedOverlappingPairCache::cleanOverlappingPair(btBroadphasePair& pair,btDispatcher* dispatcher)
-void	btOverlappingPairCache::cleanOverlappingPair(btBroadphasePair& pair)
 {
 	if (pair.m_algorithm)
 	{
 		{
-			delete pair.m_algorithm;;
+			pair.m_algorithm->~btCollisionAlgorithm();
+			dispatcher->freeCollisionAlgorithm(pair.m_algorithm);
 			pair.m_algorithm=0;
 		}
 	}
@@ -67,60 +65,20 @@ void	btOverlappingPairCache::cleanOverlappingPair(btBroadphasePair& pair)
 
 
 
-
+void	btHashedOverlappingPairCache::cleanProxyFromPairs(btBroadphaseProxy* proxy,btDispatcher* dispatcher)
-void	btOverlappingPairCache::addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
-{
-	//don't add overlap with own
-	assert(proxy0 != proxy1);
-
-	if (!needsBroadphaseCollision(proxy0,proxy1))
-		return;
-
-
-	btBroadphasePair pair(*proxy0,*proxy1);
-	
-	m_overlappingPairArray.push_back(pair);
-	gOverlappingPairs++;
-	
-}
-
-///this findPair becomes really slow. Either sort the list to speedup the query, or
-///use a different solution. It is mainly used for Removing overlapping pairs. Removal could be delayed.
-///we could keep a linked list in each proxy, and store pair in one of the proxies (with lowest memory address)
-///Also we can use a 2D bitmap, which can be useful for a future GPU implementation
- btBroadphasePair*	btOverlappingPairCache::findPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
-{
-	if (!needsBroadphaseCollision(proxy0,proxy1))
-		return 0;
-
-	btBroadphasePair tmpPair(*proxy0,*proxy1);
-	int findIndex = m_overlappingPairArray.findLinearSearch(tmpPair);
-
-	if (findIndex < m_overlappingPairArray.size())
-	{
-		//assert(it != m_overlappingPairSet.end());
-		 btBroadphasePair* pair = &m_overlappingPairArray[findIndex];
-		return pair;
-	}
-	return 0;
-}
-
-
-
-
-
-void	btOverlappingPairCache::cleanProxyFromPairs(btBroadphaseProxy* proxy)
 {
 
 	class	CleanPairCallback : public btOverlapCallback
 	{
 		btBroadphaseProxy* m_cleanProxy;
 		btOverlappingPairCache*	m_pairCache;
+		btDispatcher* m_dispatcher;
 
 	public:
-		CleanPairCallback(btBroadphaseProxy* cleanProxy,btOverlappingPairCache* pairCache)
+		CleanPairCallback(btBroadphaseProxy* cleanProxy,btOverlappingPairCache* pairCache,btDispatcher* dispatcher)
 			:m_cleanProxy(cleanProxy),
-			m_pairCache(pairCache)
+			m_pairCache(pairCache),
+			m_dispatcher(dispatcher)
 		{
 		}
 		virtual	bool	processOverlap(btBroadphasePair& pair)
@@ -128,22 +86,23 @@ void	btOverlappingPairCache::cleanProxyFromPairs(btBroadphaseProxy* proxy)
 			if ((pair.m_pProxy0 == m_cleanProxy) ||
 				(pair.m_pProxy1 == m_cleanProxy))
 			{
-				m_pairCache->cleanOverlappingPair(pair);
+				m_pairCache->cleanOverlappingPair(pair,m_dispatcher);
 			}
 			return false;
 		}
 		
 	};
 
-	CleanPairCallback cleanPairs(proxy,this);
+	CleanPairCallback cleanPairs(proxy,this,dispatcher);
 
-	processAllOverlappingPairs(&cleanPairs);
+	processAllOverlappingPairs(&cleanPairs,dispatcher);
 
 }
 
 
 
-void	btOverlappingPairCache::removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy)
+
+void	btHashedOverlappingPairCache::removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher)
 {
 
 	class	RemovePairCallback : public btOverlapCallback
@@ -166,12 +125,346 @@ void	btOverlappingPairCache::removeOverlappingPairsContainingProxy(btBroadphaseP
 
 	RemovePairCallback removeCallback(proxy);
 
-	processAllOverlappingPairs(&removeCallback);
+	processAllOverlappingPairs(&removeCallback,dispatcher);
 }
 
 
 
-void	btOverlappingPairCache::processAllOverlappingPairs(btOverlapCallback* callback)
+
+
+btBroadphasePair* btHashedOverlappingPairCache::findPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1)
+{
+	gFindPairs++;
+	if(proxy0>proxy1) btSwap(proxy0,proxy1);
+	int proxyId1 = proxy0->getUid();
+	int proxyId2 = proxy1->getUid();
+
+	/*if (proxyId1 > proxyId2) 
+		btSwap(proxyId1, proxyId2);*/
+
+	int hash = static_cast<int>(getHash(static_cast<unsigned int>(proxyId1), static_cast<unsigned int>(proxyId2)) & (m_overlappingPairArray.capacity()-1));
+
+	if (hash >= m_hashTable.size())
+	{
+		return NULL;
+	}
+
+	int index = m_hashTable[hash];
+	while (index != BT_NULL_PAIR && equalsPair(m_overlappingPairArray[index], proxyId1, proxyId2) == false)
+	{
+		index = m_next[index];
+	}
+
+	if (index == BT_NULL_PAIR)
+	{
+		return NULL;
+	}
+
+	btAssert(index < m_overlappingPairArray.size());
+
+	return &m_overlappingPairArray[index];
+}
+
+//#include <stdio.h>
+
+void	btHashedOverlappingPairCache::growTables()
+{
+
+	int newCapacity = m_overlappingPairArray.capacity();
+
+	if (m_hashTable.size() < newCapacity)
+	{
+		//grow hashtable and next table
+		int curHashtableSize = m_hashTable.size();
+
+		m_hashTable.resize(newCapacity);
+		m_next.resize(newCapacity);
+
+
+		int i;
+
+		for (i= 0; i < newCapacity; ++i)
+		{
+			m_hashTable[i] = BT_NULL_PAIR;
+		}
+		for (i = 0; i < newCapacity; ++i)
+		{
+			m_next[i] = BT_NULL_PAIR;
+		}
+
+		for(i=0;i<curHashtableSize;i++)
+		{
+	
+			const btBroadphasePair& pair = m_overlappingPairArray[i];
+			int proxyId1 = pair.m_pProxy0->getUid();
+			int proxyId2 = pair.m_pProxy1->getUid();
+			/*if (proxyId1 > proxyId2) 
+				btSwap(proxyId1, proxyId2);*/
+			int	hashValue = static_cast<int>(getHash(static_cast<unsigned int>(proxyId1),static_cast<unsigned int>(proxyId2)) & (m_overlappingPairArray.capacity()-1));	// New hash value with new mask
+			m_next[i] = m_hashTable[hashValue];
+			m_hashTable[hashValue] = i;
+		}
+
+
+	}
+}
+
+btBroadphasePair* btHashedOverlappingPairCache::internalAddPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1)
+{
+	if(proxy0>proxy1) btSwap(proxy0,proxy1);
+	int proxyId1 = proxy0->getUid();
+	int proxyId2 = proxy1->getUid();
+
+	/*if (proxyId1 > proxyId2) 
+		btSwap(proxyId1, proxyId2);*/
+
+	int	hash = static_cast<int>(getHash(static_cast<unsigned int>(proxyId1),static_cast<unsigned int>(proxyId2)) & (m_overlappingPairArray.capacity()-1));	// New hash value with new mask
+
+
+	btBroadphasePair* pair = internalFindPair(proxy0, proxy1, hash);
+	if (pair != NULL)
+	{
+		return pair;
+	}
+	/*for(int i=0;i<m_overlappingPairArray.size();++i)
+		{
+		if(	(m_overlappingPairArray[i].m_pProxy0==proxy0)&&
+			(m_overlappingPairArray[i].m_pProxy1==proxy1))
+			{
+			printf("Adding duplicated %u<>%u\r\n",proxyId1,proxyId2);
+			internalFindPair(proxy0, proxy1, hash);
+			}
+		}*/
+	int count = m_overlappingPairArray.size();
+	int oldCapacity = m_overlappingPairArray.capacity();
+	void* mem = &m_overlappingPairArray.expand();
+	int newCapacity = m_overlappingPairArray.capacity();
+
+	if (oldCapacity < newCapacity)
+	{
+		growTables();
+		//hash with new capacity
+		hash = static_cast<int>(getHash(static_cast<unsigned int>(proxyId1),static_cast<unsigned int>(proxyId2)) & (m_overlappingPairArray.capacity()-1));
+	}
+	
+	pair = new (mem) btBroadphasePair(*proxy0,*proxy1);
+//	pair->m_pProxy0 = proxy0;
+//	pair->m_pProxy1 = proxy1;
+	pair->m_algorithm = 0;
+	pair->m_userInfo = 0;
+	
+
+	m_next[count] = m_hashTable[hash];
+	m_hashTable[hash] = count;
+
+	return pair;
+}
+
+
+
+void* btHashedOverlappingPairCache::removeOverlappingPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1,btDispatcher* dispatcher)
+{
+	gRemovePairs++;
+	if(proxy0>proxy1) btSwap(proxy0,proxy1);
+	int proxyId1 = proxy0->getUid();
+	int proxyId2 = proxy1->getUid();
+
+	/*if (proxyId1 > proxyId2) 
+		btSwap(proxyId1, proxyId2);*/
+
+	int	hash = static_cast<int>(getHash(static_cast<unsigned int>(proxyId1),static_cast<unsigned int>(proxyId2)) & (m_overlappingPairArray.capacity()-1));
+
+	btBroadphasePair* pair = internalFindPair(proxy0, proxy1, hash);
+	if (pair == NULL)
+	{
+		return 0;
+	}
+
+	cleanOverlappingPair(*pair,dispatcher);
+
+	void* userData = pair->m_userInfo;
+
+	btAssert(pair->m_pProxy0->getUid() == proxyId1);
+	btAssert(pair->m_pProxy1->getUid() == proxyId2);
+
+	int pairIndex = int(pair - &m_overlappingPairArray[0]);
+	btAssert(pairIndex < m_overlappingPairArray.size());
+
+	// Remove the pair from the hash table.
+	int index = m_hashTable[hash];
+	btAssert(index != BT_NULL_PAIR);
+
+	int previous = BT_NULL_PAIR;
+	while (index != pairIndex)
+	{
+		previous = index;
+		index = m_next[index];
+	}
+
+	if (previous != BT_NULL_PAIR)
+	{
+		btAssert(m_next[previous] == pairIndex);
+		m_next[previous] = m_next[pairIndex];
+	}
+	else
+	{
+		m_hashTable[hash] = m_next[pairIndex];
+	}
+
+	// We now move the last pair into spot of the
+	// pair being removed. We need to fix the hash
+	// table indices to support the move.
+
+	int lastPairIndex = m_overlappingPairArray.size() - 1;
+
+	// If the removed pair is the last pair, we are done.
+	if (lastPairIndex == pairIndex)
+	{
+		m_overlappingPairArray.pop_back();
+		return userData;
+	}
+
+	// Remove the last pair from the hash table.
+	const btBroadphasePair* last = &m_overlappingPairArray[lastPairIndex];
+		/* missing swap here too, Nat. */ 
+	int lastHash = static_cast<int>(getHash(static_cast<unsigned int>(last->m_pProxy0->getUid()), static_cast<unsigned int>(last->m_pProxy1->getUid())) & (m_overlappingPairArray.capacity()-1));
+
+	index = m_hashTable[lastHash];
+	btAssert(index != BT_NULL_PAIR);
+
+	previous = BT_NULL_PAIR;
+	while (index != lastPairIndex)
+	{
+		previous = index;
+		index = m_next[index];
+	}
+
+	if (previous != BT_NULL_PAIR)
+	{
+		btAssert(m_next[previous] == lastPairIndex);
+		m_next[previous] = m_next[lastPairIndex];
+	}
+	else
+	{
+		m_hashTable[lastHash] = m_next[lastPairIndex];
+	}
+
+	// Copy the last pair into the remove pair's spot.
+	m_overlappingPairArray[pairIndex] = m_overlappingPairArray[lastPairIndex];
+
+	// Insert the last pair into the hash table
+	m_next[pairIndex] = m_hashTable[lastHash];
+	m_hashTable[lastHash] = pairIndex;
+
+	m_overlappingPairArray.pop_back();
+
+	return userData;
+}
+//#include <stdio.h>
+
+void	btHashedOverlappingPairCache::processAllOverlappingPairs(btOverlapCallback* callback,btDispatcher* dispatcher)
+{
+
+	int i;
+
+//	printf("m_overlappingPairArray.size()=%d\n",m_overlappingPairArray.size());
+	for (i=0;i<m_overlappingPairArray.size();)
+	{
+	
+		btBroadphasePair* pair = &m_overlappingPairArray[i];
+		if (callback->processOverlap(*pair))
+		{
+			removeOverlappingPair(pair->m_pProxy0,pair->m_pProxy1,dispatcher);
+
+			gOverlappingPairs--;
+		} else
+		{
+			i++;
+		}
+	}
+}
+
+
+
+void*	btSortedOverlappingPairCache::removeOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1, btDispatcher* dispatcher )
+{
+	if (!hasDeferredRemoval())
+	{
+		btBroadphasePair findPair(*proxy0,*proxy1);
+
+		int findIndex = m_overlappingPairArray.findLinearSearch(findPair);
+		if (findIndex < m_overlappingPairArray.size())
+		{
+			gOverlappingPairs--;
+			btBroadphasePair& pair = m_overlappingPairArray[findIndex];
+			void* userData = pair.m_userInfo;
+			cleanOverlappingPair(pair,dispatcher);
+			
+			m_overlappingPairArray.swap(findIndex,m_overlappingPairArray.capacity()-1);
+			m_overlappingPairArray.pop_back();
+			return userData;
+		}
+	}
+
+	return 0;
+}
+
+
+
+
+
+
+
+
+btBroadphasePair*	btSortedOverlappingPairCache::addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+{
+	//don't add overlap with own
+	assert(proxy0 != proxy1);
+
+	if (!needsBroadphaseCollision(proxy0,proxy1))
+		return 0;
+	
+	void* mem = &m_overlappingPairArray.expand();
+	btBroadphasePair* pair = new (mem) btBroadphasePair(*proxy0,*proxy1);
+	gOverlappingPairs++;
+	gAddedPairs++;
+	return pair;
+	
+}
+
+///this findPair becomes really slow. Either sort the list to speedup the query, or
+///use a different solution. It is mainly used for Removing overlapping pairs. Removal could be delayed.
+///we could keep a linked list in each proxy, and store pair in one of the proxies (with lowest memory address)
+///Also we can use a 2D bitmap, which can be useful for a future GPU implementation
+ btBroadphasePair*	btSortedOverlappingPairCache::findPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+{
+	if (!needsBroadphaseCollision(proxy0,proxy1))
+		return 0;
+
+	btBroadphasePair tmpPair(*proxy0,*proxy1);
+	int findIndex = m_overlappingPairArray.findLinearSearch(tmpPair);
+
+	if (findIndex < m_overlappingPairArray.size())
+	{
+		//assert(it != m_overlappingPairSet.end());
+		 btBroadphasePair* pair = &m_overlappingPairArray[findIndex];
+		return pair;
+	}
+	return 0;
+}
+
+
+
+
+
+
+
+
+
+
+//#include <stdio.h>
+
+void	btSortedOverlappingPairCache::processAllOverlappingPairs(btOverlapCallback* callback,btDispatcher* dispatcher)
 {
 
 	int i;
@@ -182,9 +475,9 @@ void	btOverlappingPairCache::processAllOverlappingPairs(btOverlapCallback* callb
 		btBroadphasePair* pair = &m_overlappingPairArray[i];
 		if (callback->processOverlap(*pair))
 		{
-			cleanOverlappingPair(*pair);
+			cleanOverlappingPair(*pair,dispatcher);
 
-			m_overlappingPairArray.swap(i,m_overlappingPairArray.size()-1);
+			m_overlappingPairArray.swap(i,m_overlappingPairArray.capacity()-1);
 			m_overlappingPairArray.pop_back();
 			gOverlappingPairs--;
 		} else
@@ -194,3 +487,93 @@ void	btOverlappingPairCache::processAllOverlappingPairs(btOverlapCallback* callb
 	}
 }
 
+
+
+
+btSortedOverlappingPairCache::btSortedOverlappingPairCache():
+	m_blockedForChanges(false),
+	m_hasDeferredRemoval(true),
+	m_overlapFilterCallback(0)
+{
+	int initialAllocatedSize= 2;
+	m_overlappingPairArray.reserve(initialAllocatedSize);
+}
+
+btSortedOverlappingPairCache::~btSortedOverlappingPairCache()
+{
+	//todo/test: show we erase/delete data, or is it automatic
+}
+
+void	btSortedOverlappingPairCache::cleanOverlappingPair(btBroadphasePair& pair,btDispatcher* dispatcher)
+{
+	if (pair.m_algorithm)
+	{
+		{
+			pair.m_algorithm->~btCollisionAlgorithm();
+			dispatcher->freeCollisionAlgorithm(pair.m_algorithm);
+			pair.m_algorithm=0;
+			gRemovePairs--;
+		}
+	}
+}
+
+
+void	btSortedOverlappingPairCache::cleanProxyFromPairs(btBroadphaseProxy* proxy,btDispatcher* dispatcher)
+{
+
+	class	CleanPairCallback : public btOverlapCallback
+	{
+		btBroadphaseProxy* m_cleanProxy;
+		btOverlappingPairCache*	m_pairCache;
+		btDispatcher* m_dispatcher;
+
+	public:
+		CleanPairCallback(btBroadphaseProxy* cleanProxy,btOverlappingPairCache* pairCache,btDispatcher* dispatcher)
+			:m_cleanProxy(cleanProxy),
+			m_pairCache(pairCache),
+			m_dispatcher(dispatcher)
+		{
+		}
+		virtual	bool	processOverlap(btBroadphasePair& pair)
+		{
+			if ((pair.m_pProxy0 == m_cleanProxy) ||
+				(pair.m_pProxy1 == m_cleanProxy))
+			{
+				m_pairCache->cleanOverlappingPair(pair,m_dispatcher);
+			}
+			return false;
+		}
+		
+	};
+
+	CleanPairCallback cleanPairs(proxy,this,dispatcher);
+
+	processAllOverlappingPairs(&cleanPairs,dispatcher);
+
+}
+
+
+void	btSortedOverlappingPairCache::removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher)
+{
+
+	class	RemovePairCallback : public btOverlapCallback
+	{
+		btBroadphaseProxy* m_obsoleteProxy;
+
+	public:
+		RemovePairCallback(btBroadphaseProxy* obsoleteProxy)
+			:m_obsoleteProxy(obsoleteProxy)
+		{
+		}
+		virtual	bool	processOverlap(btBroadphasePair& pair)
+		{
+			return ((pair.m_pProxy0 == m_obsoleteProxy) ||
+				(pair.m_pProxy1 == m_obsoleteProxy));
+		}
+		
+	};
+
+	RemovePairCallback removeCallback(proxy);
+
+	processAllOverlappingPairs(&removeCallback,dispatcher);
+}

extern/bullet2/src/BulletCollision/BroadphaseCollision/btOverlappingPairCache.h

@@ -1,4 +1,3 @@
-
 /*
 Bullet Continuous Collision Detection and Physics Library
 Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
@@ -20,9 +19,13 @@ subject to the following restrictions:
 
 #include "btBroadphaseInterface.h"
 #include "btBroadphaseProxy.h"
-#include "../../LinearMath/btPoint3.h"
+#include "btOverlappingPairCallback.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
 
+#include "LinearMath/btPoint3.h"
+#include "LinearMath/btAlignedObjectArray.h"
+class btDispatcher;
+
+typedef btAlignedObjectArray<btBroadphasePair>	btBroadphasePairArray;
 
 struct	btOverlapCallback
 {
@@ -30,6 +33,7 @@ struct	btOverlapCallback
 	{}
 	//return true for deletion of the pair
 	virtual bool	processOverlap(btBroadphasePair& pair) = 0;
+
 };
 
 struct btOverlapFilterCallback
@@ -40,38 +44,261 @@ struct btOverlapFilterCallback
 	virtual bool	needBroadphaseCollision(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1) const = 0;
 };
 
-///btOverlappingPairCache maintains the objects with overlapping AABB
+
+
+
+
+
+
+extern int gRemovePairs;
+extern int gAddedPairs;
+extern int gFindPairs;
+
+const int BT_NULL_PAIR=0xffffffff;
+
+///The btOverlappingPairCache provides an interface for overlapping pair management (add, remove, storage), used by the btBroadphaseInterface broadphases.
+///The btHashedOverlappingPairCache and btSortedOverlappingPairCache classes are two implementations.
+class btOverlappingPairCache : public btOverlappingPairCallback
+{
+public:
+	virtual ~btOverlappingPairCache() {} // this is needed so we can get to the derived class destructor
+
+	virtual btBroadphasePair*	getOverlappingPairArrayPtr() = 0;
+	
+	virtual const btBroadphasePair*	getOverlappingPairArrayPtr() const = 0;
+
+	virtual btBroadphasePairArray&	getOverlappingPairArray() = 0;
+
+	virtual	void	cleanOverlappingPair(btBroadphasePair& pair,btDispatcher* dispatcher) = 0;
+
+	virtual int getNumOverlappingPairs() const = 0;
+
+	virtual void	cleanProxyFromPairs(btBroadphaseProxy* proxy,btDispatcher* dispatcher) = 0;
+
+	virtual	void setOverlapFilterCallback(btOverlapFilterCallback* callback) = 0;
+
+	virtual void	processAllOverlappingPairs(btOverlapCallback*,btDispatcher* dispatcher) = 0;
+
+	virtual btBroadphasePair* findPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1) = 0;
+
+	virtual bool	hasDeferredRemoval() = 0;
+
+};
+
+/// Hash-space based Pair Cache, thanks to Erin Catto, Box2D, http://www.box2d.org, and Pierre Terdiman, Codercorner, http://codercorner.com
+class btHashedOverlappingPairCache : public btOverlappingPairCache
+{
+	btBroadphasePairArray	m_overlappingPairArray;
+	btOverlapFilterCallback* m_overlapFilterCallback;
+	bool		m_blockedForChanges;
+
+
+public:
+	btHashedOverlappingPairCache();
+	virtual ~btHashedOverlappingPairCache();
+
+	
+	void	removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
+
+	virtual void*	removeOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1,btDispatcher* dispatcher);
+	
+	SIMD_FORCE_INLINE bool needsBroadphaseCollision(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1) const
+	{
+		if (m_overlapFilterCallback)
+			return m_overlapFilterCallback->needBroadphaseCollision(proxy0,proxy1);
+
+		bool collides = (proxy0->m_collisionFilterGroup & proxy1->m_collisionFilterMask) != 0;
+		collides = collides && (proxy1->m_collisionFilterGroup & proxy0->m_collisionFilterMask);
+		
+		return collides;
+	}
+
+	// Add a pair and return the new pair. If the pair already exists,
+	// no new pair is created and the old one is returned.
+	virtual btBroadphasePair* 	addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+	{
+		gAddedPairs++;
+
+		if (!needsBroadphaseCollision(proxy0,proxy1))
+			return 0;
+
+		return internalAddPair(proxy0,proxy1);
+	}
+
+	
+
+	void	cleanProxyFromPairs(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
+
+	
+	virtual void	processAllOverlappingPairs(btOverlapCallback*,btDispatcher* dispatcher);
+
+	virtual btBroadphasePair*	getOverlappingPairArrayPtr()
+	{
+		return &m_overlappingPairArray[0];
+	}
+
+	const btBroadphasePair*	getOverlappingPairArrayPtr() const
+	{
+		return &m_overlappingPairArray[0];
+	}
+
+	btBroadphasePairArray&	getOverlappingPairArray()
+	{
+		return m_overlappingPairArray;
+	}
+
+	const btBroadphasePairArray&	getOverlappingPairArray() const
+	{
+		return m_overlappingPairArray;
+	}
+
+	void	cleanOverlappingPair(btBroadphasePair& pair,btDispatcher* dispatcher);
+
+
+
+	btBroadphasePair* findPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1);
+
+	int GetCount() const { return m_overlappingPairArray.size(); }
+//	btBroadphasePair* GetPairs() { return m_pairs; }
+
+	btOverlapFilterCallback* getOverlapFilterCallback()
+	{
+		return m_overlapFilterCallback;
+	}
+
+	void setOverlapFilterCallback(btOverlapFilterCallback* callback)
+	{
+		m_overlapFilterCallback = callback;
+	}
+
+	int	getNumOverlappingPairs() const
+	{
+		return m_overlappingPairArray.size();
+	}
+private:
+	
+	btBroadphasePair* 	internalAddPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
+
+	void	growTables();
+
+	SIMD_FORCE_INLINE bool equalsPair(const btBroadphasePair& pair, int proxyId1, int proxyId2)
+	{	
+		return pair.m_pProxy0->getUid() == proxyId1 && pair.m_pProxy1->getUid() == proxyId2;
+	}
+
+	/*
+	// Thomas Wang's hash, see: http://www.concentric.net/~Ttwang/tech/inthash.htm
+	// This assumes proxyId1 and proxyId2 are 16-bit.
+	SIMD_FORCE_INLINE int getHash(int proxyId1, int proxyId2)
+	{
+		int key = (proxyId2 << 16) | proxyId1;
+		key = ~key + (key << 15);
+		key = key ^ (key >> 12);
+		key = key + (key << 2);
+		key = key ^ (key >> 4);
+		key = key * 2057;
+		key = key ^ (key >> 16);
+		return key;
+	}
+	*/
+
+
+	
+	SIMD_FORCE_INLINE	unsigned int getHash(unsigned int proxyId1, unsigned int proxyId2)
+	{
+		int key = static_cast<int>(((unsigned int)proxyId1) | (((unsigned int)proxyId2) <<16));
+		// Thomas Wang's hash
+
+		key += ~(key << 15);
+		key ^=  (key >> 10);
+		key +=  (key << 3);
+		key ^=  (key >> 6);
+		key += ~(key << 11);
+		key ^=  (key >> 16);
+		return static_cast<unsigned int>(key);
+	}
+	
+
+
+
+
+	SIMD_FORCE_INLINE btBroadphasePair* internalFindPair(btBroadphaseProxy* proxy0, btBroadphaseProxy* proxy1, int hash)
+	{
+		int proxyId1 = proxy0->getUid();
+		int proxyId2 = proxy1->getUid();
+		#if 0 // wrong, 'equalsPair' use unsorted uids, copy-past devil striked again. Nat.
+		if (proxyId1 > proxyId2) 
+			btSwap(proxyId1, proxyId2);
+		#endif
+
+		int index = m_hashTable[hash];
+		
+		while( index != BT_NULL_PAIR && equalsPair(m_overlappingPairArray[index], proxyId1, proxyId2) == false)
+		{
+			index = m_next[index];
+		}
+
+		if ( index == BT_NULL_PAIR )
+		{
+			return NULL;
+		}
+
+		btAssert(index < m_overlappingPairArray.size());
+
+		return &m_overlappingPairArray[index];
+	}
+
+	virtual bool	hasDeferredRemoval()
+	{
+		return false;
+	}
+
+public:
+	
+	btAlignedObjectArray<int>	m_hashTable;
+	btAlignedObjectArray<int>	m_next;
+	
+};
+
+
+
+
+///btSortedOverlappingPairCache maintains the objects with overlapping AABB
 ///Typically managed by the Broadphase, Axis3Sweep or btSimpleBroadphase
-class	btOverlappingPairCache : public btBroadphaseInterface
+class	btSortedOverlappingPairCache : public btOverlappingPairCache
 {
 	protected:
 		//avoid brute-force finding all the time
-		btAlignedObjectArray<btBroadphasePair>	m_overlappingPairArray;
+		btBroadphasePairArray	m_overlappingPairArray;
-		
+
 		//during the dispatch, check that user doesn't destroy/create proxy
 		bool		m_blockedForChanges;
+
+		///by default, do the removal during the pair traversal
+		bool		m_hasDeferredRemoval;
 		
 		//if set, use the callback instead of the built in filter in needBroadphaseCollision
 		btOverlapFilterCallback* m_overlapFilterCallback;
+
 	public:
 			
-		btOverlappingPairCache();	
+		btSortedOverlappingPairCache();	
-		virtual ~btOverlappingPairCache();
+		virtual ~btSortedOverlappingPairCache();
 
-		virtual void	processAllOverlappingPairs(btOverlapCallback*);
+		virtual void	processAllOverlappingPairs(btOverlapCallback*,btDispatcher* dispatcher);
 
-		void	removeOverlappingPair(btBroadphasePair& pair);
+		void*	removeOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1,btDispatcher* dispatcher);
 
-		void	cleanOverlappingPair(btBroadphasePair& pair);
+		void	cleanOverlappingPair(btBroadphasePair& pair,btDispatcher* dispatcher);
 		
-		void	addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
+		btBroadphasePair*	addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
 
 		btBroadphasePair*	findPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
 			
 		
-		void	cleanProxyFromPairs(btBroadphaseProxy* proxy);
+		void	cleanProxyFromPairs(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
 
-		void	removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy);
+		void	removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
 
 
 		inline bool needsBroadphaseCollision(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1) const
@@ -84,10 +311,19 @@ class	btOverlappingPairCache : public btBroadphaseInterface
 			
 			return collides;
 		}
-			
+		
+		btBroadphasePairArray&	getOverlappingPairArray()
+		{
+			return m_overlappingPairArray;
+		}
+
+		const btBroadphasePairArray&	getOverlappingPairArray() const
+		{
+			return m_overlappingPairArray;
+		}
+
 		
 
-		virtual void	refreshOverlappingPairs() =0;
 
 		btBroadphasePair*	getOverlappingPairArrayPtr()
 		{
@@ -114,7 +350,88 @@ class	btOverlappingPairCache : public btBroadphaseInterface
 			m_overlapFilterCallback = callback;
 		}
 
+		virtual bool	hasDeferredRemoval()
+		{
+			return m_hasDeferredRemoval;
+		}
+
+
 };
+
+
+
+///btNullPairCache skips add/removal of overlapping pairs. Userful for benchmarking and testing.
+class btNullPairCache : public btOverlappingPairCache
+{
+
+	btBroadphasePairArray	m_overlappingPairArray;
+
+public:
+
+	virtual btBroadphasePair*	getOverlappingPairArrayPtr()
+	{
+		return &m_overlappingPairArray[0];
+	}
+	const btBroadphasePair*	getOverlappingPairArrayPtr() const
+	{
+		return &m_overlappingPairArray[0];
+	}
+	btBroadphasePairArray&	getOverlappingPairArray()
+	{
+		return m_overlappingPairArray;
+	}
+	
+	virtual	void	cleanOverlappingPair(btBroadphasePair& /*pair*/,btDispatcher* /*dispatcher*/)
+	{
+
+	}
+
+	virtual int getNumOverlappingPairs() const
+	{
+		return 0;
+	}
+
+	virtual void	cleanProxyFromPairs(btBroadphaseProxy* /*proxy*/,btDispatcher* /*dispatcher*/)
+	{
+
+	}
+
+	virtual	void setOverlapFilterCallback(btOverlapFilterCallback* /*callback*/)
+	{
+	}
+
+	virtual void	processAllOverlappingPairs(btOverlapCallback*,btDispatcher* /*dispatcher*/)
+	{
+	}
+
+	virtual btBroadphasePair* findPair(btBroadphaseProxy* /*proxy0*/, btBroadphaseProxy* /*proxy1*/)
+	{
+		return 0;
+	}
+
+	virtual bool	hasDeferredRemoval()
+	{
+		return true;
+	}
+
+	virtual btBroadphasePair*	addOverlappingPair(btBroadphaseProxy* /*proxy0*/,btBroadphaseProxy* /*proxy1*/)
+	{
+		return 0;
+	}
+
+	virtual void*	removeOverlappingPair(btBroadphaseProxy* /*proxy0*/,btBroadphaseProxy* /*proxy1*/,btDispatcher* /*dispatcher*/)
+	{
+		return 0;
+	}
+
+	virtual void	removeOverlappingPairsContainingProxy(btBroadphaseProxy* /*proxy0*/,btDispatcher* /*dispatcher*/)
+	{
+	}
+
+
+};
+
+
 #endif //OVERLAPPING_PAIR_CACHE_H
 
 

extern/bullet2/src/BulletCollision/BroadphaseCollision/btOverlappingPairCallback.h

@@ -0,0 +1,40 @@
+
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef OVERLAPPING_PAIR_CALLBACK_H
+#define OVERLAPPING_PAIR_CALLBACK_H
+
+class btDispatcher;
+struct  btBroadphasePair;
+
+///The btOverlappingPairCallback class is an additional optional broadphase user callback for adding/removing overlapping pairs, similar interface to btOverlappingPairCache.
+class btOverlappingPairCallback
+{
+public:
+	virtual ~btOverlappingPairCallback()
+	{
+
+	}
+	
+	virtual btBroadphasePair*	addOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1) = 0;
+
+	virtual void*	removeOverlappingPair(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1,btDispatcher* dispatcher) = 0;
+
+	virtual void	removeOverlappingPairsContainingProxy(btBroadphaseProxy* proxy0,btDispatcher* dispatcher) = 0;
+
+};
+
+#endif //OVERLAPPING_PAIR_CALLBACK_H

extern/bullet2/src/BulletCollision/BroadphaseCollision/btQuantizedBvh.h

@@ -0,0 +1,486 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef QUANTIZED_BVH_H
+#define QUANTIZED_BVH_H
+
+//#define DEBUG_CHECK_DEQUANTIZATION 1
+#ifdef DEBUG_CHECK_DEQUANTIZATION
+#ifdef __SPU__
+#define printf spu_printf
+#endif //__SPU__
+
+#include <stdio.h>
+#include <stdlib.h>
+#endif //DEBUG_CHECK_DEQUANTIZATION
+
+#include "LinearMath/btVector3.h"
+#include "LinearMath/btAlignedAllocator.h"
+
+
+//http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclang/html/vclrf__m128.asp
+
+
+//Note: currently we have 16 bytes per quantized node
+#define MAX_SUBTREE_SIZE_IN_BYTES  2048
+
+// 10 gives the potential for 1024 parts, with at most 2^21 (2097152) (minus one
+// actually) triangles each (since the sign bit is reserved
+#define MAX_NUM_PARTS_IN_BITS 10
+
+///btQuantizedBvhNode is a compressed aabb node, 16 bytes.
+///Node can be used for leafnode or internal node. Leafnodes can point to 32-bit triangle index (non-negative range).
+ATTRIBUTE_ALIGNED16	(struct) btQuantizedBvhNode
+{
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+	//12 bytes
+	unsigned short int	m_quantizedAabbMin[3];
+	unsigned short int	m_quantizedAabbMax[3];
+	//4 bytes
+	int	m_escapeIndexOrTriangleIndex;
+
+	bool isLeafNode() const
+	{
+		//skipindex is negative (internal node), triangleindex >=0 (leafnode)
+		return (m_escapeIndexOrTriangleIndex >= 0);
+	}
+	int getEscapeIndex() const
+	{
+		btAssert(!isLeafNode());
+		return -m_escapeIndexOrTriangleIndex;
+	}
+	int	getTriangleIndex() const
+	{
+		btAssert(isLeafNode());
+		// Get only the lower bits where the triangle index is stored
+		return (m_escapeIndexOrTriangleIndex&~((~0)<<(31-MAX_NUM_PARTS_IN_BITS)));
+	}
+	int	getPartId() const
+	{
+		btAssert(isLeafNode());
+		// Get only the highest bits where the part index is stored
+		return (m_escapeIndexOrTriangleIndex>>(31-MAX_NUM_PARTS_IN_BITS));
+	}
+}
+;
+
+/// btOptimizedBvhNode contains both internal and leaf node information.
+/// Total node size is 44 bytes / node. You can use the compressed version of 16 bytes.
+ATTRIBUTE_ALIGNED16 (struct) btOptimizedBvhNode
+{
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+	//32 bytes
+	btVector3	m_aabbMinOrg;
+	btVector3	m_aabbMaxOrg;
+
+	//4
+	int	m_escapeIndex;
+
+	//8
+	//for child nodes
+	int	m_subPart;
+	int	m_triangleIndex;
+	int	m_padding[5];//bad, due to alignment
+
+
+};
+
+
+///btBvhSubtreeInfo provides info to gather a subtree of limited size
+ATTRIBUTE_ALIGNED16(class) btBvhSubtreeInfo
+{
+public:
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+	//12 bytes
+	unsigned short int	m_quantizedAabbMin[3];
+	unsigned short int	m_quantizedAabbMax[3];
+	//4 bytes, points to the root of the subtree
+	int			m_rootNodeIndex;
+	//4 bytes
+	int			m_subtreeSize;
+	int			m_padding[3];
+
+	btBvhSubtreeInfo()
+	{
+		//memset(&m_padding[0], 0, sizeof(m_padding));
+	}
+
+
+	void	setAabbFromQuantizeNode(const btQuantizedBvhNode& quantizedNode)
+	{
+		m_quantizedAabbMin[0] = quantizedNode.m_quantizedAabbMin[0];
+		m_quantizedAabbMin[1] = quantizedNode.m_quantizedAabbMin[1];
+		m_quantizedAabbMin[2] = quantizedNode.m_quantizedAabbMin[2];
+		m_quantizedAabbMax[0] = quantizedNode.m_quantizedAabbMax[0];
+		m_quantizedAabbMax[1] = quantizedNode.m_quantizedAabbMax[1];
+		m_quantizedAabbMax[2] = quantizedNode.m_quantizedAabbMax[2];
+	}
+}
+;
+
+
+class btNodeOverlapCallback
+{
+public:
+	virtual ~btNodeOverlapCallback() {};
+
+	virtual void processNode(int subPart, int triangleIndex) = 0;
+};
+
+#include "LinearMath/btAlignedAllocator.h"
+#include "LinearMath/btAlignedObjectArray.h"
+
+
+
+///for code readability:
+typedef btAlignedObjectArray<btOptimizedBvhNode>	NodeArray;
+typedef btAlignedObjectArray<btQuantizedBvhNode>	QuantizedNodeArray;
+typedef btAlignedObjectArray<btBvhSubtreeInfo>		BvhSubtreeInfoArray;
+
+
+///The btQuantizedBvh class stores an AABB tree that can be quickly traversed on CPU and Cell SPU.
+///It is used by the btBvhTriangleMeshShape as midphase, and by the btMultiSapBroadphase.
+///It is recommended to use quantization for better performance and lower memory requirements.
+ATTRIBUTE_ALIGNED16(class) btQuantizedBvh
+{
+protected:
+
+	NodeArray			m_leafNodes;
+	NodeArray			m_contiguousNodes;
+
+	QuantizedNodeArray	m_quantizedLeafNodes;
+	
+	QuantizedNodeArray	m_quantizedContiguousNodes;
+	
+	int					m_curNodeIndex;
+
+
+	//quantization data
+	bool				m_useQuantization;
+	btVector3			m_bvhAabbMin;
+	btVector3			m_bvhAabbMax;
+	btVector3			m_bvhQuantization;
+public:
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+	enum btTraversalMode
+	{
+		TRAVERSAL_STACKLESS = 0,
+		TRAVERSAL_STACKLESS_CACHE_FRIENDLY,
+		TRAVERSAL_RECURSIVE
+	};
+protected:
+
+	btTraversalMode	m_traversalMode;
+	
+	BvhSubtreeInfoArray		m_SubtreeHeaders;
+
+	//This is only used for serialization so we don't have to add serialization directly to btAlignedObjectArray
+	int m_subtreeHeaderCount;
+
+
+	///two versions, one for quantized and normal nodes. This allows code-reuse while maintaining readability (no template/macro!)
+	///this might be refactored into a virtual, it is usually not calculated at run-time
+	void	setInternalNodeAabbMin(int nodeIndex, const btVector3& aabbMin)
+	{
+		if (m_useQuantization)
+		{
+			quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[0] ,aabbMin,0);
+		} else
+		{
+			m_contiguousNodes[nodeIndex].m_aabbMinOrg = aabbMin;
+
+		}
+	}
+	void	setInternalNodeAabbMax(int nodeIndex,const btVector3& aabbMax)
+	{
+		if (m_useQuantization)
+		{
+			quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[0],aabbMax,1);
+		} else
+		{
+			m_contiguousNodes[nodeIndex].m_aabbMaxOrg = aabbMax;
+		}
+	}
+
+	btVector3 getAabbMin(int nodeIndex) const
+	{
+		if (m_useQuantization)
+		{
+			return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMin[0]);
+		}
+		//non-quantized
+		return m_leafNodes[nodeIndex].m_aabbMinOrg;
+
+	}
+	btVector3 getAabbMax(int nodeIndex) const
+	{
+		if (m_useQuantization)
+		{
+			return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMax[0]);
+		} 
+		//non-quantized
+		return m_leafNodes[nodeIndex].m_aabbMaxOrg;
+		
+	}
+
+	
+	void	setInternalNodeEscapeIndex(int nodeIndex, int escapeIndex)
+	{
+		if (m_useQuantization)
+		{
+			m_quantizedContiguousNodes[nodeIndex].m_escapeIndexOrTriangleIndex = -escapeIndex;
+		} 
+		else
+		{
+			m_contiguousNodes[nodeIndex].m_escapeIndex = escapeIndex;
+		}
+
+	}
+
+	void mergeInternalNodeAabb(int nodeIndex,const btVector3& newAabbMin,const btVector3& newAabbMax) 
+	{
+		if (m_useQuantization)
+		{
+			unsigned short int quantizedAabbMin[3];
+			unsigned short int quantizedAabbMax[3];
+			quantize(quantizedAabbMin,newAabbMin,0);
+			quantize(quantizedAabbMax,newAabbMax,1);
+			for (int i=0;i<3;i++)
+			{
+				if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] > quantizedAabbMin[i])
+					m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] = quantizedAabbMin[i];
+
+				if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] < quantizedAabbMax[i])
+					m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] = quantizedAabbMax[i];
+
+			}
+		} else
+		{
+			//non-quantized
+			m_contiguousNodes[nodeIndex].m_aabbMinOrg.setMin(newAabbMin);
+			m_contiguousNodes[nodeIndex].m_aabbMaxOrg.setMax(newAabbMax);		
+		}
+	}
+
+	void	swapLeafNodes(int firstIndex,int secondIndex);
+
+	void	assignInternalNodeFromLeafNode(int internalNode,int leafNodeIndex);
+
+protected:
+
+	
+
+	void	buildTree	(int startIndex,int endIndex);
+
+	int	calcSplittingAxis(int startIndex,int endIndex);
+
+	int	sortAndCalcSplittingIndex(int startIndex,int endIndex,int splitAxis);
+	
+	void	walkStacklessTree(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const;
+
+	void	walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax, int startNodeIndex,int endNodeIndex) const;
+	void	walkStacklessQuantizedTree(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax,int startNodeIndex,int endNodeIndex) const;
+
+	///tree traversal designed for small-memory processors like PS3 SPU
+	void	walkStacklessQuantizedTreeCacheFriendly(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
+
+	///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
+	void	walkRecursiveQuantizedTreeAgainstQueryAabb(const btQuantizedBvhNode* currentNode,btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
+
+	///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
+	void	walkRecursiveQuantizedTreeAgainstQuantizedTree(const btQuantizedBvhNode* treeNodeA,const btQuantizedBvhNode* treeNodeB,btNodeOverlapCallback* nodeCallback) const;
+	
+
+#define USE_BANCHLESS 1
+#ifdef USE_BANCHLESS
+	//This block replaces the block below and uses no branches, and replaces the 8 bit return with a 32 bit return for improved performance (~3x on XBox 360)
+	SIMD_FORCE_INLINE unsigned testQuantizedAabbAgainstQuantizedAabb(unsigned short int* aabbMin1,unsigned short int* aabbMax1,const unsigned short int* aabbMin2,const unsigned short int* aabbMax2) const
+	{		
+		return static_cast<unsigned int>(btSelect((unsigned)((aabbMin1[0] <= aabbMax2[0]) & (aabbMax1[0] >= aabbMin2[0])
+			& (aabbMin1[2] <= aabbMax2[2]) & (aabbMax1[2] >= aabbMin2[2])
+			& (aabbMin1[1] <= aabbMax2[1]) & (aabbMax1[1] >= aabbMin2[1])),
+			1, 0));
+	}
+#else
+	SIMD_FORCE_INLINE bool testQuantizedAabbAgainstQuantizedAabb(unsigned short int* aabbMin1,unsigned short int* aabbMax1,const unsigned short int* aabbMin2,const unsigned short int* aabbMax2) const
+	{
+		bool overlap = true;
+		overlap = (aabbMin1[0] > aabbMax2[0] || aabbMax1[0] < aabbMin2[0]) ? false : overlap;
+		overlap = (aabbMin1[2] > aabbMax2[2] || aabbMax1[2] < aabbMin2[2]) ? false : overlap;
+		overlap = (aabbMin1[1] > aabbMax2[1] || aabbMax1[1] < aabbMin2[1]) ? false : overlap;
+		return overlap;
+	}
+#endif //USE_BANCHLESS
+
+	void	updateSubtreeHeaders(int leftChildNodexIndex,int rightChildNodexIndex);
+
+public:
+	btQuantizedBvh();
+
+	virtual ~btQuantizedBvh();
+
+	
+	///***************************************** expert/internal use only *************************
+	void	setQuantizationValues(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax,btScalar quantizationMargin=btScalar(1.0));
+	QuantizedNodeArray&	getLeafNodeArray() {			return	m_quantizedLeafNodes;	}
+	///buildInternal is expert use only: assumes that setQuantizationValues and LeafNodeArray are initialized
+	void	buildInternal();
+	///***************************************** expert/internal use only *************************
+
+	void	reportAabbOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const;
+	void	reportRayOverlappingNodex (btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget) const;
+	void	reportBoxCastOverlappingNodex(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin,const btVector3& aabbMax) const;
+
+		SIMD_FORCE_INLINE void quantize(unsigned short* out, const btVector3& point,int isMax) const
+	{
+
+		btAssert(m_useQuantization);
+
+		btAssert(point.getX() <= m_bvhAabbMax.getX());
+		btAssert(point.getY() <= m_bvhAabbMax.getY());
+		btAssert(point.getZ() <= m_bvhAabbMax.getZ());
+
+		btAssert(point.getX() >= m_bvhAabbMin.getX());
+		btAssert(point.getY() >= m_bvhAabbMin.getY());
+		btAssert(point.getZ() >= m_bvhAabbMin.getZ());
+
+		btVector3 v = (point - m_bvhAabbMin) * m_bvhQuantization;
+		///Make sure rounding is done in a way that unQuantize(quantizeWithClamp(...)) is conservative
+		///end-points always set the first bit, so that they are sorted properly (so that neighbouring AABBs overlap properly)
+		///todo: double-check this
+		if (isMax)
+		{
+			out[0] = (unsigned short) (((unsigned short)(v.getX()+btScalar(1.)) | 1));
+			out[1] = (unsigned short) (((unsigned short)(v.getY()+btScalar(1.)) | 1));
+			out[2] = (unsigned short) (((unsigned short)(v.getZ()+btScalar(1.)) | 1));
+		} else
+		{
+			out[0] = (unsigned short) (((unsigned short)(v.getX()) & 0xfffe));
+			out[1] = (unsigned short) (((unsigned short)(v.getY()) & 0xfffe));
+			out[2] = (unsigned short) (((unsigned short)(v.getZ()) & 0xfffe));
+		}
+
+
+#ifdef DEBUG_CHECK_DEQUANTIZATION
+		btVector3 newPoint = unQuantize(out);
+		if (isMax)
+		{
+			if (newPoint.getX() < point.getX())
+			{
+				printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX());
+			}
+			if (newPoint.getY() < point.getY())
+			{
+				printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY());
+			}
+			if (newPoint.getZ() < point.getZ())
+			{
+
+				printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ());
+			}
+		} else
+		{
+			if (newPoint.getX() > point.getX())
+			{
+				printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX());
+			}
+			if (newPoint.getY() > point.getY())
+			{
+				printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY());
+			}
+			if (newPoint.getZ() > point.getZ())
+			{
+				printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ());
+			}
+		}
+#endif //DEBUG_CHECK_DEQUANTIZATION
+
+	}
+
+
+	SIMD_FORCE_INLINE void quantizeWithClamp(unsigned short* out, const btVector3& point2,int isMax) const
+	{
+
+		btAssert(m_useQuantization);
+
+		btVector3 clampedPoint(point2);
+		clampedPoint.setMax(m_bvhAabbMin);
+		clampedPoint.setMin(m_bvhAabbMax);
+
+		quantize(out,clampedPoint,isMax);
+
+	}
+	
+	SIMD_FORCE_INLINE btVector3	unQuantize(const unsigned short* vecIn) const
+	{
+			btVector3	vecOut;
+			vecOut.setValue(
+			(btScalar)(vecIn[0]) / (m_bvhQuantization.getX()),
+			(btScalar)(vecIn[1]) / (m_bvhQuantization.getY()),
+			(btScalar)(vecIn[2]) / (m_bvhQuantization.getZ()));
+			vecOut += m_bvhAabbMin;
+			return vecOut;
+	}
+
+	///setTraversalMode let's you choose between stackless, recursive or stackless cache friendly tree traversal. Note this is only implemented for quantized trees.
+	void	setTraversalMode(btTraversalMode	traversalMode)
+	{
+		m_traversalMode = traversalMode;
+	}
+
+
+	SIMD_FORCE_INLINE QuantizedNodeArray&	getQuantizedNodeArray()
+	{	
+		return	m_quantizedContiguousNodes;
+	}
+
+
+	SIMD_FORCE_INLINE BvhSubtreeInfoArray&	getSubtreeInfoArray()
+	{
+		return m_SubtreeHeaders;
+	}
+
+
+	/////Calculate space needed to store BVH for serialization
+	unsigned calculateSerializeBufferSize();
+
+	/// Data buffer MUST be 16 byte aligned
+	virtual bool serialize(void *o_alignedDataBuffer, unsigned i_dataBufferSize, bool i_swapEndian);
+
+	///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place'
+	static btQuantizedBvh *deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian);
+
+	static unsigned int getAlignmentSerializationPadding();
+
+	SIMD_FORCE_INLINE bool isQuantized()
+	{
+		return m_useQuantization;
+	}
+
+private:
+	// Special "copy" constructor that allows for in-place deserialization
+	// Prevents btVector3's default constructor from being called, but doesn't inialize much else
+	// 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)
+	btQuantizedBvh(btQuantizedBvh &other, bool ownsMemory);
+
+}
+;
+
+
+#endif //QUANTIZED_BVH_H

extern/bullet2/src/BulletCollision/BroadphaseCollision/btSimpleBroadphase.cpp

@@ -14,83 +14,86 @@ subject to the following restrictions:
 */
 
 #include "btSimpleBroadphase.h"
-#include <BulletCollision/BroadphaseCollision/btDispatcher.h>
+#include "BulletCollision/BroadphaseCollision/btDispatcher.h"
-#include <BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h>
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
 
 #include "LinearMath/btVector3.h"
 #include "LinearMath/btTransform.h"
 #include "LinearMath/btMatrix3x3.h"
 #include <new>
 
+extern int gOverlappingPairs;
 
 void	btSimpleBroadphase::validate()
 {
-	for (int i=0;i<m_numProxies;i++)
+	for (int i=0;i<m_numHandles;i++)
 	{
-		for (int j=i+1;j<m_numProxies;j++)
+		for (int j=i+1;j<m_numHandles;j++)
 		{
-			assert(m_pProxies[i] != m_pProxies[j]);
+			btAssert(&m_pHandles[i] != &m_pHandles[j]);
 		}
 	}
 	
 }
 
-btSimpleBroadphase::btSimpleBroadphase(int maxProxies)
+btSimpleBroadphase::btSimpleBroadphase(int maxProxies, btOverlappingPairCache* overlappingPairCache)
-	:btOverlappingPairCache(),
+	:m_pairCache(overlappingPairCache),
-	m_firstFreeProxy(0),
+	m_ownsPairCache(false),
-	m_numProxies(0),
+	m_invalidPair(0)
-	m_maxProxies(maxProxies)
 {
 
-	m_proxies = new btSimpleBroadphaseProxy[maxProxies];
+	if (!overlappingPairCache)
-	m_freeProxies = new int[maxProxies];
-	m_pProxies = new btSimpleBroadphaseProxy*[maxProxies];
-	
-
-	int i;
-	for (i=0;i<m_maxProxies;i++)
 	{
-		m_freeProxies[i] = i;
+		void* mem = btAlignedAlloc(sizeof(btHashedOverlappingPairCache),16);
+		m_pairCache = new (mem)btHashedOverlappingPairCache();
+		m_ownsPairCache = true;
 	}
+
+	// allocate handles buffer and put all handles on free list
+	m_pHandlesRawPtr = btAlignedAlloc(sizeof(btSimpleBroadphaseProxy)*maxProxies,16);
+	m_pHandles = new(m_pHandlesRawPtr) btSimpleBroadphaseProxy[maxProxies];
+	m_maxHandles = maxProxies;
+	m_numHandles = 0;
+	m_firstFreeHandle = 0;
+	m_firstAllocatedHandle = -1;
+
+	{
+		for (int i = m_firstFreeHandle; i < maxProxies; i++)
+		{
+			m_pHandles[i].SetNextFree(i + 1);
+			m_pHandles[i].m_uniqueId = i+2;//any UID will do, we just avoid too trivial values (0,1) for debugging purposes
+			m_pHandles[i].SetNextAllocated(-1);
+		}
+		m_pHandles[maxProxies - 1].SetNextFree(0);
+		m_pHandles[maxProxies - 1].SetNextAllocated(-1);
+	
+	}
+
 }
 
 btSimpleBroadphase::~btSimpleBroadphase()
 {
-	delete[] m_proxies;
+	btAlignedFree(m_pHandlesRawPtr);
-	delete []m_freeProxies;
-	delete [] m_pProxies;
 
-	/*int i;
+	if (m_ownsPairCache)
-	for (i=m_numProxies-1;i>=0;i--)
 	{
-		BP_Proxy* proxy = m_pProxies[i]; 
+		m_pairCache->~btOverlappingPairCache();
-		destroyProxy(proxy);
+		btAlignedFree(m_pairCache);
 	}
-	*/
 }
 
 
-btBroadphaseProxy*	btSimpleBroadphase::createProxy(  const btVector3& min,  const btVector3& max,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask)
+btBroadphaseProxy*	btSimpleBroadphase::createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* /*dispatcher*/,void* multiSapProxy)
 {
-	if (m_numProxies >= m_maxProxies)
+	if (m_numHandles >= m_maxHandles)
 	{
-		assert(0);
+		btAssert(0);
 		return 0; //should never happen, but don't let the game crash ;-)
 	}
-	assert(min[0]<= max[0] && min[1]<= max[1] && min[2]<= max[2]);
+	assert(aabbMin[0]<= aabbMax[0] && aabbMin[1]<= aabbMax[1] && aabbMin[2]<= aabbMax[2]);
 
-	int freeIndex= m_freeProxies[m_firstFreeProxy];
+	int newHandleIndex = allocHandle();
-	btSimpleBroadphaseProxy* proxy = new (&m_proxies[freeIndex])btSimpleBroadphaseProxy(min,max,shapeType,userPtr,collisionFilterGroup,collisionFilterMask);
+	btSimpleBroadphaseProxy* proxy = new (&m_pHandles[newHandleIndex])btSimpleBroadphaseProxy(aabbMin,aabbMax,shapeType,userPtr,collisionFilterGroup,collisionFilterMask,multiSapProxy);
-	m_firstFreeProxy++;
-
-	btSimpleBroadphaseProxy* proxy1 = &m_proxies[0];
-		
-	int	index = int(proxy - proxy1);
-	btAssert(index == freeIndex);
-
-	m_pProxies[m_numProxies] = proxy;
-	m_numProxies++;
-	//validate();
 
 	return proxy;
 }
@@ -124,34 +127,19 @@ protected:
 	};
 };
 
-void	btSimpleBroadphase::destroyProxy(btBroadphaseProxy* proxyOrg)
+void	btSimpleBroadphase::destroyProxy(btBroadphaseProxy* proxyOrg,btDispatcher* dispatcher)
 {
 		
-		int i;
-		
 		btSimpleBroadphaseProxy* proxy0 = static_cast<btSimpleBroadphaseProxy*>(proxyOrg);
-		btSimpleBroadphaseProxy* proxy1 = &m_proxies[0];
+		freeHandle(proxy0);
-	
+
-		int index = int(proxy0 - proxy1);
+		m_pairCache->removeOverlappingPairsContainingProxy(proxyOrg,dispatcher);
-		btAssert (index < m_maxProxies);
-		m_freeProxies[--m_firstFreeProxy] = index;
 
-		removeOverlappingPairsContainingProxy(proxyOrg);
-		
-		for (i=0;i<m_numProxies;i++)
-		{
-			if (m_pProxies[i] == proxyOrg)
-			{
-				m_pProxies[i] = m_pProxies[m_numProxies-1];
-				break;
-			}
-		}
-		m_numProxies--;
 		//validate();
 		
 }
 
-void	btSimpleBroadphase::setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax)
+void	btSimpleBroadphase::setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax, btDispatcher* /*dispatcher*/)
 {
 	btSimpleBroadphaseProxy* sbp = getSimpleProxyFromProxy(proxy);
 	sbp->m_min = aabbMin;
@@ -186,37 +174,137 @@ public:
 	}
 };
 
-void	btSimpleBroadphase::refreshOverlappingPairs()
+void	btSimpleBroadphase::calculateOverlappingPairs(btDispatcher* dispatcher)
 {
 	//first check for new overlapping pairs
 	int i,j;
 
-	for (i=0;i<m_numProxies;i++)
+	if (m_firstAllocatedHandle >= 0)
 	{
-		btBroadphaseProxy* proxy0 = m_pProxies[i];
-		for (j=i+1;j<m_numProxies;j++)
-		{
-			btBroadphaseProxy* proxy1 = m_pProxies[j];
-			btSimpleBroadphaseProxy* p0 = getSimpleProxyFromProxy(proxy0);
-			btSimpleBroadphaseProxy* p1 = getSimpleProxyFromProxy(proxy1);
 
-			if (aabbOverlap(p0,p1))
+		btSimpleBroadphaseProxy* proxy0 = &m_pHandles[m_firstAllocatedHandle];
+
+		for (i=0;i<m_numHandles;i++)
+		{
+			btSimpleBroadphaseProxy* proxy1 = &m_pHandles[m_firstAllocatedHandle];
+
+			for (j=0;j<m_numHandles;j++)
 			{
-				if ( !findPair(proxy0,proxy1))
+				
+				if (proxy0 != proxy1)
 				{
-					addOverlappingPair(proxy0,proxy1);
+					btSimpleBroadphaseProxy* p0 = getSimpleProxyFromProxy(proxy0);
+					btSimpleBroadphaseProxy* p1 = getSimpleProxyFromProxy(proxy1);
+
+					if (aabbOverlap(p0,p1))
+					{
+						if ( !m_pairCache->findPair(proxy0,proxy1))
+						{
+							m_pairCache->addOverlappingPair(proxy0,proxy1);
+						}
+					} else
+					{
+					if (!m_pairCache->hasDeferredRemoval())
+					{
+						if ( m_pairCache->findPair(proxy0,proxy1))
+						{
+							m_pairCache->removeOverlappingPair(proxy0,proxy1,dispatcher);
+						}
+					}
+
+					}
 				}
+				proxy1 = &m_pHandles[proxy1->GetNextAllocated()];
+
 			}
+			proxy0 = &m_pHandles[proxy0->GetNextAllocated()];
+
+		}
+
+		if (m_ownsPairCache && m_pairCache->hasDeferredRemoval())
+		{
+			
+			btBroadphasePairArray&	overlappingPairArray = m_pairCache->getOverlappingPairArray();
+
+			//perform a sort, to find duplicates and to sort 'invalid' pairs to the end
+			overlappingPairArray.quickSort(btBroadphasePairSortPredicate());
+
+			overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+			m_invalidPair = 0;
+
+
+			btBroadphasePair previousPair;
+			previousPair.m_pProxy0 = 0;
+			previousPair.m_pProxy1 = 0;
+			previousPair.m_algorithm = 0;
+			
+			
+			for (i=0;i<overlappingPairArray.size();i++)
+			{
+			
+				btBroadphasePair& pair = overlappingPairArray[i];
+
+				bool isDuplicate = (pair == previousPair);
+
+				previousPair = pair;
+
+				bool needsRemoval = false;
+
+				if (!isDuplicate)
+				{
+					bool hasOverlap = testAabbOverlap(pair.m_pProxy0,pair.m_pProxy1);
+
+					if (hasOverlap)
+					{
+						needsRemoval = false;//callback->processOverlap(pair);
+					} else
+					{
+						needsRemoval = true;
+					}
+				} else
+				{
+					//remove duplicate
+					needsRemoval = true;
+					//should have no algorithm
+					btAssert(!pair.m_algorithm);
+				}
+				
+				if (needsRemoval)
+				{
+					m_pairCache->cleanOverlappingPair(pair,dispatcher);
+
+			//		m_overlappingPairArray.swap(i,m_overlappingPairArray.size()-1);
+			//		m_overlappingPairArray.pop_back();
+					pair.m_pProxy0 = 0;
+					pair.m_pProxy1 = 0;
+					m_invalidPair++;
+					gOverlappingPairs--;
+				} 
+				
+			}
+
+		///if you don't like to skip the invalid pairs in the array, execute following code:
+		#define CLEAN_INVALID_PAIRS 1
+		#ifdef CLEAN_INVALID_PAIRS
+
+			//perform a sort, to sort 'invalid' pairs to the end
+			overlappingPairArray.quickSort(btBroadphasePairSortPredicate());
+
+			overlappingPairArray.resize(overlappingPairArray.size() - m_invalidPair);
+			m_invalidPair = 0;
+		#endif//CLEAN_INVALID_PAIRS
 
 		}
 	}
-
-
-	CheckOverlapCallback	checkOverlap;
-
-	processAllOverlappingPairs(&checkOverlap);
-
-
 }
 
 
+bool btSimpleBroadphase::testAabbOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1)
+{
+	btSimpleBroadphaseProxy* p0 = getSimpleProxyFromProxy(proxy0);
+	btSimpleBroadphaseProxy* p1 = getSimpleProxyFromProxy(proxy1);
+	return aabbOverlap(p0,p1);
+}
+
+
+

extern/bullet2/src/BulletCollision/BroadphaseCollision/btSimpleBroadphase.h

@@ -24,35 +24,78 @@ struct btSimpleBroadphaseProxy : public btBroadphaseProxy
 {
 	btVector3	m_min;
 	btVector3	m_max;
+	int			m_nextFree;
+	int			m_nextAllocated;
+//	int			m_handleId;
+
 	
 	btSimpleBroadphaseProxy() {};
 
-	btSimpleBroadphaseProxy(const btPoint3& minpt,const btPoint3& maxpt,int shapeType,void* userPtr,short int collisionFilterGroup,short int collisionFilterMask)
+	btSimpleBroadphaseProxy(const btPoint3& minpt,const btPoint3& maxpt,int shapeType,void* userPtr,short int collisionFilterGroup,short int collisionFilterMask,void* multiSapProxy)
-	:btBroadphaseProxy(userPtr,collisionFilterGroup,collisionFilterMask),
+	:btBroadphaseProxy(userPtr,collisionFilterGroup,collisionFilterMask,multiSapProxy),
 	m_min(minpt),m_max(maxpt)		
 	{
 		(void)shapeType;
 	}
 	
+	
+	SIMD_FORCE_INLINE void SetNextFree(int next) {m_nextFree = next;}
+	SIMD_FORCE_INLINE int GetNextFree() const {return m_nextFree;}
+
+	SIMD_FORCE_INLINE void SetNextAllocated(int next) {m_nextAllocated = next;}
+	SIMD_FORCE_INLINE int GetNextAllocated() const {return m_nextAllocated;}
+
 
 };
 
-///SimpleBroadphase is a brute force aabb culling broadphase based on O(n^2) aabb checks
+///The SimpleBroadphase is just a unit-test for btAxisSweep3, bt32BitAxisSweep3, or btDbvtBroadphase, so use those classes instead.
-class btSimpleBroadphase : public btOverlappingPairCache
+///It is a brute force aabb culling broadphase based on O(n^2) aabb checks
+class btSimpleBroadphase : public btBroadphaseInterface
 {
 
 protected:
 
-	btSimpleBroadphaseProxy*	m_proxies;
+	int		m_numHandles;						// number of active handles
-	int*				m_freeProxies;
+	int		m_maxHandles;						// max number of handles
-	int				m_firstFreeProxy;
+	btSimpleBroadphaseProxy* m_pHandles;						// handles pool
+	void* m_pHandlesRawPtr;
+	int		m_firstFreeHandle;		// free handles list
+	int		m_firstAllocatedHandle;
 
-	btSimpleBroadphaseProxy** m_pProxies;
+	int allocHandle()
-	int				m_numProxies;
+	{
 
-	
+		int freeHandle = m_firstFreeHandle;
+		m_firstFreeHandle = m_pHandles[freeHandle].GetNextFree();
+		
+		m_pHandles[freeHandle].SetNextAllocated(m_firstAllocatedHandle);
+		m_firstAllocatedHandle = freeHandle;
+		
+		m_numHandles++;
+
+		return freeHandle;
+	}
+
+	void freeHandle(btSimpleBroadphaseProxy* proxy)
+	{
+		int handle = int(proxy-m_pHandles);
+		btAssert(handle >= 0 && handle < m_maxHandles);
+
+		proxy->SetNextFree(m_firstFreeHandle);
+		m_firstFreeHandle = handle;
+
+		m_firstAllocatedHandle = proxy->GetNextAllocated();
+		proxy->SetNextAllocated(-1);
+
+		m_numHandles--;
+	}
+
+
+	btOverlappingPairCache*	m_pairCache;
+	bool	m_ownsPairCache;
+
+	int	m_invalidPair;
 
-	int m_maxProxies;
 	
 	
 	inline btSimpleBroadphaseProxy*	getSimpleProxyFromProxy(btBroadphaseProxy* proxy)
@@ -67,26 +110,48 @@ protected:
 protected:
 
 
-	virtual void	refreshOverlappingPairs();
+	
+
 public:
-	btSimpleBroadphase(int maxProxies=16384);
+	btSimpleBroadphase(int maxProxies=16384,btOverlappingPairCache* overlappingPairCache=0);
 	virtual ~btSimpleBroadphase();
 
 
 		static bool	aabbOverlap(btSimpleBroadphaseProxy* proxy0,btSimpleBroadphaseProxy* proxy1);
 
 
-	virtual btBroadphaseProxy*	createProxy(  const btVector3& min,  const btVector3& max,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
+	virtual btBroadphaseProxy*	createProxy(  const btVector3& aabbMin,  const btVector3& aabbMax,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* multiSapProxy);
 
+	virtual void	calculateOverlappingPairs(btDispatcher* dispatcher);
 
-	virtual void	destroyProxy(btBroadphaseProxy* proxy);
+	virtual void	destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
-	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax);
+	virtual void	setAabb(btBroadphaseProxy* proxy,const btVector3& aabbMin,const btVector3& aabbMax, btDispatcher* dispatcher);
 		
-	
+	btOverlappingPairCache*	getOverlappingPairCache()
-	
+	{
+		return m_pairCache;
+	}
+	const btOverlappingPairCache*	getOverlappingPairCache() const
+	{
+		return m_pairCache;
+	}
+
+	bool	testAabbOverlap(btBroadphaseProxy* proxy0,btBroadphaseProxy* proxy1);
 
 
+	///getAabb returns the axis aligned bounding box in the 'global' coordinate frame
+	///will add some transform later
+	virtual void getBroadphaseAabb(btVector3& aabbMin,btVector3& aabbMax) const
+	{
+		aabbMin.setValue(-1e30f,-1e30f,-1e30f);
+		aabbMax.setValue(1e30f,1e30f,1e30f);
+	}
 
+	virtual void	printStats()
+	{
+//		printf("btSimpleBroadphase.h\n");
+//		printf("numHandles = %d, maxHandles = %d\n",m_numHandles,m_maxHandles);
+	}
 };
 
 

extern/bullet2/src/BulletCollision/CollisionDispatch/SphereTriangleDetector.cpp

@@ -26,7 +26,7 @@ m_triangle(triangle)
 
 }
 
-void	SphereTriangleDetector::getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* debugDraw)
+void	SphereTriangleDetector::getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* debugDraw,bool swapResults)
 {
 
 	(void)debugDraw;
@@ -42,7 +42,16 @@ void	SphereTriangleDetector::getClosestPoints(const ClosestPointInput& input,Res
 
 	if (collide(sphereInTr.getOrigin(),point,normal,depth,timeOfImpact))
 	{
-		output.addContactPoint(transformB.getBasis()*normal,transformB*point,depth);
+		if (swapResults)
+		{
+			btVector3 normalOnB = transformB.getBasis()*normal;
+			btVector3 normalOnA = -normalOnB;
+			btVector3 pointOnA = transformB*point+normalOnB*depth;
+			output.addContactPoint(normalOnA,pointOnA,depth);
+		} else
+		{
+			output.addContactPoint(transformB.getBasis()*normal,transformB*point,depth);
+		}
 	}
 
 }
@@ -53,6 +62,8 @@ void	SphereTriangleDetector::getClosestPoints(const ClosestPointInput& input,Res
 
 // See also geometrictools.com
 // Basic idea: D = |p - (lo + t0*lv)| where t0 = lv . (p - lo) / lv . lv
+btScalar SegmentSqrDistance(const btVector3& from, const btVector3& to,const btVector3 &p, btVector3 &nearest);
+
 btScalar SegmentSqrDistance(const btVector3& from, const btVector3& to,const btVector3 &p, btVector3 &nearest) {
 	btVector3 diff = p - from;
 	btVector3 v = to - from;

extern/bullet2/src/BulletCollision/CollisionDispatch/SphereTriangleDetector.h

@@ -16,8 +16,8 @@ subject to the following restrictions:
 #ifndef SPHERE_TRIANGLE_DETECTOR_H
 #define SPHERE_TRIANGLE_DETECTOR_H
 
-#include "../NarrowPhaseCollision/btDiscreteCollisionDetectorInterface.h"
+#include "BulletCollision/NarrowPhaseCollision/btDiscreteCollisionDetectorInterface.h"
-#include "../../LinearMath/btPoint3.h"
+#include "LinearMath/btPoint3.h"
 
 
 class btSphereShape;
@@ -28,7 +28,7 @@ class btTriangleShape;
 /// sphere-triangle to match the btDiscreteCollisionDetectorInterface
 struct SphereTriangleDetector : public btDiscreteCollisionDetectorInterface
 {
-	virtual void	getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* debugDraw);
+	virtual void	getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* debugDraw,bool swapResults=false);
 
 	SphereTriangleDetector(btSphereShape* sphere,btTriangleShape* triangle);
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btBoxBoxCollisionAlgorithm.cpp

@@ -0,0 +1,85 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "btBoxBoxCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btCollisionDispatcher.h"
+#include "BulletCollision/CollisionShapes/btBoxShape.h"
+#include "BulletCollision/CollisionDispatch/btCollisionObject.h"
+#include "btBoxBoxDetector.h"
+
+#define USE_PERSISTENT_CONTACTS 1
+
+btBoxBoxCollisionAlgorithm::btBoxBoxCollisionAlgorithm(btPersistentManifold* mf,const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* obj0,btCollisionObject* obj1)
+: btCollisionAlgorithm(ci),
+m_ownManifold(false),
+m_manifoldPtr(mf)
+{
+	if (!m_manifoldPtr && m_dispatcher->needsCollision(obj0,obj1))
+	{
+		m_manifoldPtr = m_dispatcher->getNewManifold(obj0,obj1);
+		m_ownManifold = true;
+	}
+}
+
+btBoxBoxCollisionAlgorithm::~btBoxBoxCollisionAlgorithm()
+{
+	if (m_ownManifold)
+	{
+		if (m_manifoldPtr)
+			m_dispatcher->releaseManifold(m_manifoldPtr);
+	}
+}
+
+void btBoxBoxCollisionAlgorithm::processCollision (btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut)
+{
+	if (!m_manifoldPtr)
+		return;
+
+	btCollisionObject*	col0 = body0;
+	btCollisionObject*	col1 = body1;
+	btBoxShape* box0 = (btBoxShape*)col0->getCollisionShape();
+	btBoxShape* box1 = (btBoxShape*)col1->getCollisionShape();
+
+
+
+	/// report a contact. internally this will be kept persistent, and contact reduction is done
+	resultOut->setPersistentManifold(m_manifoldPtr);
+#ifndef USE_PERSISTENT_CONTACTS	
+	m_manifoldPtr->clearManifold();
+#endif //USE_PERSISTENT_CONTACTS
+
+	btDiscreteCollisionDetectorInterface::ClosestPointInput input;
+	input.m_maximumDistanceSquared = 1e30f;
+	input.m_transformA = body0->getWorldTransform();
+	input.m_transformB = body1->getWorldTransform();
+
+	btBoxBoxDetector detector(box0,box1);
+	detector.getClosestPoints(input,*resultOut,dispatchInfo.m_debugDraw);
+
+#ifdef USE_PERSISTENT_CONTACTS
+	//  refreshContactPoints is only necessary when using persistent contact points. otherwise all points are newly added
+	if (m_ownManifold)
+	{
+		resultOut->refreshContactPoints();
+	}
+#endif //USE_PERSISTENT_CONTACTS
+
+}
+
+btScalar btBoxBoxCollisionAlgorithm::calculateTimeOfImpact(btCollisionObject* /*body0*/,btCollisionObject* /*body1*/,const btDispatcherInfo& /*dispatchInfo*/,btManifoldResult* /*resultOut*/)
+{
+	//not yet
+	return 1.f;
+}

extern/bullet2/src/BulletCollision/CollisionDispatch/btBoxBoxCollisionAlgorithm.h

@@ -0,0 +1,66 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef BOX_BOX__COLLISION_ALGORITHM_H
+#define BOX_BOX__COLLISION_ALGORITHM_H
+
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btDispatcher.h"
+#include "BulletCollision/CollisionDispatch/btCollisionCreateFunc.h"
+
+class btPersistentManifold;
+
+///box-box collision detection
+class btBoxBoxCollisionAlgorithm : public btCollisionAlgorithm
+{
+	bool	m_ownManifold;
+	btPersistentManifold*	m_manifoldPtr;
+	
+public:
+	btBoxBoxCollisionAlgorithm(const btCollisionAlgorithmConstructionInfo& ci)
+		: btCollisionAlgorithm(ci) {}
+
+	virtual void processCollision (btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
+
+	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
+
+	btBoxBoxCollisionAlgorithm(btPersistentManifold* mf,const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* body0,btCollisionObject* body1);
+
+	virtual ~btBoxBoxCollisionAlgorithm();
+
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		if (m_manifoldPtr && m_ownManifold)
+		{
+			manifoldArray.push_back(m_manifoldPtr);
+		}
+	}
+
+
+	struct CreateFunc :public 	btCollisionAlgorithmCreateFunc
+	{
+		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
+		{
+			int bbsize = sizeof(btBoxBoxCollisionAlgorithm);
+			void* ptr = ci.m_dispatcher1->allocateCollisionAlgorithm(bbsize);
+			return new(ptr) btBoxBoxCollisionAlgorithm(0,ci,body0,body1);
+		}
+	};
+
+};
+
+#endif //BOX_BOX__COLLISION_ALGORITHM_H
+

extern/bullet2/src/BulletCollision/CollisionDispatch/btBoxBoxDetector.cpp

@@ -0,0 +1,683 @@
+
+/*
+ * Box-Box collision detection re-distributed under the ZLib license with permission from Russell L. Smith
+ * Original version is from Open Dynamics Engine, Copyright (C) 2001,2002 Russell L. Smith.
+ * All rights reserved.  Email: russ@q12.org   Web: www.q12.org
+ Bullet Continuous Collision Detection and Physics Library
+ Bullet is Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+///ODE box-box collision detection is adapted to work with Bullet
+
+#include "btBoxBoxDetector.h"
+#include "BulletCollision/CollisionShapes/btBoxShape.h"
+
+#include <float.h>
+#include <string.h>
+
+btBoxBoxDetector::btBoxBoxDetector(btBoxShape* box1,btBoxShape* box2)
+: m_box1(box1),
+m_box2(box2)
+{
+
+}
+
+
+// given two boxes (p1,R1,side1) and (p2,R2,side2), collide them together and
+// generate contact points. this returns 0 if there is no contact otherwise
+// it returns the number of contacts generated.
+// `normal' returns the contact normal.
+// `depth' returns the maximum penetration depth along that normal.
+// `return_code' returns a number indicating the type of contact that was
+// detected:
+//        1,2,3 = box 2 intersects with a face of box 1
+//        4,5,6 = box 1 intersects with a face of box 2
+//        7..15 = edge-edge contact
+// `maxc' is the maximum number of contacts allowed to be generated, i.e.
+// the size of the `contact' array.
+// `contact' and `skip' are the contact array information provided to the
+// collision functions. this function only fills in the position and depth
+// fields.
+struct dContactGeom;
+#define dDOTpq(a,b,p,q) ((a)[0]*(b)[0] + (a)[p]*(b)[q] + (a)[2*(p)]*(b)[2*(q)])
+#define dInfinity FLT_MAX
+
+
+/*PURE_INLINE btScalar dDOT   (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,1,1); }
+PURE_INLINE btScalar dDOT13 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,1,3); }
+PURE_INLINE btScalar dDOT31 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,3,1); }
+PURE_INLINE btScalar dDOT33 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,3,3); }
+*/
+static btScalar dDOT   (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,1,1); }
+static btScalar dDOT44 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,4,4); }
+static btScalar dDOT41 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,4,1); }
+static btScalar dDOT14 (const btScalar *a, const btScalar *b) { return dDOTpq(a,b,1,4); }
+#define dMULTIPLYOP1_331(A,op,B,C) \
+{\
+  (A)[0] op dDOT41((B),(C)); \
+  (A)[1] op dDOT41((B+1),(C)); \
+  (A)[2] op dDOT41((B+2),(C)); \
+}
+
+#define dMULTIPLYOP0_331(A,op,B,C) \
+{ \
+  (A)[0] op dDOT((B),(C)); \
+  (A)[1] op dDOT((B+4),(C)); \
+  (A)[2] op dDOT((B+8),(C)); \
+} 
+
+#define dMULTIPLY1_331(A,B,C) dMULTIPLYOP1_331(A,=,B,C)
+#define dMULTIPLY0_331(A,B,C) dMULTIPLYOP0_331(A,=,B,C)
+
+typedef btScalar dMatrix3[4*3];
+
+void dLineClosestApproach (const btVector3& pa, const btVector3& ua,
+			   const btVector3& pb, const btVector3& ub,
+			   btScalar *alpha, btScalar *beta);
+void dLineClosestApproach (const btVector3& pa, const btVector3& ua,
+			   const btVector3& pb, const btVector3& ub,
+			   btScalar *alpha, btScalar *beta)
+{
+  btVector3 p;
+  p[0] = pb[0] - pa[0];
+  p[1] = pb[1] - pa[1];
+  p[2] = pb[2] - pa[2];
+  btScalar uaub = dDOT(ua,ub);
+  btScalar q1 =  dDOT(ua,p);
+  btScalar q2 = -dDOT(ub,p);
+  btScalar d = 1-uaub*uaub;
+  if (d <= btScalar(0.0001f)) {
+    // @@@ this needs to be made more robust
+    *alpha = 0;
+    *beta  = 0;
+  }
+  else {
+    d = 1.f/d;
+    *alpha = (q1 + uaub*q2)*d;
+    *beta  = (uaub*q1 + q2)*d;
+  }
+}
+
+
+
+// find all the intersection points between the 2D rectangle with vertices
+// at (+/-h[0],+/-h[1]) and the 2D quadrilateral with vertices (p[0],p[1]),
+// (p[2],p[3]),(p[4],p[5]),(p[6],p[7]).
+//
+// the intersection points are returned as x,y pairs in the 'ret' array.
+// the number of intersection points is returned by the function (this will
+// be in the range 0 to 8).
+
+static int intersectRectQuad2 (btScalar h[2], btScalar p[8], btScalar ret[16])
+{
+  // q (and r) contain nq (and nr) coordinate points for the current (and
+  // chopped) polygons
+  int nq=4,nr=0;
+  btScalar buffer[16];
+  btScalar *q = p;
+  btScalar *r = ret;
+  for (int dir=0; dir <= 1; dir++) {
+    // direction notation: xy[0] = x axis, xy[1] = y axis
+    for (int sign=-1; sign <= 1; sign += 2) {
+      // chop q along the line xy[dir] = sign*h[dir]
+      btScalar *pq = q;
+      btScalar *pr = r;
+      nr = 0;
+      for (int i=nq; i > 0; i--) {
+	// go through all points in q and all lines between adjacent points
+	if (sign*pq[dir] < h[dir]) {
+	  // this point is inside the chopping line
+	  pr[0] = pq[0];
+	  pr[1] = pq[1];
+	  pr += 2;
+	  nr++;
+	  if (nr & 8) {
+	    q = r;
+	    goto done;
+	  }
+	}
+	btScalar *nextq = (i > 1) ? pq+2 : q;
+	if ((sign*pq[dir] < h[dir]) ^ (sign*nextq[dir] < h[dir])) {
+	  // this line crosses the chopping line
+	  pr[1-dir] = pq[1-dir] + (nextq[1-dir]-pq[1-dir]) /
+	    (nextq[dir]-pq[dir]) * (sign*h[dir]-pq[dir]);
+	  pr[dir] = sign*h[dir];
+	  pr += 2;
+	  nr++;
+	  if (nr & 8) {
+	    q = r;
+	    goto done;
+	  }
+	}
+	pq += 2;
+      }
+      q = r;
+      r = (q==ret) ? buffer : ret;
+      nq = nr;
+    }
+  }
+ done:
+  if (q != ret) memcpy (ret,q,nr*2*sizeof(btScalar));
+  return nr;
+}
+
+
+#define M__PI 3.14159265f
+
+// given n points in the plane (array p, of size 2*n), generate m points that
+// best represent the whole set. the definition of 'best' here is not
+// predetermined - the idea is to select points that give good box-box
+// collision detection behavior. the chosen point indexes are returned in the
+// array iret (of size m). 'i0' is always the first entry in the array.
+// n must be in the range [1..8]. m must be in the range [1..n]. i0 must be
+// in the range [0..n-1].
+
+void cullPoints2 (int n, btScalar p[], int m, int i0, int iret[]);
+void cullPoints2 (int n, btScalar p[], int m, int i0, int iret[])
+{
+  // compute the centroid of the polygon in cx,cy
+  int i,j;
+  btScalar a,cx,cy,q;
+  if (n==1) {
+    cx = p[0];
+    cy = p[1];
+  }
+  else if (n==2) {
+    cx = btScalar(0.5)*(p[0] + p[2]);
+    cy = btScalar(0.5)*(p[1] + p[3]);
+  }
+  else {
+    a = 0;
+    cx = 0;
+    cy = 0;
+    for (i=0; i<(n-1); i++) {
+      q = p[i*2]*p[i*2+3] - p[i*2+2]*p[i*2+1];
+      a += q;
+      cx += q*(p[i*2]+p[i*2+2]);
+      cy += q*(p[i*2+1]+p[i*2+3]);
+    }
+    q = p[n*2-2]*p[1] - p[0]*p[n*2-1];
+    a = 1.f/(btScalar(3.0)*(a+q));
+    cx = a*(cx + q*(p[n*2-2]+p[0]));
+    cy = a*(cy + q*(p[n*2-1]+p[1]));
+  }
+
+  // compute the angle of each point w.r.t. the centroid
+  btScalar A[8];
+  for (i=0; i<n; i++) A[i] = btAtan2(p[i*2+1]-cy,p[i*2]-cx);
+
+  // search for points that have angles closest to A[i0] + i*(2*pi/m).
+  int avail[8];
+  for (i=0; i<n; i++) avail[i] = 1;
+  avail[i0] = 0;
+  iret[0] = i0;
+  iret++;
+  for (j=1; j<m; j++) {
+    a = btScalar(j)*(2*M__PI/m) + A[i0];
+    if (a > M__PI) a -= 2*M__PI;
+    btScalar maxdiff=1e9,diff;
+#ifndef dNODEBUG
+    *iret = i0;			// iret is not allowed to keep this value
+#endif
+    for (i=0; i<n; i++) {
+      if (avail[i]) {
+	diff = btFabs (A[i]-a);
+	if (diff > M__PI) diff = 2*M__PI - diff;
+	if (diff < maxdiff) {
+	  maxdiff = diff;
+	  *iret = i;
+	}
+      }
+    }
+#ifndef dNODEBUG
+    btAssert (*iret != i0);	// ensure iret got set
+#endif
+    avail[*iret] = 0;
+    iret++;
+  }
+}
+
+
+
+int dBoxBox2 (const btVector3& p1, const dMatrix3 R1,
+	     const btVector3& side1, const btVector3& p2,
+	     const dMatrix3 R2, const btVector3& side2,
+	     btVector3& normal, btScalar *depth, int *return_code,
+		 int maxc, dContactGeom * /*contact*/, int /*skip*/,btDiscreteCollisionDetectorInterface::Result& output);
+int dBoxBox2 (const btVector3& p1, const dMatrix3 R1,
+	     const btVector3& side1, const btVector3& p2,
+	     const dMatrix3 R2, const btVector3& side2,
+	     btVector3& normal, btScalar *depth, int *return_code,
+		 int maxc, dContactGeom * /*contact*/, int /*skip*/,btDiscreteCollisionDetectorInterface::Result& output)
+{
+  const btScalar fudge_factor = btScalar(1.05);
+  btVector3 p,pp,normalC;
+  const btScalar *normalR = 0;
+  btScalar A[3],B[3],R11,R12,R13,R21,R22,R23,R31,R32,R33,
+    Q11,Q12,Q13,Q21,Q22,Q23,Q31,Q32,Q33,s,s2,l;
+  int i,j,invert_normal,code;
+
+  // get vector from centers of box 1 to box 2, relative to box 1
+  p = p2 - p1;
+  dMULTIPLY1_331 (pp,R1,p);		// get pp = p relative to body 1
+
+  // get side lengths / 2
+  A[0] = side1[0]*btScalar(0.5);
+  A[1] = side1[1]*btScalar(0.5);
+  A[2] = side1[2]*btScalar(0.5);
+  B[0] = side2[0]*btScalar(0.5);
+  B[1] = side2[1]*btScalar(0.5);
+  B[2] = side2[2]*btScalar(0.5);
+
+  // Rij is R1'*R2, i.e. the relative rotation between R1 and R2
+  R11 = dDOT44(R1+0,R2+0); R12 = dDOT44(R1+0,R2+1); R13 = dDOT44(R1+0,R2+2);
+  R21 = dDOT44(R1+1,R2+0); R22 = dDOT44(R1+1,R2+1); R23 = dDOT44(R1+1,R2+2);
+  R31 = dDOT44(R1+2,R2+0); R32 = dDOT44(R1+2,R2+1); R33 = dDOT44(R1+2,R2+2);
+
+  Q11 = btFabs(R11); Q12 = btFabs(R12); Q13 = btFabs(R13);
+  Q21 = btFabs(R21); Q22 = btFabs(R22); Q23 = btFabs(R23);
+  Q31 = btFabs(R31); Q32 = btFabs(R32); Q33 = btFabs(R33);
+
+  // for all 15 possible separating axes:
+  //   * see if the axis separates the boxes. if so, return 0.
+  //   * find the depth of the penetration along the separating axis (s2)
+  //   * if this is the largest depth so far, record it.
+  // the normal vector will be set to the separating axis with the smallest
+  // depth. note: normalR is set to point to a column of R1 or R2 if that is
+  // the smallest depth normal so far. otherwise normalR is 0 and normalC is
+  // set to a vector relative to body 1. invert_normal is 1 if the sign of
+  // the normal should be flipped.
+
+#define TST(expr1,expr2,norm,cc) \
+  s2 = btFabs(expr1) - (expr2); \
+  if (s2 > 0) return 0; \
+  if (s2 > s) { \
+    s = s2; \
+    normalR = norm; \
+    invert_normal = ((expr1) < 0); \
+    code = (cc); \
+  }
+
+  s = -dInfinity;
+  invert_normal = 0;
+  code = 0;
+
+  // separating axis = u1,u2,u3
+  TST (pp[0],(A[0] + B[0]*Q11 + B[1]*Q12 + B[2]*Q13),R1+0,1);
+  TST (pp[1],(A[1] + B[0]*Q21 + B[1]*Q22 + B[2]*Q23),R1+1,2);
+  TST (pp[2],(A[2] + B[0]*Q31 + B[1]*Q32 + B[2]*Q33),R1+2,3);
+
+  // separating axis = v1,v2,v3
+  TST (dDOT41(R2+0,p),(A[0]*Q11 + A[1]*Q21 + A[2]*Q31 + B[0]),R2+0,4);
+  TST (dDOT41(R2+1,p),(A[0]*Q12 + A[1]*Q22 + A[2]*Q32 + B[1]),R2+1,5);
+  TST (dDOT41(R2+2,p),(A[0]*Q13 + A[1]*Q23 + A[2]*Q33 + B[2]),R2+2,6);
+
+  // note: cross product axes need to be scaled when s is computed.
+  // normal (n1,n2,n3) is relative to box 1.
+#undef TST
+#define TST(expr1,expr2,n1,n2,n3,cc) \
+  s2 = btFabs(expr1) - (expr2); \
+  if (s2 > 0) return 0; \
+  l = btSqrt((n1)*(n1) + (n2)*(n2) + (n3)*(n3)); \
+  if (l > 0) { \
+    s2 /= l; \
+    if (s2*fudge_factor > s) { \
+      s = s2; \
+      normalR = 0; \
+      normalC[0] = (n1)/l; normalC[1] = (n2)/l; normalC[2] = (n3)/l; \
+      invert_normal = ((expr1) < 0); \
+      code = (cc); \
+    } \
+  }
+
+  // separating axis = u1 x (v1,v2,v3)
+  TST(pp[2]*R21-pp[1]*R31,(A[1]*Q31+A[2]*Q21+B[1]*Q13+B[2]*Q12),0,-R31,R21,7);
+  TST(pp[2]*R22-pp[1]*R32,(A[1]*Q32+A[2]*Q22+B[0]*Q13+B[2]*Q11),0,-R32,R22,8);
+  TST(pp[2]*R23-pp[1]*R33,(A[1]*Q33+A[2]*Q23+B[0]*Q12+B[1]*Q11),0,-R33,R23,9);
+
+  // separating axis = u2 x (v1,v2,v3)
+  TST(pp[0]*R31-pp[2]*R11,(A[0]*Q31+A[2]*Q11+B[1]*Q23+B[2]*Q22),R31,0,-R11,10);
+  TST(pp[0]*R32-pp[2]*R12,(A[0]*Q32+A[2]*Q12+B[0]*Q23+B[2]*Q21),R32,0,-R12,11);
+  TST(pp[0]*R33-pp[2]*R13,(A[0]*Q33+A[2]*Q13+B[0]*Q22+B[1]*Q21),R33,0,-R13,12);
+
+  // separating axis = u3 x (v1,v2,v3)
+  TST(pp[1]*R11-pp[0]*R21,(A[0]*Q21+A[1]*Q11+B[1]*Q33+B[2]*Q32),-R21,R11,0,13);
+  TST(pp[1]*R12-pp[0]*R22,(A[0]*Q22+A[1]*Q12+B[0]*Q33+B[2]*Q31),-R22,R12,0,14);
+  TST(pp[1]*R13-pp[0]*R23,(A[0]*Q23+A[1]*Q13+B[0]*Q32+B[1]*Q31),-R23,R13,0,15);
+
+#undef TST
+
+  if (!code) return 0;
+
+  // if we get to this point, the boxes interpenetrate. compute the normal
+  // in global coordinates.
+  if (normalR) {
+    normal[0] = normalR[0];
+    normal[1] = normalR[4];
+    normal[2] = normalR[8];
+  }
+  else {
+    dMULTIPLY0_331 (normal,R1,normalC);
+  }
+  if (invert_normal) {
+    normal[0] = -normal[0];
+    normal[1] = -normal[1];
+    normal[2] = -normal[2];
+  }
+  *depth = -s;
+
+  // compute contact point(s)
+
+  if (code > 6) {
+    // an edge from box 1 touches an edge from box 2.
+    // find a point pa on the intersecting edge of box 1
+    btVector3 pa;
+    btScalar sign;
+    for (i=0; i<3; i++) pa[i] = p1[i];
+    for (j=0; j<3; j++) {
+      sign = (dDOT14(normal,R1+j) > 0) ? btScalar(1.0) : btScalar(-1.0);
+      for (i=0; i<3; i++) pa[i] += sign * A[j] * R1[i*4+j];
+    }
+
+    // find a point pb on the intersecting edge of box 2
+    btVector3 pb;
+    for (i=0; i<3; i++) pb[i] = p2[i];
+    for (j=0; j<3; j++) {
+      sign = (dDOT14(normal,R2+j) > 0) ? btScalar(-1.0) : btScalar(1.0);
+      for (i=0; i<3; i++) pb[i] += sign * B[j] * R2[i*4+j];
+    }
+
+    btScalar alpha,beta;
+    btVector3 ua,ub;
+    for (i=0; i<3; i++) ua[i] = R1[((code)-7)/3 + i*4];
+    for (i=0; i<3; i++) ub[i] = R2[((code)-7)%3 + i*4];
+
+    dLineClosestApproach (pa,ua,pb,ub,&alpha,&beta);
+    for (i=0; i<3; i++) pa[i] += ua[i]*alpha;
+    for (i=0; i<3; i++) pb[i] += ub[i]*beta;
+
+	{
+		
+		//contact[0].pos[i] = btScalar(0.5)*(pa[i]+pb[i]);
+		//contact[0].depth = *depth;
+		btVector3 pointInWorld;
+
+#ifdef USE_CENTER_POINT
+	    for (i=0; i<3; i++) 
+			pointInWorld[i] = (pa[i]+pb[i])*btScalar(0.5);
+		output.addContactPoint(-normal,pointInWorld,-*depth);
+#else
+		output.addContactPoint(-normal,pb,-*depth);
+#endif //
+		*return_code = code;
+	}
+    return 1;
+  }
+
+  // okay, we have a face-something intersection (because the separating
+  // axis is perpendicular to a face). define face 'a' to be the reference
+  // face (i.e. the normal vector is perpendicular to this) and face 'b' to be
+  // the incident face (the closest face of the other box).
+
+  const btScalar *Ra,*Rb,*pa,*pb,*Sa,*Sb;
+  if (code <= 3) {
+    Ra = R1;
+    Rb = R2;
+    pa = p1;
+    pb = p2;
+    Sa = A;
+    Sb = B;
+  }
+  else {
+    Ra = R2;
+    Rb = R1;
+    pa = p2;
+    pb = p1;
+    Sa = B;
+    Sb = A;
+  }
+
+  // nr = normal vector of reference face dotted with axes of incident box.
+  // anr = absolute values of nr.
+  btVector3 normal2,nr,anr;
+  if (code <= 3) {
+    normal2[0] = normal[0];
+    normal2[1] = normal[1];
+    normal2[2] = normal[2];
+  }
+  else {
+    normal2[0] = -normal[0];
+    normal2[1] = -normal[1];
+    normal2[2] = -normal[2];
+  }
+  dMULTIPLY1_331 (nr,Rb,normal2);
+  anr[0] = btFabs (nr[0]);
+  anr[1] = btFabs (nr[1]);
+  anr[2] = btFabs (nr[2]);
+
+  // find the largest compontent of anr: this corresponds to the normal
+  // for the indident face. the other axis numbers of the indicent face
+  // are stored in a1,a2.
+  int lanr,a1,a2;
+  if (anr[1] > anr[0]) {
+    if (anr[1] > anr[2]) {
+      a1 = 0;
+      lanr = 1;
+      a2 = 2;
+    }
+    else {
+      a1 = 0;
+      a2 = 1;
+      lanr = 2;
+    }
+  }
+  else {
+    if (anr[0] > anr[2]) {
+      lanr = 0;
+      a1 = 1;
+      a2 = 2;
+    }
+    else {
+      a1 = 0;
+      a2 = 1;
+      lanr = 2;
+    }
+  }
+
+  // compute center point of incident face, in reference-face coordinates
+  btVector3 center;
+  if (nr[lanr] < 0) {
+    for (i=0; i<3; i++) center[i] = pb[i] - pa[i] + Sb[lanr] * Rb[i*4+lanr];
+  }
+  else {
+    for (i=0; i<3; i++) center[i] = pb[i] - pa[i] - Sb[lanr] * Rb[i*4+lanr];
+  }
+
+  // find the normal and non-normal axis numbers of the reference box
+  int codeN,code1,code2;
+  if (code <= 3) codeN = code-1; else codeN = code-4;
+  if (codeN==0) {
+    code1 = 1;
+    code2 = 2;
+  }
+  else if (codeN==1) {
+    code1 = 0;
+    code2 = 2;
+  }
+  else {
+    code1 = 0;
+    code2 = 1;
+  }
+
+  // find the four corners of the incident face, in reference-face coordinates
+  btScalar quad[8];	// 2D coordinate of incident face (x,y pairs)
+  btScalar c1,c2,m11,m12,m21,m22;
+  c1 = dDOT14 (center,Ra+code1);
+  c2 = dDOT14 (center,Ra+code2);
+  // optimize this? - we have already computed this data above, but it is not
+  // stored in an easy-to-index format. for now it's quicker just to recompute
+  // the four dot products.
+  m11 = dDOT44 (Ra+code1,Rb+a1);
+  m12 = dDOT44 (Ra+code1,Rb+a2);
+  m21 = dDOT44 (Ra+code2,Rb+a1);
+  m22 = dDOT44 (Ra+code2,Rb+a2);
+  {
+    btScalar k1 = m11*Sb[a1];
+    btScalar k2 = m21*Sb[a1];
+    btScalar k3 = m12*Sb[a2];
+    btScalar k4 = m22*Sb[a2];
+    quad[0] = c1 - k1 - k3;
+    quad[1] = c2 - k2 - k4;
+    quad[2] = c1 - k1 + k3;
+    quad[3] = c2 - k2 + k4;
+    quad[4] = c1 + k1 + k3;
+    quad[5] = c2 + k2 + k4;
+    quad[6] = c1 + k1 - k3;
+    quad[7] = c2 + k2 - k4;
+  }
+
+  // find the size of the reference face
+  btScalar rect[2];
+  rect[0] = Sa[code1];
+  rect[1] = Sa[code2];
+
+  // intersect the incident and reference faces
+  btScalar ret[16];
+  int n = intersectRectQuad2 (rect,quad,ret);
+  if (n < 1) return 0;		// this should never happen
+
+  // convert the intersection points into reference-face coordinates,
+  // and compute the contact position and depth for each point. only keep
+  // those points that have a positive (penetrating) depth. delete points in
+  // the 'ret' array as necessary so that 'point' and 'ret' correspond.
+  btScalar point[3*8];		// penetrating contact points
+  btScalar dep[8];			// depths for those points
+  btScalar det1 = 1.f/(m11*m22 - m12*m21);
+  m11 *= det1;
+  m12 *= det1;
+  m21 *= det1;
+  m22 *= det1;
+  int cnum = 0;			// number of penetrating contact points found
+  for (j=0; j < n; j++) {
+    btScalar k1 =  m22*(ret[j*2]-c1) - m12*(ret[j*2+1]-c2);
+    btScalar k2 = -m21*(ret[j*2]-c1) + m11*(ret[j*2+1]-c2);
+    for (i=0; i<3; i++) point[cnum*3+i] =
+			  center[i] + k1*Rb[i*4+a1] + k2*Rb[i*4+a2];
+    dep[cnum] = Sa[codeN] - dDOT(normal2,point+cnum*3);
+    if (dep[cnum] >= 0) {
+      ret[cnum*2] = ret[j*2];
+      ret[cnum*2+1] = ret[j*2+1];
+      cnum++;
+    }
+  }
+  if (cnum < 1) return 0;	// this should never happen
+
+  // we can't generate more contacts than we actually have
+  if (maxc > cnum) maxc = cnum;
+  if (maxc < 1) maxc = 1;
+
+  if (cnum <= maxc) {
+    // we have less contacts than we need, so we use them all
+    for (j=0; j < cnum; j++) {
+
+		//AddContactPoint...
+
+		//dContactGeom *con = CONTACT(contact,skip*j);
+      //for (i=0; i<3; i++) con->pos[i] = point[j*3+i] + pa[i];
+      //con->depth = dep[j];
+
+		btVector3 pointInWorld;
+		for (i=0; i<3; i++) 
+			pointInWorld[i] = point[j*3+i] + pa[i];
+		output.addContactPoint(-normal,pointInWorld,-dep[j]);
+
+    }
+  }
+  else {
+    // we have more contacts than are wanted, some of them must be culled.
+    // find the deepest point, it is always the first contact.
+    int i1 = 0;
+    btScalar maxdepth = dep[0];
+    for (i=1; i<cnum; i++) {
+      if (dep[i] > maxdepth) {
+	maxdepth = dep[i];
+	i1 = i;
+      }
+    }
+
+    int iret[8];
+    cullPoints2 (cnum,ret,maxc,i1,iret);
+
+    for (j=0; j < maxc; j++) {
+//      dContactGeom *con = CONTACT(contact,skip*j);
+  //    for (i=0; i<3; i++) con->pos[i] = point[iret[j]*3+i] + pa[i];
+    //  con->depth = dep[iret[j]];
+
+		btVector3 posInWorld;
+		for (i=0; i<3; i++) 
+			posInWorld[i] = point[iret[j]*3+i] + pa[i];
+		output.addContactPoint(-normal,posInWorld,-dep[iret[j]]);
+    }
+    cnum = maxc;
+  }
+
+  *return_code = code;
+  return cnum;
+}
+
+void	btBoxBoxDetector::getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* /*debugDraw*/,bool /*swapResults*/)
+{
+	
+	const btTransform& transformA = input.m_transformA;
+	const btTransform& transformB = input.m_transformB;
+	
+	int skip = 0;
+	dContactGeom *contact = 0;
+
+	dMatrix3 R1;
+	dMatrix3 R2;
+
+	for (int j=0;j<3;j++)
+	{
+		R1[0+4*j] = transformA.getBasis()[j].x();
+		R2[0+4*j] = transformB.getBasis()[j].x();
+
+		R1[1+4*j] = transformA.getBasis()[j].y();
+		R2[1+4*j] = transformB.getBasis()[j].y();
+
+
+		R1[2+4*j] = transformA.getBasis()[j].z();
+		R2[2+4*j] = transformB.getBasis()[j].z();
+
+	}
+
+	
+
+	btVector3 normal;
+	btScalar depth;
+	int return_code;
+	int maxc = 4;
+
+
+	dBoxBox2 (transformA.getOrigin(), 
+	R1,
+	2.f*m_box1->getHalfExtentsWithMargin(),
+	transformB.getOrigin(),
+	R2, 
+	2.f*m_box2->getHalfExtentsWithMargin(),
+	normal, &depth, &return_code,
+	maxc, contact, skip,
+	output
+	);
+
+}

extern/bullet2/src/BulletCollision/CollisionDispatch/btBoxBoxDetector.h

@@ -0,0 +1,44 @@
+/*
+ * Box-Box collision detection re-distributed under the ZLib license with permission from Russell L. Smith
+ * Original version is from Open Dynamics Engine, Copyright (C) 2001,2002 Russell L. Smith.
+ * All rights reserved.  Email: russ@q12.org   Web: www.q12.org
+
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+#ifndef BOX_BOX_DETECTOR_H
+#define BOX_BOX_DETECTOR_H
+
+
+class btBoxShape;
+#include "BulletCollision/NarrowPhaseCollision/btDiscreteCollisionDetectorInterface.h"
+
+
+/// btBoxBoxDetector wraps the ODE box-box collision detector
+/// re-distributed under the Zlib license with permission from Russell L. Smith
+struct btBoxBoxDetector : public btDiscreteCollisionDetectorInterface
+{
+	btBoxShape* m_box1;
+	btBoxShape* m_box2;
+
+public:
+
+	btBoxBoxDetector(btBoxShape* box1,btBoxShape* box2);
+
+	virtual ~btBoxBoxDetector() {};
+
+	virtual void	getClosestPoints(const ClosestPointInput& input,Result& output,class btIDebugDraw* debugDraw,bool swapResults=false);
+
+};
+
+#endif //BT_BOX_BOX_DETECTOR_H

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionConfiguration.h

@@ -0,0 +1,47 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef BT_COLLISION_CONFIGURATION
+#define BT_COLLISION_CONFIGURATION
+struct btCollisionAlgorithmCreateFunc;
+
+class btStackAlloc;
+class btPoolAllocator;
+
+///btCollisionConfiguration allows to configure Bullet collision detection
+///stack allocator size, default collision algorithms and persistent manifold pool size
+///todo: describe the meaning
+class	btCollisionConfiguration
+{
+
+public:
+
+	virtual ~btCollisionConfiguration()
+	{
+	}
+
+	///memory pools
+	virtual btPoolAllocator* getPersistentManifoldPool() = 0;
+
+	virtual btPoolAllocator* getCollisionAlgorithmPool() = 0;
+
+	virtual btStackAlloc*	getStackAllocator() = 0;
+
+	virtual btCollisionAlgorithmCreateFunc* getCollisionAlgorithmCreateFunc(int proxyType0,int proxyType1) =0;
+
+};
+
+#endif //BT_COLLISION_CONFIGURATION
+

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionCreateFunc.h

@@ -16,7 +16,7 @@ subject to the following restrictions:
 #ifndef COLLISION_CREATE_FUNC
 #define COLLISION_CREATE_FUNC
 
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 typedef btAlignedObjectArray<class btCollisionObject*> btCollisionObjectArray;
 class btCollisionAlgorithm;
 class btCollisionObject;

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionDispatcher.cpp

@@ -19,69 +19,39 @@ subject to the following restrictions:
 
 
 #include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "BulletCollision/CollisionDispatch/btConvexConvexAlgorithm.h"
+
-#include "BulletCollision/CollisionDispatch/btEmptyCollisionAlgorithm.h"
-#include "BulletCollision/CollisionDispatch/btConvexConcaveCollisionAlgorithm.h"
-#include "BulletCollision/CollisionDispatch/btCompoundCollisionAlgorithm.h"
 #include "BulletCollision/CollisionShapes/btCollisionShape.h"
 #include "BulletCollision/CollisionDispatch/btCollisionObject.h"
 #include "BulletCollision/BroadphaseCollision/btOverlappingPairCache.h"
+#include "LinearMath/btPoolAllocator.h"
+#include "BulletCollision/CollisionDispatch/btCollisionConfiguration.h"
 
 int gNumManifold = 0;
 
+#ifdef BT_DEBUG
 #include <stdio.h>
+#endif
 
-	
-btCollisionDispatcher::btCollisionDispatcher(bool noDefaultAlgorithms):
-m_count(0),
-m_useIslands(true),
-m_convexConvexCreateFunc(0),
-m_convexConcaveCreateFunc(0),
-m_swappedConvexConcaveCreateFunc(0),
-m_compoundCreateFunc(0),
-m_swappedCompoundCreateFunc(0),
-m_emptyCreateFunc(0)
-{
-	(void)noDefaultAlgorithms;
-	int i;
 
-	setNearCallback(defaultNearCallback);
+btCollisionDispatcher::btCollisionDispatcher (btCollisionConfiguration* collisionConfiguration): 
-
-	m_emptyCreateFunc = new btEmptyAlgorithm::CreateFunc;
-	for (i=0;i<MAX_BROADPHASE_COLLISION_TYPES;i++)
-	{
-		for (int j=0;j<MAX_BROADPHASE_COLLISION_TYPES;j++)
-		{
-			m_doubleDispatch[i][j] = m_emptyCreateFunc;
-		}
-	}
-}
-//if you want to not link with the default collision algorithms, you can
-//define BT_EXCLUDE_DEFAULT_COLLISIONALGORITHM_REGISTRATION 
-//in your Bullet library build system
-#ifndef BT_EXCLUDE_DEFAULT_COLLISIONALGORITHM_REGISTRATION
-
-btCollisionDispatcher::btCollisionDispatcher (): 
 	m_count(0),
-	m_useIslands(true)
+	m_useIslands(true),
+	m_staticWarningReported(false),
+	m_collisionConfiguration(collisionConfiguration)
 {
 	int i;
 
 	setNearCallback(defaultNearCallback);
 	
-	//default CreationFunctions, filling the m_doubleDispatch table
+	m_collisionAlgorithmPoolAllocator = collisionConfiguration->getCollisionAlgorithmPool();
-	m_convexConvexCreateFunc = new btConvexConvexAlgorithm::CreateFunc;
+
-	m_convexConcaveCreateFunc = new btConvexConcaveCollisionAlgorithm::CreateFunc;
+	m_persistentManifoldPoolAllocator = collisionConfiguration->getPersistentManifoldPool();
-	m_swappedConvexConcaveCreateFunc = new btConvexConcaveCollisionAlgorithm::SwappedCreateFunc;
-	m_compoundCreateFunc = new btCompoundCollisionAlgorithm::CreateFunc;
-	m_swappedCompoundCreateFunc = new btCompoundCollisionAlgorithm::SwappedCreateFunc;
-	m_emptyCreateFunc = new btEmptyAlgorithm::CreateFunc;
 
 	for (i=0;i<MAX_BROADPHASE_COLLISION_TYPES;i++)
 	{
 		for (int j=0;j<MAX_BROADPHASE_COLLISION_TYPES;j++)
 		{
-			m_doubleDispatch[i][j] = internalFindCreateFunc(i,j);
+			m_doubleDispatch[i][j] = m_collisionConfiguration->getCollisionAlgorithmCreateFunc(i,j);
 			assert(m_doubleDispatch[i][j]);
 		}
 	}
@@ -89,8 +59,6 @@ btCollisionDispatcher::btCollisionDispatcher ():
 	
 };
 
-#endif //BT_EXCLUDE_DEFAULT_COLLISIONALGORITHM_REGISTRATION
-
 
 void btCollisionDispatcher::registerCollisionCreateFunc(int proxyType0, int proxyType1, btCollisionAlgorithmCreateFunc *createFunc)
 {
@@ -99,12 +67,6 @@ void btCollisionDispatcher::registerCollisionCreateFunc(int proxyType0, int prox
 
 btCollisionDispatcher::~btCollisionDispatcher()
 {
-	delete m_convexConvexCreateFunc;
-	delete m_convexConcaveCreateFunc;
-	delete m_swappedConvexConcaveCreateFunc;
-	delete m_compoundCreateFunc;
-	delete m_swappedCompoundCreateFunc;
-	delete m_emptyCreateFunc;
 }
 
 btPersistentManifold*	btCollisionDispatcher::getNewManifold(void* b0,void* b1) 
@@ -117,7 +79,18 @@ btPersistentManifold*	btCollisionDispatcher::getNewManifold(void* b0,void* b1)
 	btCollisionObject* body0 = (btCollisionObject*)b0;
 	btCollisionObject* body1 = (btCollisionObject*)b1;
 	
-	btPersistentManifold* manifold = new btPersistentManifold (body0,body1);
+	void* mem = 0;
+	
+	if (m_persistentManifoldPoolAllocator->getFreeCount())
+	{
+		mem = m_persistentManifoldPoolAllocator->allocate(sizeof(btPersistentManifold));
+	} else
+	{
+		mem = btAlignedAlloc(sizeof(btPersistentManifold),16);
+
+	}
+	btPersistentManifold* manifold = new(mem) btPersistentManifold (body0,body1,0);
+	manifold->m_index1a = m_manifoldsPtr.size();
 	m_manifoldsPtr.push_back(manifold);
 
 	return manifold;
@@ -137,13 +110,19 @@ void btCollisionDispatcher::releaseManifold(btPersistentManifold* manifold)
 	//printf("releaseManifold: gNumManifold %d\n",gNumManifold);
 	clearManifold(manifold);
 
-	///todo: this can be improved a lot, linear search might be slow part!
+	int findIndex = manifold->m_index1a;
-	int findIndex = m_manifoldsPtr.findLinearSearch(manifold);
+	btAssert(findIndex < m_manifoldsPtr.size());
-	if (findIndex < m_manifoldsPtr.size())
+	m_manifoldsPtr.swap(findIndex,m_manifoldsPtr.size()-1);
+	m_manifoldsPtr[findIndex]->m_index1a = findIndex;
+	m_manifoldsPtr.pop_back();
+
+	manifold->~btPersistentManifold();
+	if (m_persistentManifoldPoolAllocator->validPtr(manifold))
 	{
-		m_manifoldsPtr.swap(findIndex,m_manifoldsPtr.size()-1);
+		m_persistentManifoldPoolAllocator->freeMemory(manifold);
-		m_manifoldsPtr.pop_back();
+	} else
-		delete manifold;
+	{
+		btAlignedFree(manifold);
 	}
 	
 }
@@ -152,99 +131,19 @@ void btCollisionDispatcher::releaseManifold(btPersistentManifold* manifold)
 
 btCollisionAlgorithm* btCollisionDispatcher::findAlgorithm(btCollisionObject* body0,btCollisionObject* body1,btPersistentManifold* sharedManifold)
 {
-
-#ifdef USE_DISPATCH_REGISTRY_ARRAY
 	
 	btCollisionAlgorithmConstructionInfo ci;
-	ci.m_dispatcher = this;
+
+	ci.m_dispatcher1 = this;
 	ci.m_manifold = sharedManifold;
-	btCollisionAlgorithm* algo = m_doubleDispatch[body0->getCollisionShape()->getShapeType()][body1->getCollisionShape()->getShapeType()]
+	btCollisionAlgorithm* algo = m_doubleDispatch[body0->getCollisionShape()->getShapeType()][body1->getCollisionShape()->getShapeType()]->CreateCollisionAlgorithm(ci,body0,body1);
-	->CreateCollisionAlgorithm(ci,body0,body1);
+
-#else
-	btCollisionAlgorithm* algo = internalFindAlgorithm(body0,body1);
-#endif //USE_DISPATCH_REGISTRY_ARRAY
 	return algo;
 }
 
 
-#ifndef BT_EXCLUDE_DEFAULT_COLLISIONALGORITHM_REGISTRATION
-
-btCollisionAlgorithmCreateFunc* btCollisionDispatcher::internalFindCreateFunc(int proxyType0,int proxyType1)
-{
-	
-	if (btBroadphaseProxy::isConvex(proxyType0) && btBroadphaseProxy::isConvex(proxyType1))
-	{
-		return m_convexConvexCreateFunc;
-	}
-
-	if (btBroadphaseProxy::isConvex(proxyType0) && btBroadphaseProxy::isConcave(proxyType1))
-	{
-		return m_convexConcaveCreateFunc;
-	}
-
-	if (btBroadphaseProxy::isConvex(proxyType1) && btBroadphaseProxy::isConcave(proxyType0))
-	{
-		return m_swappedConvexConcaveCreateFunc;
-	}
-
-	if (btBroadphaseProxy::isCompound(proxyType0))
-	{
-		return m_compoundCreateFunc;
-	} else
-	{
-		if (btBroadphaseProxy::isCompound(proxyType1))
-		{
-			return m_swappedCompoundCreateFunc;
-		}
-	}
-
-	//failed to find an algorithm
-	return m_emptyCreateFunc;
-}
-
-#endif //BT_EXCLUDE_DEFAULT_COLLISIONALGORITHM_REGISTRATION
 
 
-#ifndef USE_DISPATCH_REGISTRY_ARRAY
-
-btCollisionAlgorithm* btCollisionDispatcher::internalFindAlgorithm(btCollisionObject* body0,btCollisionObject* body1,btPersistentManifold* sharedManifold)
-{
-	m_count++;
-	
-	btCollisionAlgorithmConstructionInfo ci;
-	ci.m_dispatcher = this;
-	
-	if (body0->getCollisionShape()->isConvex() && body1->getCollisionShape()->isConvex() )
-	{
-		return new btConvexConvexAlgorithm(sharedManifold,ci,body0,body1);
-	}
-
-	if (body0->getCollisionShape()->isConvex() && body1->getCollisionShape()->isConcave())
-	{
-		return new btConvexConcaveCollisionAlgorithm(ci,body0,body1,false);
-	}
-
-	if (body1->getCollisionShape()->isConvex() && body0->getCollisionShape()->isConcave())
-	{
-		return new btConvexConcaveCollisionAlgorithm(ci,body0,body1,true);
-	}
-
-	if (body0->getCollisionShape()->isCompound())
-	{
-		return new btCompoundCollisionAlgorithm(ci,body0,body1,false);
-	} else
-	{
-		if (body1->getCollisionShape()->isCompound())
-		{
-			return new btCompoundCollisionAlgorithm(ci,body0,body1,true);
-		}
-	}
-
-	//failed to find an algorithm
-	return new btEmptyAlgorithm(ci);
-	
-}
-#endif //USE_DISPATCH_REGISTRY_ARRAY
 
 bool	btCollisionDispatcher::needsResponse(btCollisionObject* body0,btCollisionObject* body1)
 {
@@ -264,13 +163,19 @@ bool	btCollisionDispatcher::needsCollision(btCollisionObject* body0,btCollisionO
 
 	bool needsCollision = true;
 
-	//broadphase filtering already deals with this
+#ifdef BT_DEBUG
-	if ((body0->isStaticObject() || body0->isKinematicObject()) &&
+	if (!m_staticWarningReported)
-		(body1->isStaticObject() || body1->isKinematicObject()))
 	{
-		printf("warning btCollisionDispatcher::needsCollision: static-static collision!\n");
+		//broadphase filtering already deals with this
+		if ((body0->isStaticObject() || body0->isKinematicObject()) &&
+			(body1->isStaticObject() || body1->isKinematicObject()))
+		{
+			m_staticWarningReported = true;
+			printf("warning btCollisionDispatcher::needsCollision: static-static collision!\n");
+		}
 	}
-		
+#endif //BT_DEBUG
+
 	if ((!body0->isActive()) && (!body1->isActive()))
 		needsCollision = false;
 	else if (!body0->checkCollideWith(body1))
@@ -286,23 +191,25 @@ bool	btCollisionDispatcher::needsCollision(btCollisionObject* body0,btCollisionO
 ///this is useful for the collision dispatcher.
 class btCollisionPairCallback : public btOverlapCallback
 {
-	btDispatcherInfo& m_dispatchInfo;
+	const btDispatcherInfo& m_dispatchInfo;
 	btCollisionDispatcher*	m_dispatcher;
 
 public:
 
-	btCollisionPairCallback(btDispatcherInfo& dispatchInfo,btCollisionDispatcher*	dispatcher)
+	btCollisionPairCallback(const btDispatcherInfo& dispatchInfo,btCollisionDispatcher*	dispatcher)
 	:m_dispatchInfo(dispatchInfo),
 	m_dispatcher(dispatcher)
 	{
 	}
 
-	btCollisionPairCallback& operator=(btCollisionPairCallback& other)
+	/*btCollisionPairCallback& operator=(btCollisionPairCallback& other)
 	{
 		m_dispatchInfo = other.m_dispatchInfo;
 		m_dispatcher = other.m_dispatcher;
 		return *this;
 	}
+	*/
+
 
 	virtual ~btCollisionPairCallback() {}
 
@@ -316,13 +223,14 @@ public:
 };
 
 
-void	btCollisionDispatcher::dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,btDispatcherInfo& dispatchInfo)
+
+void	btCollisionDispatcher::dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,const btDispatcherInfo& dispatchInfo,btDispatcher* dispatcher) 
 {
 	//m_blockedForChanges = true;
 
 	btCollisionPairCallback	collisionCallback(dispatchInfo,this);
 
-	pairCache->processAllOverlappingPairs(&collisionCallback);
+	pairCache->processAllOverlappingPairs(&collisionCallback,dispatcher);
 
 	//m_blockedForChanges = false;
 
@@ -332,7 +240,7 @@ void	btCollisionDispatcher::dispatchAllCollisionPairs(btOverlappingPairCache* pa
 
 
 //by default, Bullet will use this near callback
-void btCollisionDispatcher::defaultNearCallback(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, btDispatcherInfo& dispatchInfo)
+void btCollisionDispatcher::defaultNearCallback(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, const btDispatcherInfo& dispatchInfo)
 {
 		btCollisionObject* colObj0 = (btCollisionObject*)collisionPair.m_pProxy0->m_clientObject;
 		btCollisionObject* colObj1 = (btCollisionObject*)collisionPair.m_pProxy1->m_clientObject;
@@ -365,3 +273,26 @@ void btCollisionDispatcher::defaultNearCallback(btBroadphasePair& collisionPair,
 		}
 
 }
+
+
+void* btCollisionDispatcher::allocateCollisionAlgorithm(int size)
+{
+	if (m_collisionAlgorithmPoolAllocator->getFreeCount())
+	{
+		return m_collisionAlgorithmPoolAllocator->allocate(size);
+	}
+	
+	//warn user for overflow?
+	return	btAlignedAlloc(static_cast<size_t>(size), 16);
+}
+
+void btCollisionDispatcher::freeCollisionAlgorithm(void* ptr)
+{
+	if (m_collisionAlgorithmPoolAllocator->validPtr(ptr))
+	{
+		m_collisionAlgorithmPoolAllocator->freeMemory(ptr);
+	} else
+	{
+		btAlignedFree(ptr);
+	}
+}

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionDispatcher.h

@@ -16,17 +16,18 @@ subject to the following restrictions:
 #ifndef COLLISION__DISPATCHER_H
 #define COLLISION__DISPATCHER_H
 
-#include "../BroadphaseCollision/btDispatcher.h"
+#include "BulletCollision/BroadphaseCollision/btDispatcher.h"
-#include "../NarrowPhaseCollision/btPersistentManifold.h"
+#include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
 
-#include "../CollisionDispatch/btManifoldResult.h"
+#include "BulletCollision/CollisionDispatch/btManifoldResult.h"
 
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
 class btIDebugDraw;
 class btOverlappingPairCache;
-
+class btPoolAllocator;
+class btCollisionConfiguration;
 
 #include "btCollisionCreateFunc.h"
 
@@ -34,7 +35,7 @@ class btOverlappingPairCache;
 
 class btCollisionDispatcher;
 ///user can override this nearcallback for collision filtering and more finegrained control over collision detection
-typedef void (*btNearCallback)(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, btDispatcherInfo& dispatchInfo);
+typedef void (*btNearCallback)(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, const btDispatcherInfo& dispatchInfo);
 
 
 ///btCollisionDispatcher supports algorithms that handle ConvexConvex and ConvexConcave collision pairs.
@@ -46,26 +47,22 @@ class btCollisionDispatcher : public btDispatcher
 	btAlignedObjectArray<btPersistentManifold*>	m_manifoldsPtr;
 
 	bool m_useIslands;
+
+	bool	m_staticWarningReported;
 	
 	btManifoldResult	m_defaultManifoldResult;
 
 	btNearCallback		m_nearCallback;
 	
+	btPoolAllocator*	m_collisionAlgorithmPoolAllocator;
+
+	btPoolAllocator*	m_persistentManifoldPoolAllocator;
+
 	btCollisionAlgorithmCreateFunc* m_doubleDispatch[MAX_BROADPHASE_COLLISION_TYPES][MAX_BROADPHASE_COLLISION_TYPES];
 	
-	btCollisionAlgorithmCreateFunc* internalFindCreateFunc(int proxyType0,int proxyType1);
 
-	//default CreationFunctions, filling the m_doubleDispatch table
+	btCollisionConfiguration*	m_collisionConfiguration;
-	btCollisionAlgorithmCreateFunc*	m_convexConvexCreateFunc;
-	btCollisionAlgorithmCreateFunc*	m_convexConcaveCreateFunc;
-	btCollisionAlgorithmCreateFunc*	m_swappedConvexConcaveCreateFunc;
-	btCollisionAlgorithmCreateFunc*	m_compoundCreateFunc;
-	btCollisionAlgorithmCreateFunc*	m_swappedCompoundCreateFunc;
-	btCollisionAlgorithmCreateFunc*   m_emptyCreateFunc;
 
-#ifndef USE_DISPATCH_REGISTRY_ARRAY
-	btCollisionAlgorithm* internalFindAlgorithm(btCollisionObject* body0,btCollisionObject* body1,btPersistentManifold* sharedManifold = 0);
-#endif //USE_DISPATCH_REGISTRY_ARRAY
 
 public:
 
@@ -92,11 +89,7 @@ public:
 		return m_manifoldsPtr[index];
 	}
 
-	///the default constructor creates/register default collision algorithms, for convex, compound and concave shape support
+	btCollisionDispatcher (btCollisionConfiguration* collisionConfiguration);
-	btCollisionDispatcher ();
-
-	///a special constructor that doesn't create/register the default collision algorithms
-	btCollisionDispatcher(bool noDefaultAlgorithms);
 
 	virtual ~btCollisionDispatcher();
 
@@ -114,7 +107,7 @@ public:
 	
 	virtual bool	needsResponse(btCollisionObject* body0,btCollisionObject* body1);
 	
-	virtual void	dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,btDispatcherInfo& dispatchInfo);
+	virtual void	dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,const btDispatcherInfo& dispatchInfo,btDispatcher* dispatcher) ;
 
 	void	setNearCallback(btNearCallback	nearCallback)
 	{
@@ -127,7 +120,26 @@ public:
 	}
 
 	//by default, Bullet will use this near callback
-	static void  defaultNearCallback(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, btDispatcherInfo& dispatchInfo);
+	static void  defaultNearCallback(btBroadphasePair& collisionPair, btCollisionDispatcher& dispatcher, const btDispatcherInfo& dispatchInfo);
+
+	virtual	void* allocateCollisionAlgorithm(int size);
+
+	virtual	void freeCollisionAlgorithm(void* ptr);
+
+	btCollisionConfiguration*	getCollisionConfiguration()
+	{
+		return m_collisionConfiguration;
+	}
+
+	const btCollisionConfiguration*	getCollisionConfiguration() const
+	{
+		return m_collisionConfiguration;
+	}
+
+	void	setCollisionConfiguration(btCollisionConfiguration* config)
+	{
+		m_collisionConfiguration = config;
+	}
 
 };
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionObject.cpp

@@ -13,15 +13,22 @@ subject to the following restrictions:
 3. This notice may not be removed or altered from any source distribution.
 */
 
+
 #include "btCollisionObject.h"
 
 btCollisionObject::btCollisionObject()
 	:	m_broadphaseHandle(0),
 		m_collisionShape(0),
-		m_collisionFlags(0),
+		m_rootCollisionShape(0),
+		m_collisionFlags(btCollisionObject::CF_STATIC_OBJECT),
+		m_islandTag1(-1),
+		m_companionId(-1),
 		m_activationState1(1),
 		m_deactivationTime(btScalar(0.)),
+		m_friction(btScalar(0.5)),
+		m_restitution(btScalar(0.)),
 		m_userObjectPointer(0),
+		m_internalType(CO_COLLISION_OBJECT),
 		m_hitFraction(btScalar(1.)),
 		m_ccdSweptSphereRadius(btScalar(0.)),
 		m_ccdSquareMotionThreshold(btScalar(0.)),
@@ -55,3 +62,4 @@ void btCollisionObject::activate(bool forceActivation)
 }
 
 
+

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionObject.h

@@ -16,7 +16,7 @@ subject to the following restrictions:
 #ifndef COLLISION_OBJECT_H
 #define COLLISION_OBJECT_H
 
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
 
 //island management, m_activationState1
 #define ACTIVE_TAG 1
@@ -27,7 +27,8 @@ subject to the following restrictions:
 
 struct	btBroadphaseProxy;
 class	btCollisionShape;
-#include "../../LinearMath/btMotionState.h"
+#include "LinearMath/btMotionState.h"
+#include "LinearMath/btAlignedAllocator.h"
 
 
 
@@ -50,6 +51,11 @@ protected:
 	btVector3	m_interpolationAngularVelocity;
 	btBroadphaseProxy*		m_broadphaseHandle;
 	btCollisionShape*		m_collisionShape;
+	
+	///m_rootCollisionShape is temporarily used to store the original collision shape
+	///The m_collisionShape might be temporarily replaced by a child collision shape during collision detection purposes
+	///If it is NULL, the m_collisionShape is not temporarily replaced.
+	btCollisionShape*		m_rootCollisionShape;
 
 	int				m_collisionFlags;
 
@@ -65,8 +71,9 @@ protected:
 	///users can point to their objects, m_userPointer is not used by Bullet, see setUserPointer/getUserPointer
 	void*			m_userObjectPointer;
 
-	///m_internalOwner is reserved to point to Bullet's btRigidBody. Don't use this, use m_userObjectPointer instead.
+	///m_internalType is reserved to distinguish Bullet's btCollisionObject, btRigidBody, btSoftBody etc.
-	void*			m_internalOwner;
+	///do not assign your own m_internalType unless you write a new dynamics object class.
+	int				m_internalType;
 
 	///time of impact calculation
 	btScalar		m_hitFraction; 
@@ -82,13 +89,15 @@ protected:
 
 	char	m_pad[7];
 
-	virtual bool	checkCollideWithOverride(btCollisionObject* co)
+	virtual bool	checkCollideWithOverride(btCollisionObject* /* co */)
 	{
 		return true;
 	}
 
 public:
 
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
 	enum CollisionFlags
 	{
 		CF_STATIC_OBJECT= 1,
@@ -97,29 +106,35 @@ public:
 		CF_CUSTOM_MATERIAL_CALLBACK = 8//this allows per-triangle material (friction/restitution)
 	};
 
+	enum	CollisionObjectTypes
+	{
+		CO_COLLISION_OBJECT =1,
+		CO_RIGID_BODY,
+		CO_SOFT_BODY
+	};
 
-	inline bool mergesSimulationIslands() const
+	SIMD_FORCE_INLINE bool mergesSimulationIslands() const
 	{
 		///static objects, kinematic and object without contact response don't merge islands
 		return  ((m_collisionFlags & (CF_STATIC_OBJECT | CF_KINEMATIC_OBJECT | CF_NO_CONTACT_RESPONSE) )==0);
 	}
 
 
-	inline bool		isStaticObject() const {
+	SIMD_FORCE_INLINE bool		isStaticObject() const {
 		return (m_collisionFlags & CF_STATIC_OBJECT) != 0;
 	}
 
-	inline bool		isKinematicObject() const
+	SIMD_FORCE_INLINE bool		isKinematicObject() const
 	{
 		return (m_collisionFlags & CF_KINEMATIC_OBJECT) != 0;
 	}
 
-	inline bool		isStaticOrKinematicObject() const
+	SIMD_FORCE_INLINE bool		isStaticOrKinematicObject() const
 	{
 		return (m_collisionFlags & (CF_KINEMATIC_OBJECT | CF_STATIC_OBJECT)) != 0 ;
 	}
 
-	inline bool		hasContactResponse() const {
+	SIMD_FORCE_INLINE bool		hasContactResponse() const {
 		return (m_collisionFlags & CF_NO_CONTACT_RESPONSE)==0;
 	}
 
@@ -131,20 +146,35 @@ public:
 	void	setCollisionShape(btCollisionShape* collisionShape)
 	{
 		m_collisionShape = collisionShape;
+		m_rootCollisionShape = collisionShape;
 	}
 
-	const btCollisionShape*	getCollisionShape() const
+	SIMD_FORCE_INLINE const btCollisionShape*	getCollisionShape() const
 	{
 		return m_collisionShape;
 	}
 
-	btCollisionShape*	getCollisionShape()
+	SIMD_FORCE_INLINE btCollisionShape*	getCollisionShape()
 	{
 		return m_collisionShape;
 	}
 
-	
+	SIMD_FORCE_INLINE const btCollisionShape*	getRootCollisionShape() const
+	{
+		return m_rootCollisionShape;
+	}
 
+	SIMD_FORCE_INLINE btCollisionShape*	getRootCollisionShape()
+	{
+		return m_rootCollisionShape;
+	}
+
+	///Avoid using this internal API call
+	///internalSetTemporaryCollisionShape is used to temporary replace the actual collision shape by a child collision shape.
+	void	internalSetTemporaryCollisionShape(btCollisionShape* collisionShape)
+	{
+		m_collisionShape = collisionShape;
+	}
 
 	int	getActivationState() const { return m_activationState1;}
 	
@@ -186,14 +216,9 @@ public:
 	}
 
 	///reserved for Bullet internal usage
-	void*	getInternalOwner()
+	int	getInternalType() const
 	{
-		return m_internalOwner;
+		return m_internalType;
-	}
-
-	const void*	getInternalOwner() const
-	{
-		return m_internalOwner;
 	}
 
 	btTransform&	getWorldTransform()
@@ -331,6 +356,7 @@ public:
 		m_userObjectPointer = userPointer;
 	}
 
+
 	inline bool checkCollideWith(btCollisionObject* co)
 	{
 		if (m_checkCollideWith)
@@ -338,9 +364,6 @@ public:
 
 		return true;
 	}
-
+};
-
-}
-;
 
 #endif //COLLISION_OBJECT_H

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionWorld.cpp

@@ -4,8 +4,8 @@ Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
 
 This software is provided 'as-is', without any express or implied warranty.
 In no event will the authors be held liable for any damages arising from the use of this software.
-Permission is granted to anyone to use this software for any purpose, 
+Permission is granted to anyone to use this software for any purpose,
-including commercial applications, and to alter it and redistribute it freely, 
+including commercial applications, and to alter it and redistribute it freely,
 subject to the following restrictions:
 
 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.
@@ -18,59 +18,57 @@ subject to the following restrictions:
 #include "BulletCollision/CollisionDispatch/btCollisionObject.h"
 #include "BulletCollision/CollisionShapes/btCollisionShape.h"
 #include "BulletCollision/CollisionShapes/btConvexShape.h"
-
+#include "BulletCollision/NarrowPhaseCollision/btGjkEpaPenetrationDepthSolver.h"
 #include "BulletCollision/CollisionShapes/btSphereShape.h" //for raycasting
-#include "BulletCollision/CollisionShapes/btTriangleMeshShape.h" //for raycasting
+#include "BulletCollision/CollisionShapes/btBvhTriangleMeshShape.h" //for raycasting
 #include "BulletCollision/NarrowPhaseCollision/btRaycastCallback.h"
 #include "BulletCollision/CollisionShapes/btCompoundShape.h"
 #include "BulletCollision/NarrowPhaseCollision/btSubSimplexConvexCast.h"
+#include "BulletCollision/NarrowPhaseCollision/btGjkConvexCast.h"
+#include "BulletCollision/NarrowPhaseCollision/btContinuousConvexCollision.h"
+
 #include "BulletCollision/BroadphaseCollision/btBroadphaseInterface.h"
 #include "LinearMath/btAabbUtil2.h"
 #include "LinearMath/btQuickprof.h"
 #include "LinearMath/btStackAlloc.h"
 
+
 //When the user doesn't provide dispatcher or broadphase, create basic versions (and delete them in destructor)
 #include "BulletCollision/CollisionDispatch/btCollisionDispatcher.h"
 #include "BulletCollision/BroadphaseCollision/btSimpleBroadphase.h"
+#include "BulletCollision/CollisionDispatch/btCollisionConfiguration.h"
 
 
-btCollisionWorld::btCollisionWorld(btDispatcher* dispatcher,btOverlappingPairCache* pairCache, int stackSize)
+btCollisionWorld::btCollisionWorld(btDispatcher* dispatcher,btBroadphaseInterface* pairCache, btCollisionConfiguration* collisionConfiguration)
 :m_dispatcher1(dispatcher),
 m_broadphasePairCache(pairCache),
-m_ownsDispatcher(false),
+m_debugDrawer(0)
-m_ownsBroadphasePairCache(false)
 {
-	m_stackAlloc = new btStackAlloc(stackSize);
+	m_stackAlloc = collisionConfiguration->getStackAllocator();
 	m_dispatchInfo.m_stackAllocator = m_stackAlloc;
 }
 
 
 btCollisionWorld::~btCollisionWorld()
 {
-	m_stackAlloc->destroy();
-	delete m_stackAlloc;
 
 	//clean up remaining objects
 	int i;
 	for (i=0;i<m_collisionObjects.size();i++)
 	{
 		btCollisionObject* collisionObject= m_collisionObjects[i];
-		
+
 		btBroadphaseProxy* bp = collisionObject->getBroadphaseHandle();
 		if (bp)
 		{
 			//
 			// only clear the cached algorithms
 			//
-			getBroadphase()->cleanProxyFromPairs(bp);
+			getBroadphase()->getOverlappingPairCache()->cleanProxyFromPairs(bp,m_dispatcher1);
-			getBroadphase()->destroyProxy(bp);
+			getBroadphase()->destroyProxy(bp,m_dispatcher1);
 		}
 	}
 
-	if (m_ownsDispatcher)
-		delete m_dispatcher1;
-	if (m_ownsBroadphasePairCache)
-		delete m_broadphasePairCache;
 
 }
 
@@ -105,66 +103,100 @@ void	btCollisionWorld::addCollisionObject(btCollisionObject* collisionObject,sho
 			type,
 			collisionObject,
 			collisionFilterGroup,
-			collisionFilterMask
+			collisionFilterMask,
+			m_dispatcher1,0
 			))	;
 
-		
+
 
 
 
 }
 
+void	btCollisionWorld::updateAabbs()
+{
+	BT_PROFILE("updateAabbs");
+
+	btTransform predictedTrans;
+	for ( int i=0;i<m_collisionObjects.size();i++)
+	{
+		btCollisionObject* colObj = m_collisionObjects[i];
+
+		//only update aabb of active objects
+		if (colObj->isActive())
+		{
+			btPoint3 minAabb,maxAabb;
+			colObj->getCollisionShape()->getAabb(colObj->getWorldTransform(), minAabb,maxAabb);
+			btBroadphaseInterface* bp = (btBroadphaseInterface*)m_broadphasePairCache;
+
+			//moving objects should be moderately sized, probably something wrong if not
+			if ( colObj->isStaticObject() || ((maxAabb-minAabb).length2() < btScalar(1e12)))
+			{
+				bp->setAabb(colObj->getBroadphaseHandle(),minAabb,maxAabb, m_dispatcher1);
+			} else
+			{
+				//something went wrong, investigate
+				//this assert is unwanted in 3D modelers (danger of loosing work)
+				colObj->setActivationState(DISABLE_SIMULATION);
+
+				static bool reportMe = true;
+				if (reportMe && m_debugDrawer)
+				{
+					reportMe = false;
+					m_debugDrawer->reportErrorWarning("Overflow in AABB, object removed from simulation");
+					m_debugDrawer->reportErrorWarning("If you can reproduce this, please email bugs@continuousphysics.com\n");
+					m_debugDrawer->reportErrorWarning("Please include above information, your Platform, version of OS.\n");
+					m_debugDrawer->reportErrorWarning("Thanks.\n");
+				}
+			}
+		}
+	}
+
+}
 
 
 
 void	btCollisionWorld::performDiscreteCollisionDetection()
 {
+	BT_PROFILE("performDiscreteCollisionDetection");
+
 	btDispatcherInfo& dispatchInfo = getDispatchInfo();
 
-	BEGIN_PROFILE("perform Broadphase Collision Detection");
+	updateAabbs();
 
-
-	//update aabb (of all moved objects)
-
-	btVector3 aabbMin,aabbMax;
-	for (int i=0;i<m_collisionObjects.size();i++)
 	{
-		m_collisionObjects[i]->getCollisionShape()->getAabb(m_collisionObjects[i]->getWorldTransform(),aabbMin,aabbMax);
+		BT_PROFILE("calculateOverlappingPairs");
-		m_broadphasePairCache->setAabb(m_collisionObjects[i]->getBroadphaseHandle(),aabbMin,aabbMax);
+		m_broadphasePairCache->calculateOverlappingPairs(m_dispatcher1);
 	}
 
-	m_broadphasePairCache->refreshOverlappingPairs();
-
-	
-	END_PROFILE("perform Broadphase Collision Detection");
-
-	BEGIN_PROFILE("performDiscreteCollisionDetection");
 
 	btDispatcher* dispatcher = getDispatcher();
-	if (dispatcher)
+	{
-		dispatcher->dispatchAllCollisionPairs(m_broadphasePairCache,dispatchInfo);
+		BT_PROFILE("dispatchAllCollisionPairs");
-
+		if (dispatcher)
-	END_PROFILE("performDiscreteCollisionDetection");
+			dispatcher->dispatchAllCollisionPairs(m_broadphasePairCache->getOverlappingPairCache(),dispatchInfo,m_dispatcher1);
+	}
 
 }
 
 
+
 void	btCollisionWorld::removeCollisionObject(btCollisionObject* collisionObject)
 {
-	
+
-	
+
 	//bool removeFromBroadphase = false;
-	
+
 	{
-		
+
 		btBroadphaseProxy* bp = collisionObject->getBroadphaseHandle();
 		if (bp)
 		{
 			//
 			// only clear the cached algorithms
 			//
-			getBroadphase()->cleanProxyFromPairs(bp);
+			getBroadphase()->getOverlappingPairCache()->cleanProxyFromPairs(bp,m_dispatcher1);
-			getBroadphase()->destroyProxy(bp);
+			getBroadphase()->destroyProxy(bp,m_dispatcher1);
 			collisionObject->setBroadphaseHandle(0);
 		}
 	}
@@ -181,192 +213,457 @@ void	btCollisionWorld::rayTestSingle(const btTransform& rayFromTrans,const btTra
 					  btCollisionObject* collisionObject,
 					  const btCollisionShape* collisionShape,
 					  const btTransform& colObjWorldTransform,
-					  RayResultCallback& resultCallback,
+					  RayResultCallback& resultCallback)
-					  short int collisionFilterMask,
-					  bool faceNormal)
 {
-	
 	btSphereShape pointShape(btScalar(0.0));
 	pointShape.setMargin(0.f);
+	const btConvexShape* castShape = &pointShape;
 
-	objectQuerySingle(&pointShape,rayFromTrans,rayToTrans,
+	if (collisionShape->isConvex())
-					  collisionObject,
+	{
-					  collisionShape,
+		btConvexCast::CastResult castResult;
-					  colObjWorldTransform,
+		castResult.m_fraction = resultCallback.m_closestHitFraction;
-					  resultCallback,collisionFilterMask,faceNormal);
+
+		btConvexShape* convexShape = (btConvexShape*) collisionShape;
+		btVoronoiSimplexSolver	simplexSolver;
+#define USE_SUBSIMPLEX_CONVEX_CAST 1
+#ifdef USE_SUBSIMPLEX_CONVEX_CAST
+		btSubsimplexConvexCast convexCaster(castShape,convexShape,&simplexSolver);
+#else
+		//btGjkConvexCast	convexCaster(castShape,convexShape,&simplexSolver);
+		//btContinuousConvexCollision convexCaster(castShape,convexShape,&simplexSolver,0);
+#endif //#USE_SUBSIMPLEX_CONVEX_CAST
+
+		if (convexCaster.calcTimeOfImpact(rayFromTrans,rayToTrans,colObjWorldTransform,colObjWorldTransform,castResult))
+		{
+			//add hit
+			if (castResult.m_normal.length2() > btScalar(0.0001))
+			{
+				if (castResult.m_fraction < resultCallback.m_closestHitFraction)
+				{
+#ifdef USE_SUBSIMPLEX_CONVEX_CAST
+					//rotate normal into worldspace
+					castResult.m_normal = rayFromTrans.getBasis() * castResult.m_normal;
+#endif //USE_SUBSIMPLEX_CONVEX_CAST
+
+					castResult.m_normal.normalize();
+					btCollisionWorld::LocalRayResult localRayResult
+						(
+							collisionObject,
+							0,
+							castResult.m_normal,
+							castResult.m_fraction
+						);
+
+					bool normalInWorldSpace = true;
+					resultCallback.addSingleResult(localRayResult, normalInWorldSpace);
+
+				}
+			}
+		}
+	} else {
+		if (collisionShape->isConcave())
+		{
+			if (collisionShape->getShapeType()==TRIANGLE_MESH_SHAPE_PROXYTYPE)
+			{
+				///optimized version for btBvhTriangleMeshShape
+				btBvhTriangleMeshShape* triangleMesh = (btBvhTriangleMeshShape*)collisionShape;
+				btTransform worldTocollisionObject = colObjWorldTransform.inverse();
+				btVector3 rayFromLocal = worldTocollisionObject * rayFromTrans.getOrigin();
+				btVector3 rayToLocal = worldTocollisionObject * rayToTrans.getOrigin();
+
+				//ConvexCast::CastResult
+				struct BridgeTriangleRaycastCallback : public btTriangleRaycastCallback
+				{
+					btCollisionWorld::RayResultCallback* m_resultCallback;
+					btCollisionObject*	m_collisionObject;
+					btTriangleMeshShape*	m_triangleMesh;
+
+					BridgeTriangleRaycastCallback( const btVector3& from,const btVector3& to,
+						btCollisionWorld::RayResultCallback* resultCallback, btCollisionObject* collisionObject,btTriangleMeshShape*	triangleMesh):
+						btTriangleRaycastCallback(from,to),
+							m_resultCallback(resultCallback),
+							m_collisionObject(collisionObject),
+							m_triangleMesh(triangleMesh)
+						{
+						}
+
+
+					virtual btScalar reportHit(const btVector3& hitNormalLocal, btScalar hitFraction, int partId, int triangleIndex )
+					{
+						btCollisionWorld::LocalShapeInfo	shapeInfo;
+						shapeInfo.m_shapePart = partId;
+						shapeInfo.m_triangleIndex = triangleIndex;
+
+						btCollisionWorld::LocalRayResult rayResult
+						(m_collisionObject,
+							&shapeInfo,
+							hitNormalLocal,
+							hitFraction);
+
+						bool	normalInWorldSpace = false;
+						return m_resultCallback->addSingleResult(rayResult,normalInWorldSpace);
+					}
+
+				};
+
+				BridgeTriangleRaycastCallback rcb(rayFromLocal,rayToLocal,&resultCallback,collisionObject,triangleMesh);
+				rcb.m_hitFraction = resultCallback.m_closestHitFraction;
+				triangleMesh->performRaycast(&rcb,rayFromLocal,rayToLocal);
+			} else
+			{
+				btTriangleMeshShape* triangleMesh = (btTriangleMeshShape*)collisionShape;
+
+				btTransform worldTocollisionObject = colObjWorldTransform.inverse();
+
+				btVector3 rayFromLocal = worldTocollisionObject * rayFromTrans.getOrigin();
+				btVector3 rayToLocal = worldTocollisionObject * rayToTrans.getOrigin();
+
+				//ConvexCast::CastResult
+
+				struct BridgeTriangleRaycastCallback : public btTriangleRaycastCallback
+				{
+					btCollisionWorld::RayResultCallback* m_resultCallback;
+					btCollisionObject*	m_collisionObject;
+					btTriangleMeshShape*	m_triangleMesh;
+
+					BridgeTriangleRaycastCallback( const btVector3& from,const btVector3& to,
+						btCollisionWorld::RayResultCallback* resultCallback, btCollisionObject* collisionObject,btTriangleMeshShape*	triangleMesh):
+						btTriangleRaycastCallback(from,to),
+							m_resultCallback(resultCallback),
+							m_collisionObject(collisionObject),
+							m_triangleMesh(triangleMesh)
+						{
+						}
+
+
+					virtual btScalar reportHit(const btVector3& hitNormalLocal, btScalar hitFraction, int partId, int triangleIndex )
+					{
+						btCollisionWorld::LocalShapeInfo	shapeInfo;
+						shapeInfo.m_shapePart = partId;
+						shapeInfo.m_triangleIndex = triangleIndex;
+
+						btCollisionWorld::LocalRayResult rayResult
+						(m_collisionObject,
+							&shapeInfo,
+							hitNormalLocal,
+							hitFraction);
+
+						bool	normalInWorldSpace = false;
+						return m_resultCallback->addSingleResult(rayResult,normalInWorldSpace);
+
+
+					}
+
+				};
+
+
+				BridgeTriangleRaycastCallback	rcb(rayFromLocal,rayToLocal,&resultCallback,collisionObject,triangleMesh);
+				rcb.m_hitFraction = resultCallback.m_closestHitFraction;
+
+				btVector3 rayAabbMinLocal = rayFromLocal;
+				rayAabbMinLocal.setMin(rayToLocal);
+				btVector3 rayAabbMaxLocal = rayFromLocal;
+				rayAabbMaxLocal.setMax(rayToLocal);
+
+				triangleMesh->processAllTriangles(&rcb,rayAabbMinLocal,rayAabbMaxLocal);
+			}
+		} else {
+			//todo: use AABB tree or other BVH acceleration structure!
+			if (collisionShape->isCompound())
+			{
+				const btCompoundShape* compoundShape = static_cast<const btCompoundShape*>(collisionShape);
+				int i=0;
+				for (i=0;i<compoundShape->getNumChildShapes();i++)
+				{
+					btTransform childTrans = compoundShape->getChildTransform(i);
+					const btCollisionShape* childCollisionShape = compoundShape->getChildShape(i);
+					btTransform childWorldTrans = colObjWorldTransform * childTrans;
+					rayTestSingle(rayFromTrans,rayToTrans,
+						collisionObject,
+						childCollisionShape,
+						childWorldTrans,
+						resultCallback);
+
+				}
+			}
+		}
+	}
 }
 
-void	btCollisionWorld::objectQuerySingle(const btConvexShape* castShape,const btTransform& rayFromTrans,const btTransform& rayToTrans,
+void	btCollisionWorld::objectQuerySingle(const btConvexShape* castShape,const btTransform& convexFromTrans,const btTransform& convexToTrans,
 					  btCollisionObject* collisionObject,
 					  const btCollisionShape* collisionShape,
 					  const btTransform& colObjWorldTransform,
-					  RayResultCallback& resultCallback,
+					  ConvexResultCallback& resultCallback, btScalar allowedPenetration)
-					  short int collisionFilterMask,
-					  bool faceNormal)
 {
-	
-
 	if (collisionShape->isConvex())
+	{
+		btConvexCast::CastResult castResult;
+		castResult.m_allowedPenetration = allowedPenetration;
+		castResult.m_fraction = btScalar(1.);//??
+
+		btConvexShape* convexShape = (btConvexShape*) collisionShape;
+		btVoronoiSimplexSolver	simplexSolver;
+		btGjkEpaPenetrationDepthSolver	gjkEpaPenetrationSolver;
+		
+		btContinuousConvexCollision convexCaster1(castShape,convexShape,&simplexSolver,&gjkEpaPenetrationSolver);
+		//btGjkConvexCast convexCaster2(castShape,convexShape,&simplexSolver);
+		//btSubsimplexConvexCast convexCaster3(castShape,convexShape,&simplexSolver);
+
+		btConvexCast* castPtr = &convexCaster1;
+	
+	
+		
+		if (castPtr->calcTimeOfImpact(convexFromTrans,convexToTrans,colObjWorldTransform,colObjWorldTransform,castResult))
+		{
+			//add hit
+			if (castResult.m_normal.length2() > btScalar(0.0001))
 			{
-				btConvexCast::CastResult castResult;
+				if (castResult.m_fraction < resultCallback.m_closestHitFraction)
-				castResult.m_fraction = btScalar(1.);//??
-
-				btConvexShape* convexShape = (btConvexShape*) collisionShape;
-				btVoronoiSimplexSolver	simplexSolver;
-				btSubsimplexConvexCast convexCaster(castShape,convexShape,&simplexSolver);
-				//GjkConvexCast	convexCaster(castShape,convexShape,&simplexSolver);
-				//ContinuousConvexCollision convexCaster(castShape,convexShape,&simplexSolver,0);
-				
-				if (convexCaster.calcTimeOfImpact(rayFromTrans,rayToTrans,colObjWorldTransform,colObjWorldTransform,castResult))
 				{
-					//add hit
+					castResult.m_normal.normalize();
-					if (castResult.m_normal.length2() > btScalar(0.0001))
+					btCollisionWorld::LocalConvexResult localConvexResult
-					{
-						castResult.m_normal.normalize();
-						if (castResult.m_fraction < resultCallback.m_closestHitFraction)
-						{
-
-							btCollisionWorld::LocalRayResult localRayResult
 								(
-									collisionObject, 
+									collisionObject,
 									0,
 									castResult.m_normal,
+									castResult.m_hitPoint,
 									castResult.m_fraction
 								);
 
-							resultCallback.AddSingleResult(localRayResult);
+					bool normalInWorldSpace = true;
+					resultCallback.addSingleResult(localConvexResult, normalInWorldSpace);
 
-						}
-					}
 				}
 			}
-			else
+		}
+	} else {
+		if (collisionShape->isConcave())
+		{
+			if (collisionShape->getShapeType()==TRIANGLE_MESH_SHAPE_PROXYTYPE)
 			{
-				
+				btBvhTriangleMeshShape* triangleMesh = (btBvhTriangleMeshShape*)collisionShape;
-				if (collisionShape->isConcave())
+				btTransform worldTocollisionObject = colObjWorldTransform.inverse();
-					{
+				btVector3 convexFromLocal = worldTocollisionObject * convexFromTrans.getOrigin();
+				btVector3 convexToLocal = worldTocollisionObject * convexToTrans.getOrigin();
+				// rotation of box in local mesh space = MeshRotation^-1 * ConvexToRotation
+				btTransform rotationXform = btTransform(worldTocollisionObject.getBasis() * convexToTrans.getBasis());
 
-						btTriangleMeshShape* triangleMesh = (btTriangleMeshShape*)collisionShape;
+				//ConvexCast::CastResult
-						
+				struct BridgeTriangleConvexcastCallback : public btTriangleConvexcastCallback
-						btTransform worldTocollisionObject = colObjWorldTransform.inverse();
+				{
+					btCollisionWorld::ConvexResultCallback* m_resultCallback;
+					btCollisionObject*	m_collisionObject;
+					btTriangleMeshShape*	m_triangleMesh;
 
-						btVector3 rayFromLocal = worldTocollisionObject * rayFromTrans.getOrigin();
+					BridgeTriangleConvexcastCallback(const btConvexShape* castShape, const btTransform& from,const btTransform& to,
-						btVector3 rayToLocal = worldTocollisionObject * rayToTrans.getOrigin();
+						btCollisionWorld::ConvexResultCallback* resultCallback, btCollisionObject* collisionObject,btTriangleMeshShape*	triangleMesh, const btTransform& triangleToWorld):
-
+						btTriangleConvexcastCallback(castShape, from,to, triangleToWorld, triangleMesh->getMargin()),
-						//ConvexCast::CastResult
+							m_resultCallback(resultCallback),
-
+							m_collisionObject(collisionObject),
-						struct BridgeTriangleRaycastCallback : public btTriangleRaycastCallback 
+							m_triangleMesh(triangleMesh)
 						{
-							btCollisionWorld::RayResultCallback* m_resultCallback;
-							btCollisionObject*	m_collisionObject;
-							btTriangleMeshShape*	m_triangleMesh;
-
-							BridgeTriangleRaycastCallback( const btVector3& from,const btVector3& to,bool faceNormal,
-								btCollisionWorld::RayResultCallback* resultCallback, btCollisionObject* collisionObject,btTriangleMeshShape* triangleMesh):
-								btTriangleRaycastCallback(from,to,faceNormal),
-									m_resultCallback(resultCallback),
-									m_collisionObject(collisionObject),
-									m_triangleMesh(triangleMesh)
-								{
-								}
-
-
-							virtual btScalar reportHit(const btVector3& hitNormalLocal, btScalar hitFraction, int partId, int triangleIndex )
-							{
-								btCollisionWorld::LocalShapeInfo	shapeInfo;
-								shapeInfo.m_shapePart = partId;
-								shapeInfo.m_triangleIndex = triangleIndex;
-								shapeInfo.m_triangleShape = m_triangleMesh;
-								
-								btCollisionWorld::LocalRayResult rayResult
-								(m_collisionObject, 
-									&shapeInfo,
-									hitNormalLocal,
-									hitFraction);
-								
-								return m_resultCallback->AddSingleResult(rayResult);
-								
-								
-							}
-	
-						};
-
-
-						BridgeTriangleRaycastCallback	rcb(rayFromLocal,rayToLocal,faceNormal,&resultCallback,collisionObject,triangleMesh);
-						rcb.m_hitFraction = resultCallback.m_closestHitFraction;
-
-						btVector3 rayAabbMinLocal = rayFromLocal;
-						rayAabbMinLocal.setMin(rayToLocal);
-						btVector3 rayAabbMaxLocal = rayFromLocal;
-						rayAabbMaxLocal.setMax(rayToLocal);
-
-						triangleMesh->processAllTriangles(&rcb,rayAabbMinLocal,rayAabbMaxLocal);
-											
-					} else
-					{
-						//todo: use AABB tree or other BVH acceleration structure!
-						if (collisionShape->isCompound())
-						{
-							const btCompoundShape* compoundShape = static_cast<const btCompoundShape*>(collisionShape);
-							int i=0;
-							for (i=0;i<compoundShape->getNumChildShapes();i++)
-							{
-								btTransform childTrans = compoundShape->getChildTransform(i);
-								const btCollisionShape* childCollisionShape = compoundShape->getChildShape(i);
-								btTransform childWorldTrans = colObjWorldTransform * childTrans;
-								objectQuerySingle(castShape, rayFromTrans,rayToTrans,
-									collisionObject,
-									childCollisionShape,
-									childWorldTrans,
-									resultCallback, collisionFilterMask, faceNormal);
-
-							}
-
-
 						}
+
+
+					virtual btScalar reportHit(const btVector3& hitNormalLocal, const btVector3& hitPointLocal, btScalar hitFraction, int partId, int triangleIndex )
+					{
+						btCollisionWorld::LocalShapeInfo	shapeInfo;
+						shapeInfo.m_shapePart = partId;
+						shapeInfo.m_triangleIndex = triangleIndex;
+						if (hitFraction <= m_resultCallback->m_closestHitFraction)
+						{
+
+							btCollisionWorld::LocalConvexResult convexResult
+							(m_collisionObject,
+								&shapeInfo,
+								hitNormalLocal,
+								hitPointLocal,
+								hitFraction);
+
+							bool	normalInWorldSpace = true;
+
+
+							return m_resultCallback->addSingleResult(convexResult,normalInWorldSpace);
+						}
+						return hitFraction;
 					}
+
+				};
+
+				BridgeTriangleConvexcastCallback tccb(castShape, convexFromTrans,convexToTrans,&resultCallback,collisionObject,triangleMesh, colObjWorldTransform);
+				tccb.m_hitFraction = resultCallback.m_closestHitFraction;
+				btVector3 boxMinLocal, boxMaxLocal;
+				castShape->getAabb(rotationXform, boxMinLocal, boxMaxLocal);
+				triangleMesh->performConvexcast(&tccb,convexFromLocal,convexToLocal,boxMinLocal, boxMaxLocal);
+			} else
+			{
+				btBvhTriangleMeshShape* triangleMesh = (btBvhTriangleMeshShape*)collisionShape;
+				btTransform worldTocollisionObject = colObjWorldTransform.inverse();
+				btVector3 convexFromLocal = worldTocollisionObject * convexFromTrans.getOrigin();
+				btVector3 convexToLocal = worldTocollisionObject * convexToTrans.getOrigin();
+				// rotation of box in local mesh space = MeshRotation^-1 * ConvexToRotation
+				btTransform rotationXform = btTransform(worldTocollisionObject.getBasis() * convexToTrans.getBasis());
+
+				//ConvexCast::CastResult
+				struct BridgeTriangleConvexcastCallback : public btTriangleConvexcastCallback
+				{
+					btCollisionWorld::ConvexResultCallback* m_resultCallback;
+					btCollisionObject*	m_collisionObject;
+					btTriangleMeshShape*	m_triangleMesh;
+
+					BridgeTriangleConvexcastCallback(const btConvexShape* castShape, const btTransform& from,const btTransform& to,
+						btCollisionWorld::ConvexResultCallback* resultCallback, btCollisionObject* collisionObject,btTriangleMeshShape*	triangleMesh, const btTransform& triangleToWorld):
+						btTriangleConvexcastCallback(castShape, from,to, triangleToWorld, triangleMesh->getMargin()),
+							m_resultCallback(resultCallback),
+							m_collisionObject(collisionObject),
+							m_triangleMesh(triangleMesh)
+						{
+						}
+
+
+					virtual btScalar reportHit(const btVector3& hitNormalLocal, const btVector3& hitPointLocal, btScalar hitFraction, int partId, int triangleIndex )
+					{
+						btCollisionWorld::LocalShapeInfo	shapeInfo;
+						shapeInfo.m_shapePart = partId;
+						shapeInfo.m_triangleIndex = triangleIndex;
+						if (hitFraction <= m_resultCallback->m_closestHitFraction)
+						{
+
+							btCollisionWorld::LocalConvexResult convexResult
+							(m_collisionObject,
+								&shapeInfo,
+								hitNormalLocal,
+								hitPointLocal,
+								hitFraction);
+
+							bool	normalInWorldSpace = false;
+
+							return m_resultCallback->addSingleResult(convexResult,normalInWorldSpace);
+						}
+						return hitFraction;
+					}
+
+				};
+
+				BridgeTriangleConvexcastCallback tccb(castShape, convexFromTrans,convexToTrans,&resultCallback,collisionObject,triangleMesh, colObjWorldTransform);
+				tccb.m_hitFraction = resultCallback.m_closestHitFraction;
+				btVector3 boxMinLocal, boxMaxLocal;
+				castShape->getAabb(rotationXform, boxMinLocal, boxMaxLocal);
+
+				btVector3 rayAabbMinLocal = convexFromLocal;
+				rayAabbMinLocal.setMin(convexToLocal);
+				btVector3 rayAabbMaxLocal = convexFromLocal;
+				rayAabbMaxLocal.setMax(convexToLocal);
+				rayAabbMinLocal += boxMinLocal;
+				rayAabbMaxLocal += boxMaxLocal;
+				triangleMesh->processAllTriangles(&tccb,rayAabbMinLocal,rayAabbMaxLocal);
 			}
+		} else {
+			//todo: use AABB tree or other BVH acceleration structure!
+			if (collisionShape->isCompound())
+			{
+				const btCompoundShape* compoundShape = static_cast<const btCompoundShape*>(collisionShape);
+				int i=0;
+				for (i=0;i<compoundShape->getNumChildShapes();i++)
+				{
+					btTransform childTrans = compoundShape->getChildTransform(i);
+					const btCollisionShape* childCollisionShape = compoundShape->getChildShape(i);
+					btTransform childWorldTrans = colObjWorldTransform * childTrans;
+					objectQuerySingle(castShape, convexFromTrans,convexToTrans,
+						collisionObject,
+						childCollisionShape,
+						childWorldTrans,
+						resultCallback, allowedPenetration);
+				}
+			}
+		}
+	}
 }
 
-void	btCollisionWorld::rayTest(const btVector3& rayFromWorld, const btVector3& rayToWorld, RayResultCallback& resultCallback,short int collisionFilterMask, bool faceNormal)
+void	btCollisionWorld::rayTest(const btVector3& rayFromWorld, const btVector3& rayToWorld, RayResultCallback& resultCallback) const
 {
 
-	
+
 	btTransform	rayFromTrans,rayToTrans;
 	rayFromTrans.setIdentity();
 	rayFromTrans.setOrigin(rayFromWorld);
 	rayToTrans.setIdentity();
-	
+
 	rayToTrans.setOrigin(rayToWorld);
 
 	/// go over all objects, and if the ray intersects their aabb, do a ray-shape query using convexCaster (CCD)
-	
+
+	int i;
+	for (i=0;i<m_collisionObjects.size();i++)
+	{
+		///terminate further ray tests, once the closestHitFraction reached zero
+		if (resultCallback.m_closestHitFraction == btScalar(0.f))
+			break;
+
+		btCollisionObject*	collisionObject= m_collisionObjects[i];
+		//only perform raycast if filterMask matches
+		if(resultCallback.needsCollision(collisionObject->getBroadphaseHandle())) {
+			//RigidcollisionObject* collisionObject = ctrl->GetRigidcollisionObject();
+			btVector3 collisionObjectAabbMin,collisionObjectAabbMax;
+			collisionObject->getCollisionShape()->getAabb(collisionObject->getWorldTransform(),collisionObjectAabbMin,collisionObjectAabbMax);
+
+			btScalar hitLambda = resultCallback.m_closestHitFraction;
+			btVector3 hitNormal;
+			if (btRayAabb(rayFromWorld,rayToWorld,collisionObjectAabbMin,collisionObjectAabbMax,hitLambda,hitNormal))
+			{
+				rayTestSingle(rayFromTrans,rayToTrans,
+					collisionObject,
+						collisionObject->getCollisionShape(),
+						collisionObject->getWorldTransform(),
+						resultCallback);
+			}
+		}
+
+	}
+
+}
+
+void	btCollisionWorld::convexSweepTest(const btConvexShape* castShape, const btTransform& convexFromWorld, const btTransform& convexToWorld, ConvexResultCallback& resultCallback) const
+{
+	btTransform	convexFromTrans,convexToTrans;
+	convexFromTrans = convexFromWorld;
+	convexToTrans = convexToWorld;
+	btVector3 castShapeAabbMin, castShapeAabbMax;
+	/* Compute AABB that encompasses angular movement */
+	{
+		btVector3 linVel, angVel;
+		btTransformUtil::calculateVelocity (convexFromTrans, convexToTrans, 1.0, linVel, angVel);
+		btTransform R;
+		R.setIdentity ();
+		R.setRotation (convexFromTrans.getRotation());
+		castShape->calculateTemporalAabb (R, linVel, angVel, 1.0, castShapeAabbMin, castShapeAabbMax);
+	}
+
+	/// go over all objects, and if the ray intersects their aabb + cast shape aabb,
+	// do a ray-shape query using convexCaster (CCD)
 	int i;
 	for (i=0;i<m_collisionObjects.size();i++)
 	{
 		btCollisionObject*	collisionObject= m_collisionObjects[i];
 		//only perform raycast if filterMask matches
-		if(collisionObject->getBroadphaseHandle()->m_collisionFilterGroup & collisionFilterMask) { 
+		if(resultCallback.needsCollision(collisionObject->getBroadphaseHandle())) {
 			//RigidcollisionObject* collisionObject = ctrl->GetRigidcollisionObject();
 			btVector3 collisionObjectAabbMin,collisionObjectAabbMax;
 			collisionObject->getCollisionShape()->getAabb(collisionObject->getWorldTransform(),collisionObjectAabbMin,collisionObjectAabbMax);
-
+			AabbExpand (collisionObjectAabbMin, collisionObjectAabbMax, castShapeAabbMin, castShapeAabbMax);
 			btScalar hitLambda = btScalar(1.); //could use resultCallback.m_closestHitFraction, but needs testing
 			btVector3 hitNormal;
-			if (btRayAabb(rayFromWorld,rayToWorld,collisionObjectAabbMin,collisionObjectAabbMax,hitLambda,hitNormal))
+			if (btRayAabb(convexFromWorld.getOrigin(),convexToWorld.getOrigin(),collisionObjectAabbMin,collisionObjectAabbMax,hitLambda,hitNormal))
 			{
-				// before testing this object, verify that it is not filtered out
+				objectQuerySingle(castShape, convexFromTrans,convexToTrans,
-				if (resultCallback.NeedRayCast(collisionObject))
+					collisionObject,
-				{
+						collisionObject->getCollisionShape(),
-					rayTestSingle(rayFromTrans,rayToTrans,
+						collisionObject->getWorldTransform(),
-						collisionObject,
+						resultCallback,
-							collisionObject->getCollisionShape(),
+						getDispatchInfo().m_allowedCcdPenetration);
-							collisionObject->getWorldTransform(),
+			}
-							resultCallback,
-							collisionFilterMask,
-							faceNormal);
-				}
-			}	
 		}
 	}
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btCollisionWorld.h

@@ -68,12 +68,12 @@ class btStackAlloc;
 class btCollisionShape;
 class btConvexShape;
 class btBroadphaseInterface;
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
 #include "btCollisionObject.h"
 #include "btCollisionDispatcher.h" //for definition of btCollisionObjectArray
-#include "../BroadphaseCollision/btOverlappingPairCache.h"
+#include "BulletCollision/BroadphaseCollision/btOverlappingPairCache.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
 ///CollisionWorld is interface and container for the collision detection
 class btCollisionWorld
@@ -90,18 +90,22 @@ protected:
 
 	btStackAlloc*	m_stackAlloc;
 
-	btOverlappingPairCache*	m_broadphasePairCache;
+	btBroadphaseInterface*	m_broadphasePairCache;
-	
-	bool	m_ownsDispatcher;
-	bool	m_ownsBroadphasePairCache;
 
+	btIDebugDraw*	m_debugDrawer;
+
+	
 public:
 
 	//this constructor doesn't own the dispatcher and paircache/broadphase
-	btCollisionWorld(btDispatcher* dispatcher,btOverlappingPairCache* pairCache, int stackSize = 2*1024*1024);
+	btCollisionWorld(btDispatcher* dispatcher,btBroadphaseInterface* broadphasePairCache, btCollisionConfiguration* collisionConfiguration);
 
 	virtual ~btCollisionWorld();
 
+	void	setBroadphase(btBroadphaseInterface*	pairCache)
+	{
+		m_broadphasePairCache = pairCache;
+	}
 
 	btBroadphaseInterface*	getBroadphase()
 	{
@@ -110,7 +114,7 @@ public:
 
 	btOverlappingPairCache*	getPairCache()
 	{
-		return m_broadphasePairCache;
+		return m_broadphasePairCache->getOverlappingPairCache();
 	}
 
 
@@ -119,14 +123,32 @@ public:
 		return m_dispatcher1;
 	}
 
+	const btDispatcher*	getDispatcher() const
+	{
+		return m_dispatcher1;
+	}
+
+	virtual void	updateAabbs();
+
+	virtual void	setDebugDrawer(btIDebugDraw*	debugDrawer)
+	{
+			m_debugDrawer = debugDrawer;
+	}
+
+	virtual btIDebugDraw*	getDebugDrawer()
+	{
+		return m_debugDrawer;
+	}
+
+
 	///LocalShapeInfo gives extra information for complex shapes
 	///Currently, only btTriangleMeshShape is available, so it just contains triangleIndex and subpart
 	struct	LocalShapeInfo
 	{
 		int	m_shapePart;
 		int	m_triangleIndex;
-		// needed in case of compound shape
+		
-		const btCollisionShape*	m_triangleShape;
+		//const btCollisionShape*	m_shapeTemp;
 		//const btTransform*	m_shapeLocalTransform;
 	};
 
@@ -153,32 +175,43 @@ public:
 	///RayResultCallback is used to report new raycast results
 	struct	RayResultCallback
 	{
+		btScalar	m_closestHitFraction;
+		btCollisionObject*		m_collisionObject;
+		short int	m_collisionFilterGroup;
+		short int	m_collisionFilterMask;
+
 		virtual ~RayResultCallback()
 		{
 		}
-		btScalar	m_closestHitFraction;
+		bool	hasHit() const
-		bool	HasHit()
 		{
-			return (m_closestHitFraction < btScalar(1.));
+			return (m_collisionObject != 0);
 		}
 
 		RayResultCallback()
-			:m_closestHitFraction(btScalar(1.))
+			:m_closestHitFraction(btScalar(1.)),
+			m_collisionObject(0),
+			m_collisionFilterGroup(btBroadphaseProxy::DefaultFilter),
+			m_collisionFilterMask(btBroadphaseProxy::AllFilter)
 		{
 		}
-		virtual bool        NeedRayCast(btCollisionObject* object)
+
+		virtual bool needsCollision(btBroadphaseProxy* proxy0) const
 		{
-			return true;
+			bool collides = (proxy0->m_collisionFilterGroup & m_collisionFilterMask) != 0;
+			collides = collides && (m_collisionFilterGroup & proxy0->m_collisionFilterMask);
+			return collides;
 		}
-		virtual	btScalar	AddSingleResult(LocalRayResult& rayResult) = 0;
+
+
+		virtual	btScalar	addSingleResult(LocalRayResult& rayResult,bool normalInWorldSpace) = 0;
 	};
 
 	struct	ClosestRayResultCallback : public RayResultCallback
 	{
 		ClosestRayResultCallback(const btVector3&	rayFromWorld,const btVector3&	rayToWorld)
 		:m_rayFromWorld(rayFromWorld),
-		m_rayToWorld(rayToWorld),
+		m_rayToWorld(rayToWorld)
-		m_collisionObject(0)
 		{
 		}
 
@@ -187,24 +220,121 @@ public:
 
 		btVector3	m_hitNormalWorld;
 		btVector3	m_hitPointWorld;
-		btCollisionObject*	m_collisionObject;
+			
-		
+		virtual	btScalar	addSingleResult(LocalRayResult& rayResult,bool normalInWorldSpace)
-		virtual	btScalar	AddSingleResult(LocalRayResult& rayResult)
 		{
-
+			//caller already does the filter on the m_closestHitFraction
-//caller already does the filter on the m_closestHitFraction
+			btAssert(rayResult.m_hitFraction <= m_closestHitFraction);
-			assert(rayResult.m_hitFraction <= m_closestHitFraction);
 			
 			m_closestHitFraction = rayResult.m_hitFraction;
 			m_collisionObject = rayResult.m_collisionObject;
-			m_hitNormalWorld = m_collisionObject->getWorldTransform().getBasis()*rayResult.m_hitNormalLocal;
+			if (normalInWorldSpace)
+			{
+				m_hitNormalWorld = rayResult.m_hitNormalLocal;
+			} else
+			{
+				///need to transform normal into worldspace
+				m_hitNormalWorld = m_collisionObject->getWorldTransform().getBasis()*rayResult.m_hitNormalLocal;
+			}
 			m_hitPointWorld.setInterpolate3(m_rayFromWorld,m_rayToWorld,rayResult.m_hitFraction);
 			return rayResult.m_hitFraction;
 		}
 	};
 
 
-	
+	struct LocalConvexResult
+	{
+		LocalConvexResult(btCollisionObject*	hitCollisionObject, 
+			LocalShapeInfo*	localShapeInfo,
+			const btVector3&		hitNormalLocal,
+			const btVector3&		hitPointLocal,
+			btScalar hitFraction
+			)
+		:m_hitCollisionObject(hitCollisionObject),
+		m_localShapeInfo(localShapeInfo),
+		m_hitNormalLocal(hitNormalLocal),
+		m_hitPointLocal(hitPointLocal),
+		m_hitFraction(hitFraction)
+		{
+		}
+
+		btCollisionObject*		m_hitCollisionObject;
+		LocalShapeInfo*			m_localShapeInfo;
+		btVector3				m_hitNormalLocal;
+		btVector3				m_hitPointLocal;
+		btScalar				m_hitFraction;
+	};
+
+	///RayResultCallback is used to report new raycast results
+	struct	ConvexResultCallback
+	{
+		btScalar	m_closestHitFraction;
+		short int	m_collisionFilterGroup;
+		short int	m_collisionFilterMask;
+		
+		ConvexResultCallback()
+			:m_closestHitFraction(btScalar(1.)),
+			m_collisionFilterGroup(btBroadphaseProxy::DefaultFilter),
+			m_collisionFilterMask(btBroadphaseProxy::AllFilter)
+		{
+		}
+
+		virtual ~ConvexResultCallback()
+		{
+		}
+		
+		bool	hasHit() const
+		{
+			return (m_closestHitFraction < btScalar(1.));
+		}
+
+		
+
+		virtual bool needsCollision(btBroadphaseProxy* proxy0) const
+		{
+			bool collides = (proxy0->m_collisionFilterGroup & m_collisionFilterMask) != 0;
+			collides = collides && (m_collisionFilterGroup & proxy0->m_collisionFilterMask);
+			return collides;
+		}
+
+		virtual	btScalar	addSingleResult(LocalConvexResult& convexResult,bool normalInWorldSpace) = 0;
+	};
+
+	struct	ClosestConvexResultCallback : public ConvexResultCallback
+	{
+		ClosestConvexResultCallback(const btVector3&	convexFromWorld,const btVector3&	convexToWorld)
+		:m_convexFromWorld(convexFromWorld),
+		m_convexToWorld(convexToWorld),
+		m_hitCollisionObject(0)
+		{
+		}
+
+		btVector3	m_convexFromWorld;//used to calculate hitPointWorld from hitFraction
+		btVector3	m_convexToWorld;
+
+		btVector3	m_hitNormalWorld;
+		btVector3	m_hitPointWorld;
+		btCollisionObject*	m_hitCollisionObject;
+		
+		virtual	btScalar	addSingleResult(LocalConvexResult& convexResult,bool normalInWorldSpace)
+		{
+//caller already does the filter on the m_closestHitFraction
+			btAssert(convexResult.m_hitFraction <= m_closestHitFraction);
+						
+			m_closestHitFraction = convexResult.m_hitFraction;
+			m_hitCollisionObject = convexResult.m_hitCollisionObject;
+			if (normalInWorldSpace)
+			{
+				m_hitNormalWorld = convexResult.m_hitNormalLocal;
+			} else
+			{
+				///need to transform normal into worldspace
+				m_hitNormalWorld = m_hitCollisionObject->getWorldTransform().getBasis()*convexResult.m_hitNormalLocal;
+			}
+			m_hitPointWorld = convexResult.m_hitPointLocal;
+			return convexResult.m_hitFraction;
+		}
+	};
 
 	int	getNumCollisionObjects() const
 	{
@@ -213,7 +343,12 @@ public:
 
 	/// rayTest performs a raycast on all objects in the btCollisionWorld, and calls the resultCallback
 	/// This allows for several queries: first hit, all hits, any hit, dependent on the value returned by the callback.
-	void	rayTest(const btVector3& rayFromWorld, const btVector3& rayToWorld, RayResultCallback& resultCallback, short int collisionFilterMask=-1, bool faceNormal=false);
+	void	rayTest(const btVector3& rayFromWorld, const btVector3& rayToWorld, RayResultCallback& resultCallback) const; 
+
+	// convexTest performs a swept convex cast on all objects in the btCollisionWorld, and calls the resultCallback
+	// This allows for several queries: first hit, all hits, any hit, dependent on the value return by the callback.
+	void    convexSweepTest (const btConvexShape* castShape, const btTransform& from, const btTransform& to, ConvexResultCallback& resultCallback) const;
+
 
 	/// rayTestSingle performs a raycast call and calls the resultCallback. It is used internally by rayTest.
 	/// In a future implementation, we consider moving the ray test as a virtual method in btCollisionShape.
@@ -222,20 +357,16 @@ public:
 					  btCollisionObject* collisionObject,
 					  const btCollisionShape* collisionShape,
 					  const btTransform& colObjWorldTransform,
-					  RayResultCallback& resultCallback, 
+					  RayResultCallback& resultCallback);
-					  short int collisionFilterMask=-1,
-					  bool faceNormal=false);
 
 	/// objectQuerySingle performs a collision detection query and calls the resultCallback. It is used internally by rayTest.
 	static void	objectQuerySingle(const btConvexShape* castShape, const btTransform& rayFromTrans,const btTransform& rayToTrans,
 					  btCollisionObject* collisionObject,
 					  const btCollisionShape* collisionShape,
 					  const btTransform& colObjWorldTransform,
-					  RayResultCallback& resultCallback, 
+					  ConvexResultCallback& resultCallback, btScalar	allowedPenetration);
-					  short int collisionFilterMask=-1,
-					  bool faceNormal=false);
 
-	void	addCollisionObject(btCollisionObject* collisionObject,short int collisionFilterGroup=1,short int collisionFilterMask=1);
+	void	addCollisionObject(btCollisionObject* collisionObject,short int collisionFilterGroup=btBroadphaseProxy::DefaultFilter,short int collisionFilterMask=btBroadphaseProxy::AllFilter);
 
 	btCollisionObjectArray& getCollisionObjectArray()
 	{
@@ -257,6 +388,11 @@ public:
 		return m_dispatchInfo;
 	}
 
+	const btDispatcherInfo& getDispatchInfo() const
+	{
+		return m_dispatchInfo;
+	}
+
 };
 
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btCompoundCollisionAlgorithm.cpp

@@ -19,7 +19,8 @@ subject to the following restrictions:
 
 
 btCompoundCollisionAlgorithm::btCompoundCollisionAlgorithm( const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* body0,btCollisionObject* body1,bool isSwapped)
-:m_isSwapped(isSwapped)
+:btCollisionAlgorithm(ci),
+m_isSwapped(isSwapped)
 {
 	btCollisionObject* colObj = m_isSwapped? body1 : body0;
 	btCollisionObject* otherObj = m_isSwapped? body0 : body1;
@@ -32,11 +33,11 @@ btCompoundCollisionAlgorithm::btCompoundCollisionAlgorithm( const btCollisionAlg
 	m_childCollisionAlgorithms.resize(numChildren);
 	for (i=0;i<numChildren;i++)
 	{
+		btCollisionShape* tmpShape = colObj->getCollisionShape();
 		btCollisionShape* childShape = compoundShape->getChildShape(i);
-		btCollisionShape* orgShape = colObj->getCollisionShape();
+		colObj->internalSetTemporaryCollisionShape( childShape );
-		colObj->setCollisionShape( childShape );
+		m_childCollisionAlgorithms[i] = ci.m_dispatcher1->findAlgorithm(colObj,otherObj);
-		m_childCollisionAlgorithms[i] = ci.m_dispatcher->findAlgorithm(colObj,otherObj);
+		colObj->internalSetTemporaryCollisionShape( tmpShape );
-		colObj->setCollisionShape( orgShape );
 	}
 }
 
@@ -47,7 +48,8 @@ btCompoundCollisionAlgorithm::~btCompoundCollisionAlgorithm()
 	int i;
 	for (i=0;i<numChildren;i++)
 	{
-		delete m_childCollisionAlgorithms[i];
+		m_childCollisionAlgorithms[i]->~btCollisionAlgorithm();
+		m_dispatcher->freeCollisionAlgorithm(m_childCollisionAlgorithms[i]);
 	}
 }
 
@@ -75,17 +77,21 @@ void btCompoundCollisionAlgorithm::processCollision (btCollisionObject* body0,bt
 
 		//backup
 		btTransform	orgTrans = colObj->getWorldTransform();
-		btCollisionShape* orgShape = colObj->getCollisionShape();
+		btTransform	orgInterpolationTrans = colObj->getInterpolationWorldTransform();
 
 		const btTransform& childTrans = compoundShape->getChildTransform(i);
-		//btTransform	newChildWorldTrans = orgTrans*childTrans ;
+		btTransform	newChildWorldTrans = orgTrans*childTrans ;
-		colObj->setWorldTransform( orgTrans*childTrans );
+		colObj->setWorldTransform( newChildWorldTrans);
+		colObj->setInterpolationWorldTransform(newChildWorldTrans);
+
 		//the contactpoint is still projected back using the original inverted worldtrans
-		colObj->setCollisionShape( childShape );
+		btCollisionShape* tmpShape = colObj->getCollisionShape();
+		colObj->internalSetTemporaryCollisionShape( childShape );
 		m_childCollisionAlgorithms[i]->processCollision(colObj,otherObj,dispatchInfo,resultOut);
 		//revert back
-		colObj->setCollisionShape( orgShape);
+		colObj->internalSetTemporaryCollisionShape( tmpShape);
 		colObj->setWorldTransform(  orgTrans );
+		colObj->setInterpolationWorldTransform(orgInterpolationTrans);
 	}
 }
 
@@ -117,20 +123,20 @@ btScalar	btCompoundCollisionAlgorithm::calculateTimeOfImpact(btCollisionObject*
 
 		//backup
 		btTransform	orgTrans = colObj->getWorldTransform();
-		btCollisionShape* orgShape = colObj->getCollisionShape();
+	
-
 		const btTransform& childTrans = compoundShape->getChildTransform(i);
 		//btTransform	newChildWorldTrans = orgTrans*childTrans ;
 		colObj->setWorldTransform( orgTrans*childTrans );
 
-		colObj->setCollisionShape( childShape );
+		btCollisionShape* tmpShape = colObj->getCollisionShape();
+		colObj->internalSetTemporaryCollisionShape( childShape );
 		btScalar frac = m_childCollisionAlgorithms[i]->calculateTimeOfImpact(colObj,otherObj,dispatchInfo,resultOut);
 		if (frac<hitFraction)
 		{
 			hitFraction = frac;
 		}
 		//revert back
-		colObj->setCollisionShape( orgShape);
+		colObj->internalSetTemporaryCollisionShape( tmpShape);
 		colObj->setWorldTransform( orgTrans);
 	}
 	return hitFraction;

extern/bullet2/src/BulletCollision/CollisionDispatch/btCompoundCollisionAlgorithm.h

@@ -16,15 +16,16 @@ subject to the following restrictions:
 #ifndef COMPOUND_COLLISION_ALGORITHM_H
 #define COMPOUND_COLLISION_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../BroadphaseCollision/btDispatcher.h"
+#include "BulletCollision/BroadphaseCollision/btDispatcher.h"
-#include "../BroadphaseCollision/btBroadphaseInterface.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseInterface.h"
 
-#include "../NarrowPhaseCollision/btPersistentManifold.h"
+#include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
 class btDispatcher;
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
 #include "btCollisionCreateFunc.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
+class btDispatcher;
 
 /// btCompoundCollisionAlgorithm  supports collision between CompoundCollisionShapes and other collision shapes
 /// Place holder, not fully implemented yet
@@ -43,11 +44,21 @@ public:
 
 	btScalar	calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		int i;
+		for (i=0;i<m_childCollisionAlgorithms.size();i++)
+		{
+			m_childCollisionAlgorithms[i]->getAllContactManifolds(manifoldArray);
+		}
+	}
+
 	struct CreateFunc :public 	btCollisionAlgorithmCreateFunc
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btCompoundCollisionAlgorithm(ci,body0,body1,false);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btCompoundCollisionAlgorithm));
+			return new(mem) btCompoundCollisionAlgorithm(ci,body0,body1,false);
 		}
 	};
 
@@ -55,7 +66,8 @@ public:
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btCompoundCollisionAlgorithm(ci,body0,body1,true);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btCompoundCollisionAlgorithm));
+			return new(mem) btCompoundCollisionAlgorithm(ci,body0,body1,true);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexConcaveCollisionAlgorithm.cpp

@@ -29,7 +29,7 @@ subject to the following restrictions:
 btConvexConcaveCollisionAlgorithm::btConvexConcaveCollisionAlgorithm( const btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1,bool isSwapped)
 : btCollisionAlgorithm(ci),
 m_isSwapped(isSwapped),
-m_btConvexTriangleCallback(ci.m_dispatcher,body0,body1,isSwapped)
+m_btConvexTriangleCallback(ci.m_dispatcher1,body0,body1,isSwapped)
 {
 }
 
@@ -37,6 +37,13 @@ btConvexConcaveCollisionAlgorithm::~btConvexConcaveCollisionAlgorithm()
 {
 }
 
+void	btConvexConcaveCollisionAlgorithm::getAllContactManifolds(btManifoldArray&	manifoldArray)
+{
+	if (m_btConvexTriangleCallback.m_manifoldPtr)
+	{
+		manifoldArray.push_back(m_btConvexTriangleCallback.m_manifoldPtr);
+	}
+}
 
 
 btConvexTriangleCallback::btConvexTriangleCallback(btDispatcher*  dispatcher,btCollisionObject* body0,btCollisionObject* body1,bool isSwapped):
@@ -79,7 +86,7 @@ void btConvexTriangleCallback::processTriangle(btVector3* triangle,int partId, i
 	//aabb filter is already applied!	
 
 	btCollisionAlgorithmConstructionInfo ci;
-	ci.m_dispatcher = m_dispatcher;
+	ci.m_dispatcher1 = m_dispatcher;
 
 	btCollisionObject* ob = static_cast<btCollisionObject*>(m_triBody);
 
@@ -109,13 +116,11 @@ void btConvexTriangleCallback::processTriangle(btVector3* triangle,int partId, i
 	{
 		btTriangleShape tm(triangle[0],triangle[1],triangle[2]);	
 		tm.setMargin(m_collisionMarginTriangle);
-	
 		
 		btCollisionShape* tmpShape = ob->getCollisionShape();
-		ob->setCollisionShape( &tm );
+		ob->internalSetTemporaryCollisionShape( &tm );
 		
-
+		btCollisionAlgorithm* colAlgo = ci.m_dispatcher1->findAlgorithm(m_convexBody,m_triBody,m_manifoldPtr);
-		btCollisionAlgorithm* colAlgo = ci.m_dispatcher->findAlgorithm(m_convexBody,m_triBody,m_manifoldPtr);
 		///this should use the btDispatcher, so the actual registered algorithm is used
 		//		btConvexConvexAlgorithm cvxcvxalgo(m_manifoldPtr,ci,m_convexBody,m_triBody);
 
@@ -123,12 +128,11 @@ void btConvexTriangleCallback::processTriangle(btVector3* triangle,int partId, i
 	//	cvxcvxalgo.setShapeIdentifiers(-1,-1,partId,triangleIndex);
 //		cvxcvxalgo.processCollision(m_convexBody,m_triBody,*m_dispatchInfoPtr,m_resultOut);
 		colAlgo->processCollision(m_convexBody,m_triBody,*m_dispatchInfoPtr,m_resultOut);
-		delete colAlgo;
+		colAlgo->~btCollisionAlgorithm();
-		ob->setCollisionShape( tmpShape );
+		ci.m_dispatcher1->freeCollisionAlgorithm(colAlgo);
-
+		ob->internalSetTemporaryCollisionShape( tmpShape);
 	}
 
-	
 
 }
 
@@ -188,9 +192,10 @@ void btConvexConcaveCollisionAlgorithm::processCollision (btCollisionObject* bod
 
 			concaveShape->processAllTriangles( &m_btConvexTriangleCallback,m_btConvexTriangleCallback.getAabbMin(),m_btConvexTriangleCallback.getAabbMax());
 			
+			resultOut->refreshContactPoints();
 	
 		}
-
+	
 	}
 
 }

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexConcaveCollisionAlgorithm.h

@@ -16,13 +16,13 @@ subject to the following restrictions:
 #ifndef CONVEX_CONCAVE_COLLISION_ALGORITHM_H
 #define CONVEX_CONCAVE_COLLISION_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../BroadphaseCollision/btDispatcher.h"
+#include "BulletCollision/BroadphaseCollision/btDispatcher.h"
-#include "../BroadphaseCollision/btBroadphaseInterface.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseInterface.h"
-#include "../CollisionShapes/btTriangleCallback.h"
+#include "BulletCollision/CollisionShapes/btTriangleCallback.h"
-#include "../NarrowPhaseCollision/btPersistentManifold.h"
+#include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
 class btDispatcher;
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
 #include "btCollisionCreateFunc.h"
 
 ///For each triangle in the concave mesh that overlaps with the AABB of a convex (m_convexProxy), processTriangle is called.
@@ -55,11 +55,11 @@ int	m_triangleCount;
 	
 	void clearCache();
 
-	inline const btVector3& getAabbMin() const
+	SIMD_FORCE_INLINE const btVector3& getAabbMin() const
 	{
 		return m_aabbMin;
 	}
-	inline const btVector3& getAabbMax() const
+	SIMD_FORCE_INLINE const btVector3& getAabbMax() const
 	{
 		return m_aabbMax;
 	}
@@ -88,13 +88,16 @@ public:
 
 	btScalar	calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray);
+	
 	void	clearCache();
 
 	struct CreateFunc :public 	btCollisionAlgorithmCreateFunc
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btConvexConcaveCollisionAlgorithm(ci,body0,body1,false);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btConvexConcaveCollisionAlgorithm));
+			return new(mem) btConvexConcaveCollisionAlgorithm(ci,body0,body1,false);
 		}
 	};
 
@@ -102,7 +105,8 @@ public:
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btConvexConcaveCollisionAlgorithm(ci,body0,body1,true);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btConvexConcaveCollisionAlgorithm));
+			return new(mem) btConvexConcaveCollisionAlgorithm(ci,body0,body1,true);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexConvexAlgorithm.cpp

@@ -15,7 +15,7 @@ subject to the following restrictions:
 
 #include "btConvexConvexAlgorithm.h"
 
-#include <stdio.h>
+//#include <stdio.h>
 #include "BulletCollision/NarrowPhaseCollision/btDiscreteCollisionDetectorInterface.h"
 #include "BulletCollision/BroadphaseCollision/btBroadphaseInterface.h"
 #include "BulletCollision/CollisionDispatch/btCollisionObject.h"
@@ -33,7 +33,6 @@ subject to the following restrictions:
 
 
 
-#include "BulletCollision/CollisionShapes/btMinkowskiSumShape.h"
 #include "BulletCollision/NarrowPhaseCollision/btVoronoiSimplexSolver.h"
 #include "BulletCollision/CollisionShapes/btSphereShape.h"
 
@@ -48,26 +47,16 @@ subject to the following restrictions:
 
 
 
-btConvexConvexAlgorithm::CreateFunc::CreateFunc()
+
-{
-	m_ownsSolvers = true;
-	m_simplexSolver = new btVoronoiSimplexSolver();
-	m_pdSolver = new btGjkEpaPenetrationDepthSolver;
-}
 
 btConvexConvexAlgorithm::CreateFunc::CreateFunc(btSimplexSolverInterface*			simplexSolver, btConvexPenetrationDepthSolver* pdSolver)
 {
-	m_ownsSolvers = false;
 	m_simplexSolver = simplexSolver;
 	m_pdSolver = pdSolver;
 }
 
 btConvexConvexAlgorithm::CreateFunc::~CreateFunc() 
 { 
-   if (m_ownsSolvers){ 
-      delete m_simplexSolver; 
-      delete m_pdSolver; 
-   } 
 }
 
 btConvexConvexAlgorithm::btConvexConvexAlgorithm(btPersistentManifold* mf,const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* body0,btCollisionObject* body1,btSimplexSolverInterface* simplexSolver, btConvexPenetrationDepthSolver* pdSolver)
@@ -152,6 +141,11 @@ void btConvexConvexAlgorithm ::processCollision (btCollisionObject* body0,btColl
 	m_gjkPairDetector.getClosestPoints(input,*resultOut,dispatchInfo.m_debugDraw);
 #endif
 
+	if (m_ownManifold)
+	{
+		resultOut->refreshContactPoints();
+	}
+
 }
 
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexConvexAlgorithm.h

@@ -16,12 +16,13 @@ subject to the following restrictions:
 #ifndef CONVEX_CONVEX_ALGORITHM_H
 #define CONVEX_CONVEX_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../NarrowPhaseCollision/btGjkPairDetector.h"
+#include "BulletCollision/NarrowPhaseCollision/btGjkPairDetector.h"
-#include "../NarrowPhaseCollision/btPersistentManifold.h"
+#include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../NarrowPhaseCollision/btVoronoiSimplexSolver.h"
+#include "BulletCollision/NarrowPhaseCollision/btVoronoiSimplexSolver.h"
 #include "btCollisionCreateFunc.h"
+#include "btCollisionDispatcher.h"
 
 class btConvexPenetrationDepthSolver;
 
@@ -46,6 +47,14 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		///should we use m_ownManifold to avoid adding duplicates?
+		if (m_manifoldPtr && m_ownManifold)
+			manifoldArray.push_back(m_manifoldPtr);
+	}
+
+
 	void	setLowLevelOfDetail(bool useLowLevel);
 
 
@@ -58,15 +67,15 @@ public:
 	{
 		btConvexPenetrationDepthSolver*		m_pdSolver;
 		btSimplexSolverInterface*			m_simplexSolver;
-		bool	m_ownsSolvers;
 		
 		CreateFunc(btSimplexSolverInterface*			simplexSolver, btConvexPenetrationDepthSolver* pdSolver);
-		CreateFunc();
+		
 		virtual ~CreateFunc();
 
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btConvexConvexAlgorithm(ci.m_manifold,ci,body0,body1,m_simplexSolver,m_pdSolver);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btConvexConvexAlgorithm));
+			return new(mem) btConvexConvexAlgorithm(ci.m_manifold,ci,body0,body1,m_simplexSolver,m_pdSolver);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexPlaneCollisionAlgorithm.cpp

@@ -0,0 +1,108 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "btConvexPlaneCollisionAlgorithm.h"
+
+#include "BulletCollision/CollisionDispatch/btCollisionDispatcher.h"
+#include "BulletCollision/CollisionDispatch/btCollisionObject.h"
+#include "BulletCollision/CollisionShapes/btConvexShape.h"
+#include "BulletCollision/CollisionShapes/btStaticPlaneShape.h"
+
+//#include <stdio.h>
+
+btConvexPlaneCollisionAlgorithm::btConvexPlaneCollisionAlgorithm(btPersistentManifold* mf,const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* col0,btCollisionObject* col1, bool isSwapped)
+: btCollisionAlgorithm(ci),
+m_ownManifold(false),
+m_manifoldPtr(mf),
+m_isSwapped(isSwapped)
+{
+	btCollisionObject* convexObj = m_isSwapped? col1 : col0;
+	btCollisionObject* planeObj = m_isSwapped? col0 : col1;
+	
+	if (!m_manifoldPtr && m_dispatcher->needsCollision(convexObj,planeObj))
+	{
+		m_manifoldPtr = m_dispatcher->getNewManifold(convexObj,planeObj);
+		m_ownManifold = true;
+	}
+}
+
+
+btConvexPlaneCollisionAlgorithm::~btConvexPlaneCollisionAlgorithm()
+{
+	if (m_ownManifold)
+	{
+		if (m_manifoldPtr)
+			m_dispatcher->releaseManifold(m_manifoldPtr);
+	}
+}
+
+
+
+void btConvexPlaneCollisionAlgorithm::processCollision (btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut)
+{
+	(void)dispatchInfo;
+	(void)resultOut;
+	if (!m_manifoldPtr)
+		return;
+
+	btCollisionObject* convexObj = m_isSwapped? body1 : body0;
+	btCollisionObject* planeObj = m_isSwapped? body0: body1;
+
+	btConvexShape* convexShape = (btConvexShape*) convexObj->getCollisionShape();
+	btStaticPlaneShape* planeShape = (btStaticPlaneShape*) planeObj->getCollisionShape();
+
+	bool hasCollision = false;
+	const btVector3& planeNormal = planeShape->getPlaneNormal();
+	const btScalar& planeConstant = planeShape->getPlaneConstant();
+	btTransform planeInConvex;
+	planeInConvex= convexObj->getWorldTransform().inverse() * planeObj->getWorldTransform();
+	btTransform convexInPlaneTrans;
+	convexInPlaneTrans= planeObj->getWorldTransform().inverse() * convexObj->getWorldTransform();
+
+	btVector3 vtx = convexShape->localGetSupportingVertex(planeInConvex.getBasis()*-planeNormal);
+	btVector3 vtxInPlane = convexInPlaneTrans(vtx);
+	btScalar distance = (planeNormal.dot(vtxInPlane) - planeConstant);
+
+	btVector3 vtxInPlaneProjected = vtxInPlane - distance*planeNormal;
+	btVector3 vtxInPlaneWorld = planeObj->getWorldTransform() * vtxInPlaneProjected;
+
+	hasCollision = distance < m_manifoldPtr->getContactBreakingThreshold();
+	resultOut->setPersistentManifold(m_manifoldPtr);
+	if (hasCollision)
+	{
+		/// report a contact. internally this will be kept persistent, and contact reduction is done
+		btVector3 normalOnSurfaceB = planeObj->getWorldTransform().getBasis() * planeNormal;
+		btVector3 pOnB = vtxInPlaneWorld;
+		resultOut->addContactPoint(normalOnSurfaceB,pOnB,distance);
+	}
+	if (m_ownManifold)
+	{
+		if (m_manifoldPtr->getNumContacts())
+		{
+			resultOut->refreshContactPoints();
+		}
+	}
+}
+
+btScalar btConvexPlaneCollisionAlgorithm::calculateTimeOfImpact(btCollisionObject* col0,btCollisionObject* col1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut)
+{
+	(void)resultOut;
+	(void)dispatchInfo;
+	(void)col0;
+	(void)col1;
+
+	//not yet
+	return btScalar(1.);
+}

extern/bullet2/src/BulletCollision/CollisionDispatch/btConvexPlaneCollisionAlgorithm.h

@@ -0,0 +1,71 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef CONVEX_PLANE_COLLISION_ALGORITHM_H
+#define CONVEX_PLANE_COLLISION_ALGORITHM_H
+
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/CollisionDispatch/btCollisionCreateFunc.h"
+class btPersistentManifold;
+#include "btCollisionDispatcher.h"
+
+#include "LinearMath/btVector3.h"
+
+/// btSphereBoxCollisionAlgorithm  provides sphere-box collision detection.
+/// Other features are frame-coherency (persistent data) and collision response.
+class btConvexPlaneCollisionAlgorithm : public btCollisionAlgorithm
+{
+	bool	m_ownManifold;
+	btPersistentManifold*	m_manifoldPtr;
+	bool	m_isSwapped;
+	
+public:
+
+	btConvexPlaneCollisionAlgorithm(btPersistentManifold* mf,const btCollisionAlgorithmConstructionInfo& ci,btCollisionObject* col0,btCollisionObject* col1, bool isSwapped);
+
+	virtual ~btConvexPlaneCollisionAlgorithm();
+
+	virtual void processCollision (btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
+
+	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
+
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		if (m_manifoldPtr && m_ownManifold)
+		{
+			manifoldArray.push_back(m_manifoldPtr);
+		}
+	}
+
+	struct CreateFunc :public 	btCollisionAlgorithmCreateFunc
+	{
+		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
+		{
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btConvexPlaneCollisionAlgorithm));
+			if (!m_swapped)
+			{
+				return new(mem) btConvexPlaneCollisionAlgorithm(0,ci,body0,body1,false);
+			} else
+			{
+				return new(mem) btConvexPlaneCollisionAlgorithm(0,ci,body0,body1,true);
+			}
+		}
+	};
+
+};
+
+#endif //CONVEX_PLANE_COLLISION_ALGORITHM_H
+

extern/bullet2/src/BulletCollision/CollisionDispatch/btDefaultCollisionConfiguration.cpp

@@ -0,0 +1,281 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "btDefaultCollisionConfiguration.h"
+
+#include "BulletCollision/CollisionDispatch/btConvexConvexAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btEmptyCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btConvexConcaveCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btCompoundCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btConvexPlaneCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btBoxBoxCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btSphereSphereCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btSphereBoxCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btSphereTriangleCollisionAlgorithm.h"
+#include "BulletCollision/NarrowPhaseCollision/btGjkEpaPenetrationDepthSolver.h"
+#include "BulletCollision/NarrowPhaseCollision/btMinkowskiPenetrationDepthSolver.h"
+#include "BulletCollision/NarrowPhaseCollision/btVoronoiSimplexSolver.h"
+
+
+
+#include "LinearMath/btStackAlloc.h"
+#include "LinearMath/btPoolAllocator.h"
+
+
+
+
+
+btDefaultCollisionConfiguration::btDefaultCollisionConfiguration(const btDefaultCollisionConstructionInfo& constructionInfo)
+//btDefaultCollisionConfiguration::btDefaultCollisionConfiguration(btStackAlloc*	stackAlloc,btPoolAllocator*	persistentManifoldPool,btPoolAllocator*	collisionAlgorithmPool)
+{
+
+	void* mem = btAlignedAlloc(sizeof(btVoronoiSimplexSolver),16);
+	m_simplexSolver = new (mem)btVoronoiSimplexSolver();
+	
+#define USE_EPA 1
+#ifdef USE_EPA
+	mem = btAlignedAlloc(sizeof(btGjkEpaPenetrationDepthSolver),16);
+	m_pdSolver = new (mem)btGjkEpaPenetrationDepthSolver;
+#else
+	mem = btAlignedAlloc(sizeof(btMinkowskiPenetrationDepthSolver),16);
+	m_pdSolver = new (mem)btMinkowskiPenetrationDepthSolver;
+#endif//USE_EPA	
+	
+
+	//default CreationFunctions, filling the m_doubleDispatch table
+	mem = btAlignedAlloc(sizeof(btConvexConvexAlgorithm::CreateFunc),16);
+	m_convexConvexCreateFunc = new(mem) btConvexConvexAlgorithm::CreateFunc(m_simplexSolver,m_pdSolver);
+	mem = btAlignedAlloc(sizeof(btConvexConcaveCollisionAlgorithm::CreateFunc),16);
+	m_convexConcaveCreateFunc = new (mem)btConvexConcaveCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc(sizeof(btConvexConcaveCollisionAlgorithm::CreateFunc),16);
+	m_swappedConvexConcaveCreateFunc = new (mem)btConvexConcaveCollisionAlgorithm::SwappedCreateFunc;
+	mem = btAlignedAlloc(sizeof(btCompoundCollisionAlgorithm::CreateFunc),16);
+	m_compoundCreateFunc = new (mem)btCompoundCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc(sizeof(btCompoundCollisionAlgorithm::SwappedCreateFunc),16);
+	m_swappedCompoundCreateFunc = new (mem)btCompoundCollisionAlgorithm::SwappedCreateFunc;
+	mem = btAlignedAlloc(sizeof(btEmptyAlgorithm::CreateFunc),16);
+	m_emptyCreateFunc = new(mem) btEmptyAlgorithm::CreateFunc;
+	
+	mem = btAlignedAlloc(sizeof(btSphereSphereCollisionAlgorithm::CreateFunc),16);
+	m_sphereSphereCF = new(mem) btSphereSphereCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc(sizeof(btSphereBoxCollisionAlgorithm::CreateFunc),16);
+	m_sphereBoxCF = new(mem) btSphereBoxCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc(sizeof(btSphereBoxCollisionAlgorithm::CreateFunc),16);
+	m_boxSphereCF = new (mem)btSphereBoxCollisionAlgorithm::CreateFunc;
+	m_boxSphereCF->m_swapped = true;
+	mem = btAlignedAlloc(sizeof(btSphereTriangleCollisionAlgorithm::CreateFunc),16);
+	m_sphereTriangleCF = new (mem)btSphereTriangleCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc(sizeof(btSphereTriangleCollisionAlgorithm::CreateFunc),16);
+	m_triangleSphereCF = new (mem)btSphereTriangleCollisionAlgorithm::CreateFunc;
+	m_triangleSphereCF->m_swapped = true;
+	
+	mem = btAlignedAlloc(sizeof(btBoxBoxCollisionAlgorithm::CreateFunc),16);
+	m_boxBoxCF = new(mem)btBoxBoxCollisionAlgorithm::CreateFunc;
+
+	//convex versus plane
+	mem = btAlignedAlloc (sizeof(btConvexPlaneCollisionAlgorithm::CreateFunc),16);
+	m_convexPlaneCF = new (mem) btConvexPlaneCollisionAlgorithm::CreateFunc;
+	mem = btAlignedAlloc (sizeof(btConvexPlaneCollisionAlgorithm::CreateFunc),16);
+	m_planeConvexCF = new (mem) btConvexPlaneCollisionAlgorithm::CreateFunc;
+	m_planeConvexCF->m_swapped = true;
+	
+	///calculate maximum element size, big enough to fit any collision algorithm in the memory pool
+	int maxSize = sizeof(btConvexConvexAlgorithm);
+	int maxSize2 = sizeof(btConvexConcaveCollisionAlgorithm);
+	int maxSize3 = sizeof(btCompoundCollisionAlgorithm);
+	int maxSize4 = sizeof(btEmptyAlgorithm);
+	
+	int	collisionAlgorithmMaxElementSize = btMax(maxSize,maxSize2);
+	collisionAlgorithmMaxElementSize = btMax(collisionAlgorithmMaxElementSize,maxSize3);
+	collisionAlgorithmMaxElementSize = btMax(collisionAlgorithmMaxElementSize,maxSize4);
+
+	if (constructionInfo.m_stackAlloc)
+	{
+		m_ownsStackAllocator = false;
+		this->m_stackAlloc = constructionInfo.m_stackAlloc;
+	} else
+	{
+		m_ownsStackAllocator = true;
+		void* mem = btAlignedAlloc(sizeof(btStackAlloc),16);
+		m_stackAlloc = new(mem)btStackAlloc(constructionInfo.m_defaultStackAllocatorSize);
+	}
+		
+	if (constructionInfo.m_persistentManifoldPool)
+	{
+		m_ownsPersistentManifoldPool = false;
+		m_persistentManifoldPool = constructionInfo.m_persistentManifoldPool;
+	} else
+	{
+		m_ownsPersistentManifoldPool = true;
+		void* mem = btAlignedAlloc(sizeof(btPoolAllocator),16);
+		m_persistentManifoldPool = new (mem) btPoolAllocator(sizeof(btPersistentManifold),constructionInfo.m_defaultMaxPersistentManifoldPoolSize);
+	}
+	
+	if (constructionInfo.m_collisionAlgorithmPool)
+	{
+		m_ownsCollisionAlgorithmPool = false;
+		m_collisionAlgorithmPool = constructionInfo.m_collisionAlgorithmPool;
+	} else
+	{
+		m_ownsCollisionAlgorithmPool = true;
+		void* mem = btAlignedAlloc(sizeof(btPoolAllocator),16);
+		m_collisionAlgorithmPool = new(mem) btPoolAllocator(collisionAlgorithmMaxElementSize,constructionInfo.m_defaultMaxCollisionAlgorithmPoolSize);
+	}
+
+
+}
+
+btDefaultCollisionConfiguration::~btDefaultCollisionConfiguration()
+{
+	if (m_ownsStackAllocator)
+	{
+		m_stackAlloc->destroy();
+		m_stackAlloc->~btStackAlloc();
+		btAlignedFree(m_stackAlloc);
+	}
+	if (m_ownsCollisionAlgorithmPool)
+	{
+		m_collisionAlgorithmPool->~btPoolAllocator();
+		btAlignedFree(m_collisionAlgorithmPool);
+	}
+	if (m_ownsPersistentManifoldPool)
+	{
+		m_persistentManifoldPool->~btPoolAllocator();
+		btAlignedFree(m_persistentManifoldPool);
+	}
+
+	m_convexConvexCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree(	m_convexConvexCreateFunc);
+
+	m_convexConcaveCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_convexConcaveCreateFunc);
+	m_swappedConvexConcaveCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_swappedConvexConcaveCreateFunc);
+
+	m_compoundCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_compoundCreateFunc);
+
+	m_swappedCompoundCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_swappedCompoundCreateFunc);
+
+	m_emptyCreateFunc->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_emptyCreateFunc);
+
+	m_sphereSphereCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_sphereSphereCF);
+
+	m_sphereBoxCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_sphereBoxCF);
+	m_boxSphereCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_boxSphereCF);
+	m_sphereTriangleCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_sphereTriangleCF);
+	m_triangleSphereCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_triangleSphereCF);
+	m_boxBoxCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_boxBoxCF);
+
+	m_convexPlaneCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_convexPlaneCF);
+	m_planeConvexCF->~btCollisionAlgorithmCreateFunc();
+	btAlignedFree( m_planeConvexCF);
+
+	m_simplexSolver->~btVoronoiSimplexSolver();
+	btAlignedFree(m_simplexSolver);
+
+	m_pdSolver->~btConvexPenetrationDepthSolver();
+	
+	btAlignedFree(m_pdSolver);
+
+
+}
+
+
+btCollisionAlgorithmCreateFunc* btDefaultCollisionConfiguration::getCollisionAlgorithmCreateFunc(int proxyType0,int proxyType1)
+{
+
+
+
+	if ((proxyType0 == SPHERE_SHAPE_PROXYTYPE) && (proxyType1==SPHERE_SHAPE_PROXYTYPE))
+	{
+		return	m_sphereSphereCF;
+	}
+
+	if ((proxyType0 == SPHERE_SHAPE_PROXYTYPE) && (proxyType1==BOX_SHAPE_PROXYTYPE))
+	{
+		return	m_sphereBoxCF;
+	}
+
+	if ((proxyType0 == BOX_SHAPE_PROXYTYPE ) && (proxyType1==SPHERE_SHAPE_PROXYTYPE))
+	{
+		return	m_boxSphereCF;
+	}
+
+	if ((proxyType0 == SPHERE_SHAPE_PROXYTYPE ) && (proxyType1==TRIANGLE_SHAPE_PROXYTYPE))
+	{
+		return	m_sphereTriangleCF;
+	}
+
+	if ((proxyType0 == TRIANGLE_SHAPE_PROXYTYPE  ) && (proxyType1==SPHERE_SHAPE_PROXYTYPE))
+	{
+		return	m_triangleSphereCF;
+	} 
+
+	if ((proxyType0 == BOX_SHAPE_PROXYTYPE) && (proxyType1 == BOX_SHAPE_PROXYTYPE))
+	{
+		return m_boxBoxCF;
+	}
+	
+	if (btBroadphaseProxy::isConvex(proxyType0) && (proxyType1 == STATIC_PLANE_PROXYTYPE))
+	{
+		return m_convexPlaneCF;
+	}
+
+	if (btBroadphaseProxy::isConvex(proxyType1) && (proxyType0 == STATIC_PLANE_PROXYTYPE))
+	{
+		return m_planeConvexCF;
+	}
+	
+
+
+	if (btBroadphaseProxy::isConvex(proxyType0) && btBroadphaseProxy::isConvex(proxyType1))
+	{
+		return m_convexConvexCreateFunc;
+	}
+
+	if (btBroadphaseProxy::isConvex(proxyType0) && btBroadphaseProxy::isConcave(proxyType1))
+	{
+		return m_convexConcaveCreateFunc;
+	}
+
+	if (btBroadphaseProxy::isConvex(proxyType1) && btBroadphaseProxy::isConcave(proxyType0))
+	{
+		return m_swappedConvexConcaveCreateFunc;
+	}
+
+	if (btBroadphaseProxy::isCompound(proxyType0))
+	{
+		return m_compoundCreateFunc;
+	} else
+	{
+		if (btBroadphaseProxy::isCompound(proxyType1))
+		{
+			return m_swappedCompoundCreateFunc;
+		}
+	}
+
+	//failed to find an algorithm
+	return m_emptyCreateFunc;
+}

extern/bullet2/src/BulletCollision/CollisionDispatch/btDefaultCollisionConfiguration.h

@@ -0,0 +1,112 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef BT_DEFAULT_COLLISION_CONFIGURATION
+#define BT_DEFAULT_COLLISION_CONFIGURATION
+
+#include "btCollisionConfiguration.h"
+class btVoronoiSimplexSolver;
+class btConvexPenetrationDepthSolver;
+
+struct	btDefaultCollisionConstructionInfo
+{
+	btStackAlloc*		m_stackAlloc;
+	btPoolAllocator*	m_persistentManifoldPool;
+	btPoolAllocator*	m_collisionAlgorithmPool;
+	int					m_defaultMaxPersistentManifoldPoolSize;
+	int					m_defaultMaxCollisionAlgorithmPoolSize;
+	int					m_defaultStackAllocatorSize;
+
+	btDefaultCollisionConstructionInfo()
+		:m_stackAlloc(0),
+		m_persistentManifoldPool(0),
+		m_collisionAlgorithmPool(0),
+		m_defaultMaxPersistentManifoldPoolSize(65535),
+		m_defaultMaxCollisionAlgorithmPoolSize(65535),
+		m_defaultStackAllocatorSize(5*1024*1024)
+	{
+	}
+};
+
+
+
+///btCollisionConfiguration allows to configure Bullet collision detection
+///stack allocator, pool memory allocators
+///todo: describe the meaning
+class	btDefaultCollisionConfiguration : public btCollisionConfiguration
+{
+
+	int	m_persistentManifoldPoolSize;
+	
+	btStackAlloc*	m_stackAlloc;
+	bool	m_ownsStackAllocator;
+
+	btPoolAllocator*	m_persistentManifoldPool;
+	bool	m_ownsPersistentManifoldPool;
+
+	btPoolAllocator*	m_collisionAlgorithmPool;
+	bool	m_ownsCollisionAlgorithmPool;
+
+	//default simplex/penetration depth solvers
+	btVoronoiSimplexSolver*	m_simplexSolver;
+	btConvexPenetrationDepthSolver*	m_pdSolver;
+	
+	//default CreationFunctions, filling the m_doubleDispatch table
+	btCollisionAlgorithmCreateFunc*	m_convexConvexCreateFunc;
+	btCollisionAlgorithmCreateFunc*	m_convexConcaveCreateFunc;
+	btCollisionAlgorithmCreateFunc*	m_swappedConvexConcaveCreateFunc;
+	btCollisionAlgorithmCreateFunc*	m_compoundCreateFunc;
+	btCollisionAlgorithmCreateFunc*	m_swappedCompoundCreateFunc;
+	btCollisionAlgorithmCreateFunc* m_emptyCreateFunc;
+	btCollisionAlgorithmCreateFunc* m_sphereSphereCF;
+	btCollisionAlgorithmCreateFunc* m_sphereBoxCF;
+	btCollisionAlgorithmCreateFunc* m_boxSphereCF;
+	btCollisionAlgorithmCreateFunc* m_boxBoxCF;
+	btCollisionAlgorithmCreateFunc*	m_sphereTriangleCF;
+	btCollisionAlgorithmCreateFunc*	m_triangleSphereCF;
+	btCollisionAlgorithmCreateFunc*	m_planeConvexCF;
+	btCollisionAlgorithmCreateFunc*	m_convexPlaneCF;
+	
+public:
+
+
+	btDefaultCollisionConfiguration(const btDefaultCollisionConstructionInfo& constructionInfo = btDefaultCollisionConstructionInfo());
+
+	virtual ~btDefaultCollisionConfiguration();
+
+		///memory pools
+	virtual btPoolAllocator* getPersistentManifoldPool()
+	{
+		return m_persistentManifoldPool;
+	}
+
+	virtual btPoolAllocator* getCollisionAlgorithmPool()
+	{
+		return m_collisionAlgorithmPool;
+	}
+
+	virtual btStackAlloc*	getStackAllocator()
+	{
+		return m_stackAlloc;
+	}
+
+
+	virtual btCollisionAlgorithmCreateFunc* getCollisionAlgorithmCreateFunc(int proxyType0,int proxyType1);
+
+
+};
+
+#endif //BT_DEFAULT_COLLISION_CONFIGURATION
+

extern/bullet2/src/BulletCollision/CollisionDispatch/btEmptyCollisionAlgorithm.h

@@ -15,8 +15,9 @@ subject to the following restrictions:
 
 #ifndef EMPTY_ALGORITH
 #define EMPTY_ALGORITH
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
 #include "btCollisionCreateFunc.h"
+#include "btCollisionDispatcher.h"
 
 #define ATTRIBUTE_ALIGNED(a)
 
@@ -33,13 +34,18 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+	}
+
 	struct CreateFunc :public 	btCollisionAlgorithmCreateFunc
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
 			(void)body0;
 			(void)body1;
-			return new btEmptyAlgorithm(ci);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btEmptyAlgorithm));
+			return new(mem) btEmptyAlgorithm(ci);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btManifoldResult.cpp

@@ -79,12 +79,30 @@ void btManifoldResult::addContactPoint(const btVector3& normalOnBInWorld,const b
 	}
 
 	btManifoldPoint newPt(localA,localB,normalOnBInWorld,depth);
-
+	newPt.m_positionWorldOnA = pointA;
+	newPt.m_positionWorldOnB = pointInWorld;
+	
 	int insertIndex = m_manifoldPtr->getCacheEntry(newPt);
 
 	newPt.m_combinedFriction = calculateCombinedFriction(m_body0,m_body1);
 	newPt.m_combinedRestitution = calculateCombinedRestitution(m_body0,m_body1);
 
+   //BP mod, store contact triangles.
+   newPt.m_partId0 = m_partId0;
+   newPt.m_partId1 = m_partId1;
+   newPt.m_index0  = m_index0;
+   newPt.m_index1  = m_index1;
+	
+	///todo, check this for any side effects
+	if (insertIndex >= 0)
+	{
+		//const btManifoldPoint& oldPoint = m_manifoldPtr->getContactPoint(insertIndex);
+		m_manifoldPtr->replaceContactPoint(newPt,insertIndex);
+	} else
+	{
+		insertIndex = m_manifoldPtr->addManifoldPoint(newPt);
+	}
+	
 	//User can override friction and/or restitution
 	if (gContactAddedCallback &&
 		//and if either of the two bodies requires custom material
@@ -94,16 +112,8 @@ void btManifoldResult::addContactPoint(const btVector3& normalOnBInWorld,const b
 		//experimental feature info, for per-triangle material etc.
 		btCollisionObject* obj0 = isSwapped? m_body1 : m_body0;
 		btCollisionObject* obj1 = isSwapped? m_body0 : m_body1;
-		(*gContactAddedCallback)(newPt,obj0,m_partId0,m_index0,obj1,m_partId1,m_index1);
+		(*gContactAddedCallback)(m_manifoldPtr->getContactPoint(insertIndex),obj0,m_partId0,m_index0,obj1,m_partId1,m_index1);
 	}
 
-	if (insertIndex >= 0)
-	{
-		//const btManifoldPoint& oldPoint = m_manifoldPtr->getContactPoint(insertIndex);
-		m_manifoldPtr->replaceContactPoint(newPt,insertIndex);
-	} else
-	{
-		m_manifoldPtr->AddManifoldPoint(newPt);
-	}
 }
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btManifoldResult.h

@@ -18,12 +18,12 @@ subject to the following restrictions:
 #define MANIFOLD_RESULT_H
 
 class btCollisionObject;
-class btPersistentManifold;
+#include "BulletCollision/NarrowPhaseCollision/btPersistentManifold.h"
 class btManifoldPoint;
 
 #include "BulletCollision/NarrowPhaseCollision/btDiscreteCollisionDetectorInterface.h"
 
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
 
 typedef bool (*ContactAddedCallback)(btManifoldPoint& cp,	const btCollisionObject* colObj0,int partId0,int index0,const btCollisionObject* colObj1,int partId1,int index1);
 extern ContactAddedCallback		gContactAddedCallback;
@@ -60,6 +60,15 @@ public:
 		m_manifoldPtr = manifoldPtr;
 	}
 
+	const btPersistentManifold*	getPersistentManifold() const
+	{
+		return m_manifoldPtr;
+	}
+	btPersistentManifold*	getPersistentManifold()
+	{
+		return m_manifoldPtr;
+	}
+
 	virtual void setShapeIdentifiers(int partId0,int index0,	int partId1,int index1)
 	{
 			m_partId0=partId0;
@@ -70,6 +79,22 @@ public:
 
 	virtual void addContactPoint(const btVector3& normalOnBInWorld,const btVector3& pointInWorld,btScalar depth);
 
+	SIMD_FORCE_INLINE	void refreshContactPoints()
+	{
+		btAssert(m_manifoldPtr);
+		if (!m_manifoldPtr->getNumContacts())
+			return;
+
+		bool isSwapped = m_manifoldPtr->getBody0() != m_body0;
+
+		if (isSwapped)
+		{
+			m_manifoldPtr->refreshContactPoints(m_rootTransB,m_rootTransA);
+		} else
+		{
+			m_manifoldPtr->refreshContactPoints(m_rootTransA,m_rootTransB);
+		}
+	}
 
 
 };

extern/bullet2/src/BulletCollision/CollisionDispatch/btSimulationIslandManager.cpp

@@ -1,3 +1,17 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
 
 
 #include "LinearMath/btScalar.h"
@@ -7,7 +21,7 @@
 #include "BulletCollision/CollisionDispatch/btCollisionObject.h"
 #include "BulletCollision/CollisionDispatch/btCollisionWorld.h"
 
-#include <stdio.h>
+//#include <stdio.h>
 #include "LinearMath/btQuickprof.h"
 
 btSimulationIslandManager::btSimulationIslandManager()
@@ -25,17 +39,17 @@ void btSimulationIslandManager::initUnionFind(int n)
 }
 		
 
-void btSimulationIslandManager::findUnions(btDispatcher* dispatcher)
+void btSimulationIslandManager::findUnions(btDispatcher* /* dispatcher */,btCollisionWorld* colWorld)
 {
 	
 	{
-		for (int i=0;i<dispatcher->getNumManifolds();i++)
+		btBroadphasePair* pairPtr = colWorld->getPairCache()->getOverlappingPairArrayPtr();
-		{
-			const btPersistentManifold* manifold = dispatcher->getManifoldByIndexInternal(i);
-			//static objects (invmass btScalar(0.)) don't merge !
 
-			 const  btCollisionObject* colObj0 = static_cast<const btCollisionObject*>(manifold->getBody0());
+		for (int i=0;i<colWorld->getPairCache()->getNumOverlappingPairs();i++)
-			 const  btCollisionObject* colObj1 = static_cast<const btCollisionObject*>(manifold->getBody1());
+		{
+			const btBroadphasePair& collisionPair = pairPtr[i];
+			btCollisionObject* colObj0 = (btCollisionObject*)collisionPair.m_pProxy0->m_clientObject;
+			btCollisionObject* colObj1 = (btCollisionObject*)collisionPair.m_pProxy1->m_clientObject;
 
 			if (((colObj0) && ((colObj0)->mergesSimulationIslands())) &&
 				((colObj1) && ((colObj1)->mergesSimulationIslands())))
@@ -71,7 +85,7 @@ void	btSimulationIslandManager::updateActivationState(btCollisionWorld* colWorld
 	}
 	// do the union find
 	
-	findUnions(dispatcher);
+	findUnions(dispatcher,colWorld);
 	
 
 	
@@ -129,30 +143,13 @@ class btPersistentManifoldSortPredicate
 };
 
 
-
+void btSimulationIslandManager::buildIslands(btDispatcher* dispatcher,btCollisionObjectArray& collisionObjects)
-
-
-//
-// todo: this is random access, it can be walked 'cache friendly'!
-//
-void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,btCollisionObjectArray& collisionObjects, IslandCallback* callback)
 {
 
+	BT_PROFILE("islandUnionFindAndQuickSort");
 	
-	
+	m_islandmanifold.resize(0);
-	/*if (0)
-	{
-		int maxNumManifolds = dispatcher->getNumManifolds();
-		btCollisionDispatcher* colDis = (btCollisionDispatcher*)dispatcher;
-		btPersistentManifold** manifold = colDis->getInternalManifoldPointer();
-		callback->ProcessIsland(&collisionObjects[0],collisionObjects.size(),manifold,maxNumManifolds, 0);
-		return;
-	}
-	*/
 
-
-	BEGIN_PROFILE("islandUnionFindAndHeapSort");
-	
 	//we are going to sort the unionfind array, and store the element id in the size
 	//afterwards, we clean unionfind, to make sure no-one uses it anymore
 	
@@ -183,7 +180,7 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 			btCollisionObject* colObj0 = collisionObjects[i];
 			if ((colObj0->getIslandTag() != islandId) && (colObj0->getIslandTag() != -1))
 			{
-				printf("error in island management\n");
+//				printf("error in island management\n");
 			}
 
 			assert((colObj0->getIslandTag() == islandId) || (colObj0->getIslandTag() == -1));
@@ -210,7 +207,7 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 				btCollisionObject* colObj0 = collisionObjects[i];
 				if ((colObj0->getIslandTag() != islandId) && (colObj0->getIslandTag() != -1))
 				{
-					printf("error in island management\n");
+//					printf("error in island management\n");
 				}
 
 				assert((colObj0->getIslandTag() == islandId) || (colObj0->getIslandTag() == -1));
@@ -231,7 +228,7 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 				btCollisionObject* colObj0 = collisionObjects[i];
 				if ((colObj0->getIslandTag() != islandId) && (colObj0->getIslandTag() != -1))
 				{
-					printf("error in island management\n");
+//					printf("error in island management\n");
 				}
 
 				assert((colObj0->getIslandTag() == islandId) || (colObj0->getIslandTag() == -1));
@@ -247,11 +244,17 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 		}
 	}
 
-	btAlignedObjectArray<btPersistentManifold*>  islandmanifold;
+	
 	int i;
 	int maxNumManifolds = dispatcher->getNumManifolds();
-	islandmanifold.reserve(maxNumManifolds);
 
+#define SPLIT_ISLANDS 1
+#ifdef SPLIT_ISLANDS
+
+	
+#endif //SPLIT_ISLANDS
+
+	
 	for (i=0;i<maxNumManifolds ;i++)
 	{
 		 btPersistentManifold* manifold = dispatcher->getManifoldByIndexInternal(i);
@@ -265,29 +268,52 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 		{
 		
 			//kinematic objects don't merge islands, but wake up all connected objects
-			if (colObj0->isStaticOrKinematicObject() && colObj0->getActivationState() != ISLAND_SLEEPING)
+			if (colObj0->isKinematicObject() && colObj0->getActivationState() != ISLAND_SLEEPING)
 			{
 				colObj1->activate();
 			}
-			if (colObj1->isStaticOrKinematicObject() && colObj1->getActivationState() != ISLAND_SLEEPING)
+			if (colObj1->isKinematicObject() && colObj1->getActivationState() != ISLAND_SLEEPING)
 			{
 				colObj0->activate();
 			}
-
+#ifdef SPLIT_ISLANDS
-			//filtering for response
+	//		//filtering for response
 			if (dispatcher->needsResponse(colObj0,colObj1))
-				islandmanifold.push_back(manifold);
+				m_islandmanifold.push_back(manifold);
+#endif //SPLIT_ISLANDS
 		}
 	}
+}
 
-	int numManifolds = int (islandmanifold.size());
 
+
+//
+// todo: this is random access, it can be walked 'cache friendly'!
+//
+void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,btCollisionObjectArray& collisionObjects, IslandCallback* callback)
+{
+
+	buildIslands(dispatcher,collisionObjects);
+
+	int endIslandIndex=1;
+	int startIslandIndex;
+	int numElem = getUnionFind().getNumElements();
+
+	BT_PROFILE("processIslands");
+
+#ifndef SPLIT_ISLANDS
+	btPersistentManifold** manifold = dispatcher->getInternalManifoldPointer();
+	
+	callback->ProcessIsland(&collisionObjects[0],collisionObjects.size(),manifold,maxNumManifolds, -1);
+#else
 	// Sort manifolds, based on islands
 	// Sort the vector using predicate and std::sort
 	//std::sort(islandmanifold.begin(), islandmanifold.end(), btPersistentManifoldSortPredicate);
 
+	int numManifolds = int (m_islandmanifold.size());
+
 	//we should do radix sort, it it much faster (O(n) instead of O (n log2(n))
-	islandmanifold.heapSort(btPersistentManifoldSortPredicate());
+	m_islandmanifold.quickSort(btPersistentManifoldSortPredicate());
 
 	//now process all active islands (sets of manifolds for now)
 
@@ -296,10 +322,9 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 
 	//int islandId;
 
-	END_PROFILE("islandUnionFindAndHeapSort");
+	
-
-	btAlignedObjectArray<btCollisionObject*>	islandBodies;
 
+//	printf("Start Islands\n");
 
 	//traverse the simulation islands, and call the solver, unless all objects are sleeping/deactivated
 	for ( startIslandIndex=0;startIslandIndex<numElem;startIslandIndex = endIslandIndex)
@@ -313,7 +338,7 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
                 {
                         int i = getUnionFind().getElement(endIslandIndex).m_sz;
                         btCollisionObject* colObj0 = collisionObjects[i];
-						islandBodies.push_back(colObj0);
+						m_islandBodies.push_back(colObj0);
                         if (!colObj0->isActive())
                                 islandSleeping = true;
                 }
@@ -325,12 +350,12 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 
 		if (startManifoldIndex<numManifolds)
 		{
-			int curIslandId = getIslandId(islandmanifold[startManifoldIndex]);
+			int curIslandId = getIslandId(m_islandmanifold[startManifoldIndex]);
 			if (curIslandId == islandId)
 			{
-				startManifold = &islandmanifold[startManifoldIndex];
+				startManifold = &m_islandmanifold[startManifoldIndex];
 			
-				for (endManifoldIndex = startManifoldIndex+1;(endManifoldIndex<numManifolds) && (islandId == getIslandId(islandmanifold[endManifoldIndex]));endManifoldIndex++)
+				for (endManifoldIndex = startManifoldIndex+1;(endManifoldIndex<numManifolds) && (islandId == getIslandId(m_islandmanifold[endManifoldIndex]));endManifoldIndex++)
 				{
 
 				}
@@ -342,7 +367,8 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 
 		if (!islandSleeping)
 		{
-			callback->ProcessIsland(&islandBodies[0],islandBodies.size(),startManifold,numIslandManifolds, islandId);
+			callback->ProcessIsland(&m_islandBodies[0],m_islandBodies.size(),startManifold,numIslandManifolds, islandId);
+//			printf("Island callback of size:%d bodies, %d manifolds\n",islandBodies.size(),numIslandManifolds);
 		}
 		
 		if (numIslandManifolds)
@@ -350,8 +376,9 @@ void btSimulationIslandManager::buildAndProcessIslands(btDispatcher* dispatcher,
 			startManifoldIndex = endManifoldIndex;
 		}
 
-		islandBodies.resize(0);
+		m_islandBodies.resize(0);
 	}
+#endif //SPLIT_ISLANDS
+
 
-	
 }

extern/bullet2/src/BulletCollision/CollisionDispatch/btSimulationIslandManager.h

@@ -16,18 +16,26 @@ subject to the following restrictions:
 #ifndef SIMULATION_ISLAND_MANAGER_H
 #define SIMULATION_ISLAND_MANAGER_H
 
-#include "../CollisionDispatch/btUnionFind.h"
+#include "BulletCollision/CollisionDispatch/btUnionFind.h"
 #include "btCollisionCreateFunc.h"
+#include "LinearMath/btAlignedObjectArray.h"
+
 
 class btCollisionObject;
 class btCollisionWorld;
 class btDispatcher;
+class btPersistentManifold;
+
 
 ///SimulationIslandManager creates and handles simulation islands, using btUnionFind
 class btSimulationIslandManager
 {
 	btUnionFind m_unionFind;
 
+	btAlignedObjectArray<btPersistentManifold*>  m_islandmanifold;
+	btAlignedObjectArray<btCollisionObject* >  m_islandBodies;
+	
+	
 public:
 	btSimulationIslandManager();
 	virtual ~btSimulationIslandManager();
@@ -42,7 +50,7 @@ public:
 	virtual	void	storeIslandActivationState(btCollisionWorld* world);
 
 
-	void	findUnions(btDispatcher* dispatcher);
+	void	findUnions(btDispatcher* dispatcher,btCollisionWorld* colWorld);
 
 	
 
@@ -55,6 +63,8 @@ public:
 
 	void	buildAndProcessIslands(btDispatcher* dispatcher,btCollisionObjectArray& collisionObjects, IslandCallback* callback);
 
+	void buildIslands(btDispatcher* dispatcher,btCollisionObjectArray& collisionObjects);
+
 };
 
 #endif //SIMULATION_ISLAND_MANAGER_H

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereBoxCollisionAlgorithm.cpp

@@ -68,18 +68,25 @@ void btSphereBoxCollisionAlgorithm::processCollision (btCollisionObject* body0,b
 	
 	btScalar dist = getSphereDistance(boxObj,pOnBox,pOnSphere,sphereCenter,radius);
 
+	resultOut->setPersistentManifold(m_manifoldPtr);
+
 	if (dist < SIMD_EPSILON)
 	{
 		btVector3 normalOnSurfaceB = (pOnBox- pOnSphere).normalize();
 
 		/// report a contact. internally this will be kept persistent, and contact reduction is done
 
-		resultOut->setPersistentManifold(m_manifoldPtr);
 		resultOut->addContactPoint(normalOnSurfaceB,pOnBox,dist);
 		
 	}
 
-	
+	if (m_ownManifold)
+	{
+		if (m_manifoldPtr->getNumContacts())
+		{
+			resultOut->refreshContactPoints();
+		}
+	}
 
 }
 
@@ -102,8 +109,8 @@ btScalar btSphereBoxCollisionAlgorithm::getSphereDistance(btCollisionObject* box
 	btVector3 bounds[2];
 	btBoxShape* boxShape= (btBoxShape*)boxObj->getCollisionShape();
 	
-	bounds[0] = -boxShape->getHalfExtents();
+	bounds[0] = -boxShape->getHalfExtentsWithoutMargin();
-	bounds[1] = boxShape->getHalfExtents();
+	bounds[1] = boxShape->getHalfExtentsWithoutMargin();
 
 	margins = boxShape->getMargin();//also add sphereShape margin?
 
@@ -209,6 +216,10 @@ btScalar btSphereBoxCollisionAlgorithm::getSpherePenetration( btCollisionObject*
 	btVector3	p0, tmp, prel, n[6], normal;
 	btScalar   fSep = btScalar(-10000000.0), fSepThis;
 
+	// set p0 and normal to a default value to shup up GCC
+	p0.setValue(btScalar(0.), btScalar(0.), btScalar(0.));
+	normal.setValue(btScalar(0.), btScalar(0.), btScalar(0.));
+
 	n[0].setValue( btScalar(-1.0),  btScalar(0.0),  btScalar(0.0) );
 	n[1].setValue(  btScalar(0.0), btScalar(-1.0),  btScalar(0.0) );
 	n[2].setValue(  btScalar(0.0),  btScalar(0.0), btScalar(-1.0) );

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereBoxCollisionAlgorithm.h

@@ -16,11 +16,13 @@ subject to the following restrictions:
 #ifndef SPHERE_BOX_COLLISION_ALGORITHM_H
 #define SPHERE_BOX_COLLISION_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../CollisionDispatch/btCollisionCreateFunc.h"
+#include "BulletCollision/CollisionDispatch/btCollisionCreateFunc.h"
 class btPersistentManifold;
-#include "../../LinearMath/btVector3.h"
+#include "btCollisionDispatcher.h"
+
+#include "LinearMath/btVector3.h"
 
 /// btSphereBoxCollisionAlgorithm  provides sphere-box collision detection.
 /// Other features are frame-coherency (persistent data) and collision response.
@@ -40,6 +42,14 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		if (m_manifoldPtr && m_ownManifold)
+		{
+			manifoldArray.push_back(m_manifoldPtr);
+		}
+	}
+
 	btScalar getSphereDistance( btCollisionObject* boxObj,btVector3& v3PointOnBox, btVector3& v3PointOnSphere, const btVector3& v3SphereCenter, btScalar fRadius );
 
 	btScalar getSpherePenetration( btCollisionObject* boxObj, btVector3& v3PointOnBox, btVector3& v3PointOnSphere, const btVector3& v3SphereCenter, btScalar fRadius, const btVector3& aabbMin, const btVector3& aabbMax);
@@ -48,12 +58,13 @@ public:
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btSphereBoxCollisionAlgorithm));
 			if (!m_swapped)
 			{
-				return new btSphereBoxCollisionAlgorithm(0,ci,body0,body1,false);
+				return new(mem) btSphereBoxCollisionAlgorithm(0,ci,body0,body1,false);
 			} else
 			{
-				return new btSphereBoxCollisionAlgorithm(0,ci,body0,body1,true);
+				return new(mem) btSphereBoxCollisionAlgorithm(0,ci,body0,body1,true);
 			}
 		}
 	};

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereSphereCollisionAlgorithm.cpp

@@ -46,6 +46,8 @@ void btSphereSphereCollisionAlgorithm::processCollision (btCollisionObject* col0
 	if (!m_manifoldPtr)
 		return;
 
+	resultOut->setPersistentManifold(m_manifoldPtr);
+
 	btSphereShape* sphere0 = (btSphereShape*)col0->getCollisionShape();
 	btSphereShape* sphere1 = (btSphereShape*)col1->getCollisionShape();
 
@@ -54,23 +56,34 @@ void btSphereSphereCollisionAlgorithm::processCollision (btCollisionObject* col0
 	btScalar radius0 = sphere0->getRadius();
 	btScalar radius1 = sphere1->getRadius();
 
+	//m_manifoldPtr->clearManifold(); //don't do this, it disables warmstarting
+
 	///iff distance positive, don't generate a new contact
 	if ( len > (radius0+radius1))
+	{
 		return;
-
+	}
 	///distance (negative means penetration)
 	btScalar dist = len - (radius0+radius1);
 
-	btVector3 normalOnSurfaceB = diff / len;
+	btVector3 normalOnSurfaceB(1,0,0);
+	if (len > SIMD_EPSILON)
+	{
+		normalOnSurfaceB = diff / len;
+	}
+
 	///point on A (worldspace)
 	btVector3 pos0 = col0->getWorldTransform().getOrigin() - radius0 * normalOnSurfaceB;
 	///point on B (worldspace)
 	btVector3 pos1 = col1->getWorldTransform().getOrigin() + radius1* normalOnSurfaceB;
 
 	/// report a contact. internally this will be kept persistent, and contact reduction is done
-	resultOut->setPersistentManifold(m_manifoldPtr);
+	
+	
 	resultOut->addContactPoint(normalOnSurfaceB,pos1,dist);
 
+	//no resultOut->refreshContactPoints(); needed, because of clearManifold (all points are new)
+
 }
 
 btScalar btSphereSphereCollisionAlgorithm::calculateTimeOfImpact(btCollisionObject* col0,btCollisionObject* col1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut)

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereSphereCollisionAlgorithm.h

@@ -16,9 +16,11 @@ subject to the following restrictions:
 #ifndef SPHERE_SPHERE_COLLISION_ALGORITHM_H
 #define SPHERE_SPHERE_COLLISION_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../CollisionDispatch/btCollisionCreateFunc.h"
+#include "BulletCollision/CollisionDispatch/btCollisionCreateFunc.h"
+#include "btCollisionDispatcher.h"
+
 class btPersistentManifold;
 
 /// btSphereSphereCollisionAlgorithm  provides sphere-sphere collision detection.
@@ -39,6 +41,13 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		if (m_manifoldPtr && m_ownManifold)
+		{
+			manifoldArray.push_back(m_manifoldPtr);
+		}
+	}
 	
 	virtual ~btSphereSphereCollisionAlgorithm();
 
@@ -46,7 +55,8 @@ public:
 	{
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
-			return new btSphereSphereCollisionAlgorithm(0,ci,body0,body1);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btSphereSphereCollisionAlgorithm));
+			return new(mem) btSphereSphereCollisionAlgorithm(0,ci,body0,body1);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereTriangleCollisionAlgorithm.cpp

@@ -48,8 +48,11 @@ void btSphereTriangleCollisionAlgorithm::processCollision (btCollisionObject* co
 	if (!m_manifoldPtr)
 		return;
 
-	btSphereShape* sphere = (btSphereShape*)col0->getCollisionShape();
+	btCollisionObject* sphereObj = m_swapped? col1 : col0;
-	btTriangleShape* triangle = (btTriangleShape*)col1->getCollisionShape();
+	btCollisionObject* triObj = m_swapped? col0 : col1;
+
+	btSphereShape* sphere = (btSphereShape*)sphereObj->getCollisionShape();
+	btTriangleShape* triangle = (btTriangleShape*)triObj->getCollisionShape();
 	
 	/// report a contact. internally this will be kept persistent, and contact reduction is done
 	resultOut->setPersistentManifold(m_manifoldPtr);
@@ -57,11 +60,16 @@ void btSphereTriangleCollisionAlgorithm::processCollision (btCollisionObject* co
 	
 	btDiscreteCollisionDetectorInterface::ClosestPointInput input;
 	input.m_maximumDistanceSquared = btScalar(1e30);//todo: tighter bounds
-	input.m_transformA = col0->getWorldTransform();
+	input.m_transformA = sphereObj->getWorldTransform();
-	input.m_transformB = col1->getWorldTransform();
+	input.m_transformB = triObj->getWorldTransform();
 
-	detector.getClosestPoints(input,*resultOut,dispatchInfo.m_debugDraw);
+	bool swapResults = m_swapped && !m_ownManifold;
 
+	detector.getClosestPoints(input,*resultOut,dispatchInfo.m_debugDraw,swapResults);
+
+	if (m_ownManifold)
+		resultOut->refreshContactPoints();
+	
 }
 
 btScalar btSphereTriangleCollisionAlgorithm::calculateTimeOfImpact(btCollisionObject* col0,btCollisionObject* col1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut)

extern/bullet2/src/BulletCollision/CollisionDispatch/btSphereTriangleCollisionAlgorithm.h

@@ -16,10 +16,11 @@ subject to the following restrictions:
 #ifndef SPHERE_TRIANGLE_COLLISION_ALGORITHM_H
 #define SPHERE_TRIANGLE_COLLISION_ALGORITHM_H
 
-#include "../BroadphaseCollision/btCollisionAlgorithm.h"
+#include "BulletCollision/BroadphaseCollision/btCollisionAlgorithm.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../CollisionDispatch/btCollisionCreateFunc.h"
+#include "BulletCollision/CollisionDispatch/btCollisionCreateFunc.h"
 class btPersistentManifold;
+#include "btCollisionDispatcher.h"
 
 /// btSphereSphereCollisionAlgorithm  provides sphere-sphere collision detection.
 /// Other features are frame-coherency (persistent data) and collision response.
@@ -40,6 +41,13 @@ public:
 
 	virtual btScalar calculateTimeOfImpact(btCollisionObject* body0,btCollisionObject* body1,const btDispatcherInfo& dispatchInfo,btManifoldResult* resultOut);
 
+	virtual	void	getAllContactManifolds(btManifoldArray&	manifoldArray)
+	{
+		if (m_manifoldPtr && m_ownManifold)
+		{
+			manifoldArray.push_back(m_manifoldPtr);
+		}
+	}
 	
 	virtual ~btSphereTriangleCollisionAlgorithm();
 
@@ -49,7 +57,9 @@ public:
 		virtual	btCollisionAlgorithm* CreateCollisionAlgorithm(btCollisionAlgorithmConstructionInfo& ci, btCollisionObject* body0,btCollisionObject* body1)
 		{
 			
-				return new btSphereTriangleCollisionAlgorithm(ci.m_manifold,ci,body0,body1,m_swapped);
+			void* mem = ci.m_dispatcher1->allocateCollisionAlgorithm(sizeof(btSphereTriangleCollisionAlgorithm));
+
+			return new(mem) btSphereTriangleCollisionAlgorithm(ci.m_manifold,ci,body0,body1,m_swapped);
 		}
 	};
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btUnionFind.cpp

@@ -18,6 +18,7 @@ subject to the following restrictions:
 
 
 
+
 btUnionFind::~btUnionFind()
 {
 	Free();
@@ -76,8 +77,7 @@ void	btUnionFind::sortIslands()
 	
 	 // Sort the vector using predicate and std::sort
 	  //std::sort(m_elements.begin(), m_elements.end(), btUnionFindElementSortPredicate);
-	//perhaps use radix sort?
+	  m_elements.quickSort(btUnionFindElementSortPredicate());
-	  m_elements.heapSort(btUnionFindElementSortPredicate());
 
 }
 

extern/bullet2/src/BulletCollision/CollisionDispatch/btUnionFind.h

@@ -16,7 +16,7 @@ subject to the following restrictions:
 #ifndef UNION_FIND_H
 #define UNION_FIND_H
 
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
 	#define USE_PATH_COMPRESSION 1
 
@@ -46,11 +46,11 @@ class btUnionFind
 
 	  void	reset(int N);
 
-	  inline int	getNumElements() const
+	  SIMD_FORCE_INLINE int	getNumElements() const
 	  {
 		  return int(m_elements.size());
 	  }
-	  inline bool  isRoot(int x) const
+	  SIMD_FORCE_INLINE bool  isRoot(int x) const
 	  {
 		  return (x == m_elements[x].m_id);
 	  }

extern/bullet2/src/BulletCollision/CollisionShapes/btBoxShape.cpp

@@ -15,16 +15,13 @@ subject to the following restrictions:
 
 #include "btBoxShape.h"
 
-btVector3 btBoxShape::getHalfExtents() const
+
-{
-	return m_implicitShapeDimensions * m_localScaling;
-}
 //{ 
 
 
 void btBoxShape::getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const
 {
-	btVector3 halfExtents = getHalfExtents();
+	const btVector3& halfExtents = getHalfExtentsWithoutMargin();
 
 	btMatrix3x3 abs_b = t.getBasis().absolute();  
 	btPoint3 center = t.getOrigin();
@@ -40,10 +37,10 @@ void btBoxShape::getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabb
 }
 
 
-void	btBoxShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btBoxShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	//btScalar margin = btScalar(0.);
-	btVector3 halfExtents = getHalfExtents();
+	btVector3 halfExtents = getHalfExtentsWithMargin();
 
 	btScalar lx=btScalar(2.)*(halfExtents.x());
 	btScalar ly=btScalar(2.)*(halfExtents.y());

extern/bullet2/src/BulletCollision/CollisionShapes/btBoxShape.h

@@ -18,11 +18,11 @@ subject to the following restrictions:
 
 #include "btPolyhedralConvexShape.h"
 #include "btCollisionMargin.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h"
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h"
-#include "../../LinearMath/btPoint3.h"
+#include "LinearMath/btPoint3.h"
-#include "../../LinearMath/btSimdMinMax.h"
+#include "LinearMath/btMinMax.h"
 
-///btBoxShape implements both a feature based (vertex/edge/plane) and implicit (getSupportingVertex) Box
+///The btBoxShape is a box primitive around the origin, its sides axis aligned with length specified by half extents, in local shape coordinates. When used as part of a btCollisionObject or btRigidBody it will be an oriented box in world space.
 class btBoxShape: public btPolyhedralConvexShape
 {
 
@@ -31,47 +31,52 @@ class btBoxShape: public btPolyhedralConvexShape
 
 public:
 
-	btVector3 getHalfExtents() const;
+	btVector3 getHalfExtentsWithMargin() const
-		
+	{
+		btVector3 halfExtents = getHalfExtentsWithoutMargin();
+		btVector3 margin(getMargin(),getMargin(),getMargin());
+		halfExtents += margin;
+		return halfExtents;
+	}
+	
+	const btVector3& getHalfExtentsWithoutMargin() const
+	{
+		return m_implicitShapeDimensions;//changed in Bullet 2.63: assume the scaling and margin are included
+	}
+	
+
 	virtual int	getShapeType() const { return BOX_SHAPE_PROXYTYPE;}
 
 	virtual btVector3	localGetSupportingVertex(const btVector3& vec) const
 	{
+		btVector3 halfExtents = getHalfExtentsWithoutMargin();
+		btVector3 margin(getMargin(),getMargin(),getMargin());
+		halfExtents += margin;
 		
-		btVector3 halfExtents = getHalfExtents();
+		return btVector3(btFsels(vec.x(), halfExtents.x(), -halfExtents.x()),
-		
+			btFsels(vec.y(), halfExtents.y(), -halfExtents.y()),
-		btVector3 supVertex;
+			btFsels(vec.z(), halfExtents.z(), -halfExtents.z()));
-		supVertex = btPoint3(vec.x() < btScalar(0.0) ? -halfExtents.x() : halfExtents.x(),
-                     vec.y() < btScalar(0.0) ? -halfExtents.y() : halfExtents.y(),
-                     vec.z() < btScalar(0.0) ? -halfExtents.z() : halfExtents.z()); 
-  
-		return supVertex;
 	}
 
-	virtual inline btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec)const
+	SIMD_FORCE_INLINE  btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec)const
 	{
-		btVector3 halfExtents = getHalfExtents();
+		const btVector3& halfExtents = getHalfExtentsWithoutMargin();
-		btVector3 margin(getMargin(),getMargin(),getMargin());
+		
-		halfExtents -= margin;
+		return btVector3(btFsels(vec.x(), halfExtents.x(), -halfExtents.x()),
-
+			btFsels(vec.y(), halfExtents.y(), -halfExtents.y()),
-		return btVector3(vec.x() < btScalar(0.0) ? -halfExtents.x() : halfExtents.x(),
+			btFsels(vec.z(), halfExtents.z(), -halfExtents.z()));
-                    vec.y() < btScalar(0.0) ? -halfExtents.y() : halfExtents.y(),
-                    vec.z() < btScalar(0.0) ? -halfExtents.z() : halfExtents.z()); 
 	}
 
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const
 	{
-		btVector3 halfExtents = getHalfExtents();
+		const btVector3& halfExtents = getHalfExtentsWithoutMargin();
-		btVector3 margin(getMargin(),getMargin(),getMargin());
+	
-		halfExtents -= margin;
-
-
 		for (int i=0;i<numVectors;i++)
 		{
 			const btVector3& vec = vectors[i];
-			supportVerticesOut[i].setValue(vec.x() < btScalar(0.0) ? -halfExtents.x() : halfExtents.x(),
+			supportVerticesOut[i].setValue(btFsels(vec.x(), halfExtents.x(), -halfExtents.x()),
-                    vec.y() < btScalar(0.0) ? -halfExtents.y() : halfExtents.y(),
+				btFsels(vec.y(), halfExtents.y(), -halfExtents.y()),
-                    vec.z() < btScalar(0.0) ? -halfExtents.z() : halfExtents.z()); 
+				btFsels(vec.z(), halfExtents.z(), -halfExtents.z())); 
 		}
 
 	}
@@ -79,14 +84,38 @@ public:
 
 	btBoxShape( const btVector3& boxHalfExtents)
 	{
-		m_implicitShapeDimensions = boxHalfExtents;
+		btVector3 margin(getMargin(),getMargin(),getMargin());
+		m_implicitShapeDimensions = (boxHalfExtents * m_localScaling) - margin;
 	};
-	
+
+	virtual void setMargin(btScalar collisionMargin)
+	{
+		//correct the m_implicitShapeDimensions for the margin
+		btVector3 oldMargin(getMargin(),getMargin(),getMargin());
+		btVector3 implicitShapeDimensionsWithMargin = m_implicitShapeDimensions+oldMargin;
+		
+		btConvexInternalShape::setMargin(collisionMargin);
+		btVector3 newMargin(getMargin(),getMargin(),getMargin());
+		m_implicitShapeDimensions = implicitShapeDimensionsWithMargin - newMargin;
+
+	}
+	virtual void	setLocalScaling(const btVector3& scaling)
+	{
+		btVector3 oldMargin(getMargin(),getMargin(),getMargin());
+		btVector3 implicitShapeDimensionsWithMargin = m_implicitShapeDimensions+oldMargin;
+		btVector3 unScaledImplicitShapeDimensionsWithMargin = implicitShapeDimensionsWithMargin / m_localScaling;
+
+		btConvexInternalShape::setLocalScaling(scaling);
+
+		m_implicitShapeDimensions = (unScaledImplicitShapeDimensionsWithMargin * m_localScaling) - oldMargin;
+
+	}
+
 	virtual void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
 
 	
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	virtual void getPlane(btVector3& planeNormal,btPoint3& planeSupport,int i ) const
 	{
@@ -116,7 +145,7 @@ public:
 
 	virtual void getVertex(int i,btVector3& vtx) const
 	{
-		btVector3 halfExtents = getHalfExtents();
+		btVector3 halfExtents = getHalfExtentsWithoutMargin();
 
 		vtx = btVector3(
 				halfExtents.x() * (1-(i&1)) - halfExtents.x() * (i&1),
@@ -127,7 +156,7 @@ public:
 
 	virtual void	getPlaneEquation(btVector4& plane,int i) const
 	{
-		btVector3 halfExtents = getHalfExtents();
+		btVector3 halfExtents = getHalfExtentsWithoutMargin();
 
 		switch (i)
 		{
@@ -234,7 +263,7 @@ public:
 	
 	virtual	bool isInside(const btPoint3& pt,btScalar tolerance) const
 	{
-		btVector3 halfExtents = getHalfExtents();
+		btVector3 halfExtents = getHalfExtentsWithoutMargin();
 
 		//btScalar minDist = 2*tolerance;
 		
@@ -250,7 +279,7 @@ public:
 
 
 	//debugging
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "Box";
 	}
@@ -291,3 +320,4 @@ public:
 
 #endif //OBB_BOX_MINKOWSKI_H
 
+

extern/bullet2/src/BulletCollision/CollisionShapes/btBvhTriangleMeshShape.cpp

@@ -18,32 +18,56 @@ subject to the following restrictions:
 #include "BulletCollision/CollisionShapes/btBvhTriangleMeshShape.h"
 #include "BulletCollision/CollisionShapes/btOptimizedBvh.h"
 
-
 ///Bvh Concave triangle mesh is a static-triangle mesh shape with Bounding Volume Hierarchy optimization.
 ///Uses an interface to access the triangles to allow for sharing graphics/physics triangles.
-btBvhTriangleMeshShape::btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression)
+btBvhTriangleMeshShape::btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression, bool buildBvh)
-:btTriangleMeshShape(meshInterface),m_useQuantizedAabbCompression(useQuantizedAabbCompression)
+:btTriangleMeshShape(meshInterface),
+m_bvh(0),
+m_useQuantizedAabbCompression(useQuantizedAabbCompression),
+m_ownsBvh(false)
 {
 	//construct bvh from meshInterface
 #ifndef DISABLE_BVH
 
-	m_bvh = new btOptimizedBvh();
 	btVector3 bvhAabbMin,bvhAabbMax;
-	meshInterface->calculateAabbBruteForce(bvhAabbMin,bvhAabbMax);
+	if(meshInterface->hasPremadeAabb())
-	m_bvh->build(meshInterface,m_useQuantizedAabbCompression,bvhAabbMin,bvhAabbMax);
+	{
+		meshInterface->getPremadeAabb(&bvhAabbMin, &bvhAabbMax);
+	}
+	else
+	{
+		meshInterface->calculateAabbBruteForce(bvhAabbMin,bvhAabbMax);
+	}
+	
+	if (buildBvh)
+	{
+		void* mem = btAlignedAlloc(sizeof(btOptimizedBvh),16);
+		m_bvh = new (mem) btOptimizedBvh();
+		m_bvh->build(meshInterface,m_useQuantizedAabbCompression,bvhAabbMin,bvhAabbMax);
+		m_ownsBvh = true;
+	}
 
 #endif //DISABLE_BVH
 
 }
 
-btBvhTriangleMeshShape::btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression,const btVector3& bvhAabbMin,const btVector3& bvhAabbMax)
+btBvhTriangleMeshShape::btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression,const btVector3& bvhAabbMin,const btVector3& bvhAabbMax,bool buildBvh)
-:btTriangleMeshShape(meshInterface),m_useQuantizedAabbCompression(useQuantizedAabbCompression)
+:btTriangleMeshShape(meshInterface),
+m_bvh(0),
+m_useQuantizedAabbCompression(useQuantizedAabbCompression),
+m_ownsBvh(false)
 {
 	//construct bvh from meshInterface
 #ifndef DISABLE_BVH
 
-	m_bvh = new btOptimizedBvh();
+	if (buildBvh)
-	m_bvh->build(meshInterface,m_useQuantizedAabbCompression,bvhAabbMin,bvhAabbMax);
+	{
+		void* mem = btAlignedAlloc(sizeof(btOptimizedBvh),16);
+		m_bvh = new (mem) btOptimizedBvh();
+		
+		m_bvh->build(meshInterface,m_useQuantizedAabbCompression,bvhAabbMin,bvhAabbMax);
+		m_ownsBvh = true;
+	}
 
 #endif //DISABLE_BVH
 
@@ -58,16 +82,140 @@ void	btBvhTriangleMeshShape::partialRefitTree(const btVector3& aabbMin,const btV
 }
 
 
-void	btBvhTriangleMeshShape::refitTree()
+void	btBvhTriangleMeshShape::refitTree(const btVector3& aabbMin,const btVector3& aabbMax)
 {
-	m_bvh->refit( m_meshInterface );
+	m_bvh->refit( m_meshInterface, aabbMin,aabbMax );
 	
 	recalcLocalAabb();
 }
 
 btBvhTriangleMeshShape::~btBvhTriangleMeshShape()
 {
-	delete m_bvh;
+	if (m_ownsBvh)
+	{
+		m_bvh->~btOptimizedBvh();
+		btAlignedFree(m_bvh);
+	}
+}
+
+void	btBvhTriangleMeshShape::performRaycast (btTriangleCallback* callback, const btVector3& raySource, const btVector3& rayTarget)
+{
+	struct	MyNodeOverlapCallback : public btNodeOverlapCallback
+	{
+		btStridingMeshInterface*	m_meshInterface;
+		btTriangleCallback* m_callback;
+
+		MyNodeOverlapCallback(btTriangleCallback* callback,btStridingMeshInterface* meshInterface)
+			:m_meshInterface(meshInterface),
+			m_callback(callback)
+		{
+		}
+				
+		virtual void processNode(int nodeSubPart, int nodeTriangleIndex)
+		{
+			btVector3 m_triangle[3];
+			const unsigned char *vertexbase;
+			int numverts;
+			PHY_ScalarType type;
+			int stride;
+			const unsigned char *indexbase;
+			int indexstride;
+			int numfaces;
+			PHY_ScalarType indicestype;
+
+			m_meshInterface->getLockedReadOnlyVertexIndexBase(
+				&vertexbase,
+				numverts,
+				type,
+				stride,
+				&indexbase,
+				indexstride,
+				numfaces,
+				indicestype,
+				nodeSubPart);
+
+			unsigned int* gfxbase = (unsigned int*)(indexbase+nodeTriangleIndex*indexstride);
+			btAssert(indicestype==PHY_INTEGER||indicestype==PHY_SHORT);
+	
+			const btVector3& meshScaling = m_meshInterface->getScaling();
+			for (int j=2;j>=0;j--)
+			{
+				int graphicsindex = indicestype==PHY_SHORT?((unsigned short*)gfxbase)[j]:gfxbase[j];
+
+				btScalar* graphicsbase = (btScalar*)(vertexbase+graphicsindex*stride);
+
+				m_triangle[j] = btVector3(graphicsbase[0]*meshScaling.getX(),graphicsbase[1]*meshScaling.getY(),graphicsbase[2]*meshScaling.getZ());		
+			}
+
+			/* Perform ray vs. triangle collision here */
+			m_callback->processTriangle(m_triangle,nodeSubPart,nodeTriangleIndex);
+			m_meshInterface->unLockReadOnlyVertexBase(nodeSubPart);
+		}
+	};
+
+	MyNodeOverlapCallback	myNodeCallback(callback,m_meshInterface);
+
+	m_bvh->reportRayOverlappingNodex(&myNodeCallback,raySource,rayTarget);
+}
+
+void	btBvhTriangleMeshShape::performConvexcast (btTriangleCallback* callback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax)
+{
+	struct	MyNodeOverlapCallback : public btNodeOverlapCallback
+	{
+		btStridingMeshInterface*	m_meshInterface;
+		btTriangleCallback* m_callback;
+
+		MyNodeOverlapCallback(btTriangleCallback* callback,btStridingMeshInterface* meshInterface)
+			:m_meshInterface(meshInterface),
+			m_callback(callback)
+		{
+		}
+				
+		virtual void processNode(int nodeSubPart, int nodeTriangleIndex)
+		{
+			btVector3 m_triangle[3];
+			const unsigned char *vertexbase;
+			int numverts;
+			PHY_ScalarType type;
+			int stride;
+			const unsigned char *indexbase;
+			int indexstride;
+			int numfaces;
+			PHY_ScalarType indicestype;
+
+			m_meshInterface->getLockedReadOnlyVertexIndexBase(
+				&vertexbase,
+				numverts,
+				type,
+				stride,
+				&indexbase,
+				indexstride,
+				numfaces,
+				indicestype,
+				nodeSubPart);
+
+			unsigned int* gfxbase = (unsigned int*)(indexbase+nodeTriangleIndex*indexstride);
+			btAssert(indicestype==PHY_INTEGER||indicestype==PHY_SHORT);
+	
+			const btVector3& meshScaling = m_meshInterface->getScaling();
+			for (int j=2;j>=0;j--)
+			{
+				int graphicsindex = indicestype==PHY_SHORT?((unsigned short*)gfxbase)[j]:gfxbase[j];
+
+				btScalar* graphicsbase = (btScalar*)(vertexbase+graphicsindex*stride);
+
+				m_triangle[j] = btVector3(graphicsbase[0]*meshScaling.getX(),graphicsbase[1]*meshScaling.getY(),graphicsbase[2]*meshScaling.getZ());		
+			}
+
+			/* Perform ray vs. triangle collision here */
+			m_callback->processTriangle(m_triangle,nodeSubPart,nodeTriangleIndex);
+			m_meshInterface->unLockReadOnlyVertexBase(nodeSubPart);
+		}
+	};
+
+	MyNodeOverlapCallback	myNodeCallback(callback,m_meshInterface);
+
+	m_bvh->reportBoxCastOverlappingNodex (&myNodeCallback, raySource, rayTarget, aabbMin, aabbMax);
 }
 
 //perform bvh tree traversal and report overlapping triangles to 'callback'
@@ -118,13 +266,14 @@ void	btBvhTriangleMeshShape::processAllTriangles(btTriangleCallback* callback,co
 				indicestype,
 				nodeSubPart);
 
-			int* gfxbase = (int*)(indexbase+nodeTriangleIndex*indexstride);
+			unsigned int* gfxbase = (unsigned int*)(indexbase+nodeTriangleIndex*indexstride);
+			btAssert(indicestype==PHY_INTEGER||indicestype==PHY_SHORT);
 	
 			const btVector3& meshScaling = m_meshInterface->getScaling();
 			for (int j=2;j>=0;j--)
 			{
 				
-				int graphicsindex = gfxbase[j];
+				int graphicsindex = indicestype==PHY_SHORT?((unsigned short*)gfxbase)[j]:gfxbase[j];
 
 
 #ifdef DEBUG_TRIANGLE_MESH
@@ -157,17 +306,37 @@ void	btBvhTriangleMeshShape::processAllTriangles(btTriangleCallback* callback,co
 
 }
 
-
+void   btBvhTriangleMeshShape::setLocalScaling(const btVector3& scaling)
-void	btBvhTriangleMeshShape::setLocalScaling(const btVector3& scaling)
 {
-	if ((getLocalScaling() -scaling).length2() > SIMD_EPSILON)
+   if ((getLocalScaling() -scaling).length2() > SIMD_EPSILON)
-	{
+   {
-		btTriangleMeshShape::setLocalScaling(scaling);
+      btTriangleMeshShape::setLocalScaling(scaling);
-		delete m_bvh;
+      if (m_ownsBvh)
-		///m_localAabbMin/m_localAabbMax is already re-calculated in btTriangleMeshShape. We could just scale aabb, but this needs some more work
+      {
-		m_bvh = new btOptimizedBvh();
+         m_bvh->~btOptimizedBvh();
-		//rebuild the bvh...
+         btAlignedFree(m_bvh);
-		m_bvh->build(m_meshInterface,m_useQuantizedAabbCompression,m_localAabbMin,m_localAabbMax);
+      }
-
+      ///m_localAabbMin/m_localAabbMax is already re-calculated in btTriangleMeshShape. We could just scale aabb, but this needs some more work
-	}
+      void* mem = btAlignedAlloc(sizeof(btOptimizedBvh),16);
+      m_bvh = new(mem) btOptimizedBvh();
+      //rebuild the bvh...
+      m_bvh->build(m_meshInterface,m_useQuantizedAabbCompression,m_localAabbMin,m_localAabbMax);
+      m_ownsBvh = true;
+   }
 }
+
+void   btBvhTriangleMeshShape::setOptimizedBvh(btOptimizedBvh* bvh, const btVector3& scaling)
+{
+   btAssert(!m_bvh);
+   btAssert(!m_ownsBvh);
+
+   m_bvh = bvh;
+   m_ownsBvh = false;
+   // update the scaling without rebuilding the bvh
+   if ((getLocalScaling() -scaling).length2() > SIMD_EPSILON)
+   {
+      btTriangleMeshShape::setLocalScaling(scaling);
+   }
+}
+
+

extern/bullet2/src/BulletCollision/CollisionShapes/btBvhTriangleMeshShape.h

@@ -18,45 +18,55 @@ subject to the following restrictions:
 
 #include "btTriangleMeshShape.h"
 #include "btOptimizedBvh.h"
+#include "LinearMath/btAlignedAllocator.h"
 
-///Bvh Concave triangle mesh is a static-triangle mesh shape with Bounding Volume Hierarchy optimization.
+
-///Uses an interface to access the triangles to allow for sharing graphics/physics triangles.
+///The btBvhTriangleMeshShape is a static-triangle mesh shape with several optimizations, such as bounding volume hierarchy and cache friendly traversal for PlayStation 3 Cell SPU. It is recommended to enable useQuantizedAabbCompression for better memory usage.
+///It takes a triangle mesh as input, for example a btTriangleMesh or btTriangleIndexVertexArray. The btBvhTriangleMeshShape class allows for triangle mesh deformations by a refit or partialRefit method.
+///Instead of building the bounding volume hierarchy acceleration structure, it is also possible to serialize (save) and deserialize (load) the structure from disk.
+///See Demos\ConcaveDemo\ConcavePhysicsDemo.cpp for an example.
 ATTRIBUTE_ALIGNED16(class) btBvhTriangleMeshShape : public btTriangleMeshShape
 {
 
 	btOptimizedBvh*	m_bvh;
 	bool m_useQuantizedAabbCompression;
-	bool m_pad[12];////need padding due to alignment
+	bool m_ownsBvh;
+	bool m_pad[11];////need padding due to alignment
 
 public:
 
-	btBvhTriangleMeshShape() :btTriangleMeshShape(0) {};
+	BT_DECLARE_ALIGNED_ALLOCATOR();
-	btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression);
+
+	btBvhTriangleMeshShape() :btTriangleMeshShape(0),m_bvh(0),m_ownsBvh(false) {};
+	btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression, bool buildBvh = true);
 
 	///optionally pass in a larger bvh aabb, used for quantization. This allows for deformations within this aabb
-	btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression,const btVector3& bvhAabbMin,const btVector3& bvhAabbMax);
+	btBvhTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression,const btVector3& bvhAabbMin,const btVector3& bvhAabbMax, bool buildBvh = true);
 	
 	virtual ~btBvhTriangleMeshShape();
 
-	
+	bool getOwnsBvh () const
-	/*
+	{
+		return m_ownsBvh;
+	}
+
 	virtual int	getShapeType() const
 	{
 		return TRIANGLE_MESH_SHAPE_PROXYTYPE;
 	}
-	*/
+	
-
+	void performRaycast (btTriangleCallback* callback, const btVector3& raySource, const btVector3& rayTarget);
-
+	void performConvexcast (btTriangleCallback* callback, const btVector3& boxSource, const btVector3& boxTarget, const btVector3& boxMin, const btVector3& boxMax);
 
 	virtual void	processAllTriangles(btTriangleCallback* callback,const btVector3& aabbMin,const btVector3& aabbMax) const;
 
-	void	refitTree();
+	void	refitTree(const btVector3& aabbMin,const btVector3& aabbMax);
 
 	///for a fast incremental refit of parts of the tree. Note: the entire AABB of the tree will become more conservative, it never shrinks
 	void	partialRefitTree(const btVector3& aabbMin,const btVector3& aabbMax);
 
 	//debugging
-	virtual char*	getName()const {return "BVHTRIANGLEMESH";}
+	virtual const char*	getName()const {return "BVHTRIANGLEMESH";}
 
 
 	virtual void	setLocalScaling(const btVector3& scaling);
@@ -65,6 +75,10 @@ public:
 	{
 		return m_bvh;
 	}
+
+
+	void	setOptimizedBvh(btOptimizedBvh* bvh, const btVector3& localScaling=btVector3(1,1,1));
+
 	bool	usesQuantizedAabbCompression() const
 	{
 		return	m_useQuantizedAabbCompression;

extern/bullet2/src/BulletCollision/CollisionShapes/btCapsuleShape.cpp

@@ -21,6 +21,7 @@ subject to the following restrictions:
 
 btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 {
+	m_upAxis = 1;
 	m_implicitShapeDimensions.setValue(radius,0.5f*height,radius);
 }
 
@@ -50,7 +51,9 @@ btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 
 
 	{
-		btVector3 pos(0,getHalfHeight(),0);
+		btVector3 pos(0,0,0);
+		pos[getUpAxis()] = getHalfHeight();
+
 		vtx = pos +vec*m_localScaling*(radius) - vec * getMargin();
 		newDot = vec.dot(vtx);
 		if (newDot > maxDot)
@@ -60,7 +63,9 @@ btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 		}
 	}
 	{
-		btVector3 pos(0,-getHalfHeight(),0);
+		btVector3 pos(0,0,0);
+		pos[getUpAxis()] = -getHalfHeight();
+
 		vtx = pos +vec*m_localScaling*(radius) - vec * getMargin();
 		newDot = vec.dot(vtx);
 		if (newDot > maxDot)
@@ -88,7 +93,8 @@ btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 		btVector3 vtx;
 		btScalar newDot;
 		{
-			btVector3 pos(0,getHalfHeight(),0);
+			btVector3 pos(0,0,0);
+			pos[getUpAxis()] = getHalfHeight();
 			vtx = pos +vec*m_localScaling*(radius) - vec * getMargin();
 			newDot = vec.dot(vtx);
 			if (newDot > maxDot)
@@ -98,7 +104,8 @@ btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 			}
 		}
 		{
-			btVector3 pos(0,-getHalfHeight(),0);
+			btVector3 pos(0,0,0);
+			pos[getUpAxis()] = -getHalfHeight();
 			vtx = pos +vec*m_localScaling*(radius) - vec * getMargin();
 			newDot = vec.dot(vtx);
 			if (newDot > maxDot)
@@ -112,7 +119,7 @@ btCapsuleShape::btCapsuleShape(btScalar radius, btScalar height)
 }
 
 
-void	btCapsuleShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btCapsuleShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	//as an approximation, take the inertia of the box that bounds the spheres
 
@@ -122,7 +129,8 @@ void	btCapsuleShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
 	
 	btScalar radius = getRadius();
 
-	btVector3 halfExtents(radius,radius+getHalfHeight(),radius);
+	btVector3 halfExtents(radius,radius,radius);
+	halfExtents[getUpAxis()]+=getHalfHeight();
 
 	btScalar margin = CONVEX_DISTANCE_MARGIN;
 
@@ -140,6 +148,22 @@ void	btCapsuleShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
 
 }
 
+btCapsuleShapeX::btCapsuleShapeX(btScalar radius,btScalar height)
+{
+	m_upAxis = 0;
+	m_implicitShapeDimensions.setValue(0.5f*height, radius,radius);
+}
+
+
+
+
+
+
+btCapsuleShapeZ::btCapsuleShapeZ(btScalar radius,btScalar height)
+{
+	m_upAxis = 2;
+	m_implicitShapeDimensions.setValue(radius,radius,0.5f*height);
+}
 
 
 

extern/bullet2/src/BulletCollision/CollisionShapes/btCapsuleShape.h

@@ -16,20 +16,27 @@ subject to the following restrictions:
 #ifndef BT_CAPSULE_SHAPE_H
 #define BT_CAPSULE_SHAPE_H
 
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
 
-///btCapsuleShape represents a capsule around the Y axis
+///The btCapsuleShape represents a capsule around the Y axis, there is also the btCapsuleShapeX aligned around the X axis and btCapsuleShapeZ around the Z axis.
-///A more general solution that can represent capsules is the btMultiSphereShape
+///The total height is height+2*radius, so the height is just the height between the center of each 'sphere' of the capsule caps.
-class btCapsuleShape : public btConvexShape
+///The btCapsuleShape is a convex hull of two spheres. The btMultiSphereShape is a more general collision shape that takes the convex hull of multiple sphere, so it can also represent a capsule when just using two spheres.
+class btCapsuleShape : public btConvexInternalShape
 {
+protected:
+	int	m_upAxis;
+
+protected:
+	///only used for btCapsuleShapeZ and btCapsuleShapeX subclasses.
+	btCapsuleShape() {};
 
 public:
 	btCapsuleShape(btScalar radius,btScalar height);
 
 	///CollisionShape Interface
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	/// btConvexShape Interface
 	virtual btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec)const;
@@ -38,23 +45,75 @@ public:
 	
 	virtual int	getShapeType() const { return CAPSULE_SHAPE_PROXYTYPE; }
 
-	virtual char*	getName()const 
+	virtual void getAabb (const btTransform& t, btVector3& aabbMin, btVector3& aabbMax) const
+	{
+			btVector3 halfExtents(getRadius(),getRadius(),getRadius());
+			halfExtents[m_upAxis] = getRadius() + getHalfHeight();
+			btMatrix3x3 abs_b = t.getBasis().absolute();  
+			btPoint3 center = t.getOrigin();
+			btVector3 extent = btVector3(abs_b[0].dot(halfExtents),abs_b[1].dot(halfExtents),abs_b[2].dot(halfExtents));		  
+			extent += btVector3(getMargin(),getMargin(),getMargin());
+			aabbMin = center - extent;
+			aabbMax = center + extent;
+	}
+
+	virtual const char*	getName()const 
 	{
 		return "CapsuleShape";
 	}
 
+	int	getUpAxis() const
+	{
+		return m_upAxis;
+	}
+
 	btScalar	getRadius() const
 	{
-		return m_implicitShapeDimensions.getX();
+		int radiusAxis = (m_upAxis+2)%3;
+		return m_implicitShapeDimensions[radiusAxis];
 	}
 
 	btScalar	getHalfHeight() const
 	{
-		return m_implicitShapeDimensions.getY();
+		return m_implicitShapeDimensions[m_upAxis];
 	}
 
 };
 
+///btCapsuleShapeX represents a capsule around the Z axis
+///the total height is height+2*radius, so the height is just the height between the center of each 'sphere' of the capsule caps.
+class btCapsuleShapeX : public btCapsuleShape
+{
+public:
+
+	btCapsuleShapeX(btScalar radius,btScalar height);
+		
+	//debugging
+	virtual const char*	getName()const
+	{
+		return "CapsuleX";
+	}
+
+	
+
+};
+
+///btCapsuleShapeZ represents a capsule around the Z axis
+///the total height is height+2*radius, so the height is just the height between the center of each 'sphere' of the capsule caps.
+class btCapsuleShapeZ : public btCapsuleShape
+{
+public:
+	btCapsuleShapeZ(btScalar radius,btScalar height);
+
+		//debugging
+	virtual const char*	getName()const
+	{
+		return "CapsuleZ";
+	}
+
+	
+};
+
 
 
 #endif //BT_CAPSULE_SHAPE_H

extern/bullet2/src/BulletCollision/CollisionShapes/btCollisionShape.cpp

@@ -21,7 +21,12 @@ subject to the following restrictions:
   can be used by probes that are checking whether the
   library is actually installed.
 */
-extern "C" void btBulletCollisionProbe () {}
+extern "C" 
+{
+void btBulletCollisionProbe ();
+
+void btBulletCollisionProbe () {}
+}
 
 
 
@@ -46,7 +51,7 @@ btScalar	btCollisionShape::getAngularMotionDisc() const
 	return disc;
 }
 
-void btCollisionShape::calculateTemporalAabb(const btTransform& curTrans,const btVector3& linvel,const btVector3& angvel,btScalar timeStep, btVector3& temporalAabbMin,btVector3& temporalAabbMax)
+void btCollisionShape::calculateTemporalAabb(const btTransform& curTrans,const btVector3& linvel,const btVector3& angvel,btScalar timeStep, btVector3& temporalAabbMin,btVector3& temporalAabbMax) const
 {
 	//start with static aabb
 	getAabb(curTrans,temporalAabbMin,temporalAabbMax);

extern/bullet2/src/BulletCollision/CollisionShapes/btCollisionShape.h

@@ -16,18 +16,21 @@ subject to the following restrictions:
 #ifndef COLLISION_SHAPE_H
 #define COLLISION_SHAPE_H
 
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
-#include "../../LinearMath/btMatrix3x3.h"
+#include "LinearMath/btMatrix3x3.h"
-#include "../../LinearMath/btPoint3.h"
+#include "LinearMath/btPoint3.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" //for the shape types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" //for the shape types
 
-///btCollisionShape provides interface for collision shapes that can be shared among btCollisionObjects.
+///The btCollisionShape class provides an interface for collision shapes that can be shared among btCollisionObjects.
 class btCollisionShape
 {
+
+	void* m_userPointer;
+
 public:
 
-	btCollisionShape() 
+	btCollisionShape() : m_userPointer(0)
 	{
 	}
 	virtual ~btCollisionShape()
@@ -45,30 +48,30 @@ public:
 
 	///calculateTemporalAabb calculates the enclosing aabb for the moving object over interval [0..timeStep)
 	///result is conservative
-	void calculateTemporalAabb(const btTransform& curTrans,const btVector3& linvel,const btVector3& angvel,btScalar timeStep, btVector3& temporalAabbMin,btVector3& temporalAabbMax);
+	void calculateTemporalAabb(const btTransform& curTrans,const btVector3& linvel,const btVector3& angvel,btScalar timeStep, btVector3& temporalAabbMin,btVector3& temporalAabbMax) const;
 
 #ifndef __SPU__
 
-	inline bool	isPolyhedral() const
+	SIMD_FORCE_INLINE bool	isPolyhedral() const
 	{
 		return btBroadphaseProxy::isPolyhedral(getShapeType());
 	}
 
-	inline bool	isConvex() const
+	SIMD_FORCE_INLINE bool	isConvex() const
 	{
 		return btBroadphaseProxy::isConvex(getShapeType());
 	}
-	inline bool	isConcave() const
+	SIMD_FORCE_INLINE bool	isConcave() const
 	{
 		return btBroadphaseProxy::isConcave(getShapeType());
 	}
-	inline bool	isCompound() const
+	SIMD_FORCE_INLINE bool	isCompound() const
 	{
 		return btBroadphaseProxy::isCompound(getShapeType());
 	}
 
 	///isInfinite is used to catch simulation error (aabb check)
-	inline bool isInfinite() const
+	SIMD_FORCE_INLINE bool isInfinite() const
 	{
 		return btBroadphaseProxy::isInfinite(getShapeType());
 	}
@@ -76,11 +79,11 @@ public:
 	virtual int		getShapeType() const=0;
 	virtual void	setLocalScaling(const btVector3& scaling) =0;
 	virtual const btVector3& getLocalScaling() const =0;
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) = 0;
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const = 0;
 
 
 //debugging support
-	virtual char*	getName()const =0 ;
+	virtual const char*	getName()const =0 ;
 #endif //__SPU__
 
 	
@@ -88,6 +91,18 @@ public:
 	virtual void	setMargin(btScalar margin) = 0;
 	virtual btScalar	getMargin() const = 0;
 
+	
+	///optional user data pointer
+	void	setUserPointer(void* userPtr)
+	{
+		m_userPointer = userPtr;
+	}
+
+	void*	getUserPointer() const
+	{
+		return m_userPointer;
+	}
+
 };	
 
 #endif //COLLISION_SHAPE_H

extern/bullet2/src/BulletCollision/CollisionShapes/btCompoundShape.cpp

@@ -35,8 +35,15 @@ btCompoundShape::~btCompoundShape()
 
 void	btCompoundShape::addChildShape(const btTransform& localTransform,btCollisionShape* shape)
 {
-	m_childTransforms.push_back(localTransform);
+	//m_childTransforms.push_back(localTransform);
-	m_childShapes.push_back(shape);
+	//m_childShapes.push_back(shape);
+	btCompoundShapeChild child;
+	child.m_transform = localTransform;
+	child.m_childShape = shape;
+	child.m_childShapeType = shape->getShapeType();
+	child.m_childMargin = shape->getMargin();
+
+	m_children.push_back(child);
 
 	//extend the local aabbMin/aabbMax
 	btVector3 localAabbMin,localAabbMax;
@@ -55,9 +62,53 @@ void	btCompoundShape::addChildShape(const btTransform& localTransform,btCollisio
 	}
 }
 
+void btCompoundShape::removeChildShape(btCollisionShape* shape)
+{
+   bool done_removing;
+   
+   // Find the children containing the shape specified, and remove those children.
+   do
+   {
+      done_removing = true;
+      
+      for(int i = 0; i < m_children.size(); i++)
+      {
+         if(m_children[i].m_childShape == shape)
+         {
+            m_children.remove(m_children[i]);
+            done_removing = false;  // Do another iteration pass after removing from the vector
+            break;
+         }
+      }
+   }
+   while (!done_removing);
+   
+   recalculateLocalAabb();
+}
 
-
+void btCompoundShape::recalculateLocalAabb()
-	///getAabb's default implementation is brute force, expected derived classes to implement a fast dedicated version
+{
+   // Recalculate the local aabb
+   // Brute force, it iterates over all the shapes left.
+   m_localAabbMin = btVector3(btScalar(1e30),btScalar(1e30),btScalar(1e30));
+   m_localAabbMax = btVector3(btScalar(-1e30),btScalar(-1e30),btScalar(-1e30));
+   
+   //extend the local aabbMin/aabbMax
+   for (int j = 0; j < m_children.size(); j++)
+   {
+      btVector3 localAabbMin,localAabbMax;
+      m_children[j].m_childShape->getAabb(m_children[j].m_transform, localAabbMin, localAabbMax);
+      for (int i=0;i<3;i++)
+      {
+         if (m_localAabbMin[i] > localAabbMin[i])
+             m_localAabbMin[i] = localAabbMin[i];
+         if (m_localAabbMax[i] < localAabbMax[i])
+             m_localAabbMax[i] = localAabbMax[i];
+      }
+   }
+}
+	
+   ///getAabb's default implementation is brute force, expected derived classes to implement a fast dedicated version
 void btCompoundShape::getAabb(const btTransform& trans,btVector3& aabbMin,btVector3& aabbMax) const
 {
 	btVector3 localHalfExtents = btScalar(0.5)*(m_localAabbMax-m_localAabbMin);
@@ -76,7 +127,7 @@ void btCompoundShape::getAabb(const btTransform& trans,btVector3& aabbMin,btVect
 	aabbMax = center + extent;
 }
 
-void	btCompoundShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btCompoundShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	//approximation: take the inertia from the aabb for now
 	btTransform ident;

extern/bullet2/src/BulletCollision/CollisionShapes/btCompoundShape.h

@@ -18,58 +18,95 @@ subject to the following restrictions:
 
 #include "btCollisionShape.h"
 
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
-#include "../../LinearMath/btMatrix3x3.h"
+#include "LinearMath/btMatrix3x3.h"
 #include "btCollisionMargin.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
 class btOptimizedBvh;
 
-/// btCompoundShape allows to store multiple other btCollisionShapes
+
-/// This allows for concave collision objects. This is more general then the Static Concave btTriangleMeshShape.
+ATTRIBUTE_ALIGNED16(struct) btCompoundShapeChild
-class btCompoundShape	: public btCollisionShape
 {
-	btAlignedObjectArray<btTransform>		m_childTransforms;
+	BT_DECLARE_ALIGNED_ALLOCATOR();
-	btAlignedObjectArray<btCollisionShape*>	m_childShapes;
+
+	btTransform			m_transform;
+	btCollisionShape*	m_childShape;
+	int					m_childShapeType;
+	btScalar			m_childMargin;
+};
+
+SIMD_FORCE_INLINE bool operator==(const btCompoundShapeChild& c1, const btCompoundShapeChild& c2)
+{
+   return  ( c1.m_transform      == c2.m_transform &&
+             c1.m_childShape     == c2.m_childShape &&
+             c1.m_childShapeType == c2.m_childShapeType &&
+             c1.m_childMargin    == c2.m_childMargin );
+}
+
+/// btCompoundShape allows to store multiple other btCollisionShapes
+/// This allows for moving concave collision objects. This is more general then the static concave btBvhTriangleMeshShape.
+ATTRIBUTE_ALIGNED16(class) btCompoundShape	: public btCollisionShape
+{
+	//btAlignedObjectArray<btTransform>		m_childTransforms;
+	//btAlignedObjectArray<btCollisionShape*>	m_childShapes;
+	btAlignedObjectArray<btCompoundShapeChild> m_children;
 	btVector3						m_localAabbMin;
 	btVector3						m_localAabbMax;
 
 	btOptimizedBvh*					m_aabbTree;
 
 public:
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
 	btCompoundShape();
 
 	virtual ~btCompoundShape();
 
 	void	addChildShape(const btTransform& localTransform,btCollisionShape* shape);
 
+   /** Remove all children shapes that contain the specified shape. */
+	virtual void removeChildShape(btCollisionShape* shape);
+
+	  
+	
+
 	int		getNumChildShapes() const
 	{
-		return int (m_childShapes.size());
+		return int (m_children.size());
 	}
 
 	btCollisionShape* getChildShape(int index)
 	{
-		return m_childShapes[index];
+		return m_children[index].m_childShape;
 	}
 	const btCollisionShape* getChildShape(int index) const
 	{
-		return m_childShapes[index];
+		return m_children[index].m_childShape;
 	}
 
-	btTransform&	getChildTransform(int index)
+	btTransform	getChildTransform(int index)
 	{
-		return m_childTransforms[index];
+		return m_children[index].m_transform;
 	}
-	const btTransform&	getChildTransform(int index) const
+	const btTransform	getChildTransform(int index) const
 	{
-		return m_childTransforms[index];
+		return m_children[index].m_transform;
+	}
+
+
+	btCompoundShapeChild* getChildList()
+	{
+		return &m_children[0];
 	}
 
 	///getAabb's default implementation is brute force, expected derived classes to implement a fast dedicated version
-	void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
+	virtual	void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
-
+	
+   /** Re-calculate the local Aabb. Is called at the end of removeChildShapes. 
+       Use this yourself if you modify the children or their transforms. */
+	virtual void recalculateLocalAabb(); 
 
 	virtual void	setLocalScaling(const btVector3& scaling)
 	{
@@ -80,7 +117,7 @@ public:
 		return m_localScaling;
 	}
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 	
 	virtual int	getShapeType() const { return COMPOUND_SHAPE_PROXYTYPE;}
 
@@ -92,7 +129,7 @@ public:
 	{
 		return m_collisionMargin;
 	}
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "Compound";
 	}

extern/bullet2/src/BulletCollision/CollisionShapes/btConcaveShape.h

@@ -17,12 +17,12 @@ subject to the following restrictions:
 #define CONCAVE_SHAPE_H
 
 #include "btCollisionShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 #include "btTriangleCallback.h"
 
 
-///Concave shape proves an interface concave shapes that can produce triangles that overlapping a given AABB.
+///The btConcaveShape class provides an interface for non-moving (static) concave shapes.
-///Static triangle mesh, infinite plane, height field/landscapes are example that implement this interface.
+///It has been implemented by the btStaticPlaneShape, btBvhTriangleMeshShape and btHeightfieldTerrainShape.
 class btConcaveShape : public btCollisionShape
 {
 protected:

extern/bullet2/src/BulletCollision/CollisionShapes/btConeShape.h

@@ -16,11 +16,11 @@ subject to the following restrictions:
 #ifndef CONE_MINKOWSKI_H
 #define CONE_MINKOWSKI_H
 
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
-///btConeShape implements a Cone shape, around the Y axis
+///The btConeShape implements a cone shape primitive, centered around the origin and aligned with the Y axis. The btConeShapeX is aligned around the X axis and btConeShapeZ around the Z axis.
-class btConeShape : public btConvexShape
+class btConeShape : public btConvexInternalShape
 
 {
 
@@ -42,7 +42,7 @@ public:
 	btScalar getHeight() const { return m_height;}
 
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia)
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const
 	{
 		btTransform identity;
 		identity.setIdentity();
@@ -72,7 +72,7 @@ public:
 
 		virtual int	getShapeType() const { return CONE_SHAPE_PROXYTYPE; }
 
-		virtual char*	getName()const 
+		virtual const char*	getName()const 
 		{
 			return "Cone";
 		}

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexHullShape.h

@@ -17,18 +17,17 @@ subject to the following restrictions:
 #define CONVEX_HULL_SHAPE_H
 
 #include "btPolyhedralConvexShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
-#include "../../LinearMath/btAlignedObjectArray.h"
+#include "LinearMath/btAlignedObjectArray.h"
 
-///ConvexHullShape implements an implicit (getSupportingVertex) Convex Hull of a Point Cloud (vertices)
+///The btConvexHullShape implements an implicit convex hull of an array of vertices.
-///No connectivity is needed. localGetSupportingVertex iterates linearly though all vertices.
+///Bullet provides a general and fast collision detector for convex shapes based on GJK and EPA using localGetSupportingVertex.
-///on modern hardware, due to cache coherency this isn't that bad. Complex algorithms tend to trash the cash.
-///(memory is much slower then the cpu)
 ATTRIBUTE_ALIGNED16(class) btConvexHullShape : public btPolyhedralConvexShape
 {
 	btAlignedObjectArray<btPoint3>	m_points;
 
 public:
+	BT_DECLARE_ALIGNED_ALLOCATOR();
 
 	
 	///this constructor optionally takes in a pointer to points. Each point is assumed to be 3 consecutive btScalar (x,y,z), the striding defines the number of bytes between each point, in memory.
@@ -43,7 +42,12 @@ public:
 		return &m_points[0];
 	}
 
-	int getNumPoints()
+	const btPoint3* getPoints() const
+	{
+		return &m_points[0];
+	}
+
+	int getNumPoints() const 
 	{
 		return m_points.size();
 	}
@@ -56,7 +60,7 @@ public:
 	virtual int	getShapeType()const { return CONVEX_HULL_SHAPE_PROXYTYPE; }
 
 	//debugging
-	virtual char*	getName()const {return "Convex";}
+	virtual const char*	getName()const {return "Convex";}
 
 	
 	virtual int	getNumVertices() const;

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexInternalShape.cpp

@@ -0,0 +1,78 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+
+#include "btConvexInternalShape.h"
+
+
+btConvexInternalShape::btConvexInternalShape()
+: m_localScaling(btScalar(1.),btScalar(1.),btScalar(1.)),
+m_collisionMargin(CONVEX_DISTANCE_MARGIN)
+{
+}
+
+
+void	btConvexInternalShape::setLocalScaling(const btVector3& scaling)
+{
+	m_localScaling = scaling;
+}
+
+
+
+void	btConvexInternalShape::getAabbSlow(const btTransform& trans,btVector3&minAabb,btVector3&maxAabb) const
+{
+
+	btScalar margin = getMargin();
+	for (int i=0;i<3;i++)
+	{
+		btVector3 vec(btScalar(0.),btScalar(0.),btScalar(0.));
+		vec[i] = btScalar(1.);
+
+		btVector3 sv = localGetSupportingVertex(vec*trans.getBasis());
+
+		btVector3 tmp = trans(sv);
+		maxAabb[i] = tmp[i]+margin;
+		vec[i] = btScalar(-1.);
+		tmp = trans(localGetSupportingVertex(vec*trans.getBasis()));
+		minAabb[i] = tmp[i]-margin;
+	}
+};
+
+
+btVector3	btConvexInternalShape::localGetSupportingVertex(const btVector3& vec)const
+{
+#ifndef __SPU__
+
+	 btVector3	supVertex = localGetSupportingVertexWithoutMargin(vec);
+
+	if ( getMargin()!=btScalar(0.) )
+	{
+		btVector3 vecnorm = vec;
+		if (vecnorm .length2() < (SIMD_EPSILON*SIMD_EPSILON))
+		{
+			vecnorm.setValue(btScalar(-1.),btScalar(-1.),btScalar(-1.));
+		} 
+		vecnorm.normalize();
+		supVertex+= getMargin() * vecnorm;
+	}
+	return supVertex;
+
+#else
+	return btVector3(0,0,0);
+#endif //__SPU__
+
+ }
+
+

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexInternalShape.h

@@ -0,0 +1,98 @@
+
+#ifndef BT_CONVEX_INTERNAL_SHAPE_H
+#define BT_CONVEX_INTERNAL_SHAPE_H
+
+#include "btConvexShape.h"
+
+///The btConvexInternalShape is an internal base class, shared by most convex shape implementations.
+class btConvexInternalShape : public btConvexShape
+{
+
+	protected:
+
+	//local scaling. collisionMargin is not scaled !
+	btVector3	m_localScaling;
+
+	btVector3	m_implicitShapeDimensions;
+	
+	btScalar	m_collisionMargin;
+
+	btScalar	m_padding;
+
+public:
+
+	btConvexInternalShape();
+
+	virtual ~btConvexInternalShape()
+	{
+
+	}
+
+
+	virtual btVector3	localGetSupportingVertex(const btVector3& vec)const;
+#ifndef __SPU__
+	virtual btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec) const= 0;
+	
+	//notice that the vectors should be unit length
+	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const= 0;
+#endif //#ifndef __SPU__
+
+	const btVector3& getImplicitShapeDimensions() const
+	{
+		return m_implicitShapeDimensions;
+	}
+
+	///getAabb's default implementation is brute force, expected derived classes to implement a fast dedicated version
+	void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const
+	{
+		getAabbSlow(t,aabbMin,aabbMax);
+	}
+
+
+	
+	virtual void getAabbSlow(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
+
+
+	virtual void	setLocalScaling(const btVector3& scaling);
+	virtual const btVector3& getLocalScaling() const 
+	{
+		return m_localScaling;
+	}
+
+	const btVector3& getLocalScalingNV() const 
+	{
+		return m_localScaling;
+	}
+
+	virtual void	setMargin(btScalar margin)
+	{
+		m_collisionMargin = margin;
+	}
+	virtual btScalar	getMargin() const
+	{
+		return m_collisionMargin;
+	}
+
+	btScalar	getMarginNV() const
+	{
+		return m_collisionMargin;
+	}
+
+	virtual int		getNumPreferredPenetrationDirections() const
+	{
+		return 0;
+	}
+	
+	virtual void	getPreferredPenetrationDirection(int index, btVector3& penetrationVector) const
+	{
+		(void)penetrationVector;
+		(void)index;
+		btAssert(0);
+	}
+
+
+	
+};
+
+
+#endif //BT_CONVEX_INTERNAL_SHAPE_H

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexShape.cpp

@@ -16,62 +16,3 @@ subject to the following restrictions:
 #include "btConvexShape.h"
 
 
-btConvexShape::btConvexShape()
-: m_localScaling(btScalar(1.),btScalar(1.),btScalar(1.)),
-m_collisionMargin(CONVEX_DISTANCE_MARGIN)
-{
-}
-
-
-void	btConvexShape::setLocalScaling(const btVector3& scaling)
-{
-	m_localScaling = scaling;
-}
-
-
-
-void	btConvexShape::getAabbSlow(const btTransform& trans,btVector3&minAabb,btVector3&maxAabb) const
-{
-
-	btScalar margin = getMargin();
-	for (int i=0;i<3;i++)
-	{
-		btVector3 vec(btScalar(0.),btScalar(0.),btScalar(0.));
-		vec[i] = btScalar(1.);
-
-		btVector3 sv = localGetSupportingVertex(vec*trans.getBasis());
-
-		btVector3 tmp = trans(sv);
-		maxAabb[i] = tmp[i]+margin;
-		vec[i] = btScalar(-1.);
-		tmp = trans(localGetSupportingVertex(vec*trans.getBasis()));
-		minAabb[i] = tmp[i]-margin;
-	}
-};
-
-
-btVector3	btConvexShape::localGetSupportingVertex(const btVector3& vec)const
-{
-#ifndef __SPU__
-
-	 btVector3	supVertex = localGetSupportingVertexWithoutMargin(vec);
-
-	if ( getMargin()!=btScalar(0.) )
-	{
-		btVector3 vecnorm = vec;
-		if (vecnorm .length2() < (SIMD_EPSILON*SIMD_EPSILON))
-		{
-			vecnorm.setValue(btScalar(-1.),btScalar(-1.),btScalar(-1.));
-		} 
-		vecnorm.normalize();
-		supVertex+= getMargin() * vecnorm;
-	}
-	return supVertex;
-
-#else
-	return btVector3(0,0,0);
-#endif //__SPU__
-
- }
-
-

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexShape.h

@@ -18,37 +18,25 @@ subject to the following restrictions:
 
 #include "btCollisionShape.h"
 
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
-#include "../../LinearMath/btMatrix3x3.h"
+#include "LinearMath/btMatrix3x3.h"
 #include "btCollisionMargin.h"
+#include "LinearMath/btAlignedAllocator.h"
 
 //todo: get rid of this btConvexCastResult thing!
 struct btConvexCastResult;
 #define MAX_PREFERRED_PENETRATION_DIRECTIONS 10
 
-/// btConvexShape is an abstract shape interface.
+/// The btConvexShape is an abstract shape interface, implemented by all convex shapes such as btBoxShape, btConvexHullShape etc.
-/// The explicit part provides plane-equations, the implicit part provides GetClosestPoint interface.
+/// It describes general convex shapes using the localGetSupportingVertex interface, used by collision detectors such as btGjkPairDetector.
-/// used in combination with GJK or btConvexCast
 ATTRIBUTE_ALIGNED16(class) btConvexShape : public btCollisionShape
 {
 
-protected:
-
-	//local scaling. collisionMargin is not scaled !
-	btVector3	m_localScaling;
-
-	btVector3	m_implicitShapeDimensions;
-	
-	btScalar	m_collisionMargin;
-
-	btScalar	m_padding[2];
-
-
-
 
 public:
-	btConvexShape();
+
+	BT_DECLARE_ALIGNED_ALLOCATOR();
 
 	virtual ~btConvexShape()
 	{
@@ -56,7 +44,7 @@ public:
 	}
 
 
-	virtual btVector3	localGetSupportingVertex(const btVector3& vec)const;
+	virtual btVector3	localGetSupportingVertex(const btVector3& vec)const =0;
 #ifndef __SPU__
 	virtual btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec) const= 0;
 	
@@ -64,63 +52,24 @@ public:
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const= 0;
 #endif //#ifndef __SPU__
 
-	const btVector3& getImplicitShapeDimensions() const
-	{
-		return m_implicitShapeDimensions;
-	}
 
 	///getAabb's default implementation is brute force, expected derived classes to implement a fast dedicated version
-	void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const
+	void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const =0;
-	{
-		getAabbSlow(t,aabbMin,aabbMax);
-	}
 
+	virtual void getAabbSlow(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const =0;
 
+	virtual void	setLocalScaling(const btVector3& scaling) =0;
+	virtual const btVector3& getLocalScaling() const =0;
+
+	virtual void	setMargin(btScalar margin)=0;
+
+	virtual btScalar	getMargin() const=0;
+
+	virtual int		getNumPreferredPenetrationDirections() const=0;
 	
-	virtual void getAabbSlow(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
+	virtual void	getPreferredPenetrationDirection(int index, btVector3& penetrationVector) const=0;
 
-
+};
-	virtual void	setLocalScaling(const btVector3& scaling);
-	virtual const btVector3& getLocalScaling() const 
-	{
-		return m_localScaling;
-	}
-
-	const btVector3& getLocalScalingNV() const 
-	{
-		return m_localScaling;
-	}
-
-	virtual void	setMargin(btScalar margin)
-	{
-		m_collisionMargin = margin;
-	}
-	virtual btScalar	getMargin() const
-	{
-		return m_collisionMargin;
-	}
-
-	btScalar	getMarginNV() const
-	{
-		return m_collisionMargin;
-	}
-
-	virtual int		getNumPreferredPenetrationDirections() const
-	{
-		return 0;
-	}
-	
-	virtual void	getPreferredPenetrationDirection(int index, btVector3& penetrationVector) const
-	{
-		(void)penetrationVector;
-		(void)index;
-		btAssert(0);
-	}
-
-
-
-}
-;
 
 
 

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexTriangleMeshShape.cpp

@@ -19,10 +19,11 @@ subject to the following restrictions:
 #include "BulletCollision/CollisionShapes/btStridingMeshInterface.h"
 
 
-btConvexTriangleMeshShape ::btConvexTriangleMeshShape (btStridingMeshInterface* meshInterface)
+btConvexTriangleMeshShape ::btConvexTriangleMeshShape (btStridingMeshInterface* meshInterface, bool calcAabb)
 :m_stridingMesh(meshInterface)
 {
-	recalcLocalAabb();
+	if ( calcAabb )
+		recalcLocalAabb();
 }
 
 

extern/bullet2/src/BulletCollision/CollisionShapes/btConvexTriangleMeshShape.h

@@ -3,20 +3,24 @@
 
 
 #include "btPolyhedralConvexShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
 
-/// btConvexTriangleMeshShape is a convex hull of a triangle mesh. If you just have a point cloud, you can use btConvexHullShape instead.
+/// The btConvexTriangleMeshShape is a convex hull of a triangle mesh, but the performance is not as good as btConvexHullShape.
-/// It uses the btStridingMeshInterface instead of a point cloud. This can avoid the duplication of the triangle mesh data.
+/// A small benefit of this class is that it uses the btStridingMeshInterface, so you can avoid the duplication of the triangle mesh data. Nevertheless, most users should use the much better performing btConvexHullShape instead.
 class btConvexTriangleMeshShape : public btPolyhedralConvexShape
 {
 
 	class btStridingMeshInterface*	m_stridingMesh;
 
 public:
-	btConvexTriangleMeshShape(btStridingMeshInterface* meshInterface);
+	btConvexTriangleMeshShape(btStridingMeshInterface* meshInterface, bool calcAabb = true);
 
-	class btStridingMeshInterface*	getStridingMesh()
+	class btStridingMeshInterface*	getMeshInterface()
+	{
+		return m_stridingMesh;
+	}
+	const class btStridingMeshInterface* getMeshInterface() const
 	{
 		return m_stridingMesh;
 	}
@@ -28,7 +32,7 @@ public:
 	virtual int	getShapeType()const { return CONVEX_TRIANGLEMESH_SHAPE_PROXYTYPE; }
 
 	//debugging
-	virtual char*	getName()const {return "ConvexTrimesh";}
+	virtual const char*	getName()const {return "ConvexTrimesh";}
 	
 	virtual int	getNumVertices() const;
 	virtual int getNumEdges() const;
@@ -49,3 +53,4 @@ public:
 #endif //CONVEX_TRIANGLEMESH_SHAPE_H
 
 
+

extern/bullet2/src/BulletCollision/CollisionShapes/btCylinderShape.cpp

@@ -45,7 +45,7 @@ void btCylinderShape::getAabb(const btTransform& t,btVector3& aabbMin,btVector3&
 }
 
 
-inline btVector3 CylinderLocalSupportX(const btVector3& halfExtents,const btVector3& v) 
+SIMD_FORCE_INLINE btVector3 CylinderLocalSupportX(const btVector3& halfExtents,const btVector3& v) 
 {
 const int cylinderUpAxis = 0;
 const int XX = 1;
@@ -163,24 +163,24 @@ const int ZZ = 1;
 
 btVector3	btCylinderShapeX::localGetSupportingVertexWithoutMargin(const btVector3& vec)const
 {
-	return CylinderLocalSupportX(getHalfExtents(),vec);
+	return CylinderLocalSupportX(getHalfExtentsWithoutMargin(),vec);
 }
 
 
 btVector3	btCylinderShapeZ::localGetSupportingVertexWithoutMargin(const btVector3& vec)const
 {
-	return CylinderLocalSupportZ(getHalfExtents(),vec);
+	return CylinderLocalSupportZ(getHalfExtentsWithoutMargin(),vec);
 }
 btVector3	btCylinderShape::localGetSupportingVertexWithoutMargin(const btVector3& vec)const
 {
-	return CylinderLocalSupportY(getHalfExtents(),vec);
+	return CylinderLocalSupportY(getHalfExtentsWithoutMargin(),vec);
 }
 
 void	btCylinderShape::batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const
 {
 	for (int i=0;i<numVectors;i++)
 	{
-		supportVerticesOut[i] = CylinderLocalSupportY(getHalfExtents(),vectors[i]);
+		supportVerticesOut[i] = CylinderLocalSupportY(getHalfExtentsWithoutMargin(),vectors[i]);
 	}
 }
 
@@ -188,7 +188,7 @@ void	btCylinderShapeZ::batchedUnitVectorGetSupportingVertexWithoutMargin(const b
 {
 	for (int i=0;i<numVectors;i++)
 	{
-		supportVerticesOut[i] = CylinderLocalSupportZ(getHalfExtents(),vectors[i]);
+		supportVerticesOut[i] = CylinderLocalSupportZ(getHalfExtentsWithoutMargin(),vectors[i]);
 	}
 }
 
@@ -199,7 +199,7 @@ void	btCylinderShapeX::batchedUnitVectorGetSupportingVertexWithoutMargin(const b
 {
 	for (int i=0;i<numVectors;i++)
 	{
-		supportVerticesOut[i] = CylinderLocalSupportX(getHalfExtents(),vectors[i]);
+		supportVerticesOut[i] = CylinderLocalSupportX(getHalfExtentsWithoutMargin(),vectors[i]);
 	}
 }
 

extern/bullet2/src/BulletCollision/CollisionShapes/btCylinderShape.h

@@ -17,10 +17,10 @@ subject to the following restrictions:
 #define CYLINDER_MINKOWSKI_H
 
 #include "btBoxShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
 
-/// implements cylinder shape interface
+/// The btCylinderShape class implements a cylinder shape primitive, centered around the origin. Its central axis aligned with the Y axis. btCylinderShapeX is aligned with the X axis and btCylinderShapeZ around the Z axis.
 class btCylinderShape : public btBoxShape
 
 {
@@ -60,7 +60,7 @@ public:
 
 
 	//use box inertia
-	//	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	//	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	virtual int	getShapeType() const
 	{
@@ -74,11 +74,11 @@ public:
 
 	virtual btScalar getRadius() const
 	{
-		return getHalfExtents().getX();
+		return getHalfExtentsWithMargin().getX();
 	}
 
 	//debugging
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "CylinderY";
 	}
@@ -96,14 +96,14 @@ public:
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const;
 	
 		//debugging
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "CylinderX";
 	}
 
 	virtual btScalar getRadius() const
 	{
-		return getHalfExtents().getY();
+		return getHalfExtentsWithMargin().getY();
 	}
 
 };
@@ -121,14 +121,14 @@ public:
 		return 2;
 	}
 		//debugging
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "CylinderZ";
 	}
 
 	virtual btScalar getRadius() const
 	{
-		return getHalfExtents().getX();
+		return getHalfExtentsWithMargin().getX();
 	}
 
 };

extern/bullet2/src/BulletCollision/CollisionShapes/btEmptyShape.cpp

@@ -40,7 +40,7 @@ void btEmptyShape::getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aa
 
 }
 
-void	btEmptyShape::calculateLocalInertia(btScalar ,btVector3& )
+void	btEmptyShape::calculateLocalInertia(btScalar ,btVector3& ) const
 {
 	btAssert(0);
 }

extern/bullet2/src/BulletCollision/CollisionShapes/btEmptyShape.h

@@ -18,16 +18,16 @@ subject to the following restrictions:
 
 #include "btConcaveShape.h"
 
-#include "../../LinearMath/btVector3.h"
+#include "LinearMath/btVector3.h"
-#include "../../LinearMath/btTransform.h"
+#include "LinearMath/btTransform.h"
-#include "../../LinearMath/btMatrix3x3.h"
+#include "LinearMath/btMatrix3x3.h"
 #include "btCollisionMargin.h"
 
 
 
 
-/// btEmptyShape is a collision shape without actual collision detection. 
+/// The btEmptyShape is a collision shape without actual collision detection shape, so most users should ignore this class.
-///It can be replaced by another shape during runtime
+/// It can be replaced by another shape during runtime, but the inertia tensor should be recomputed.
 class btEmptyShape	: public btConcaveShape
 {
 public:
@@ -49,12 +49,12 @@ public:
 		return m_localScaling;
 	}
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 	
 	virtual int	getShapeType() const { return EMPTY_SHAPE_PROXYTYPE;}
 
 	
-	virtual char*	getName()const
+	virtual const char*	getName()const
 	{
 		return "Empty";
 	}

extern/bullet2/src/BulletCollision/CollisionShapes/btHeightfieldTerrainShape.cpp

@@ -18,16 +18,18 @@ subject to the following restrictions:
 #include "LinearMath/btTransformUtil.h"
 
 
-btHeightfieldTerrainShape::btHeightfieldTerrainShape(int width,int length,void* heightfieldData,btScalar maxHeight,int upAxis,bool useFloatData,bool flipQuadEdges)
+btHeightfieldTerrainShape::btHeightfieldTerrainShape(int heightStickWidth, int heightStickLength,void* heightfieldData,btScalar maxHeight,int upAxis,bool useFloatData,bool flipQuadEdges)
-:m_localScaling(btScalar(1.),btScalar(1.),btScalar(1.)),
+: m_heightStickWidth(heightStickWidth),
-m_width(width),
+m_heightStickLength(heightStickLength),
-m_length(length),
-m_heightfieldDataUnknown(heightfieldData),
 m_maxHeight(maxHeight),
-m_upAxis(upAxis),
+m_width((btScalar)heightStickWidth-1),
+m_length((btScalar)heightStickLength-1),
+m_heightfieldDataUnknown(heightfieldData),
 m_useFloatData(useFloatData),
 m_flipQuadEdges(flipQuadEdges),
-m_useDiamondSubdivision(false)
+m_useDiamondSubdivision(false),
+m_upAxis(upAxis),
+m_localScaling(btScalar(1.),btScalar(1.),btScalar(1.))
 {
 
 
@@ -43,25 +45,25 @@ m_useDiamondSubdivision(false)
 	case 0:
 		{
 			halfExtents.setValue(
-				m_maxHeight,
+				btScalar(m_maxHeight),
-				m_width,
+				btScalar(m_width), //?? don't know if this should change
-				m_length);
+				btScalar(m_length));
 			break;
 		}
 	case 1:
 		{
 			halfExtents.setValue(
-				m_width,
+				btScalar(m_width),
-				m_maxHeight,
+				btScalar(m_maxHeight),
-				m_length);
+				btScalar(m_length));
 			break;
 		};
 	case 2:
 		{
 			halfExtents.setValue(
-				m_width,
+				btScalar(m_width),
-				m_length,
+				btScalar(m_length),
-				m_maxHeight
+				btScalar(m_maxHeight)
 			);
 			break;
 		}
@@ -114,11 +116,11 @@ btScalar	btHeightfieldTerrainShape::getHeightFieldValue(int x,int y) const
 	btScalar val = 0.f;
 	if (m_useFloatData)
 	{
-		val = m_heightfieldDataFloat[(y*m_width)+x];
+		val = m_heightfieldDataFloat[(y*m_heightStickWidth)+x];
 	} else
 	{
 		//assume unsigned short int
-		unsigned char heightFieldValue = m_heightfieldDataUnsignedChar[(y*m_width)+x];
+		unsigned char heightFieldValue = m_heightfieldDataUnsignedChar[(y*m_heightStickWidth)+x];
 		val = heightFieldValue* (m_maxHeight/btScalar(65535));
 	}
 	return val;
@@ -133,8 +135,8 @@ void	btHeightfieldTerrainShape::getVertex(int x,int y,btVector3& vertex) const
 
 	btAssert(x>=0);
 	btAssert(y>=0);
-	btAssert(x<m_width);
+	btAssert(x<m_heightStickWidth);
-	btAssert(y<m_length);
+	btAssert(y<m_heightStickLength);
 
 
 	btScalar	height = getHeightFieldValue(x,y);
@@ -145,25 +147,25 @@ void	btHeightfieldTerrainShape::getVertex(int x,int y,btVector3& vertex) const
 		{
 		vertex.setValue(
 			height,
-			(-m_width/2 ) + x,
+			(-m_width/btScalar(2.0)) + x,
-			(-m_length/2 ) + y
+			(-m_length/btScalar(2.0) ) + y
 			);
 			break;
 		}
 	case 1:
 		{
 			vertex.setValue(
-			(-m_width/2 ) + x,
+			(-m_width/btScalar(2.0)) + x,
 			height,
-			(-m_length/2 ) + y
+			(-m_length/btScalar(2.0)) + y
 			);
 			break;
 		};
 	case 2:
 		{
 			vertex.setValue(
-			(-m_width/2 ) + x,
+			(-m_width/btScalar(2.0)) + x,
-			(-m_length/2 ) + y,
+			(-m_length/btScalar(2.0)) + y,
 			height
 			);
 			break;
@@ -180,21 +182,20 @@ void	btHeightfieldTerrainShape::getVertex(int x,int y,btVector3& vertex) const
 }
 
 
-void btHeightfieldTerrainShape::quantizeWithClamp(int* out, const btVector3& point) const
+void btHeightfieldTerrainShape::quantizeWithClamp(int* out, const btVector3& point,int /*isMax*/) const
 {
-	
-
 	btVector3 clampedPoint(point);
 	clampedPoint.setMax(m_localAabbMin);
 	clampedPoint.setMin(m_localAabbMax);
 
-	btVector3 v = (clampedPoint );// * m_quantization;
+	btVector3 v = (clampedPoint);// - m_bvhAabbMin) * m_bvhQuantization;
-
-	out[0] = (int)(v.getX());
-	out[1] = (int)(v.getY());
-	out[2] = (int)(v.getZ());
-	//correct for
 
+	//TODO: optimization: check out how to removed this btFabs
+		
+	out[0] = (int)(v.getX() + v.getX() / btFabs(v.getX())* btScalar(0.5) );
+	out[1] = (int)(v.getY() + v.getY() / btFabs(v.getY())* btScalar(0.5) );
+	out[2] = (int)(v.getZ() + v.getZ() / btFabs(v.getZ())* btScalar(0.5) );
+	
 }
 
 
@@ -212,24 +213,24 @@ void	btHeightfieldTerrainShape::processAllTriangles(btTriangleCallback* callback
 	btVector3	localAabbMin = aabbMin*btVector3(1.f/m_localScaling[0],1.f/m_localScaling[1],1.f/m_localScaling[2]);
 	btVector3	localAabbMax = aabbMax*btVector3(1.f/m_localScaling[0],1.f/m_localScaling[1],1.f/m_localScaling[2]);
 	
-	quantizeWithClamp(quantizedAabbMin, localAabbMin);
+	quantizeWithClamp(quantizedAabbMin, localAabbMin,0);
-	quantizeWithClamp(quantizedAabbMax, localAabbMax);
+	quantizeWithClamp(quantizedAabbMax, localAabbMax,1);
 	
 	
 
 	int startX=0;
-	int endX=m_width-1;
+	int endX=m_heightStickWidth-1;
 	int startJ=0;
-	int endJ=m_length-1;
+	int endJ=m_heightStickLength-1;
 
 	switch (m_upAxis)
 	{
 	case 0:
 		{
-			quantizedAabbMin[1]+=m_width/2-1;
+			quantizedAabbMin[1]+=m_heightStickWidth/2-1;
-			quantizedAabbMax[1]+=m_width/2+1;
+			quantizedAabbMax[1]+=m_heightStickWidth/2+1;
-			quantizedAabbMin[2]+=m_length/2-1;
+			quantizedAabbMin[2]+=m_heightStickLength/2-1;
-			quantizedAabbMax[2]+=m_length/2+1;
+			quantizedAabbMax[2]+=m_heightStickLength/2+1;
 
 			if (quantizedAabbMin[1]>startX)
 				startX = quantizedAabbMin[1];
@@ -243,10 +244,10 @@ void	btHeightfieldTerrainShape::processAllTriangles(btTriangleCallback* callback
 		}
 	case 1:
 		{
-			quantizedAabbMin[0]+=m_width/2-1;
+			quantizedAabbMin[0]+=m_heightStickWidth/2-1;
-			quantizedAabbMax[0]+=m_width/2+1;
+			quantizedAabbMax[0]+=m_heightStickWidth/2+1;
-			quantizedAabbMin[2]+=m_length/2-1;
+			quantizedAabbMin[2]+=m_heightStickLength/2-1;
-			quantizedAabbMax[2]+=m_length/2+1;
+			quantizedAabbMax[2]+=m_heightStickLength/2+1;
 
 			if (quantizedAabbMin[0]>startX)
 				startX = quantizedAabbMin[0];
@@ -260,10 +261,10 @@ void	btHeightfieldTerrainShape::processAllTriangles(btTriangleCallback* callback
 		};
 	case 2:
 		{
-			quantizedAabbMin[0]+=m_width/2-1;
+			quantizedAabbMin[0]+=m_heightStickWidth/2-1;
-			quantizedAabbMax[0]+=m_width/2+1;
+			quantizedAabbMax[0]+=m_heightStickWidth/2+1;
-			quantizedAabbMin[1]+=m_length/2-1;
+			quantizedAabbMin[1]+=m_heightStickLength/2-1;
-			quantizedAabbMax[1]+=m_length/2+1;
+			quantizedAabbMax[1]+=m_heightStickLength/2+1;
 
 			if (quantizedAabbMin[0]>startX)
 				startX = quantizedAabbMin[0];
@@ -290,7 +291,7 @@ void	btHeightfieldTerrainShape::processAllTriangles(btTriangleCallback* callback
 		for(int x=startX; x<endX; x++)
 		{
 			btVector3 vertices[3];
-			if (m_flipQuadEdges || (m_useDiamondSubdivision && ((j+x) & 1)))
+			if (m_flipQuadEdges || (m_useDiamondSubdivision && !((j+x) & 1)))
 			{
         //first triangle
         getVertex(x,j,vertices[0]);
@@ -322,7 +323,7 @@ void	btHeightfieldTerrainShape::processAllTriangles(btTriangleCallback* callback
 
 }
 
-void	btHeightfieldTerrainShape::calculateLocalInertia(btScalar ,btVector3& inertia)
+void	btHeightfieldTerrainShape::calculateLocalInertia(btScalar ,btVector3& inertia) const
 {
 	//moving concave objects not supported
 	

extern/bullet2/src/BulletCollision/CollisionShapes/btHeightfieldTerrainShape.h

@@ -18,7 +18,8 @@ subject to the following restrictions:
 
 #include "btConcaveShape.h"
 
-///btHeightfieldTerrainShape simulates a 2D heightfield terrain 
+///The btHeightfieldTerrainShape simulates a 2D heightfield terrain collision shape. You can also use the more general btBvhTriangleMeshShape instead.
+///An example implementation of btHeightfieldTerrainShape is provided in Demos/VehicleDemo/VehicleDemo.cpp
 class btHeightfieldTerrainShape : public btConcaveShape
 {
 protected:
@@ -26,9 +27,11 @@ protected:
 	btVector3	m_localAabbMax;
 	
 	///terrain data
-	int	m_width;
+	int	m_heightStickWidth;
-	int m_length;
+	int m_heightStickLength;
 	btScalar	m_maxHeight;
+	btScalar m_width;
+	btScalar m_length;
 	union
 	{
 		unsigned char*	m_heightfieldDataUnsignedChar;
@@ -45,7 +48,7 @@ protected:
 	btVector3	m_localScaling;
 
 	virtual btScalar	getHeightFieldValue(int x,int y) const;
-	void		quantizeWithClamp(int* out, const btVector3& point) const;
+	void		quantizeWithClamp(int* out, const btVector3& point,int isMax) const;
 	void		getVertex(int x,int y,btVector3& vertex) const;
 
 	inline bool testQuantizedAabbAgainstQuantizedAabb(int* aabbMin1, int* aabbMax1,const  int* aabbMin2,const  int* aabbMax2) const
@@ -58,7 +61,7 @@ protected:
 	}
 
 public:
-	btHeightfieldTerrainShape(int width,int height,void* heightfieldData, btScalar maxHeight,int upAxis,bool useFloatData,bool flipQuadEdges);
+	btHeightfieldTerrainShape(int heightStickWidth,int heightStickHeight,void* heightfieldData, btScalar maxHeight,int upAxis,bool useFloatData,bool flipQuadEdges);
 
 	virtual ~btHeightfieldTerrainShape();
 
@@ -74,14 +77,14 @@ public:
 
 	virtual void	processAllTriangles(btTriangleCallback* callback,const btVector3& aabbMin,const btVector3& aabbMax) const;
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	virtual void	setLocalScaling(const btVector3& scaling);
 	
 	virtual const btVector3& getLocalScaling() const;
 	
 	//debugging
-	virtual char*	getName()const {return "HEIGHTFIELD";}
+	virtual const char*	getName()const {return "HEIGHTFIELD";}
 
 };
 

extern/bullet2/src/BulletCollision/CollisionShapes/btMaterial.h

@@ -0,0 +1,34 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2008 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+/// This file was created by Alex Silverman
+
+#ifndef MATERIAL_H
+#define MATERIAL_H
+
+// Material class to be used by btMultimaterialTriangleMeshShape to store triangle properties
+class btMaterial
+{
+    // public members so that materials can change due to world events
+public:
+    btScalar m_friction;
+    btScalar m_restitution;
+    int pad[2];
+
+    btMaterial(){}
+    btMaterial(btScalar fric, btScalar rest) { m_friction = fric; m_restitution = rest; }
+};
+
+#endif // MATERIAL_H

extern/bullet2/src/BulletCollision/CollisionShapes/btMinkowskiSumShape.cpp

@@ -26,9 +26,9 @@ m_shapeB(shapeB)
 
 btVector3 btMinkowskiSumShape::localGetSupportingVertexWithoutMargin(const btVector3& vec)const
 {
-	btVector3 supVertexA = m_transA(m_shapeA->localGetSupportingVertexWithoutMargin(vec*m_transA.getBasis()));
+	btVector3 supVertexA = m_transA(m_shapeA->localGetSupportingVertexWithoutMargin(-vec*m_transA.getBasis()));
 	btVector3 supVertexB = m_transB(m_shapeB->localGetSupportingVertexWithoutMargin(vec*m_transB.getBasis()));
-	return supVertexA + supVertexB;
+	return  supVertexA - supVertexB;
 }
 
 void	btMinkowskiSumShape::batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const
@@ -49,7 +49,7 @@ btScalar	btMinkowskiSumShape::getMargin() const
 }
 
 
-void	btMinkowskiSumShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btMinkowskiSumShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	(void)mass;
 	btAssert(0);

extern/bullet2/src/BulletCollision/CollisionShapes/btMinkowskiSumShape.h

@@ -16,11 +16,11 @@ subject to the following restrictions:
 #ifndef MINKOWSKI_SUM_SHAPE_H
 #define MINKOWSKI_SUM_SHAPE_H
 
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
-/// btMinkowskiSumShape represents implicit (getSupportingVertex) based minkowski sum of two convex implicit shapes.
+/// The btMinkowskiSumShape is only for advanced users. This shape represents implicit based minkowski sum of two convex implicit shapes.
-class btMinkowskiSumShape : public btConvexShape
+class btMinkowskiSumShape : public btConvexInternalShape
 {
 
 	btTransform	m_transA;
@@ -37,7 +37,7 @@ public:
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const;
 
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	void	setTransformA(const btTransform&	transA) { m_transA = transA;}
 	void	setTransformB(const btTransform&	transB) { m_transB = transB;}
@@ -53,7 +53,7 @@ public:
 	const btConvexShape*	getShapeA() const { return m_shapeA;}
 	const btConvexShape*	getShapeB() const { return m_shapeB;}
 
-	virtual char*	getName()const 
+	virtual const char*	getName()const 
 	{
 		return "MinkowskiSum";
 	}

extern/bullet2/src/BulletCollision/CollisionShapes/btMultiSphereShape.cpp

@@ -47,7 +47,7 @@ btMultiSphereShape::btMultiSphereShape (const btVector3& inertiaHalfExtents,cons
 
 	btVector3 vec = vec0;
 	btScalar lenSqr = vec.length2();
-	if (lenSqr < btScalar(0.0001))
+	if (lenSqr < (SIMD_EPSILON*SIMD_EPSILON))
 	{
 		vec.setValue(1,0,0);
 	} else
@@ -116,7 +116,7 @@ btMultiSphereShape::btMultiSphereShape (const btVector3& inertiaHalfExtents,cons
 
 
 
-void	btMultiSphereShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btMultiSphereShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	//as an approximation, take the inertia of the box that bounds the spheres
 

extern/bullet2/src/BulletCollision/CollisionShapes/btMultiSphereShape.h

@@ -16,13 +16,14 @@ subject to the following restrictions:
 #ifndef MULTI_SPHERE_MINKOWSKI_H
 #define MULTI_SPHERE_MINKOWSKI_H
 
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
 #define MAX_NUM_SPHERES 5
 
-///btMultiSphereShape represents implicit convex hull of a collection of spheres (using getSupportingVertex)
+///The btMultiSphereShape represents the convex hull of a collection of spheres. You can create special capsules or other smooth volumes.
-class btMultiSphereShape : public btConvexShape
+///It is possible to animate the spheres for deformation.
+class btMultiSphereShape : public btConvexInternalShape
 
 {
 	
@@ -39,7 +40,7 @@ public:
 	btMultiSphereShape (const btVector3& inertiaHalfExtents,const btVector3* positions,const btScalar* radi,int numSpheres);
 
 	///CollisionShape Interface
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	/// btConvexShape Interface
 	virtual btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec)const;
@@ -63,7 +64,7 @@ public:
 
 	virtual int	getShapeType() const { return MULTI_SPHERE_SHAPE_PROXYTYPE; }
 
-	virtual char*	getName()const 
+	virtual const char*	getName()const 
 	{
 		return "MultiSphere";
 	}

extern/bullet2/src/BulletCollision/CollisionShapes/btMultimaterialTriangleMeshShape.cpp

@@ -0,0 +1,45 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2008 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+/// This file was created by Alex Silverman
+
+#include "BulletCollision/CollisionShapes/btMultimaterialTriangleMeshShape.h"
+#include "BulletCollision/CollisionShapes/btTriangleIndexVertexMaterialArray.h"
+//#include "BulletCollision/CollisionShapes/btOptimizedBvh.h"
+
+
+///Obtains the material for a specific triangle
+const btMaterial * btMultimaterialTriangleMeshShape::getMaterialProperties(int partID, int triIndex)
+{
+    const unsigned char * materialBase = 0;
+    int numMaterials;
+    PHY_ScalarType materialType;
+    int materialStride;
+    const unsigned char * triangleMaterialBase = 0;
+    int numTriangles;
+    int triangleMaterialStride;
+    PHY_ScalarType triangleType;
+
+    ((btTriangleIndexVertexMaterialArray*)m_meshInterface)->getLockedReadOnlyMaterialBase(&materialBase, numMaterials, materialType, materialStride,
+        &triangleMaterialBase, numTriangles, triangleMaterialStride, triangleType, partID);
+
+    // return the pointer to the place with the friction for the triangle
+    // TODO: This depends on whether it's a moving mesh or not
+    // BUG IN GIMPACT
+    //return (btScalar*)(&materialBase[triangleMaterialBase[(triIndex-1) * triangleMaterialStride] * materialStride]);
+    int * matInd = (int *)(&(triangleMaterialBase[(triIndex * triangleMaterialStride)]));
+    btMaterial *matVal = (btMaterial *)(&(materialBase[*matInd * materialStride]));
+    return (matVal);
+}

extern/bullet2/src/BulletCollision/CollisionShapes/btMultimaterialTriangleMeshShape.h

@@ -0,0 +1,124 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2008 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+/// This file was created by Alex Silverman
+
+#ifndef BVH_TRIANGLE_MATERIAL_MESH_SHAPE_H
+#define BVH_TRIANGLE_MATERIAL_MESH_SHAPE_H
+
+#include "btBvhTriangleMeshShape.h"
+#include "btMaterial.h"
+
+///The BvhTriangleMaterialMeshShape extends the btBvhTriangleMeshShape. Its main contribution is the interface into a material array, which allows per-triangle friction and restitution.
+ATTRIBUTE_ALIGNED16(class) btMultimaterialTriangleMeshShape : public btBvhTriangleMeshShape
+{
+    btAlignedObjectArray <btMaterial*> m_materialList;
+    int ** m_triangleMaterials;
+
+public:
+
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+    btMultimaterialTriangleMeshShape(): btBvhTriangleMeshShape() {}
+    btMultimaterialTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression, bool buildBvh = true):
+        btBvhTriangleMeshShape(meshInterface, useQuantizedAabbCompression, buildBvh)
+        {
+            btVector3 m_triangle[3];
+            const unsigned char *vertexbase;
+            int numverts;
+            PHY_ScalarType type;
+            int stride;
+            const unsigned char *indexbase;
+            int indexstride;
+            int numfaces;
+            PHY_ScalarType indicestype;
+
+            //m_materialLookup = (int**)(btAlignedAlloc(sizeof(int*) * meshInterface->getNumSubParts(), 16));
+
+            for(int i = 0; i < meshInterface->getNumSubParts(); i++)
+            {
+                m_meshInterface->getLockedReadOnlyVertexIndexBase(
+                    &vertexbase,
+                    numverts,
+                    type,
+                    stride,
+                    &indexbase,
+                    indexstride,
+                    numfaces,
+                    indicestype,
+                    i);
+                //m_materialLookup[i] = (int*)(btAlignedAlloc(sizeof(int) * numfaces, 16));
+            }
+        }
+
+	///optionally pass in a larger bvh aabb, used for quantization. This allows for deformations within this aabb
+	btMultimaterialTriangleMeshShape(btStridingMeshInterface* meshInterface, bool useQuantizedAabbCompression,const btVector3& bvhAabbMin,const btVector3& bvhAabbMax, bool buildBvh = true):
+        btBvhTriangleMeshShape(meshInterface, useQuantizedAabbCompression, bvhAabbMin, bvhAabbMax, buildBvh)
+        {
+            btVector3 m_triangle[3];
+            const unsigned char *vertexbase;
+            int numverts;
+            PHY_ScalarType type;
+            int stride;
+            const unsigned char *indexbase;
+            int indexstride;
+            int numfaces;
+            PHY_ScalarType indicestype;
+
+            //m_materialLookup = (int**)(btAlignedAlloc(sizeof(int*) * meshInterface->getNumSubParts(), 16));
+
+            for(int i = 0; i < meshInterface->getNumSubParts(); i++)
+            {
+                m_meshInterface->getLockedReadOnlyVertexIndexBase(
+                    &vertexbase,
+                    numverts,
+                    type,
+                    stride,
+                    &indexbase,
+                    indexstride,
+                    numfaces,
+                    indicestype,
+                    i);
+                //m_materialLookup[i] = (int*)(btAlignedAlloc(sizeof(int) * numfaces * 2, 16));
+            }
+        }
+	
+    virtual ~btMultimaterialTriangleMeshShape()
+    {
+/*
+        for(int i = 0; i < m_meshInterface->getNumSubParts(); i++)
+        {
+            btAlignedFree(m_materialValues[i]);
+            m_materialLookup[i] = NULL;
+        }
+        btAlignedFree(m_materialValues);
+        m_materialLookup = NULL;
+*/
+    }
+	virtual int	getShapeType() const
+	{
+		return MULTIMATERIAL_TRIANGLE_MESH_PROXYTYPE;
+	}
+	
+	//debugging
+	virtual const char*	getName()const {return "MULTIMATERIALTRIANGLEMESH";}
+
+    ///Obtains the material for a specific triangle
+    const btMaterial * getMaterialProperties(int partID, int triIndex);
+
+}
+;
+
+#endif //BVH_TRIANGLE_MATERIAL_MESH_SHAPE_H

extern/bullet2/src/BulletCollision/CollisionShapes/btOptimizedBvh.cpp

@@ -19,13 +19,12 @@ subject to the following restrictions:
 #include "LinearMath/btIDebugDraw.h"
 
 
-
+btOptimizedBvh::btOptimizedBvh()
-btOptimizedBvh::btOptimizedBvh() : m_useQuantization(false), 
-					m_traversalMode(TRAVERSAL_STACKLESS_CACHE_FRIENDLY)
-					//m_traversalMode(TRAVERSAL_STACKLESS)
-					//m_traversalMode(TRAVERSAL_RECURSIVE)
 { 
+}
 
+btOptimizedBvh::~btOptimizedBvh()
+{
 }
 
 
@@ -80,7 +79,7 @@ void btOptimizedBvh::build(btStridingMeshInterface* triangles, bool useQuantized
 	struct	QuantizedNodeTriangleCallback : public btInternalTriangleIndexCallback
 	{
 		QuantizedNodeArray&	m_triangleNodes;
-		const btOptimizedBvh* m_optimizedTree; // for quantization
+		const btQuantizedBvh* m_optimizedTree; // for quantization
 
 		QuantizedNodeTriangleCallback& operator=(QuantizedNodeTriangleCallback& other)
 		{
@@ -89,14 +88,16 @@ void btOptimizedBvh::build(btStridingMeshInterface* triangles, bool useQuantized
 			return *this;
 		}
 
-		QuantizedNodeTriangleCallback(QuantizedNodeArray&	triangleNodes,const btOptimizedBvh* tree)
+		QuantizedNodeTriangleCallback(QuantizedNodeArray&	triangleNodes,const btQuantizedBvh* tree)
 			:m_triangleNodes(triangleNodes),m_optimizedTree(tree)
 		{
 		}
 
 		virtual void internalProcessTriangleIndex(btVector3* triangle,int partId,int  triangleIndex)
 		{
-			btAssert(partId==0);
+			// The partId and triangle index must fit in the same (positive) integer
+			btAssert(partId < (1<<MAX_NUM_PARTS_IN_BITS));
+			btAssert(triangleIndex < (1<<(31-MAX_NUM_PARTS_IN_BITS)));
 			//negative indices are reserved for escapeIndex
 			btAssert(triangleIndex>=0);
 
@@ -111,10 +112,29 @@ void btOptimizedBvh::build(btStridingMeshInterface* triangles, bool useQuantized
 			aabbMin.setMin(triangle[2]);
 			aabbMax.setMax(triangle[2]);
 
-			m_optimizedTree->quantizeWithClamp(&node.m_quantizedAabbMin[0],aabbMin);
+			//PCK: add these checks for zero dimensions of aabb
-			m_optimizedTree->quantizeWithClamp(&node.m_quantizedAabbMax[0],aabbMax);
+			const btScalar MIN_AABB_DIMENSION = btScalar(0.002);
+			const btScalar MIN_AABB_HALF_DIMENSION = btScalar(0.001);
+			if (aabbMax.x() - aabbMin.x() < MIN_AABB_DIMENSION)
+			{
+				aabbMax.setX(aabbMax.x() + MIN_AABB_HALF_DIMENSION);
+				aabbMin.setX(aabbMin.x() - MIN_AABB_HALF_DIMENSION);
+			}
+			if (aabbMax.y() - aabbMin.y() < MIN_AABB_DIMENSION)
+			{
+				aabbMax.setY(aabbMax.y() + MIN_AABB_HALF_DIMENSION);
+				aabbMin.setY(aabbMin.y() - MIN_AABB_HALF_DIMENSION);
+			}
+			if (aabbMax.z() - aabbMin.z() < MIN_AABB_DIMENSION)
+			{
+				aabbMax.setZ(aabbMax.z() + MIN_AABB_HALF_DIMENSION);
+				aabbMin.setZ(aabbMin.z() - MIN_AABB_HALF_DIMENSION);
+			}
 
-			node.m_escapeIndexOrTriangleIndex = triangleIndex;
+			m_optimizedTree->quantize(&node.m_quantizedAabbMin[0],aabbMin,0);
+			m_optimizedTree->quantize(&node.m_quantizedAabbMax[0],aabbMax,1);
+
+			node.m_escapeIndexOrTriangleIndex = (partId<<(31-MAX_NUM_PARTS_IN_BITS)) | triangleIndex;
 
 			m_triangleNodes.push_back(node);
 		}
@@ -170,10 +190,45 @@ void btOptimizedBvh::build(btStridingMeshInterface* triangles, bool useQuantized
 		subtree.m_rootNodeIndex = 0;
 		subtree.m_subtreeSize = m_quantizedContiguousNodes[0].isLeafNode() ? 1 : m_quantizedContiguousNodes[0].getEscapeIndex();
 	}
+
+	//PCK: update the copy of the size
+	m_subtreeHeaderCount = m_SubtreeHeaders.size();
+
+	//PCK: clear m_quantizedLeafNodes and m_leafNodes, they are temporary
+	m_quantizedLeafNodes.clear();
+	m_leafNodes.clear();
 }
 
 
 
+
+void	btOptimizedBvh::refit(btStridingMeshInterface* meshInterface,const btVector3& aabbMin,const btVector3& aabbMax)
+{
+	if (m_useQuantization)
+	{
+
+		setQuantizationValues(aabbMin,aabbMax);
+
+		updateBvhNodes(meshInterface,0,m_curNodeIndex,0);
+
+		///now update all subtree headers
+
+		int i;
+		for (i=0;i<m_SubtreeHeaders.size();i++)
+		{
+			btBvhSubtreeInfo& subtree = m_SubtreeHeaders[i];
+			subtree.setAabbFromQuantizeNode(m_quantizedContiguousNodes[subtree.m_rootNodeIndex]);
+		}
+
+	} else
+	{
+
+	}
+}
+
+
+
+
 void	btOptimizedBvh::refitPartial(btStridingMeshInterface* meshInterface,const btVector3& aabbMin,const btVector3& aabbMax)
 {
 	//incrementally initialize quantization values
@@ -193,16 +248,17 @@ void	btOptimizedBvh::refitPartial(btStridingMeshInterface* meshInterface,const b
 	unsigned short	quantizedQueryAabbMin[3];
 	unsigned short	quantizedQueryAabbMax[3];
 
-	quantizeWithClamp(&quantizedQueryAabbMin[0],aabbMin);
+	quantize(&quantizedQueryAabbMin[0],aabbMin,0);
-	quantizeWithClamp(&quantizedQueryAabbMax[0],aabbMax);
+	quantize(&quantizedQueryAabbMax[0],aabbMax,1);
 
 	int i;
 	for (i=0;i<this->m_SubtreeHeaders.size();i++)
 	{
 		btBvhSubtreeInfo& subtree = m_SubtreeHeaders[i];
 
-		bool overlap = testQuantizedAabbAgainstQuantizedAabb(quantizedQueryAabbMin,quantizedQueryAabbMax,subtree.m_quantizedAabbMin,subtree.m_quantizedAabbMax);
+		//PCK: unsigned instead of bool
-		if (overlap)
+		unsigned overlap = testQuantizedAabbAgainstQuantizedAabb(quantizedQueryAabbMin,quantizedQueryAabbMax,subtree.m_quantizedAabbMin,subtree.m_quantizedAabbMax);
+		if (overlap != 0)
 		{
 			updateBvhNodes(meshInterface,subtree.m_rootNodeIndex,subtree.m_rootNodeIndex+subtree.m_subtreeSize,i);
 
@@ -212,36 +268,23 @@ void	btOptimizedBvh::refitPartial(btStridingMeshInterface* meshInterface,const b
 	
 }
 
-///just for debugging, to visualize the individual patches/subtrees
-#ifdef DEBUG_PATCH_COLORS
-btVector3 color[4]=
-{
-	btVector3(255,0,0),
-	btVector3(0,255,0),
-	btVector3(0,0,255),
-	btVector3(0,255,255)
-};
-#endif //DEBUG_PATCH_COLORS
-
-
 void	btOptimizedBvh::updateBvhNodes(btStridingMeshInterface* meshInterface,int firstNode,int endNode,int index)
 {
 	(void)index;
 
 	btAssert(m_useQuantization);
 
-	int nodeSubPart=0;
+	int curNodeSubPart=-1;
 
 	//get access info to trianglemesh data
-		const unsigned char *vertexbase;
+		const unsigned char *vertexbase = 0;
-		int numverts;
+		int numverts = 0;
-		PHY_ScalarType type;
+		PHY_ScalarType type = PHY_INTEGER;
-		int stride;
+		int stride = 0;
-		const unsigned char *indexbase;
+		const unsigned char *indexbase = 0;
-		int indexstride;
+		int indexstride = 0;
-		int numfaces;
+		int numfaces = 0;
-		PHY_ScalarType indicestype;
+		PHY_ScalarType indicestype = PHY_INTEGER;
-		meshInterface->getLockedReadOnlyVertexIndexBase(&vertexbase,numverts,	type,stride,&indexbase,indexstride,numfaces,indicestype,nodeSubPart);
 
 		btVector3	triangleVerts[3];
 		btVector3	aabbMin,aabbMax;
@@ -256,16 +299,24 @@ void	btOptimizedBvh::updateBvhNodes(btStridingMeshInterface* meshInterface,int f
 			if (curNode.isLeafNode())
 			{
 				//recalc aabb from triangle data
+				int nodeSubPart = curNode.getPartId();
 				int nodeTriangleIndex = curNode.getTriangleIndex();
+				if (nodeSubPart != curNodeSubPart)
+				{
+					if (curNodeSubPart >= 0)
+						meshInterface->unLockReadOnlyVertexBase(curNodeSubPart);
+					meshInterface->getLockedReadOnlyVertexIndexBase(&vertexbase,numverts,	type,stride,&indexbase,indexstride,numfaces,indicestype,nodeSubPart);
+					btAssert(indicestype==PHY_INTEGER||indicestype==PHY_SHORT);
+				}
 				//triangles->getLockedReadOnlyVertexIndexBase(vertexBase,numVerts,
 
-				int* gfxbase = (int*)(indexbase+nodeTriangleIndex*indexstride);
+				unsigned int* gfxbase = (unsigned int*)(indexbase+nodeTriangleIndex*indexstride);
 				
 				
 				for (int j=2;j>=0;j--)
 				{
 					
-					int graphicsindex = gfxbase[j];
+					int graphicsindex = indicestype==PHY_SHORT?((unsigned short*)gfxbase)[j]:gfxbase[j];
 					btScalar* graphicsbase = (btScalar*)(vertexbase+graphicsindex*stride);
 #ifdef DEBUG_PATCH_COLORS
 					btVector3 mycolor = color[index&3];
@@ -292,8 +343,8 @@ void	btOptimizedBvh::updateBvhNodes(btStridingMeshInterface* meshInterface,int f
 				aabbMin.setMin(triangleVerts[2]);
 				aabbMax.setMax(triangleVerts[2]);
 
-				quantizeWithClamp(&curNode.m_quantizedAabbMin[0],aabbMin);
+				quantize(&curNode.m_quantizedAabbMin[0],aabbMin,0);
-				quantizeWithClamp(&curNode.m_quantizedAabbMax[0],aabbMax);
+				quantize(&curNode.m_quantizedAabbMax[0],aabbMax,1);
 				
 			} else
 			{
@@ -321,525 +372,17 @@ void	btOptimizedBvh::updateBvhNodes(btStridingMeshInterface* meshInterface,int f
 
 		}
 
-		meshInterface->unLockReadOnlyVertexBase(nodeSubPart);
+		if (curNodeSubPart >= 0)
+			meshInterface->unLockReadOnlyVertexBase(curNodeSubPart);
 
 		
 }
 
-void	btOptimizedBvh::setQuantizationValues(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax,btScalar quantizationMargin)
+///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place'
+btOptimizedBvh* btOptimizedBvh::deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian)
 {
-	//enlarge the AABB to avoid division by zero when initializing the quantization values
+	btQuantizedBvh* bvh = btQuantizedBvh::deSerializeInPlace(i_alignedDataBuffer,i_dataBufferSize,i_swapEndian);
-	btVector3 clampValue(quantizationMargin,quantizationMargin,quantizationMargin);
-	m_bvhAabbMin = bvhAabbMin - clampValue;
-	m_bvhAabbMax = bvhAabbMax + clampValue;
-	btVector3 aabbSize = m_bvhAabbMax - m_bvhAabbMin;
-	m_bvhQuantization = btVector3(btScalar(65535.0),btScalar(65535.0),btScalar(65535.0)) / aabbSize;
-}
-
-
-void	btOptimizedBvh::refit(btStridingMeshInterface* meshInterface)
-{
-	if (m_useQuantization)
-	{
-		//calculate new aabb
-		btVector3 aabbMin,aabbMax;
-		meshInterface->calculateAabbBruteForce(aabbMin,aabbMax);
-
-		setQuantizationValues(aabbMin,aabbMax);
-
-		updateBvhNodes(meshInterface,0,m_curNodeIndex,0);
-
-		///now update all subtree headers
-
-		int i;
-		for (i=0;i<m_SubtreeHeaders.size();i++)
-		{
-			btBvhSubtreeInfo& subtree = m_SubtreeHeaders[i];
-			subtree.setAabbFromQuantizeNode(m_quantizedContiguousNodes[subtree.m_rootNodeIndex]);
-		}
-
-	} else
-	{
-
-	}
-}
-
-
-
-btOptimizedBvh::~btOptimizedBvh()
-{
-}
-
-#ifdef DEBUG_TREE_BUILDING
-int gStackDepth = 0;
-int gMaxStackDepth = 0;
-#endif //DEBUG_TREE_BUILDING
-
-void	btOptimizedBvh::buildTree	(int startIndex,int endIndex)
-{
-#ifdef DEBUG_TREE_BUILDING
-	gStackDepth++;
-	if (gStackDepth > gMaxStackDepth)
-		gMaxStackDepth = gStackDepth;
-#endif //DEBUG_TREE_BUILDING
-
-
-	int splitAxis, splitIndex, i;
-	int numIndices =endIndex-startIndex;
-	int curIndex = m_curNodeIndex;
-
-	assert(numIndices>0);
-
-	if (numIndices==1)
-	{
-#ifdef DEBUG_TREE_BUILDING
-		gStackDepth--;
-#endif //DEBUG_TREE_BUILDING
-		
-		assignInternalNodeFromLeafNode(m_curNodeIndex,startIndex);
-
-		m_curNodeIndex++;
-		return;	
-	}
-	//calculate Best Splitting Axis and where to split it. Sort the incoming 'leafNodes' array within range 'startIndex/endIndex'.
 	
-	splitAxis = calcSplittingAxis(startIndex,endIndex);
+	//we don't add additional data so just do a static upcast
-
+	return static_cast<btOptimizedBvh*>(bvh);
-	splitIndex = sortAndCalcSplittingIndex(startIndex,endIndex,splitAxis);
-
-	int internalNodeIndex = m_curNodeIndex;
-	
-	setInternalNodeAabbMax(m_curNodeIndex,btVector3(btScalar(-1e30),btScalar(-1e30),btScalar(-1e30)));
-	setInternalNodeAabbMin(m_curNodeIndex,btVector3(btScalar(1e30),btScalar(1e30),btScalar(1e30)));
-	
-	for (i=startIndex;i<endIndex;i++)
-	{
-		mergeInternalNodeAabb(m_curNodeIndex,getAabbMin(i),getAabbMax(i));
-	}
-
-	m_curNodeIndex++;
-	
-
-	//internalNode->m_escapeIndex;
-	
-	int leftChildNodexIndex = m_curNodeIndex;
-
-	//build left child tree
-	buildTree(startIndex,splitIndex);
-
-	int rightChildNodexIndex = m_curNodeIndex;
-	//build right child tree
-	buildTree(splitIndex,endIndex);
-
-#ifdef DEBUG_TREE_BUILDING
-	gStackDepth--;
-#endif //DEBUG_TREE_BUILDING
-
-	int escapeIndex = m_curNodeIndex - curIndex;
-
-	if (m_useQuantization)
-	{
-		//escapeIndex is the number of nodes of this subtree
-		const int sizeQuantizedNode =sizeof(btQuantizedBvhNode);
-		const int treeSizeInBytes = escapeIndex * sizeQuantizedNode;
-		if (treeSizeInBytes > MAX_SUBTREE_SIZE_IN_BYTES)
-		{
-			updateSubtreeHeaders(leftChildNodexIndex,rightChildNodexIndex);
-		}
-	}
-
-	setInternalNodeEscapeIndex(internalNodeIndex,escapeIndex);
-
-}
-
-void	btOptimizedBvh::updateSubtreeHeaders(int leftChildNodexIndex,int rightChildNodexIndex)
-{
-	btAssert(m_useQuantization);
-
-	btQuantizedBvhNode& leftChildNode = m_quantizedContiguousNodes[leftChildNodexIndex];
-	int leftSubTreeSize = leftChildNode.isLeafNode() ? 1 : leftChildNode.getEscapeIndex();
-	int leftSubTreeSizeInBytes =  leftSubTreeSize * sizeof(btQuantizedBvhNode);
-	
-	btQuantizedBvhNode& rightChildNode = m_quantizedContiguousNodes[rightChildNodexIndex];
-	int rightSubTreeSize = rightChildNode.isLeafNode() ? 1 : rightChildNode.getEscapeIndex();
-	int rightSubTreeSizeInBytes =  rightSubTreeSize * sizeof(btQuantizedBvhNode);
-
-	if(leftSubTreeSizeInBytes <= MAX_SUBTREE_SIZE_IN_BYTES)
-	{
-		btBvhSubtreeInfo& subtree = m_SubtreeHeaders.expand();
-		subtree.setAabbFromQuantizeNode(leftChildNode);
-		subtree.m_rootNodeIndex = leftChildNodexIndex;
-		subtree.m_subtreeSize = leftSubTreeSize;
-	}
-
-	if(rightSubTreeSizeInBytes <= MAX_SUBTREE_SIZE_IN_BYTES)
-	{
-		btBvhSubtreeInfo& subtree = m_SubtreeHeaders.expand();
-		subtree.setAabbFromQuantizeNode(rightChildNode);
-		subtree.m_rootNodeIndex = rightChildNodexIndex;
-		subtree.m_subtreeSize = rightSubTreeSize;
-	}
-}
-
-
-int	btOptimizedBvh::sortAndCalcSplittingIndex(int startIndex,int endIndex,int splitAxis)
-{
-	int i;
-	int splitIndex =startIndex;
-	int numIndices = endIndex - startIndex;
-	btScalar splitValue;
-
-	btVector3 means(btScalar(0.),btScalar(0.),btScalar(0.));
-	for (i=startIndex;i<endIndex;i++)
-	{
-		btVector3 center = btScalar(0.5)*(getAabbMax(i)+getAabbMin(i));
-		means+=center;
-	}
-	means *= (btScalar(1.)/(btScalar)numIndices);
-	
-	splitValue = means[splitAxis];
-	
-	//sort leafNodes so all values larger then splitValue comes first, and smaller values start from 'splitIndex'.
-	for (i=startIndex;i<endIndex;i++)
-	{
-		btVector3 center = btScalar(0.5)*(getAabbMax(i)+getAabbMin(i));
-		if (center[splitAxis] > splitValue)
-		{
-			//swap
-			swapLeafNodes(i,splitIndex);
-			splitIndex++;
-		}
-	}
-
-	//if the splitIndex causes unbalanced trees, fix this by using the center in between startIndex and endIndex
-	//otherwise the tree-building might fail due to stack-overflows in certain cases.
-	//unbalanced1 is unsafe: it can cause stack overflows
-	//bool unbalanced1 = ((splitIndex==startIndex) || (splitIndex == (endIndex-1)));
-
-	//unbalanced2 should work too: always use center (perfect balanced trees)	
-	//bool unbalanced2 = true;
-
-	//this should be safe too:
-	int rangeBalancedIndices = numIndices/3;
-	bool unbalanced = ((splitIndex<=(startIndex+rangeBalancedIndices)) || (splitIndex >=(endIndex-1-rangeBalancedIndices)));
-	
-	if (unbalanced)
-	{
-		splitIndex = startIndex+ (numIndices>>1);
-	}
-
-	bool unbal = (splitIndex==startIndex) || (splitIndex == (endIndex));
-	btAssert(!unbal);
-
-	return splitIndex;
-}
-
-
-int	btOptimizedBvh::calcSplittingAxis(int startIndex,int endIndex)
-{
-	int i;
-
-	btVector3 means(btScalar(0.),btScalar(0.),btScalar(0.));
-	btVector3 variance(btScalar(0.),btScalar(0.),btScalar(0.));
-	int numIndices = endIndex-startIndex;
-
-	for (i=startIndex;i<endIndex;i++)
-	{
-		btVector3 center = btScalar(0.5)*(getAabbMax(i)+getAabbMin(i));
-		means+=center;
-	}
-	means *= (btScalar(1.)/(btScalar)numIndices);
-		
-	for (i=startIndex;i<endIndex;i++)
-	{
-		btVector3 center = btScalar(0.5)*(getAabbMax(i)+getAabbMin(i));
-		btVector3 diff2 = center-means;
-		diff2 = diff2 * diff2;
-		variance += diff2;
-	}
-	variance *= (btScalar(1.)/	((btScalar)numIndices-1)	);
-	
-	return variance.maxAxis();
-}
-
-
-
-void	btOptimizedBvh::reportAabbOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const
-{
-	//either choose recursive traversal (walkTree) or stackless (walkStacklessTree)
-
-
-	if (m_useQuantization)
-	{
-		///quantize query AABB
-		unsigned short int quantizedQueryAabbMin[3];
-		unsigned short int quantizedQueryAabbMax[3];
-		quantizeWithClamp(quantizedQueryAabbMin,aabbMin);
-		quantizeWithClamp(quantizedQueryAabbMax,aabbMax);
-
-		switch (m_traversalMode)
-		{
-		case TRAVERSAL_STACKLESS:
-				walkStacklessQuantizedTree(nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax,0,m_curNodeIndex);
-			break;
-		case TRAVERSAL_STACKLESS_CACHE_FRIENDLY:
-				walkStacklessQuantizedTreeCacheFriendly(nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax);
-			break;
-		case TRAVERSAL_RECURSIVE:
-			{
-				const btQuantizedBvhNode* rootNode = &m_quantizedContiguousNodes[0];
-				walkRecursiveQuantizedTreeAgainstQueryAabb(rootNode,nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax);
-			}
-			break;
-		default:
-			//unsupported
-			btAssert(0);
-		}
-	} else
-	{
-		walkStacklessTree(nodeCallback,aabbMin,aabbMax);
-	}
-}
-
-
-int maxIterations = 0;
-
-void	btOptimizedBvh::walkStacklessTree(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const
-{
-	btAssert(!m_useQuantization);
-
-	const btOptimizedBvhNode* rootNode = &m_contiguousNodes[0];
-	int escapeIndex, curIndex = 0;
-	int walkIterations = 0;
-	bool aabbOverlap, isLeafNode;
-
-	while (curIndex < m_curNodeIndex)
-	{
-		//catch bugs in tree data
-		assert (walkIterations < m_curNodeIndex);
-
-		walkIterations++;
-		aabbOverlap = TestAabbAgainstAabb2(aabbMin,aabbMax,rootNode->m_aabbMinOrg,rootNode->m_aabbMaxOrg);
-		isLeafNode = rootNode->m_escapeIndex == -1;
-		
-		if (isLeafNode && aabbOverlap)
-		{
-			nodeCallback->processNode(rootNode->m_subPart,rootNode->m_triangleIndex);
-		} 
-		
-		if (aabbOverlap || isLeafNode)
-		{
-			rootNode++;
-			curIndex++;
-		} else
-		{
-			escapeIndex = rootNode->m_escapeIndex;
-			rootNode += escapeIndex;
-			curIndex += escapeIndex;
-		}
-	}
-	if (maxIterations < walkIterations)
-		maxIterations = walkIterations;
-
-}
-
-/*
-///this was the original recursive traversal, before we optimized towards stackless traversal
-void	btOptimizedBvh::walkTree(btOptimizedBvhNode* rootNode,btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const
-{
-	bool isLeafNode, aabbOverlap = TestAabbAgainstAabb2(aabbMin,aabbMax,rootNode->m_aabbMin,rootNode->m_aabbMax);
-	if (aabbOverlap)
-	{
-		isLeafNode = (!rootNode->m_leftChild && !rootNode->m_rightChild);
-		if (isLeafNode)
-		{
-			nodeCallback->processNode(rootNode);
-		} else
-		{
-			walkTree(rootNode->m_leftChild,nodeCallback,aabbMin,aabbMax);
-			walkTree(rootNode->m_rightChild,nodeCallback,aabbMin,aabbMax);
-		}
-	}
-
-}
-*/
-
-void btOptimizedBvh::walkRecursiveQuantizedTreeAgainstQueryAabb(const btQuantizedBvhNode* currentNode,btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const
-{
-	btAssert(m_useQuantization);
-	
-	bool aabbOverlap, isLeafNode;
-
-	aabbOverlap = testQuantizedAabbAgainstQuantizedAabb(quantizedQueryAabbMin,quantizedQueryAabbMax,currentNode->m_quantizedAabbMin,currentNode->m_quantizedAabbMax);
-	isLeafNode = currentNode->isLeafNode();
-		
-	if (aabbOverlap)
-	{
-		if (isLeafNode)
-		{
-			nodeCallback->processNode(0,currentNode->getTriangleIndex());
-		} else
-		{
-			//process left and right children
-			const btQuantizedBvhNode* leftChildNode = currentNode+1;
-			walkRecursiveQuantizedTreeAgainstQueryAabb(leftChildNode,nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax);
-
-			const btQuantizedBvhNode* rightChildNode = leftChildNode->isLeafNode() ? leftChildNode+1:leftChildNode+leftChildNode->getEscapeIndex();
-			walkRecursiveQuantizedTreeAgainstQueryAabb(rightChildNode,nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax);
-		}
-	}		
-}
-
-
-
-
-
-
-
-void	btOptimizedBvh::walkStacklessQuantizedTree(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax,int startNodeIndex,int endNodeIndex) const
-{
-	btAssert(m_useQuantization);
-	
-	int curIndex = startNodeIndex;
-	int walkIterations = 0;
-	int subTreeSize = endNodeIndex - startNodeIndex;
-
-	const btQuantizedBvhNode* rootNode = &m_quantizedContiguousNodes[startNodeIndex];
-	int escapeIndex;
-	
-	bool aabbOverlap, isLeafNode;
-
-	while (curIndex < endNodeIndex)
-	{
-
-//#define VISUALLY_ANALYZE_BVH 1
-#ifdef VISUALLY_ANALYZE_BVH
-		//some code snippet to debugDraw aabb, to visually analyze bvh structure
-		static int drawPatch = 0;
-		//need some global access to a debugDrawer
-		extern btIDebugDraw* debugDrawerPtr;
-		if (curIndex==drawPatch)
-		{
-			btVector3 aabbMin,aabbMax;
-			aabbMin = unQuantize(rootNode->m_quantizedAabbMin);
-			aabbMax = unQuantize(rootNode->m_quantizedAabbMax);
-			btVector3	color(1,0,0);
-			debugDrawerPtr->drawAabb(aabbMin,aabbMax,color);
-		}
-#endif//VISUALLY_ANALYZE_BVH
-
-		//catch bugs in tree data
-		assert (walkIterations < subTreeSize);
-
-		walkIterations++;
-		aabbOverlap = testQuantizedAabbAgainstQuantizedAabb(quantizedQueryAabbMin,quantizedQueryAabbMax,rootNode->m_quantizedAabbMin,rootNode->m_quantizedAabbMax);
-		isLeafNode = rootNode->isLeafNode();
-		
-		if (isLeafNode && aabbOverlap)
-		{
-			nodeCallback->processNode(0,rootNode->getTriangleIndex());
-		} 
-		
-		if (aabbOverlap || isLeafNode)
-		{
-			rootNode++;
-			curIndex++;
-		} else
-		{
-			escapeIndex = rootNode->getEscapeIndex();
-			rootNode += escapeIndex;
-			curIndex += escapeIndex;
-		}
-	}
-	if (maxIterations < walkIterations)
-		maxIterations = walkIterations;
-
-}
-
-//This traversal can be called from Playstation 3 SPU
-void	btOptimizedBvh::walkStacklessQuantizedTreeCacheFriendly(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const
-{
-	btAssert(m_useQuantization);
-
-	int i;
-
-
-	for (i=0;i<this->m_SubtreeHeaders.size();i++)
-	{
-		const btBvhSubtreeInfo& subtree = m_SubtreeHeaders[i];
-
-		bool overlap = testQuantizedAabbAgainstQuantizedAabb(quantizedQueryAabbMin,quantizedQueryAabbMax,subtree.m_quantizedAabbMin,subtree.m_quantizedAabbMax);
-		if (overlap)
-		{
-			walkStacklessQuantizedTree(nodeCallback,quantizedQueryAabbMin,quantizedQueryAabbMax,
-				subtree.m_rootNodeIndex,
-				subtree.m_rootNodeIndex+subtree.m_subtreeSize);
-		}
-	}
-}
-
-
-
-
-void	btOptimizedBvh::reportSphereOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const
-{
-	(void)nodeCallback;
-	(void)aabbMin;
-	(void)aabbMax;
-	//not yet, please use aabb
-	btAssert(0);
-}
-
-
-void btOptimizedBvh::quantizeWithClamp(unsigned short* out, const btVector3& point) const
-{
-
-	btAssert(m_useQuantization);
-
-	btVector3 clampedPoint(point);
-	clampedPoint.setMax(m_bvhAabbMin);
-	clampedPoint.setMin(m_bvhAabbMax);
-
-	btVector3 v = (clampedPoint - m_bvhAabbMin) * m_bvhQuantization;
-	out[0] = (unsigned short)(v.getX()+0.5f);
-	out[1] = (unsigned short)(v.getY()+0.5f);
-	out[2] = (unsigned short)(v.getZ()+0.5f);		
-}
-
-btVector3	btOptimizedBvh::unQuantize(const unsigned short* vecIn) const
-{
-	btVector3	vecOut;
-	vecOut.setValue(
-		(btScalar)(vecIn[0]) / (m_bvhQuantization.getX()),
-		(btScalar)(vecIn[1]) / (m_bvhQuantization.getY()),
-		(btScalar)(vecIn[2]) / (m_bvhQuantization.getZ()));
-	vecOut += m_bvhAabbMin;
-	return vecOut;
-}
-
-
-void	btOptimizedBvh::swapLeafNodes(int i,int splitIndex)
-{
-	if (m_useQuantization)
-	{
-			btQuantizedBvhNode tmp = m_quantizedLeafNodes[i];
-			m_quantizedLeafNodes[i] = m_quantizedLeafNodes[splitIndex];
-			m_quantizedLeafNodes[splitIndex] = tmp;
-	} else
-	{
-			btOptimizedBvhNode tmp = m_leafNodes[i];
-			m_leafNodes[i] = m_leafNodes[splitIndex];
-			m_leafNodes[splitIndex] = tmp;
-	}
-}
-
-void	btOptimizedBvh::assignInternalNodeFromLeafNode(int internalNode,int leafNodeIndex)
-{
-	if (m_useQuantization)
-	{
-		m_quantizedContiguousNodes[internalNode] = m_quantizedLeafNodes[leafNodeIndex];
-	} else
-	{
-		m_contiguousNodes[internalNode] = m_leafNodes[leafNodeIndex];
-	}
 }

extern/bullet2/src/BulletCollision/CollisionShapes/btOptimizedBvh.h

@@ -16,315 +16,48 @@ subject to the following restrictions:
 #ifndef OPTIMIZED_BVH_H
 #define OPTIMIZED_BVH_H
 
-
+#include "BulletCollision/BroadphaseCollision/btQuantizedBvh.h"
-#include "../../LinearMath/btVector3.h"
-
-
-//http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclang/html/vclrf__m128.asp
-
-
 
 class btStridingMeshInterface;
 
-//Note: currently we have 16 bytes per quantized node
-#define MAX_SUBTREE_SIZE_IN_BYTES  2048
 
-
+///The btOptimizedBvh extends the btQuantizedBvh to create AABB tree for triangle meshes, through the btStridingMeshInterface.
-///btQuantizedBvhNode is a compressed aabb node, 16 bytes.
+ATTRIBUTE_ALIGNED16(class) btOptimizedBvh : public btQuantizedBvh
-///Node can be used for leafnode or internal node. Leafnodes can point to 32-bit triangle index (non-negative range).
-ATTRIBUTE_ALIGNED16	(struct) btQuantizedBvhNode
 {
 	
-	//12 bytes
-	unsigned short int	m_quantizedAabbMin[3];
-	unsigned short int	m_quantizedAabbMax[3];
-	//4 bytes
-	int	m_escapeIndexOrTriangleIndex;
-
-	bool isLeafNode() const
-	{
-		//skipindex is negative (internal node), triangleindex >=0 (leafnode)
-		return (m_escapeIndexOrTriangleIndex >= 0);
-	}
-	int getEscapeIndex() const
-	{
-		btAssert(!isLeafNode());
-		return -m_escapeIndexOrTriangleIndex;
-	}
-	int	getTriangleIndex() const
-	{
-		btAssert(isLeafNode());
-		return m_escapeIndexOrTriangleIndex;
-	}
-}
-;
-
-/// btOptimizedBvhNode contains both internal and leaf node information.
-/// Total node size is 44 bytes / node. You can use the compressed version of 16 bytes.
-ATTRIBUTE_ALIGNED16 (struct) btOptimizedBvhNode
-{
-	//32 bytes
-	btVector3	m_aabbMinOrg;
-	btVector3	m_aabbMaxOrg;
-
-	//4
-	int	m_escapeIndex;
-
-	//8
-	//for child nodes
-	int	m_subPart;
-	int	m_triangleIndex;
-	int	m_padding[5];//bad, due to alignment
-
-
-};
-
-
-///btBvhSubtreeInfo provides info to gather a subtree of limited size
-ATTRIBUTE_ALIGNED16(class) btBvhSubtreeInfo
-{
 public:
-	//12 bytes
+	BT_DECLARE_ALIGNED_ALLOCATOR();
-	unsigned short int	m_quantizedAabbMin[3];
-	unsigned short int	m_quantizedAabbMax[3];
-	//4 bytes, points to the root of the subtree
-	int			m_rootNodeIndex;
-	//4 bytes
-	int			m_subtreeSize;
-	int			m_padding[3];
-
-
-	void	setAabbFromQuantizeNode(const btQuantizedBvhNode& quantizedNode)
-	{
-		m_quantizedAabbMin[0] = quantizedNode.m_quantizedAabbMin[0];
-		m_quantizedAabbMin[1] = quantizedNode.m_quantizedAabbMin[1];
-		m_quantizedAabbMin[2] = quantizedNode.m_quantizedAabbMin[2];
-		m_quantizedAabbMax[0] = quantizedNode.m_quantizedAabbMax[0];
-		m_quantizedAabbMax[1] = quantizedNode.m_quantizedAabbMax[1];
-		m_quantizedAabbMax[2] = quantizedNode.m_quantizedAabbMax[2];
-	}
-}
-;
-
-
-class btNodeOverlapCallback
-{
-public:
-	virtual ~btNodeOverlapCallback() {};
-
-	virtual void processNode(int subPart, int triangleIndex) = 0;
-};
-
-#include "../../LinearMath/btAlignedAllocator.h"
-#include "../../LinearMath/btAlignedObjectArray.h"
-
-
-
-///for code readability:
-typedef btAlignedObjectArray<btOptimizedBvhNode>	NodeArray;
-typedef btAlignedObjectArray<btQuantizedBvhNode>	QuantizedNodeArray;
-typedef btAlignedObjectArray<btBvhSubtreeInfo>		BvhSubtreeInfoArray;
-
-
-///OptimizedBvh store an AABB tree that can be quickly traversed on CPU (and SPU,GPU in future)
-ATTRIBUTE_ALIGNED16(class) btOptimizedBvh
-{
-	NodeArray			m_leafNodes;
-	NodeArray			m_contiguousNodes;
-
-	QuantizedNodeArray	m_quantizedLeafNodes;
-	
-	QuantizedNodeArray	m_quantizedContiguousNodes;
-	
-	int					m_curNodeIndex;
-
-
-	//quantization data
-	bool				m_useQuantization;
-	btVector3			m_bvhAabbMin;
-	btVector3			m_bvhAabbMax;
-	btVector3			m_bvhQuantization;
-
-	enum btTraversalMode
-	{
-		TRAVERSAL_STACKLESS = 0,
-		TRAVERSAL_STACKLESS_CACHE_FRIENDLY,
-		TRAVERSAL_RECURSIVE
-	};
-
-	btTraversalMode	m_traversalMode;
-
-	
-
-
-	BvhSubtreeInfoArray		m_SubtreeHeaders;
-
-
-	///two versions, one for quantized and normal nodes. This allows code-reuse while maintaining readability (no template/macro!)
-	///this might be refactored into a virtual, it is usually not calculated at run-time
-	void	setInternalNodeAabbMin(int nodeIndex, const btVector3& aabbMin)
-	{
-		if (m_useQuantization)
-		{
-			quantizeWithClamp(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[0] ,aabbMin);
-		} else
-		{
-			m_contiguousNodes[nodeIndex].m_aabbMinOrg = aabbMin;
-
-		}
-	}
-	void	setInternalNodeAabbMax(int nodeIndex,const btVector3& aabbMax)
-	{
-		if (m_useQuantization)
-		{
-			quantizeWithClamp(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[0],aabbMax);
-		} else
-		{
-			m_contiguousNodes[nodeIndex].m_aabbMaxOrg = aabbMax;
-		}
-	}
-
-	btVector3 getAabbMin(int nodeIndex) const
-	{
-		if (m_useQuantization)
-		{
-			return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMin[0]);
-		}
-		//non-quantized
-		return m_leafNodes[nodeIndex].m_aabbMinOrg;
-
-	}
-	btVector3 getAabbMax(int nodeIndex) const
-	{
-		if (m_useQuantization)
-		{
-			return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMax[0]);
-		} 
-		//non-quantized
-		return m_leafNodes[nodeIndex].m_aabbMaxOrg;
-		
-	}
-
-	void	setQuantizationValues(const btVector3& bvhAabbMin,const btVector3& bvhAabbMax,btScalar quantizationMargin=btScalar(1.0));
-	
-	void	setInternalNodeEscapeIndex(int nodeIndex, int escapeIndex)
-	{
-		if (m_useQuantization)
-		{
-			m_quantizedContiguousNodes[nodeIndex].m_escapeIndexOrTriangleIndex = -escapeIndex;
-		} 
-		else
-		{
-			m_contiguousNodes[nodeIndex].m_escapeIndex = escapeIndex;
-		}
-
-	}
-
-	void mergeInternalNodeAabb(int nodeIndex,const btVector3& newAabbMin,const btVector3& newAabbMax) 
-	{
-		if (m_useQuantization)
-		{
-			unsigned short int quantizedAabbMin[3];
-			unsigned short int quantizedAabbMax[3];
-			quantizeWithClamp(quantizedAabbMin,newAabbMin);
-			quantizeWithClamp(quantizedAabbMax,newAabbMax);
-			for (int i=0;i<3;i++)
-			{
-				if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] > quantizedAabbMin[i])
-					m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] = quantizedAabbMin[i];
-
-				if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] < quantizedAabbMax[i])
-					m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] = quantizedAabbMax[i];
-
-			}
-		} else
-		{
-			//non-quantized
-			m_contiguousNodes[nodeIndex].m_aabbMinOrg.setMin(newAabbMin);
-			m_contiguousNodes[nodeIndex].m_aabbMaxOrg.setMax(newAabbMax);		
-		}
-	}
-
-	void	swapLeafNodes(int firstIndex,int secondIndex);
-
-	void	assignInternalNodeFromLeafNode(int internalNode,int leafNodeIndex);
 
 protected:
 
-	
-
-	void	buildTree	(int startIndex,int endIndex);
-
-	int	calcSplittingAxis(int startIndex,int endIndex);
-
-	int	sortAndCalcSplittingIndex(int startIndex,int endIndex,int splitAxis);
-	
-	void	walkStacklessTree(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const;
-
-	void	walkStacklessQuantizedTree(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax,int startNodeIndex,int endNodeIndex) const;
-
-	///tree traversal designed for small-memory processors like PS3 SPU
-	void	walkStacklessQuantizedTreeCacheFriendly(btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
-
-	///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
-	void	walkRecursiveQuantizedTreeAgainstQueryAabb(const btQuantizedBvhNode* currentNode,btNodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
-
-	///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
-	void	walkRecursiveQuantizedTreeAgainstQuantizedTree(const btQuantizedBvhNode* treeNodeA,const btQuantizedBvhNode* treeNodeB,btNodeOverlapCallback* nodeCallback) const;
-	
-
-	inline bool testQuantizedAabbAgainstQuantizedAabb(unsigned short int* aabbMin1,unsigned short int* aabbMax1,const unsigned short int* aabbMin2,const unsigned short int* aabbMax2) const
-	{
-		bool overlap = true;
-		overlap = (aabbMin1[0] > aabbMax2[0] || aabbMax1[0] < aabbMin2[0]) ? false : overlap;
-		overlap = (aabbMin1[2] > aabbMax2[2] || aabbMax1[2] < aabbMin2[2]) ? false : overlap;
-		overlap = (aabbMin1[1] > aabbMax2[1] || aabbMax1[1] < aabbMin2[1]) ? false : overlap;
-		return overlap;
-	}
-
-	void	updateSubtreeHeaders(int leftChildNodexIndex,int rightChildNodexIndex);
-
 public:
+
 	btOptimizedBvh();
 
 	virtual ~btOptimizedBvh();
 
 	void	build(btStridingMeshInterface* triangles,bool useQuantizedAabbCompression, const btVector3& bvhAabbMin, const btVector3& bvhAabbMax);
 
-	void	reportAabbOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const;
+	void	refit(btStridingMeshInterface* triangles,const btVector3& aabbMin,const btVector3& aabbMax);
-
-	void	reportSphereOverlappingNodex(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const;
-
-	void quantizeWithClamp(unsigned short* out, const btVector3& point) const;
-	
-	btVector3	unQuantize(const unsigned short* vecIn) const;
-
-	///setTraversalMode let's you choose between stackless, recursive or stackless cache friendly tree traversal. Note this is only implemented for quantized trees.
-	void	setTraversalMode(btTraversalMode	traversalMode)
-	{
-		m_traversalMode = traversalMode;
-	}
-
-	void	refit(btStridingMeshInterface* triangles);
 
 	void	refitPartial(btStridingMeshInterface* triangles,const btVector3& aabbMin, const btVector3& aabbMax);
 
 	void	updateBvhNodes(btStridingMeshInterface* meshInterface,int firstNode,int endNode,int index);
 
-
+	/// Data buffer MUST be 16 byte aligned
-	QuantizedNodeArray&	getQuantizedNodeArray()
+	virtual bool serialize(void *o_alignedDataBuffer, unsigned i_dataBufferSize, bool i_swapEndian)
-	{	
-		return	m_quantizedContiguousNodes;
-	}
-
-	BvhSubtreeInfoArray&	getSubtreeInfoArray()
 	{
-		return m_SubtreeHeaders;
+		return btQuantizedBvh::serialize(o_alignedDataBuffer,i_dataBufferSize,i_swapEndian);
+
 	}
 
-}
+	///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place'
-;
+	static btOptimizedBvh *deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian);
+
+
+};
 
 
 #endif //OPTIMIZED_BVH_H
 
+

extern/bullet2/src/BulletCollision/CollisionShapes/btPolyhedralConvexShape.cpp

@@ -13,7 +13,7 @@ subject to the following restrictions:
 3. This notice may not be removed or altered from any source distribution.
 */
 
-#include <BulletCollision/CollisionShapes/btPolyhedralConvexShape.h>
+#include "BulletCollision/CollisionShapes/btPolyhedralConvexShape.h"
 
 btPolyhedralConvexShape::btPolyhedralConvexShape()
 :m_localAabbMin(1,1,1),
@@ -95,7 +95,7 @@ void	btPolyhedralConvexShape::batchedUnitVectorGetSupportingVertexWithoutMargin(
 
 
 
-void	btPolyhedralConvexShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btPolyhedralConvexShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	//not yet, return box inertia
 
@@ -128,10 +128,46 @@ void btPolyhedralConvexShape::getAabb(const btTransform& trans,btVector3& aabbMi
 
 
 
+void	btPolyhedralConvexShape::setLocalScaling(const btVector3& scaling)
+{
+	btConvexInternalShape::setLocalScaling(scaling);
+	recalcLocalAabb();
+}
 
 void	btPolyhedralConvexShape::recalcLocalAabb()
 {
 	m_isLocalAabbValid = true;
+	
+	#if 1
+	static const btVector3 _directions[] =
+	{
+		btVector3( 1.,  0.,  0.),
+		btVector3( 0.,  1.,  0.),
+		btVector3( 0.,  0.,  1.),
+		btVector3( -1., 0.,  0.),
+		btVector3( 0., -1.,  0.),
+		btVector3( 0.,  0., -1.)
+	};
+	
+	btVector3 _supporting[] =
+	{
+		btVector3( 0., 0., 0.),
+		btVector3( 0., 0., 0.),
+		btVector3( 0., 0., 0.),
+		btVector3( 0., 0., 0.),
+		btVector3( 0., 0., 0.),
+		btVector3( 0., 0., 0.)
+	};
+	
+	batchedUnitVectorGetSupportingVertexWithoutMargin(_directions, _supporting, 6);
+	
+	for ( int i = 0; i < 3; ++i )
+	{
+		m_localAabbMax[i] = _supporting[i][i] + m_collisionMargin;
+		m_localAabbMin[i] = _supporting[i + 3][i] - m_collisionMargin;
+	}
+	
+	#else
 
 	for (int i=0;i<3;i++)
 	{
@@ -143,6 +179,8 @@ void	btPolyhedralConvexShape::recalcLocalAabb()
 		tmp = localGetSupportingVertex(vec);
 		m_localAabbMin[i] = tmp[i]-m_collisionMargin;
 	}
+	#endif
 }
 
 
+

extern/bullet2/src/BulletCollision/CollisionShapes/btPolyhedralConvexShape.h

@@ -16,13 +16,13 @@ subject to the following restrictions:
 #ifndef BU_SHAPE
 #define BU_SHAPE
 
-#include "../../LinearMath/btPoint3.h"
+#include "LinearMath/btPoint3.h"
-#include "../../LinearMath/btMatrix3x3.h"
+#include "LinearMath/btMatrix3x3.h"
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
 
 
-///PolyhedralConvexShape is an interface class for feature based (vertex/edge/face) convex shapes.
+///The btPolyhedralConvexShape is an internal interface class for polyhedral convex shapes.
-class btPolyhedralConvexShape : public btConvexShape
+class btPolyhedralConvexShape : public btConvexInternalShape
 {
 
 protected:
@@ -38,7 +38,7 @@ public:
 	virtual btVector3	localGetSupportingVertexWithoutMargin(const btVector3& vec)const;
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const;
 	
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 
 	inline void getNonvirtualAabb(const btTransform& trans,btVector3& aabbMin,btVector3& aabbMax, btScalar margin) const
@@ -73,6 +73,8 @@ public:
 	
 	virtual void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
 
+	virtual void	setLocalScaling(const btVector3& scaling);
+
 	void	recalcLocalAabb();
 
 	virtual int	getNumVertices() const = 0 ;

extern/bullet2/src/BulletCollision/CollisionShapes/btShapeHull.cpp

@@ -0,0 +1,164 @@
+/*
+btbtShapeHull implemented by John McCutchan.
+
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2008 Erwin Coumans  http://bulletphysics.com
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose,
+including commercial applications, and to alter it and redistribute it freely,
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "btShapeHull.h"
+#include "LinearMath/btConvexHull.h"
+
+#define NUM_UNITSPHERE_POINTS 42
+
+static btVector3 btUnitSpherePoints[NUM_UNITSPHERE_POINTS+MAX_PREFERRED_PENETRATION_DIRECTIONS*2] = 
+{
+	btVector3(btScalar(0.000000) , btScalar(-0.000000),btScalar(-1.000000)),
+	btVector3(btScalar(0.723608) , btScalar(-0.525725),btScalar(-0.447219)),
+	btVector3(btScalar(-0.276388) , btScalar(-0.850649),btScalar(-0.447219)),
+	btVector3(btScalar(-0.894426) , btScalar(-0.000000),btScalar(-0.447216)),
+	btVector3(btScalar(-0.276388) , btScalar(0.850649),btScalar(-0.447220)),
+	btVector3(btScalar(0.723608) , btScalar(0.525725),btScalar(-0.447219)),
+	btVector3(btScalar(0.276388) , btScalar(-0.850649),btScalar(0.447220)),
+	btVector3(btScalar(-0.723608) , btScalar(-0.525725),btScalar(0.447219)),
+	btVector3(btScalar(-0.723608) , btScalar(0.525725),btScalar(0.447219)),
+	btVector3(btScalar(0.276388) , btScalar(0.850649),btScalar(0.447219)),
+	btVector3(btScalar(0.894426) , btScalar(0.000000),btScalar(0.447216)),
+	btVector3(btScalar(-0.000000) , btScalar(0.000000),btScalar(1.000000)),
+	btVector3(btScalar(0.425323) , btScalar(-0.309011),btScalar(-0.850654)),
+	btVector3(btScalar(-0.162456) , btScalar(-0.499995),btScalar(-0.850654)),
+	btVector3(btScalar(0.262869) , btScalar(-0.809012),btScalar(-0.525738)),
+	btVector3(btScalar(0.425323) , btScalar(0.309011),btScalar(-0.850654)),
+	btVector3(btScalar(0.850648) , btScalar(-0.000000),btScalar(-0.525736)),
+	btVector3(btScalar(-0.525730) , btScalar(-0.000000),btScalar(-0.850652)),
+	btVector3(btScalar(-0.688190) , btScalar(-0.499997),btScalar(-0.525736)),
+	btVector3(btScalar(-0.162456) , btScalar(0.499995),btScalar(-0.850654)),
+	btVector3(btScalar(-0.688190) , btScalar(0.499997),btScalar(-0.525736)),
+	btVector3(btScalar(0.262869) , btScalar(0.809012),btScalar(-0.525738)),
+	btVector3(btScalar(0.951058) , btScalar(0.309013),btScalar(0.000000)),
+	btVector3(btScalar(0.951058) , btScalar(-0.309013),btScalar(0.000000)),
+	btVector3(btScalar(0.587786) , btScalar(-0.809017),btScalar(0.000000)),
+	btVector3(btScalar(0.000000) , btScalar(-1.000000),btScalar(0.000000)),
+	btVector3(btScalar(-0.587786) , btScalar(-0.809017),btScalar(0.000000)),
+	btVector3(btScalar(-0.951058) , btScalar(-0.309013),btScalar(-0.000000)),
+	btVector3(btScalar(-0.951058) , btScalar(0.309013),btScalar(-0.000000)),
+	btVector3(btScalar(-0.587786) , btScalar(0.809017),btScalar(-0.000000)),
+	btVector3(btScalar(-0.000000) , btScalar(1.000000),btScalar(-0.000000)),
+	btVector3(btScalar(0.587786) , btScalar(0.809017),btScalar(-0.000000)),
+	btVector3(btScalar(0.688190) , btScalar(-0.499997),btScalar(0.525736)),
+	btVector3(btScalar(-0.262869) , btScalar(-0.809012),btScalar(0.525738)),
+	btVector3(btScalar(-0.850648) , btScalar(0.000000),btScalar(0.525736)),
+	btVector3(btScalar(-0.262869) , btScalar(0.809012),btScalar(0.525738)),
+	btVector3(btScalar(0.688190) , btScalar(0.499997),btScalar(0.525736)),
+	btVector3(btScalar(0.525730) , btScalar(0.000000),btScalar(0.850652)),
+	btVector3(btScalar(0.162456) , btScalar(-0.499995),btScalar(0.850654)),
+	btVector3(btScalar(-0.425323) , btScalar(-0.309011),btScalar(0.850654)),
+	btVector3(btScalar(-0.425323) , btScalar(0.309011),btScalar(0.850654)),
+	btVector3(btScalar(0.162456) , btScalar(0.499995),btScalar(0.850654))
+};
+
+btShapeHull::btShapeHull (const btConvexShape* shape)
+{
+	m_shape = shape;
+	m_vertices.clear ();
+	m_indices.clear();
+	m_numIndices = 0;
+}
+
+btShapeHull::~btShapeHull ()
+{
+	m_indices.clear();	
+	m_vertices.clear ();
+}
+
+bool
+btShapeHull::buildHull (btScalar /*margin*/)
+{
+	int numSampleDirections = NUM_UNITSPHERE_POINTS;
+	{
+		int numPDA = m_shape->getNumPreferredPenetrationDirections();
+		if (numPDA)
+		{
+			for (int i=0;i<numPDA;i++)
+			{
+				btVector3 norm;
+				m_shape->getPreferredPenetrationDirection(i,norm);
+				btUnitSpherePoints[numSampleDirections] = norm;
+				numSampleDirections++;
+			}
+		}
+	}
+
+	btVector3 supportPoints[NUM_UNITSPHERE_POINTS+MAX_PREFERRED_PENETRATION_DIRECTIONS*2];
+	int i;
+	for (i = 0; i < numSampleDirections; i++)
+	{
+		supportPoints[i] = m_shape->localGetSupportingVertex(btUnitSpherePoints[i]);
+	}
+
+	HullDesc hd;
+	hd.mFlags = QF_TRIANGLES;
+	hd.mVcount = static_cast<unsigned int>(numSampleDirections);
+
+#ifdef BT_USE_DOUBLE_PRECISION
+	hd.mVertices = &supportPoints[0];
+	hd.mVertexStride = sizeof(btVector3);
+#else
+	hd.mVertices = &supportPoints[0];
+	hd.mVertexStride = sizeof (btVector3);
+#endif
+
+	HullLibrary hl;
+	HullResult hr;
+	if (hl.CreateConvexHull (hd, hr) == QE_FAIL)
+	{
+		return false;
+	}
+
+	m_vertices.resize (static_cast<int>(hr.mNumOutputVertices));
+
+
+	for (i = 0; i < static_cast<int>(hr.mNumOutputVertices); i++)
+	{
+		m_vertices[i] = hr.m_OutputVertices[i];
+	}
+	m_numIndices = hr.mNumIndices;
+	m_indices.resize(static_cast<int>(m_numIndices));
+	for (i = 0; i < static_cast<int>(m_numIndices); i++)
+	{
+		m_indices[i] = hr.m_Indices[i];
+	}
+
+	// free temporary hull result that we just copied
+	hl.ReleaseResult (hr);
+
+	return true;
+}
+
+int
+btShapeHull::numTriangles () const
+{
+	return static_cast<int>(m_numIndices / 3);
+}
+
+int
+btShapeHull::numVertices () const
+{
+	return m_vertices.size ();
+}
+
+int
+btShapeHull::numIndices () const
+{
+	return static_cast<int>(m_numIndices);
+}
+

extern/bullet2/src/BulletCollision/CollisionShapes/btShapeHull.h

@@ -0,0 +1,56 @@
+/*
+btShapeHull implemented by John McCutchan.
+
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2008 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose,
+including commercial applications, and to alter it and redistribute it freely,
+subject to the following restrictions:
+
+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.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef _SHAPE_HULL_H
+#define _SHAPE_HULL_H
+
+#include "LinearMath/btAlignedObjectArray.h"
+#include "BulletCollision/CollisionShapes/btConvexShape.h"
+
+
+///The btShapeHull class takes a btConvexShape, builds a simplified convex hull using btConvexHull and provides triangle indices and vertices.
+///It can be useful for to simplify a complex convex object and for visualization of a non-polyhedral convex object.
+///It approximates the convex hull using the supporting vertex of 42 directions.
+class btShapeHull
+{
+public:
+	btShapeHull (const btConvexShape* shape);
+	~btShapeHull ();
+
+	bool buildHull (btScalar margin);
+
+	int numTriangles () const;
+	int numVertices () const;
+	int numIndices () const;
+
+	const btVector3* getVertexPointer() const
+	{
+		return &m_vertices[0];
+	}
+	const unsigned int* getIndexPointer() const
+	{
+		return &m_indices[0];
+	}
+
+protected:
+	btAlignedObjectArray<btVector3> m_vertices;
+	btAlignedObjectArray<unsigned int> m_indices;
+	unsigned int m_numIndices;
+	const btConvexShape* m_shape;
+};
+
+#endif //_SHAPE_HULL_H

extern/bullet2/src/BulletCollision/CollisionShapes/btSphereShape.cpp

@@ -68,7 +68,7 @@ void btSphereShape::getAabb(const btTransform& t,btVector3& aabbMin,btVector3& a
 
 
 
-void	btSphereShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btSphereShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	btScalar elem = btScalar(0.4) * mass * getMargin()*getMargin();
 	inertia.setValue(elem,elem,elem);

extern/bullet2/src/BulletCollision/CollisionShapes/btSphereShape.h

@@ -16,15 +16,17 @@ subject to the following restrictions:
 #ifndef SPHERE_MINKOWSKI_H
 #define SPHERE_MINKOWSKI_H
 
-#include "btConvexShape.h"
+#include "btConvexInternalShape.h"
-#include "../BroadphaseCollision/btBroadphaseProxy.h" // for the types
+#include "BulletCollision/BroadphaseCollision/btBroadphaseProxy.h" // for the types
 
-///btSphereShape implements an implicit (getSupportingVertex) Sphere
+///The btSphereShape implements an implicit sphere, centered around a local origin with radius.
-ATTRIBUTE_ALIGNED16(class) btSphereShape : public btConvexShape
+ATTRIBUTE_ALIGNED16(class) btSphereShape : public btConvexInternalShape
 
 {
 	
 public:
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
 	btSphereShape (btScalar radius);
 	
 	
@@ -34,26 +36,26 @@ public:
 	virtual void	batchedUnitVectorGetSupportingVertexWithoutMargin(const btVector3* vectors,btVector3* supportVerticesOut,int numVectors) const;
 
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	virtual void getAabb(const btTransform& t,btVector3& aabbMin,btVector3& aabbMax) const;
 
 	virtual int	getShapeType() const { return SPHERE_SHAPE_PROXYTYPE; }
 
-	btScalar	getRadius() const { return m_implicitShapeDimensions.getX();}
+	btScalar	getRadius() const { return m_implicitShapeDimensions.getX() * m_localScaling.getX();}
 
 	//debugging
-	virtual char*	getName()const {return "SPHERE";}
+	virtual const char*	getName()const {return "SPHERE";}
 
 	virtual void	setMargin(btScalar margin)
 	{
-		btConvexShape::setMargin(margin);
+		btConvexInternalShape::setMargin(margin);
 	}
 	virtual btScalar	getMargin() const
 	{
 		//to improve gjk behaviour, use radius+margin as the full margin, so never get into the penetration case
 		//this means, non-uniform scaling is not supported anymore
-		return m_localScaling.getX() * getRadius() + btConvexShape::getMargin();
+		return getRadius();
 	}
 
 

extern/bullet2/src/BulletCollision/CollisionShapes/btStaticPlaneShape.cpp

@@ -19,10 +19,11 @@ subject to the following restrictions:
 
 
 btStaticPlaneShape::btStaticPlaneShape(const btVector3& planeNormal,btScalar planeConstant)
-:m_planeNormal(planeNormal),
+:m_planeNormal(planeNormal.normalized()),
 m_planeConstant(planeConstant),
 m_localScaling(btScalar(0.),btScalar(0.),btScalar(0.))
 {
+	//	btAssert( btFuzzyZero(m_planeNormal.length() - btScalar(1.)) );
 }
 
 
@@ -86,7 +87,7 @@ void	btStaticPlaneShape::processAllTriangles(btTriangleCallback* callback,const
 
 }
 
-void	btStaticPlaneShape::calculateLocalInertia(btScalar mass,btVector3& inertia)
+void	btStaticPlaneShape::calculateLocalInertia(btScalar mass,btVector3& inertia) const
 {
 	(void)mass;
 

extern/bullet2/src/BulletCollision/CollisionShapes/btStaticPlaneShape.h

@@ -19,8 +19,7 @@ subject to the following restrictions:
 #include "btConcaveShape.h"
 
 
-///StaticPlaneShape simulates an 'infinite' plane by dynamically reporting triangles approximated by intersection of the plane with the AABB.
+///The btStaticPlaneShape simulates an infinite non-moving (static) collision plane.
-///Assumed is that the other objects is not also infinite, so a reasonable sized AABB.
 class btStaticPlaneShape : public btConcaveShape
 {
 protected:
@@ -46,14 +45,23 @@ public:
 
 	virtual void	processAllTriangles(btTriangleCallback* callback,const btVector3& aabbMin,const btVector3& aabbMax) const;
 
-	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia);
+	virtual void	calculateLocalInertia(btScalar mass,btVector3& inertia) const;
 
 	virtual void	setLocalScaling(const btVector3& scaling);
 	virtual const btVector3& getLocalScaling() const;
 	
+	const btVector3&	getPlaneNormal() const
+	{
+		return	m_planeNormal;
+	}
+
+	const btScalar&	getPlaneConstant() const
+	{
+		return	m_planeConstant;
+	}
 
 	//debugging
-	virtual char*	getName()const {return "STATICPLANE";}
+	virtual const char*	getName()const {return "STATICPLANE";}
 
 
 };