Bundled VHACD library for convex decomposition.

Modified both MeshInstance tools as well as importer to use it instead of QuickHull.

32 files changed, 17866 insertions, 0 deletions

diff --git a/thirdparty/vhacd/inc/FloatMath.h b/thirdparty/vhacd/inc/FloatMath.h
new file mode 100644
index 0000000000..37b07cb69f
--- /dev/null
+++ b/thirdparty/vhacd/inc/FloatMath.h
@@ -0,0 +1,525 @@
+#ifndef FLOAT_MATH_LIB_H
+
+#define FLOAT_MATH_LIB_H
+
+
+#include <float.h>
+#include <stdint.h>
+
+namespace FLOAT_MATH
+{
+
+enum FM_ClipState
+{
+  FMCS_XMIN       = (1<<0),
+  FMCS_XMAX       = (1<<1),
+  FMCS_YMIN       = (1<<2),
+  FMCS_YMAX       = (1<<3),
+  FMCS_ZMIN       = (1<<4),
+  FMCS_ZMAX       = (1<<5),
+};
+
+enum FM_Axis
+{
+  FM_XAXIS   = (1<<0),
+  FM_YAXIS   = (1<<1),
+  FM_ZAXIS   = (1<<2)
+};
+
+enum LineSegmentType
+{
+  LS_START,
+  LS_MIDDLE,
+  LS_END
+};
+
+
+const float FM_PI = 3.1415926535897932384626433832795028841971693993751f;
+const float FM_DEG_TO_RAD = ((2.0f * FM_PI) / 360.0f);
+const float FM_RAD_TO_DEG = (360.0f / (2.0f * FM_PI));
+
+//***************** Float versions
+//***
+//*** vectors are assumed to be 3 floats or 3 doubles representing X, Y, Z
+//*** quaternions are assumed to be 4 floats or 4 doubles representing X,Y,Z,W
+//*** matrices are assumed to be 16 floats or 16 doubles representing a standard D3D or OpenGL style 4x4 matrix
+//*** bounding volumes are expressed as two sets of 3 floats/double representing bmin(x,y,z) and bmax(x,y,z)
+//*** Plane equations are assumed to be 4 floats or 4 doubles representing Ax,By,Cz,D
+
+FM_Axis fm_getDominantAxis(const float normal[3]);
+FM_Axis fm_getDominantAxis(const double normal[3]);
+
+void fm_decomposeTransform(const float local_transform[16],float trans[3],float rot[4],float scale[3]);
+void fm_decomposeTransform(const double local_transform[16],double trans[3],double rot[4],double scale[3]);
+
+void  fm_multiplyTransform(const float *pA,const float *pB,float *pM);
+void  fm_multiplyTransform(const double *pA,const double *pB,double *pM);
+
+void  fm_inverseTransform(const float matrix[16],float inverse_matrix[16]);
+void  fm_inverseTransform(const double matrix[16],double inverse_matrix[16]);
+
+void  fm_identity(float matrix[16]); // set 4x4 matrix to identity.
+void  fm_identity(double matrix[16]); // set 4x4 matrix to identity.
+
+void  fm_inverseRT(const float matrix[16], const float pos[3], float t[3]); // inverse rotate translate the point.
+void  fm_inverseRT(const double matrix[16],const double pos[3],double t[3]); // inverse rotate translate the point.
+
+void  fm_transform(const float matrix[16], const float pos[3], float t[3]); // rotate and translate this point.
+void  fm_transform(const double matrix[16],const double pos[3],double t[3]); // rotate and translate this point.
+
+float  fm_getDeterminant(const float matrix[16]);
+double fm_getDeterminant(const double matrix[16]);
+
+void fm_getSubMatrix(int32_t ki,int32_t kj,float pDst[16],const float matrix[16]);
+void fm_getSubMatrix(int32_t ki,int32_t kj,double pDst[16],const float matrix[16]);
+
+void  fm_rotate(const float matrix[16],const float pos[3],float t[3]); // only rotate the point by a 4x4 matrix, don't translate.
+void  fm_rotate(const double matri[16],const double pos[3],double t[3]); // only rotate the point by a 4x4 matrix, don't translate.
+
+void  fm_eulerToMatrix(float ax,float ay,float az,float matrix[16]); // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+void  fm_eulerToMatrix(double ax,double ay,double az,double matrix[16]); // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+
+void  fm_getAABB(uint32_t vcount,const float *points,uint32_t pstride,float bmin[3],float bmax[3]);
+void  fm_getAABB(uint32_t vcount,const double *points,uint32_t pstride,double bmin[3],double bmax[3]);
+
+void  fm_getAABBCenter(const float bmin[3],const float bmax[3],float center[3]);
+void  fm_getAABBCenter(const double bmin[3],const double bmax[3],double center[3]);
+
+void fm_transformAABB(const float bmin[3],const float bmax[3],const float matrix[16],float tbmin[3],float tbmax[3]);
+void fm_transformAABB(const double bmin[3],const double bmax[3],const double matrix[16],double tbmin[3],double tbmax[3]);
+
+void  fm_eulerToQuat(float x,float y,float z,float quat[4]); // convert euler angles to quaternion.
+void  fm_eulerToQuat(double x,double y,double z,double quat[4]); // convert euler angles to quaternion.
+
+void  fm_quatToEuler(const float quat[4],float &ax,float &ay,float &az);
+void  fm_quatToEuler(const double quat[4],double &ax,double &ay,double &az);
+
+void  fm_eulerToQuat(const float euler[3],float quat[4]); // convert euler angles to quaternion. Angles must be radians not degrees!
+void  fm_eulerToQuat(const double euler[3],double quat[4]); // convert euler angles to quaternion.
+
+void  fm_scale(float x,float y,float z,float matrix[16]); // apply scale to the matrix.
+void  fm_scale(double x,double y,double z,double matrix[16]); // apply scale to the matrix.
+
+void  fm_eulerToQuatDX(float x,float y,float z,float quat[4]); // convert euler angles to quaternion using the fucked up DirectX method
+void  fm_eulerToQuatDX(double x,double y,double z,double quat[4]); // convert euler angles to quaternion using the fucked up DirectX method
+
+void  fm_eulerToMatrixDX(float x,float y,float z,float matrix[16]); // convert euler angles to quaternion using the fucked up DirectX method.
+void  fm_eulerToMatrixDX(double x,double y,double z,double matrix[16]); // convert euler angles to quaternion using the fucked up DirectX method.
+
+void  fm_quatToMatrix(const float quat[4],float matrix[16]); // convert quaterinion rotation to matrix, translation set to zero.
+void  fm_quatToMatrix(const double quat[4],double matrix[16]); // convert quaterinion rotation to matrix, translation set to zero.
+
+void  fm_quatRotate(const float quat[4],const float v[3],float r[3]); // rotate a vector directly by a quaternion.
+void  fm_quatRotate(const double quat[4],const double v[3],double r[3]); // rotate a vector directly by a quaternion.
+
+void  fm_getTranslation(const float matrix[16],float t[3]);
+void  fm_getTranslation(const double matrix[16],double t[3]);
+
+void  fm_setTranslation(const float *translation,float matrix[16]);
+void  fm_setTranslation(const double *translation,double matrix[16]);
+
+void  fm_multiplyQuat(const float *qa,const float *qb,float *quat);
+void  fm_multiplyQuat(const double *qa,const double *qb,double *quat);
+
+void  fm_matrixToQuat(const float matrix[16],float quat[4]); // convert the 3x3 portion of a 4x4 matrix into a quaterion as x,y,z,w
+void  fm_matrixToQuat(const double matrix[16],double quat[4]); // convert the 3x3 portion of a 4x4 matrix into a quaterion as x,y,z,w
+
+float fm_sphereVolume(float radius); // return's the volume of a sphere of this radius (4/3 PI * R cubed )
+double fm_sphereVolume(double radius); // return's the volume of a sphere of this radius (4/3 PI * R cubed )
+
+float fm_cylinderVolume(float radius,float h);
+double fm_cylinderVolume(double radius,double h);
+
+float fm_capsuleVolume(float radius,float h);
+double fm_capsuleVolume(double radius,double h);
+
+float fm_distance(const float p1[3],const float p2[3]);
+double fm_distance(const double p1[3],const double p2[3]);
+
+float fm_distanceSquared(const float p1[3],const float p2[3]);
+double fm_distanceSquared(const double p1[3],const double p2[3]);
+
+float fm_distanceSquaredXZ(const float p1[3],const float p2[3]);
+double fm_distanceSquaredXZ(const double p1[3],const double p2[3]);
+
+float fm_computePlane(const float p1[3],const float p2[3],const float p3[3],float *n); // return D
+double fm_computePlane(const double p1[3],const double p2[3],const double p3[3],double *n); // return D
+
+float fm_distToPlane(const float plane[4],const float pos[3]); // computes the distance of this point from the plane.
+double fm_distToPlane(const double plane[4],const double pos[3]); // computes the distance of this point from the plane.
+
+float fm_dot(const float p1[3],const float p2[3]);
+double fm_dot(const double p1[3],const double p2[3]);
+
+void  fm_cross(float cross[3],const float a[3],const float b[3]);
+void  fm_cross(double cross[3],const double a[3],const double b[3]);
+
+float  fm_computeNormalVector(float n[3],const float p1[3],const float p2[3]); // as P2-P1 normalized.
+double  fm_computeNormalVector(double n[3],const double p1[3],const double p2[3]); // as P2-P1 normalized.
+
+bool  fm_computeWindingOrder(const float p1[3],const float p2[3],const float p3[3]); // returns true if the triangle is clockwise.
+bool  fm_computeWindingOrder(const double p1[3],const double p2[3],const double p3[3]); // returns true if the triangle is clockwise.
+
+float  fm_normalize(float n[3]); // normalize this vector and return the distance
+double  fm_normalize(double n[3]); // normalize this vector and return the distance
+
+float  fm_normalizeQuat(float n[4]); // normalize this quat
+double  fm_normalizeQuat(double n[4]); // normalize this quat
+
+void  fm_matrixMultiply(const float A[16],const float B[16],float dest[16]);
+void  fm_matrixMultiply(const double A[16],const double B[16],double dest[16]);
+
+void  fm_composeTransform(const float position[3],const float quat[4],const float scale[3],float matrix[16]);
+void  fm_composeTransform(const double position[3],const double quat[4],const double scale[3],double matrix[16]);
+
+float fm_computeArea(const float p1[3],const float p2[3],const float p3[3]);
+double fm_computeArea(const double p1[3],const double p2[3],const double p3[3]);
+
+void  fm_lerp(const float p1[3],const float p2[3],float dest[3],float lerpValue);
+void  fm_lerp(const double p1[3],const double p2[3],double dest[3],double lerpValue);
+
+bool  fm_insideTriangleXZ(const float test[3],const float p1[3],const float p2[3],const float p3[3]);
+bool  fm_insideTriangleXZ(const double test[3],const double p1[3],const double p2[3],const double p3[3]);
+
+bool  fm_insideAABB(const float pos[3],const float bmin[3],const float bmax[3]);
+bool  fm_insideAABB(const double pos[3],const double bmin[3],const double bmax[3]);
+
+bool  fm_insideAABB(const float obmin[3],const float obmax[3],const float tbmin[3],const float tbmax[3]); // test if bounding box tbmin/tmbax is fully inside obmin/obmax
+bool  fm_insideAABB(const double obmin[3],const double obmax[3],const double tbmin[3],const double tbmax[3]); // test if bounding box tbmin/tmbax is fully inside obmin/obmax
+
+uint32_t fm_clipTestPoint(const float bmin[3],const float bmax[3],const float pos[3]);
+uint32_t fm_clipTestPoint(const double bmin[3],const double bmax[3],const double pos[3]);
+
+uint32_t fm_clipTestPointXZ(const float bmin[3],const float bmax[3],const float pos[3]); // only tests X and Z, not Y
+uint32_t fm_clipTestPointXZ(const double bmin[3],const double bmax[3],const double pos[3]); // only tests X and Z, not Y
+
+
+uint32_t fm_clipTestAABB(const float bmin[3],const float bmax[3],const float p1[3],const float p2[3],const float p3[3],uint32_t &andCode);
+uint32_t fm_clipTestAABB(const double bmin[3],const double bmax[3],const double p1[3],const double p2[3],const double p3[3],uint32_t &andCode);
+
+
+bool     fm_lineTestAABBXZ(const float p1[3],const float p2[3],const float bmin[3],const float bmax[3],float &time);
+bool     fm_lineTestAABBXZ(const double p1[3],const double p2[3],const double bmin[3],const double bmax[3],double &time);
+
+bool     fm_lineTestAABB(const float p1[3],const float p2[3],const float bmin[3],const float bmax[3],float &time);
+bool     fm_lineTestAABB(const double p1[3],const double p2[3],const double bmin[3],const double bmax[3],double &time);
+
+
+void  fm_initMinMax(const float p[3],float bmin[3],float bmax[3]);
+void  fm_initMinMax(const double p[3],double bmin[3],double bmax[3]);
+
+void  fm_initMinMax(float bmin[3],float bmax[3]);
+void  fm_initMinMax(double bmin[3],double bmax[3]);
+
+void  fm_minmax(const float p[3],float bmin[3],float bmax[3]); // accumulate to a min-max value
+void  fm_minmax(const double p[3],double bmin[3],double bmax[3]); // accumulate to a min-max value
+
+// Computes the diagonal length of the bounding box and then inflates the bounding box on all sides
+// by the ratio provided.
+void fm_inflateMinMax(float bmin[3], float bmax[3], float ratio);
+void fm_inflateMinMax(double bmin[3], double bmax[3], double ratio);
+
+float fm_solveX(const float plane[4],float y,float z); // solve for X given this plane equation and the other two components.
+double fm_solveX(const double plane[4],double y,double z); // solve for X given this plane equation and the other two components.
+
+float fm_solveY(const float plane[4],float x,float z); // solve for Y given this plane equation and the other two components.
+double fm_solveY(const double plane[4],double x,double z); // solve for Y given this plane equation and the other two components.
+
+float fm_solveZ(const float plane[4],float x,float y); // solve for Z given this plane equation and the other two components.
+double fm_solveZ(const double plane[4],double x,double y); // solve for Z given this plane equation and the other two components.
+
+bool  fm_computeBestFitPlane(uint32_t vcount,     // number of input data points
+	const float *points,     // starting address of points array.
+	uint32_t vstride,    // stride between input points.
+	const float *weights,    // *optional point weighting values.
+	uint32_t wstride,    // weight stride for each vertex.
+	float plane[4],		// Best fit plane equation
+	float center[3]);  // Best fit weighted center of input points
+
+bool  fm_computeBestFitPlane(uint32_t vcount,     // number of input data points
+	const double *points,     // starting address of points array.
+	uint32_t vstride,    // stride between input points.
+	const double *weights,    // *optional point weighting values.
+	uint32_t wstride,    // weight stride for each vertex.
+	double plane[4],
+	double center[3]); 
+
+// Computes the average center of a set of data points
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+						 const float *points,     // starting address of points array.
+						 float *center);
+
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+						 const double *points,     // starting address of points array.
+						 double *center);
+
+// Compute centroid of a triangle mesh; takes area of each triangle into account
+// weighted average
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+						const float *points,     // starting address of points array.
+						uint32_t triangleCount,
+						const uint32_t *indices,
+						float *center);
+
+// Compute centroid of a triangle mesh; takes area of each triangle into account
+// weighted average
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+	const double *points,     // starting address of points array.
+	uint32_t triangleCount,
+	const uint32_t *indices,
+	double *center);
+
+
+float  fm_computeBestFitAABB(uint32_t vcount,const float *points,uint32_t pstride,float bmin[3],float bmax[3]); // returns the diagonal distance
+double fm_computeBestFitAABB(uint32_t vcount,const double *points,uint32_t pstride,double bmin[3],double bmax[3]); // returns the diagonal distance
+
+float  fm_computeBestFitSphere(uint32_t vcount,const float *points,uint32_t pstride,float center[3]);
+double  fm_computeBestFitSphere(uint32_t vcount,const double *points,uint32_t pstride,double center[3]);
+
+bool fm_lineSphereIntersect(const float center[3],float radius,const float p1[3],const float p2[3],float intersect[3]);
+bool fm_lineSphereIntersect(const double center[3],double radius,const double p1[3],const double p2[3],double intersect[3]);
+
+bool fm_intersectRayAABB(const float bmin[3],const float bmax[3],const float pos[3],const float dir[3],float intersect[3]);
+bool fm_intersectLineSegmentAABB(const float bmin[3],const float bmax[3],const float p1[3],const float p2[3],float intersect[3]);
+
+bool fm_lineIntersectsTriangle(const float rayStart[3],const float rayEnd[3],const float p1[3],const float p2[3],const float p3[3],float sect[3]);
+bool fm_lineIntersectsTriangle(const double rayStart[3],const double rayEnd[3],const double p1[3],const double p2[3],const double p3[3],double sect[3]);
+
+bool fm_rayIntersectsTriangle(const float origin[3],const float dir[3],const float v0[3],const float v1[3],const float v2[3],float &t);
+bool fm_rayIntersectsTriangle(const double origin[3],const double dir[3],const double v0[3],const double v1[3],const double v2[3],double &t);
+
+bool fm_raySphereIntersect(const float center[3],float radius,const float pos[3],const float dir[3],float distance,float intersect[3]);
+bool fm_raySphereIntersect(const double center[3],double radius,const double pos[3],const double dir[3],double distance,double intersect[3]);
+
+void fm_catmullRom(float out_vector[3],const float p1[3],const float p2[3],const float p3[3],const float *p4, const float s);
+void fm_catmullRom(double out_vector[3],const double p1[3],const double p2[3],const double p3[3],const double *p4, const double s);
+
+bool fm_intersectAABB(const float bmin1[3],const float bmax1[3],const float bmin2[3],const float bmax2[3]);
+bool fm_intersectAABB(const double bmin1[3],const double bmax1[3],const double bmin2[3],const double bmax2[3]);
+
+
+// computes the rotation quaternion to go from unit-vector v0 to unit-vector v1
+void fm_rotationArc(const float v0[3],const float v1[3],float quat[4]);
+void fm_rotationArc(const double v0[3],const double v1[3],double quat[4]);
+
+float  fm_distancePointLineSegment(const float Point[3],const float LineStart[3],const float LineEnd[3],float intersection[3],LineSegmentType &type,float epsilon);
+double fm_distancePointLineSegment(const double Point[3],const double LineStart[3],const double LineEnd[3],double intersection[3],LineSegmentType &type,double epsilon);
+
+
+bool fm_colinear(const double p1[3],const double p2[3],const double p3[3],double epsilon=0.999);               // true if these three points in a row are co-linear
+bool fm_colinear(const float  p1[3],const float  p2[3],const float p3[3],float epsilon=0.999f);
+
+bool fm_colinear(const float a1[3],const float a2[3],const float b1[3],const float b2[3],float epsilon=0.999f);  // true if these two line segments are co-linear.
+bool fm_colinear(const double a1[3],const double a2[3],const double b1[3],const double b2[3],double epsilon=0.999);  // true if these two line segments are co-linear.
+
+enum IntersectResult
+{
+  IR_DONT_INTERSECT,
+  IR_DO_INTERSECT,
+  IR_COINCIDENT,
+  IR_PARALLEL,
+};
+
+IntersectResult fm_intersectLineSegments2d(const float a1[3], const float a2[3], const float b1[3], const float b2[3], float intersectionPoint[3]);
+IntersectResult fm_intersectLineSegments2d(const double a1[3],const double a2[3],const double b1[3],const double b2[3],double intersectionPoint[3]);
+
+IntersectResult fm_intersectLineSegments2dTime(const float a1[3], const float a2[3], const float b1[3], const float b2[3],float &t1,float &t2);
+IntersectResult fm_intersectLineSegments2dTime(const double a1[3],const double a2[3],const double b1[3],const double b2[3],double &t1,double &t2);
+
+// Plane-Triangle splitting
+
+enum PlaneTriResult
+{
+  PTR_ON_PLANE,
+  PTR_FRONT,
+  PTR_BACK,
+  PTR_SPLIT,
+};
+
+PlaneTriResult fm_planeTriIntersection(const float plane[4],    // the plane equation in Ax+By+Cz+D format
+									const float *triangle, // the source triangle.
+									uint32_t tstride,  // stride in bytes of the input and output *vertices*
+									float        epsilon,  // the co-planer epsilon value.
+									float       *front,    // the triangle in front of the
+									uint32_t &fcount,  // number of vertices in the 'front' triangle
+									float       *back,     // the triangle in back of the plane
+									uint32_t &bcount); // the number of vertices in the 'back' triangle.
+
+
+PlaneTriResult fm_planeTriIntersection(const double plane[4],    // the plane equation in Ax+By+Cz+D format
+									const double *triangle, // the source triangle.
+									uint32_t tstride,  // stride in bytes of the input and output *vertices*
+									double        epsilon,  // the co-planer epsilon value.
+									double       *front,    // the triangle in front of the
+									uint32_t &fcount,  // number of vertices in the 'front' triangle
+									double       *back,     // the triangle in back of the plane
+									uint32_t &bcount); // the number of vertices in the 'back' triangle.
+
+
+bool fm_intersectPointPlane(const float p1[3],const float p2[3],float *split,const float plane[4]);
+bool fm_intersectPointPlane(const double p1[3],const double p2[3],double *split,const double plane[4]);
+
+PlaneTriResult fm_getSidePlane(const float p[3],const float plane[4],float epsilon);
+PlaneTriResult fm_getSidePlane(const double p[3],const double plane[4],double epsilon);
+
+
+void fm_computeBestFitOBB(uint32_t vcount,const float *points,uint32_t pstride,float *sides,float matrix[16],bool bruteForce=true);
+void fm_computeBestFitOBB(uint32_t vcount,const double *points,uint32_t pstride,double *sides,double matrix[16],bool bruteForce=true);
+
+void fm_computeBestFitOBB(uint32_t vcount,const float *points,uint32_t pstride,float *sides,float pos[3],float quat[4],bool bruteForce=true);
+void fm_computeBestFitOBB(uint32_t vcount,const double *points,uint32_t pstride,double *sides,double pos[3],double quat[4],bool bruteForce=true);
+
+void fm_computeBestFitABB(uint32_t vcount,const float *points,uint32_t pstride,float *sides,float pos[3]);
+void fm_computeBestFitABB(uint32_t vcount,const double *points,uint32_t pstride,double *sides,double pos[3]);
+
+
+//** Note, if the returned capsule height is less than zero, then you must represent it is a sphere of size radius.
+void fm_computeBestFitCapsule(uint32_t vcount,const float *points,uint32_t pstride,float &radius,float &height,float matrix[16],bool bruteForce=true);
+void fm_computeBestFitCapsule(uint32_t vcount,const double *points,uint32_t pstride,float &radius,float &height,double matrix[16],bool bruteForce=true);
+
+
+void fm_planeToMatrix(const float plane[4],float matrix[16]); // convert a plane equation to a 4x4 rotation matrix.  Reference vector is 0,1,0
+void fm_planeToQuat(const float plane[4],float quat[4],float pos[3]); // convert a plane equation to a quaternion and translation
+
+void fm_planeToMatrix(const double plane[4],double matrix[16]); // convert a plane equation to a 4x4 rotation matrix
+void fm_planeToQuat(const double plane[4],double quat[4],double pos[3]); // convert a plane equation to a quaternion and translation
+
+inline void fm_doubleToFloat3(const double p[3],float t[3]) { t[0] = (float) p[0]; t[1] = (float)p[1]; t[2] = (float)p[2]; };
+inline void fm_floatToDouble3(const float p[3],double t[3]) { t[0] = (double)p[0]; t[1] = (double)p[1]; t[2] = (double)p[2]; };
+
+
+void  fm_eulerMatrix(float ax,float ay,float az,float matrix[16]); // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+void  fm_eulerMatrix(double ax,double ay,double az,double matrix[16]); // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+
+
+float  fm_computeMeshVolume(const float *vertices,uint32_t tcount,const uint32_t *indices);
+double fm_computeMeshVolume(const double *vertices,uint32_t tcount,const uint32_t *indices);
+
+
+#define FM_DEFAULT_GRANULARITY 0.001f  // 1 millimeter is the default granularity
+
+class fm_VertexIndex
+{
+public:
+  virtual uint32_t          getIndex(const float pos[3],bool &newPos) = 0;  // get welded index for this float vector[3]
+  virtual uint32_t          getIndex(const double pos[3],bool &newPos) = 0;  // get welded index for this double vector[3]
+  virtual const float *   getVerticesFloat(void) const = 0;
+  virtual const double *  getVerticesDouble(void) const = 0;
+  virtual const float *   getVertexFloat(uint32_t index) const = 0;
+  virtual const double *  getVertexDouble(uint32_t index) const = 0;
+  virtual uint32_t          getVcount(void) const = 0;
+  virtual bool            isDouble(void) const = 0;
+  virtual bool            saveAsObj(const char *fname,uint32_t tcount,uint32_t *indices) = 0;
+};
+
+fm_VertexIndex * fm_createVertexIndex(double granularity,bool snapToGrid); // create an indexed vertex system for doubles
+fm_VertexIndex * fm_createVertexIndex(float granularity,bool snapToGrid);  // create an indexed vertext system for floats
+void             fm_releaseVertexIndex(fm_VertexIndex *vindex);
+
+
+class fm_Triangulate
+{
+public:
+  virtual const double *       triangulate3d(uint32_t pcount,
+											 const double *points,
+											 uint32_t vstride,
+											 uint32_t &tcount,
+											 bool consolidate,
+											 double epsilon) = 0;
+
+  virtual const float  *       triangulate3d(uint32_t pcount,
+											 const float  *points,
+											 uint32_t vstride,
+											 uint32_t &tcount,
+											 bool consolidate,
+											 float epsilon) = 0;
+};
+
+fm_Triangulate * fm_createTriangulate(void);
+void             fm_releaseTriangulate(fm_Triangulate *t);
+
+
+const float * fm_getPoint(const float *points,uint32_t pstride,uint32_t index);
+const double * fm_getPoint(const double *points,uint32_t pstride,uint32_t index);
+
+bool   fm_insideTriangle(float Ax, float Ay,float Bx, float By,float Cx, float Cy,float Px, float Py);
+bool   fm_insideTriangle(double Ax, double Ay,double Bx, double By,double Cx, double Cy,double Px, double Py);
+float  fm_areaPolygon2d(uint32_t pcount,const float *points,uint32_t pstride);
+double fm_areaPolygon2d(uint32_t pcount,const double *points,uint32_t pstride);
+
+bool  fm_pointInsidePolygon2d(uint32_t pcount,const float *points,uint32_t pstride,const float *point,uint32_t xindex=0,uint32_t yindex=1);
+bool  fm_pointInsidePolygon2d(uint32_t pcount,const double *points,uint32_t pstride,const double *point,uint32_t xindex=0,uint32_t yindex=1);
+
+uint32_t fm_consolidatePolygon(uint32_t pcount,const float *points,uint32_t pstride,float *dest,float epsilon=0.999999f); // collapses co-linear edges.
+uint32_t fm_consolidatePolygon(uint32_t pcount,const double *points,uint32_t pstride,double *dest,double epsilon=0.999999); // collapses co-linear edges.
+
+
+bool fm_computeSplitPlane(uint32_t vcount,const double *vertices,uint32_t tcount,const uint32_t *indices,double *plane);
+bool fm_computeSplitPlane(uint32_t vcount,const float *vertices,uint32_t tcount,const uint32_t *indices,float *plane);
+
+void fm_nearestPointInTriangle(const float *pos,const float *p1,const float *p2,const float *p3,float *nearest);
+void fm_nearestPointInTriangle(const double *pos,const double *p1,const double *p2,const double *p3,double *nearest);
+
+float  fm_areaTriangle(const float *p1,const float *p2,const float *p3);
+double fm_areaTriangle(const double *p1,const double *p2,const double *p3);
+
+void fm_subtract(const float *A,const float *B,float *diff); // compute A-B and store the result in 'diff'
+void fm_subtract(const double *A,const double *B,double *diff); // compute A-B and store the result in 'diff'
+
+void fm_multiply(float *A,float scaler);
+void fm_multiply(double *A,double scaler);
+
+void fm_add(const float *A,const float *B,float *sum);
+void fm_add(const double *A,const double *B,double *sum);
+
+void fm_copy3(const float *source,float *dest);
+void fm_copy3(const double *source,double *dest);
+
+// re-indexes an indexed triangle mesh but drops unused vertices.  The output_indices can be the same pointer as the input indices.
+// the output_vertices can point to the input vertices if you desire.  The output_vertices buffer should be at least the same size
+// is the input buffer.  The routine returns the new vertex count after re-indexing.
+uint32_t  fm_copyUniqueVertices(uint32_t vcount,const float *input_vertices,float *output_vertices,uint32_t tcount,const uint32_t *input_indices,uint32_t *output_indices);
+uint32_t  fm_copyUniqueVertices(uint32_t vcount,const double *input_vertices,double *output_vertices,uint32_t tcount,const uint32_t *input_indices,uint32_t *output_indices);
+
+bool    fm_isMeshCoplanar(uint32_t tcount,const uint32_t *indices,const float *vertices,bool doubleSided); // returns true if this collection of indexed triangles are co-planar!
+bool    fm_isMeshCoplanar(uint32_t tcount,const uint32_t *indices,const double *vertices,bool doubleSided); // returns true if this collection of indexed triangles are co-planar!
+
+bool    fm_samePlane(const float p1[4],const float p2[4],float normalEpsilon=0.01f,float dEpsilon=0.001f,bool doubleSided=false); // returns true if these two plane equations are identical within an epsilon
+bool    fm_samePlane(const double p1[4],const double p2[4],double normalEpsilon=0.01,double dEpsilon=0.001,bool doubleSided=false);
+
+void    fm_OBBtoAABB(const float obmin[3],const float obmax[3],const float matrix[16],float abmin[3],float abmax[3]);
+
+// a utility class that will tessellate a mesh.
+class fm_Tesselate
+{
+public:
+  virtual const uint32_t * tesselate(fm_VertexIndex *vindex,uint32_t tcount,const uint32_t *indices,float longEdge,uint32_t maxDepth,uint32_t &outcount) = 0;
+};
+
+fm_Tesselate * fm_createTesselate(void);
+void           fm_releaseTesselate(fm_Tesselate *t);
+
+void fm_computeMeanNormals(uint32_t vcount,       // the number of vertices
+						   const float *vertices,     // the base address of the vertex position data.
+						   uint32_t vstride,      // the stride between position data.
+						   float *normals,            // the base address  of the destination for mean vector normals
+						   uint32_t nstride,      // the stride between normals
+						   uint32_t tcount,       // the number of triangles
+						   const uint32_t *indices);     // the triangle indices
+
+void fm_computeMeanNormals(uint32_t vcount,       // the number of vertices
+						   const double *vertices,     // the base address of the vertex position data.
+						   uint32_t vstride,      // the stride between position data.
+						   double *normals,            // the base address  of the destination for mean vector normals
+						   uint32_t nstride,      // the stride between normals
+						   uint32_t tcount,       // the number of triangles
+						   const uint32_t *indices);     // the triangle indices
+
+
+bool fm_isValidTriangle(const float *p1,const float *p2,const float *p3,float epsilon=0.00001f);
+bool fm_isValidTriangle(const double *p1,const double *p2,const double *p3,double epsilon=0.00001f);
+
+
+}; // end of namespace
+
+#endif
diff --git a/thirdparty/vhacd/inc/btAlignedAllocator.h b/thirdparty/vhacd/inc/btAlignedAllocator.h
new file mode 100644
index 0000000000..4f534e43f0
--- /dev/null
+++ b/thirdparty/vhacd/inc/btAlignedAllocator.h
@@ -0,0 +1,111 @@
+/*
+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_ALIGNED_ALLOCATOR
+#define BT_ALIGNED_ALLOCATOR
+
+///we probably replace this with our own aligned memory allocator
+///so we replace _aligned_malloc and _aligned_free with our own
+///that is better portable and more predictable
+
+#include "btScalar.h"
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+//#define BT_DEBUG_MEMORY_ALLOCATIONS 1
+#ifdef BT_DEBUG_MEMORY_ALLOCATIONS
+
+#define btAlignedAlloc(a, b) \
+	btAlignedAllocInternal(a, b, __LINE__, __FILE__)
+
+#define btAlignedFree(ptr) \
+	btAlignedFreeInternal(ptr, __LINE__, __FILE__)
+
+void *btAlignedAllocInternal(size_t size, int32_t alignment, int32_t line, char *filename);
+
+void btAlignedFreeInternal(void *ptr, int32_t line, char *filename);
+
+#else
+void *btAlignedAllocInternal(size_t size, int32_t alignment);
+void btAlignedFreeInternal(void *ptr);
+
+#define btAlignedAlloc(size, alignment) btAlignedAllocInternal(size, alignment)
+#define btAlignedFree(ptr) btAlignedFreeInternal(ptr)
+
+#endif
+typedef int32_t size_type;
+
+typedef void *(btAlignedAllocFunc)(size_t size, int32_t alignment);
+typedef void(btAlignedFreeFunc)(void *memblock);
+typedef void *(btAllocFunc)(size_t size);
+typedef void(btFreeFunc)(void *memblock);
+
+///The developer can let all Bullet memory allocations go through a custom memory allocator, using btAlignedAllocSetCustom
+void btAlignedAllocSetCustom(btAllocFunc *allocFunc, btFreeFunc *freeFunc);
+///If the developer has already an custom aligned allocator, then btAlignedAllocSetCustomAligned can be used. The default aligned allocator pre-allocates extra memory using the non-aligned allocator, and instruments it.
+void btAlignedAllocSetCustomAligned(btAlignedAllocFunc *allocFunc, btAlignedFreeFunc *freeFunc);
+
+///The btAlignedAllocator is a portable class for aligned memory allocations.
+///Default implementations for unaligned and aligned allocations can be overridden by a custom allocator using btAlignedAllocSetCustom and btAlignedAllocSetCustomAligned.
+template <typename T, unsigned Alignment>
+class btAlignedAllocator {
+
+	typedef btAlignedAllocator<T, Alignment> self_type;
+
+public:
+	//just going down a list:
+	btAlignedAllocator() {}
+	/*
+	btAlignedAllocator( const self_type & ) {}
+	*/
+
+	template <typename Other>
+	btAlignedAllocator(const btAlignedAllocator<Other, Alignment> &) {}
+
+	typedef const T *const_pointer;
+	typedef const T &const_reference;
+	typedef T *pointer;
+	typedef T &reference;
+	typedef T value_type;
+
+	pointer address(reference ref) const { return &ref; }
+	const_pointer address(const_reference ref) const { return &ref; }
+	pointer allocate(size_type n, const_pointer *hint = 0) {
+		(void)hint;
+		return reinterpret_cast<pointer>(btAlignedAlloc(sizeof(value_type) * n, Alignment));
+	}
+	void construct(pointer ptr, const value_type &value) { new (ptr) value_type(value); }
+	void deallocate(pointer ptr) {
+		btAlignedFree(reinterpret_cast<void *>(ptr));
+	}
+	void destroy(pointer ptr) { ptr->~value_type(); }
+
+	template <typename O>
+	struct rebind {
+		typedef btAlignedAllocator<O, Alignment> other;
+	};
+	template <typename O>
+	self_type &operator=(const btAlignedAllocator<O, Alignment> &) { return *this; }
+
+	friend bool operator==(const self_type &, const self_type &) { return true; }
+};
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_ALIGNED_ALLOCATOR
diff --git a/thirdparty/vhacd/inc/btAlignedObjectArray.h b/thirdparty/vhacd/inc/btAlignedObjectArray.h
new file mode 100644
index 0000000000..d82449e8fd
--- /dev/null
+++ b/thirdparty/vhacd/inc/btAlignedObjectArray.h
@@ -0,0 +1,420 @@
+/*
+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_OBJECT_ARRAY__
+#define BT_OBJECT_ARRAY__
+
+#include "btAlignedAllocator.h"
+#include "btScalar.h" // has definitions like SIMD_FORCE_INLINE
+
+///If the platform doesn't support placement new, you can disable BT_USE_PLACEMENT_NEW
+///then the btAlignedObjectArray doesn't support objects with virtual methods, and non-trivial constructors/destructors
+///You can enable BT_USE_MEMCPY, then swapping elements in the array will use memcpy instead of operator=
+///see discussion here: http://continuousphysics.com/Bullet/phpBB2/viewtopic.php?t=1231 and
+///http://www.continuousphysics.com/Bullet/phpBB2/viewtopic.php?t=1240
+
+#define BT_USE_PLACEMENT_NEW 1
+//#define BT_USE_MEMCPY 1 //disable, because it is cumbersome to find out for each platform where memcpy is defined. It can be in <memory.h> or <string.h> or otherwise...
+#define BT_ALLOW_ARRAY_COPY_OPERATOR // enabling this can accidently perform deep copies of data if you are not careful
+
+#ifdef BT_USE_MEMCPY
+#include <memory.h>
+#include <string.h>
+#endif //BT_USE_MEMCPY
+
+#ifdef BT_USE_PLACEMENT_NEW
+#include <new> //for placement new
+#endif //BT_USE_PLACEMENT_NEW
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+///The btAlignedObjectArray template class uses a subset of the stl::vector interface for its methods
+///It is developed to replace stl::vector to avoid portability issues, including STL alignment issues to add SIMD/SSE data
+template <typename T>
+//template <class T>
+class btAlignedObjectArray {
+	btAlignedAllocator<T, 16> m_allocator;
+
+	int32_t m_size;
+	int32_t m_capacity;
+	T *m_data;
+	//PCK: added this line
+	bool m_ownsMemory;
+
+#ifdef BT_ALLOW_ARRAY_COPY_OPERATOR
+public:
+	SIMD_FORCE_INLINE btAlignedObjectArray<T> &operator=(const btAlignedObjectArray<T> &other) {
+		copyFromArray(other);
+		return *this;
+	}
+#else //BT_ALLOW_ARRAY_COPY_OPERATOR
+private:
+	SIMD_FORCE_INLINE btAlignedObjectArray<T> &operator=(const btAlignedObjectArray<T> &other);
+#endif //BT_ALLOW_ARRAY_COPY_OPERATOR
+
+protected:
+	SIMD_FORCE_INLINE int32_t allocSize(int32_t size) {
+		return (size ? size * 2 : 1);
+	}
+	SIMD_FORCE_INLINE void copy(int32_t start, int32_t end, T *dest) const {
+		int32_t i;
+		for (i = start; i < end; ++i)
+#ifdef BT_USE_PLACEMENT_NEW
+			new (&dest[i]) T(m_data[i]);
+#else
+			dest[i] = m_data[i];
+#endif //BT_USE_PLACEMENT_NEW
+	}
+
+	SIMD_FORCE_INLINE void init() {
+		//PCK: added this line
+		m_ownsMemory = true;
+		m_data = 0;
+		m_size = 0;
+		m_capacity = 0;
+	}
+	SIMD_FORCE_INLINE void destroy(int32_t first, int32_t last) {
+		int32_t i;
+		for (i = first; i < last; i++) {
+			m_data[i].~T();
+		}
+	}
+
+	SIMD_FORCE_INLINE void *allocate(int32_t size) {
+		if (size)
+			return m_allocator.allocate(size);
+		return 0;
+	}
+
+	SIMD_FORCE_INLINE void deallocate() {
+		if (m_data) {
+			//PCK: enclosed the deallocation in this block
+			if (m_ownsMemory) {
+				m_allocator.deallocate(m_data);
+			}
+			m_data = 0;
+		}
+	}
+
+public:
+	btAlignedObjectArray() {
+		init();
+	}
+
+	~btAlignedObjectArray() {
+		clear();
+	}
+
+	///Generally it is best to avoid using the copy constructor of an btAlignedObjectArray, and use a (const) reference to the array instead.
+	btAlignedObjectArray(const btAlignedObjectArray &otherArray) {
+		init();
+
+		int32_t otherSize = otherArray.size();
+		resize(otherSize);
+		otherArray.copy(0, otherSize, m_data);
+	}
+
+	/// return the number of elements in the array
+	SIMD_FORCE_INLINE int32_t size() const {
+		return m_size;
+	}
+
+	SIMD_FORCE_INLINE const T &at(int32_t n) const {
+		btAssert(n >= 0);
+		btAssert(n < size());
+		return m_data[n];
+	}
+
+	SIMD_FORCE_INLINE T &at(int32_t n) {
+		btAssert(n >= 0);
+		btAssert(n < size());
+		return m_data[n];
+	}
+
+	SIMD_FORCE_INLINE const T &operator[](int32_t n) const {
+		btAssert(n >= 0);
+		btAssert(n < size());
+		return m_data[n];
+	}
+
+	SIMD_FORCE_INLINE T &operator[](int32_t n) {
+		btAssert(n >= 0);
+		btAssert(n < size());
+		return m_data[n];
+	}
+
+	///clear the array, deallocated memory. Generally it is better to use array.resize(0), to reduce performance overhead of run-time memory (de)allocations.
+	SIMD_FORCE_INLINE void clear() {
+		destroy(0, size());
+
+		deallocate();
+
+		init();
+	}
+
+	SIMD_FORCE_INLINE void pop_back() {
+		btAssert(m_size > 0);
+		m_size--;
+		m_data[m_size].~T();
+	}
+
+	///resize changes the number of elements in the array. If the new size is larger, the new elements will be constructed using the optional second argument.
+	///when the new number of elements is smaller, the destructor will be called, but memory will not be freed, to reduce performance overhead of run-time memory (de)allocations.
+	SIMD_FORCE_INLINE void resize(int32_t newsize, const T &fillData = T()) {
+		int32_t curSize = size();
+
+		if (newsize < curSize) {
+			for (int32_t i = newsize; i < curSize; i++) {
+				m_data[i].~T();
+			}
+		} else {
+			if (newsize > size()) {
+				reserve(newsize);
+			}
+#ifdef BT_USE_PLACEMENT_NEW
+			for (int32_t i = curSize; i < newsize; i++) {
+				new (&m_data[i]) T(fillData);
+			}
+#endif //BT_USE_PLACEMENT_NEW
+		}
+
+		m_size = newsize;
+	}
+
+	SIMD_FORCE_INLINE T &expandNonInitializing() {
+		int32_t sz = size();
+		if (sz == capacity()) {
+			reserve(allocSize(size()));
+		}
+		m_size++;
+
+		return m_data[sz];
+	}
+
+	SIMD_FORCE_INLINE T &expand(const T &fillValue = T()) {
+		int32_t sz = size();
+		if (sz == capacity()) {
+			reserve(allocSize(size()));
+		}
+		m_size++;
+#ifdef BT_USE_PLACEMENT_NEW
+		new (&m_data[sz]) T(fillValue); //use the in-place new (not really allocating heap memory)
+#endif
+
+		return m_data[sz];
+	}
+
+	SIMD_FORCE_INLINE void push_back(const T &_Val) {
+		int32_t sz = size();
+		if (sz == capacity()) {
+			reserve(allocSize(size()));
+		}
+
+#ifdef BT_USE_PLACEMENT_NEW
+		new (&m_data[m_size]) T(_Val);
+#else
+		m_data[size()] = _Val;
+#endif //BT_USE_PLACEMENT_NEW
+
+		m_size++;
+	}
+
+	/// return the pre-allocated (reserved) elements, this is at least as large as the total number of elements,see size() and reserve()
+	SIMD_FORCE_INLINE int32_t capacity() const {
+		return m_capacity;
+	}
+
+	SIMD_FORCE_INLINE void reserve(int32_t _Count) { // determine new minimum length of allocated storage
+		if (capacity() < _Count) { // not enough room, reallocate
+			T *s = (T *)allocate(_Count);
+
+			copy(0, size(), s);
+
+			destroy(0, size());
+
+			deallocate();
+
+			//PCK: added this line
+			m_ownsMemory = true;
+
+			m_data = s;
+
+			m_capacity = _Count;
+		}
+	}
+
+	class less {
+	public:
+		bool operator()(const T &a, const T &b) {
+			return (a < b);
+		}
+	};
+
+	template <typename L>
+	void quickSortInternal(const L &CompareFunc, int32_t lo, int32_t hi) {
+		//  lo is the lower index, hi is the upper index
+		//  of the region of array a that is to be sorted
+		int32_t i = lo, j = hi;
+		T x = m_data[(lo + hi) / 2];
+
+		//  partition
+		do {
+			while (CompareFunc(m_data[i], x))
+				i++;
+			while (CompareFunc(x, m_data[j]))
+				j--;
+			if (i <= j) {
+				swap(i, j);
+				i++;
+				j--;
+			}
+		} while (i <= j);
+
+		//  recursion
+		if (lo < j)
+			quickSortInternal(CompareFunc, lo, j);
+		if (i < hi)
+			quickSortInternal(CompareFunc, i, hi);
+	}
+
+	template <typename L>
+	void quickSort(const L &CompareFunc) {
+		//don't sort 0 or 1 elements
+		if (size() > 1) {
+			quickSortInternal(CompareFunc, 0, size() - 1);
+		}
+	}
+
+	///heap sort from http://www.csse.monash.edu.au/~lloyd/tildeAlgDS/Sort/Heap/
+	template <typename L>
+	void downHeap(T *pArr, int32_t k, int32_t n, const L &CompareFunc) {
+		/*  PRE: a[k+1..N] is a heap */
+		/* POST:  a[k..N]  is a heap */
+
+		T temp = pArr[k - 1];
+		/* k has child(s) */
+		while (k <= n / 2) {
+			int32_t child = 2 * k;
+
+			if ((child < n) && CompareFunc(pArr[child - 1], pArr[child])) {
+				child++;
+			}
+			/* pick larger child */
+			if (CompareFunc(temp, pArr[child - 1])) {
+				/* move child up */
+				pArr[k - 1] = pArr[child - 1];
+				k = child;
+			} else {
+				break;
+			}
+		}
+		pArr[k - 1] = temp;
+	} /*downHeap*/
+
+	void swap(int32_t index0, int32_t index1) {
+#ifdef BT_USE_MEMCPY
+		char temp[sizeof(T)];
+		memcpy(temp, &m_data[index0], sizeof(T));
+		memcpy(&m_data[index0], &m_data[index1], sizeof(T));
+		memcpy(&m_data[index1], temp, sizeof(T));
+#else
+		T temp = m_data[index0];
+		m_data[index0] = m_data[index1];
+		m_data[index1] = temp;
+#endif //BT_USE_PLACEMENT_NEW
+	}
+
+	template <typename L>
+	void heapSort(const L &CompareFunc) {
+		/* sort a[0..N-1],  N.B. 0 to N-1 */
+		int32_t k;
+		int32_t n = m_size;
+		for (k = n / 2; k > 0; k--) {
+			downHeap(m_data, k, n, CompareFunc);
+		}
+
+		/* a[1..N] is now a heap */
+		while (n >= 1) {
+			swap(0, n - 1); /* largest of a[0..n-1] */
+
+			n = n - 1;
+			/* restore a[1..i-1] heap */
+			downHeap(m_data, 1, n, CompareFunc);
+		}
+	}
+
+	///non-recursive binary search, assumes sorted array
+	int32_t findBinarySearch(const T &key) const {
+		int32_t first = 0;
+		int32_t last = size() - 1;
+
+		//assume sorted array
+		while (first <= last) {
+			int32_t mid = (first + last) / 2; // compute mid point.
+			if (key > m_data[mid])
+				first = mid + 1; // repeat search in top half.
+			else if (key < m_data[mid])
+				last = mid - 1; // repeat search in bottom half.
+			else
+				return mid; // found it. return position /////
+		}
+		return size(); // failed to find key
+	}
+
+	int32_t findLinearSearch(const T &key) const {
+		int32_t index = size();
+		int32_t i;
+
+		for (i = 0; i < size(); i++) {
+			if (m_data[i] == key) {
+				index = i;
+				break;
+			}
+		}
+		return index;
+	}
+
+	void remove(const T &key) {
+
+		int32_t findIndex = findLinearSearch(key);
+		if (findIndex < size()) {
+			swap(findIndex, size() - 1);
+			pop_back();
+		}
+	}
+
+	//PCK: whole function
+	void initializeFromBuffer(void *buffer, int32_t size, int32_t capacity) {
+		clear();
+		m_ownsMemory = false;
+		m_data = (T *)buffer;
+		m_size = size;
+		m_capacity = capacity;
+	}
+
+	void copyFromArray(const btAlignedObjectArray &otherArray) {
+		int32_t otherSize = otherArray.size();
+		resize(otherSize);
+		otherArray.copy(0, otherSize, m_data);
+	}
+};
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_OBJECT_ARRAY__
diff --git a/thirdparty/vhacd/inc/btConvexHullComputer.h b/thirdparty/vhacd/inc/btConvexHullComputer.h
new file mode 100644
index 0000000000..f10a3cf86f
--- /dev/null
+++ b/thirdparty/vhacd/inc/btConvexHullComputer.h
@@ -0,0 +1,100 @@
+/*
+Copyright (c) 2011 Ole Kniemeyer, MAXON, www.maxon.net
+
+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_CONVEX_HULL_COMPUTER_H
+#define BT_CONVEX_HULL_COMPUTER_H
+
+#include "btAlignedObjectArray.h"
+#include "btVector3.h"
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+/// Convex hull implementation based on Preparata and Hong
+/// See http://code.google.com/p/bullet/issues/detail?id=275
+/// Ole Kniemeyer, MAXON Computer GmbH
+class btConvexHullComputer {
+private:
+	btScalar compute(const void *coords, bool doubleCoords, int32_t stride, int32_t count, btScalar shrink, btScalar shrinkClamp);
+
+public:
+	class Edge {
+	private:
+		int32_t next;
+		int32_t reverse;
+		int32_t targetVertex;
+
+		friend class btConvexHullComputer;
+
+	public:
+		int32_t getSourceVertex() const {
+			return (this + reverse)->targetVertex;
+		}
+
+		int32_t getTargetVertex() const {
+			return targetVertex;
+		}
+
+		const Edge *getNextEdgeOfVertex() const // clockwise list of all edges of a vertex
+		{
+			return this + next;
+		}
+
+		const Edge *getNextEdgeOfFace() const // counter-clockwise list of all edges of a face
+		{
+			return (this + reverse)->getNextEdgeOfVertex();
+		}
+
+		const Edge *getReverseEdge() const {
+			return this + reverse;
+		}
+	};
+
+	// Vertices of the output hull
+	btAlignedObjectArray<btVector3> vertices;
+
+	// Edges of the output hull
+	btAlignedObjectArray<Edge> edges;
+
+	// Faces of the convex hull. Each entry is an index into the "edges" array pointing to an edge of the face. Faces are planar n-gons
+	btAlignedObjectArray<int32_t> faces;
+
+	/*
+		Compute convex hull of "count" vertices stored in "coords". "stride" is the difference in bytes
+		between the addresses of consecutive vertices. If "shrink" is positive, the convex hull is shrunken
+		by that amount (each face is moved by "shrink" length units towards the center along its normal).
+		If "shrinkClamp" is positive, "shrink" is clamped to not exceed "shrinkClamp * innerRadius", where "innerRadius"
+		is the minimum distance of a face to the center of the convex hull.
+
+		The returned value is the amount by which the hull has been shrunken. If it is negative, the amount was so large
+		that the resulting convex hull is empty.
+
+		The output convex hull can be found in the member variables "vertices", "edges", "faces".
+		*/
+	btScalar compute(const float *coords, int32_t stride, int32_t count, btScalar shrink, btScalar shrinkClamp) {
+		return compute(coords, false, stride, count, shrink, shrinkClamp);
+	}
+
+	// same as above, but double precision
+	btScalar compute(const double *coords, int32_t stride, int32_t count, btScalar shrink, btScalar shrinkClamp) {
+		return compute(coords, true, stride, count, shrink, shrinkClamp);
+	}
+};
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_CONVEX_HULL_COMPUTER_H
diff --git a/thirdparty/vhacd/inc/btMinMax.h b/thirdparty/vhacd/inc/btMinMax.h
new file mode 100644
index 0000000000..a5be258667
--- /dev/null
+++ b/thirdparty/vhacd/inc/btMinMax.h
@@ -0,0 +1,66 @@
+/*
+Copyright (c) 2003-2006 Gino van den Bergen / 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_GEN_MINMAX_H
+#define BT_GEN_MINMAX_H
+
+#include "btScalar.h"
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+template <class T>
+SIMD_FORCE_INLINE const T &btMin(const T &a, const T &b) {
+	return a < b ? a : b;
+}
+
+template <class T>
+SIMD_FORCE_INLINE const T &btMax(const T &a, const T &b) {
+	return a > b ? a : b;
+}
+
+template <class T>
+SIMD_FORCE_INLINE const T &btClamped(const T &a, const T &lb, const T &ub) {
+	return a < lb ? lb : (ub < a ? ub : a);
+}
+
+template <class T>
+SIMD_FORCE_INLINE void btSetMin(T &a, const T &b) {
+	if (b < a) {
+		a = b;
+	}
+}
+
+template <class T>
+SIMD_FORCE_INLINE void btSetMax(T &a, const T &b) {
+	if (a < b) {
+		a = b;
+	}
+}
+
+template <class T>
+SIMD_FORCE_INLINE void btClamp(T &a, const T &lb, const T &ub) {
+	if (a < lb) {
+		a = lb;
+	} else if (ub < a) {
+		a = ub;
+	}
+}
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_GEN_MINMAX_H
diff --git a/thirdparty/vhacd/inc/btScalar.h b/thirdparty/vhacd/inc/btScalar.h
new file mode 100644
index 0000000000..a9f836f9f2
--- /dev/null
+++ b/thirdparty/vhacd/inc/btScalar.h
@@ -0,0 +1,565 @@
+/*
+Copyright (c) 2003-2009 Erwin Coumans  http://bullet.googlecode.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.
+*/
+
+#ifndef BT_SCALAR_H
+#define BT_SCALAR_H
+
+#ifdef BT_MANAGED_CODE
+//Aligned data types not supported in managed code
+#pragma unmanaged
+#endif
+
+#include <float.h>
+#include <math.h>
+#include <stdint.h>
+#include <stdlib.h> //size_t for MSVC 6.0
+
+/* SVN $Revision$ on $Date$ from http://bullet.googlecode.com*/
+#define BT_BULLET_VERSION 279
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+inline int32_t btGetVersion() {
+	return BT_BULLET_VERSION;
+}
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#if defined(DEBUG) || defined(_DEBUG)
+#define BT_DEBUG
+#endif
+
+#ifdef _WIN32
+
+#if defined(__MINGW32__) || defined(__CYGWIN__) || (defined(_MSC_VER) && _MSC_VER < 1300)
+
+#define SIMD_FORCE_INLINE inline
+#define ATTRIBUTE_ALIGNED16(a) a
+#define ATTRIBUTE_ALIGNED64(a) a
+#define ATTRIBUTE_ALIGNED128(a) a
+#else
+//#define BT_HAS_ALIGNED_ALLOCATOR
+#pragma warning(disable : 4324) // disable padding warning
+//			#pragma warning(disable:4530) // Disable the exception disable but used in MSCV Stl warning.
+//			#pragma warning(disable:4996) //Turn off warnings about deprecated C routines
+//			#pragma warning(disable:4786) // Disable the "debug name too long" warning
+
+#define SIMD_FORCE_INLINE __forceinline
+#define ATTRIBUTE_ALIGNED16(a) __declspec(align(16)) a
+#define ATTRIBUTE_ALIGNED64(a) __declspec(align(64)) a
+#define ATTRIBUTE_ALIGNED128(a) __declspec(align(128)) a
+#ifdef _XBOX
+#define BT_USE_VMX128
+
+#include <ppcintrinsics.h>
+#define BT_HAVE_NATIVE_FSEL
+#define btFsel(a, b, c) __fsel((a), (b), (c))
+#else
+
+#if (defined(_WIN32) && (_MSC_VER) && _MSC_VER >= 1400) && (!defined(BT_USE_DOUBLE_PRECISION))
+#define BT_USE_SSE
+#include <emmintrin.h>
+#endif
+
+#endif //_XBOX
+
+#endif //__MINGW32__
+
+#include <assert.h>
+#ifdef BT_DEBUG
+#define btAssert assert
+#else
+#define btAssert(x)
+#endif
+//btFullAssert is optional, slows down a lot
+#define btFullAssert(x)
+
+#define btLikely(_c) _c
+#define btUnlikely(_c) _c
+
+#else
+
+#if defined(__CELLOS_LV2__)
+#define SIMD_FORCE_INLINE inline __attribute__((always_inline))
+#define ATTRIBUTE_ALIGNED16(a) a __attribute__((aligned(16)))
+#define ATTRIBUTE_ALIGNED64(a) a __attribute__((aligned(64)))
+#define ATTRIBUTE_ALIGNED128(a) a __attribute__((aligned(128)))
+#ifndef assert
+#include <assert.h>
+#endif
+#ifdef BT_DEBUG
+#ifdef __SPU__
+#include <spu_printf.h>
+#define printf spu_printf
+#define btAssert(x)                                                \
+	{                                                              \
+		if (!(x)) {                                                \
+			printf("Assert " __FILE__ ":%u (" #x ")\n", __LINE__); \
+			spu_hcmpeq(0, 0);                                      \
+		}                                                          \
+	}
+#else
+#define btAssert assert
+#endif
+
+#else
+#define btAssert(x)
+#endif
+//btFullAssert is optional, slows down a lot
+#define btFullAssert(x)
+
+#define btLikely(_c) _c
+#define btUnlikely(_c) _c
+
+#else
+
+#ifdef USE_LIBSPE2
+
+#define SIMD_FORCE_INLINE __inline
+#define ATTRIBUTE_ALIGNED16(a) a __attribute__((aligned(16)))
+#define ATTRIBUTE_ALIGNED64(a) a __attribute__((aligned(64)))
+#define ATTRIBUTE_ALIGNED128(a) a __attribute__((aligned(128)))
+#ifndef assert
+#include <assert.h>
+#endif
+#ifdef BT_DEBUG
+#define btAssert assert
+#else
+#define btAssert(x)
+#endif
+//btFullAssert is optional, slows down a lot
+#define btFullAssert(x)
+
+#define btLikely(_c) __builtin_expect((_c), 1)
+#define btUnlikely(_c) __builtin_expect((_c), 0)
+
+#else
+//non-windows systems
+
+#if (defined(__APPLE__) && defined(__i386__) && (!defined(BT_USE_DOUBLE_PRECISION)))
+#define BT_USE_SSE
+#include <emmintrin.h>
+
+#define SIMD_FORCE_INLINE inline
+///@todo: check out alignment methods for other platforms/compilers
+#define ATTRIBUTE_ALIGNED16(a) a __attribute__((aligned(16)))
+#define ATTRIBUTE_ALIGNED64(a) a __attribute__((aligned(64)))
+#define ATTRIBUTE_ALIGNED128(a) a __attribute__((aligned(128)))
+#ifndef assert
+#include <assert.h>
+#endif
+
+#if defined(DEBUG) || defined(_DEBUG)
+#define btAssert assert
+#else
+#define btAssert(x)
+#endif
+
+//btFullAssert is optional, slows down a lot
+#define btFullAssert(x)
+#define btLikely(_c) _c
+#define btUnlikely(_c) _c
+
+#else
+
+#define SIMD_FORCE_INLINE inline
+///@todo: check out alignment methods for other platforms/compilers
+///#define ATTRIBUTE_ALIGNED16(a) a __attribute__ ((aligned (16)))
+///#define ATTRIBUTE_ALIGNED64(a) a __attribute__ ((aligned (64)))
+///#define ATTRIBUTE_ALIGNED128(a) a __attribute__ ((aligned (128)))
+#define ATTRIBUTE_ALIGNED16(a) a
+#define ATTRIBUTE_ALIGNED64(a) a
+#define ATTRIBUTE_ALIGNED128(a) a
+#ifndef assert
+#include <assert.h>
+#endif
+
+#if defined(DEBUG) || defined(_DEBUG)
+#define btAssert assert
+#else
+#define btAssert(x)
+#endif
+
+//btFullAssert is optional, slows down a lot
+#define btFullAssert(x)
+#define btLikely(_c) _c
+#define btUnlikely(_c) _c
+#endif //__APPLE__
+
+#endif // LIBSPE2
+
+#endif //__CELLOS_LV2__
+#endif
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+///The btScalar type abstracts floating point numbers, to easily switch between double and single floating point precision.
+#if defined(BT_USE_DOUBLE_PRECISION)
+typedef double btScalar;
+//this number could be bigger in double precision
+#define BT_LARGE_FLOAT 1e30
+#else
+typedef float btScalar;
+//keep BT_LARGE_FLOAT*BT_LARGE_FLOAT < FLT_MAX
+#define BT_LARGE_FLOAT 1e18f
+#endif
+
+#define BT_DECLARE_ALIGNED_ALLOCATOR()                                                                     \
+	SIMD_FORCE_INLINE void *operator new(size_t sizeInBytes) { return btAlignedAlloc(sizeInBytes, 16); }   \
+	SIMD_FORCE_INLINE void operator delete(void *ptr) { btAlignedFree(ptr); }                              \
+	SIMD_FORCE_INLINE void *operator new(size_t, void *ptr) { return ptr; }                                \
+	SIMD_FORCE_INLINE void operator delete(void *, void *) {}                                              \
+	SIMD_FORCE_INLINE void *operator new[](size_t sizeInBytes) { return btAlignedAlloc(sizeInBytes, 16); } \
+	SIMD_FORCE_INLINE void operator delete[](void *ptr) { btAlignedFree(ptr); }                            \
+	SIMD_FORCE_INLINE void *operator new[](size_t, void *ptr) { return ptr; }                              \
+	SIMD_FORCE_INLINE void operator delete[](void *, void *) {}
+
+#if defined(BT_USE_DOUBLE_PRECISION) || defined(BT_FORCE_DOUBLE_FUNCTIONS)
+
+SIMD_FORCE_INLINE btScalar btSqrt(btScalar x) {
+	return sqrt(x);
+}
+SIMD_FORCE_INLINE btScalar btFabs(btScalar x) {
+	return fabs(x);
+}
+SIMD_FORCE_INLINE btScalar btCos(btScalar x) {
+	return cos(x);
+}
+SIMD_FORCE_INLINE btScalar btSin(btScalar x) {
+	return sin(x);
+}
+SIMD_FORCE_INLINE btScalar btTan(btScalar x) {
+	return tan(x);
+}
+SIMD_FORCE_INLINE btScalar btAcos(btScalar x) {
+	if (x < btScalar(-1))
+		x = btScalar(-1);
+	if (x > btScalar(1))
+		x = btScalar(1);
+	return acos(x);
+}
+SIMD_FORCE_INLINE btScalar btAsin(btScalar x) {
+	if (x < btScalar(-1))
+		x = btScalar(-1);
+	if (x > btScalar(1))
+		x = btScalar(1);
+	return asin(x);
+}
+SIMD_FORCE_INLINE btScalar btAtan(btScalar x) {
+	return atan(x);
+}
+SIMD_FORCE_INLINE btScalar btAtan2(btScalar x, btScalar y) {
+	return atan2(x, y);
+}
+SIMD_FORCE_INLINE btScalar btExp(btScalar x) {
+	return exp(x);
+}
+SIMD_FORCE_INLINE btScalar btLog(btScalar x) {
+	return log(x);
+}
+SIMD_FORCE_INLINE btScalar btPow(btScalar x, btScalar y) {
+	return pow(x, y);
+}
+SIMD_FORCE_INLINE btScalar btFmod(btScalar x, btScalar y) {
+	return fmod(x, y);
+}
+
+#else
+
+SIMD_FORCE_INLINE btScalar btSqrt(btScalar y) {
+#ifdef USE_APPROXIMATION
+	double x, z, tempf;
+	unsigned long *tfptr = ((unsigned long *)&tempf) + 1;
+
+	tempf = y;
+	*tfptr = (0xbfcdd90a - *tfptr) >> 1; /* estimate of 1/sqrt(y) */
+	x = tempf;
+	z = y * btScalar(0.5);
+	x = (btScalar(1.5) * x) - (x * x) * (x * z); /* iteration formula     */
+	x = (btScalar(1.5) * x) - (x * x) * (x * z);
+	x = (btScalar(1.5) * x) - (x * x) * (x * z);
+	x = (btScalar(1.5) * x) - (x * x) * (x * z);
+	x = (btScalar(1.5) * x) - (x * x) * (x * z);
+	return x * y;
+#else
+	return sqrtf(y);
+#endif
+}
+SIMD_FORCE_INLINE btScalar btFabs(btScalar x) {
+	return fabsf(x);
+}
+SIMD_FORCE_INLINE btScalar btCos(btScalar x) {
+	return cosf(x);
+}
+SIMD_FORCE_INLINE btScalar btSin(btScalar x) {
+	return sinf(x);
+}
+SIMD_FORCE_INLINE btScalar btTan(btScalar x) {
+	return tanf(x);
+}
+SIMD_FORCE_INLINE btScalar btAcos(btScalar x) {
+	if (x < btScalar(-1))
+		x = btScalar(-1);
+	if (x > btScalar(1))
+		x = btScalar(1);
+	return acosf(x);
+}
+SIMD_FORCE_INLINE btScalar btAsin(btScalar x) {
+	if (x < btScalar(-1))
+		x = btScalar(-1);
+	if (x > btScalar(1))
+		x = btScalar(1);
+	return asinf(x);
+}
+SIMD_FORCE_INLINE btScalar btAtan(btScalar x) {
+	return atanf(x);
+}
+SIMD_FORCE_INLINE btScalar btAtan2(btScalar x, btScalar y) {
+	return atan2f(x, y);
+}
+SIMD_FORCE_INLINE btScalar btExp(btScalar x) {
+	return expf(x);
+}
+SIMD_FORCE_INLINE btScalar btLog(btScalar x) {
+	return logf(x);
+}
+SIMD_FORCE_INLINE btScalar btPow(btScalar x, btScalar y) {
+	return powf(x, y);
+}
+SIMD_FORCE_INLINE btScalar btFmod(btScalar x, btScalar y) {
+	return fmodf(x, y);
+}
+
+#endif
+
+#define SIMD_2_PI btScalar(6.283185307179586232)
+#define SIMD_PI (SIMD_2_PI * btScalar(0.5))
+#define SIMD_HALF_PI (SIMD_2_PI * btScalar(0.25))
+#define SIMD_RADS_PER_DEG (SIMD_2_PI / btScalar(360.0))
+#define SIMD_DEGS_PER_RAD (btScalar(360.0) / SIMD_2_PI)
+#define SIMDSQRT12 btScalar(0.7071067811865475244008443621048490)
+
+#define btRecipSqrt(x) ((btScalar)(btScalar(1.0) / btSqrt(btScalar(x)))) /* reciprocal square root */
+
+#ifdef BT_USE_DOUBLE_PRECISION
+#define SIMD_EPSILON DBL_EPSILON
+#define SIMD_INFINITY DBL_MAX
+#else
+#define SIMD_EPSILON FLT_EPSILON
+#define SIMD_INFINITY FLT_MAX
+#endif
+
+SIMD_FORCE_INLINE btScalar btAtan2Fast(btScalar y, btScalar x) {
+	btScalar coeff_1 = SIMD_PI / 4.0f;
+	btScalar coeff_2 = 3.0f * coeff_1;
+	btScalar abs_y = btFabs(y);
+	btScalar angle;
+	if (x >= 0.0f) {
+		btScalar r = (x - abs_y) / (x + abs_y);
+		angle = coeff_1 - coeff_1 * r;
+	} else {
+		btScalar r = (x + abs_y) / (abs_y - x);
+		angle = coeff_2 - coeff_1 * r;
+	}
+	return (y < 0.0f) ? -angle : angle;
+}
+
+SIMD_FORCE_INLINE bool btFuzzyZero(btScalar x) {
+	return btFabs(x) < SIMD_EPSILON;
+}
+
+SIMD_FORCE_INLINE bool btEqual(btScalar a, btScalar eps) {
+	return (((a) <= eps) && !((a) < -eps));
+}
+SIMD_FORCE_INLINE bool btGreaterEqual(btScalar a, btScalar eps) {
+	return (!((a) <= eps));
+}
+
+SIMD_FORCE_INLINE int32_t btIsNegative(btScalar x) {
+	return x < btScalar(0.0) ? 1 : 0;
+}
+
+SIMD_FORCE_INLINE btScalar btRadians(btScalar x) {
+	return x * SIMD_RADS_PER_DEG;
+}
+SIMD_FORCE_INLINE btScalar btDegrees(btScalar x) {
+	return x * SIMD_DEGS_PER_RAD;
+}
+
+#define BT_DECLARE_HANDLE(name) \
+	typedef struct name##__ {   \
+		int32_t unused;         \
+	} * name
+
+#ifndef btFsel
+SIMD_FORCE_INLINE btScalar btFsel(btScalar a, btScalar b, btScalar c) {
+	return a >= 0 ? b : c;
+}
+#endif
+#define btFsels(a, b, c) (btScalar) btFsel(a, b, c)
+
+SIMD_FORCE_INLINE bool btMachineIsLittleEndian() {
+	long int i = 1;
+	const char *p = (const char *)&i;
+	if (p[0] == 1) // Lowest address contains the least significant byte
+		return true;
+	else
+		return false;
+}
+
+///btSelect avoids branches, which makes performance much better for consoles like Playstation 3 and XBox 360
+///Thanks Phil Knight. See also http://www.cellperformance.com/articles/2006/04/more_techniques_for_eliminatin_1.html
+SIMD_FORCE_INLINE unsigned btSelect(unsigned condition, unsigned valueIfConditionNonZero, unsigned valueIfConditionZero) {
+	// Set testNz to 0xFFFFFFFF if condition is nonzero, 0x00000000 if condition is zero
+	// Rely on positive value or'ed with its negative having sign bit on
+	// and zero value or'ed with its negative (which is still zero) having sign bit off
+	// Use arithmetic shift right, shifting the sign bit through all 32 bits
+	unsigned testNz = (unsigned)(((int32_t)condition | -(int32_t)condition) >> 31);
+	unsigned testEqz = ~testNz;
+	return ((valueIfConditionNonZero & testNz) | (valueIfConditionZero & testEqz));
+}
+SIMD_FORCE_INLINE int32_t btSelect(unsigned condition, int32_t valueIfConditionNonZero, int32_t valueIfConditionZero) {
+	unsigned testNz = (unsigned)(((int32_t)condition | -(int32_t)condition) >> 31);
+	unsigned testEqz = ~testNz;
+	return static_cast<int32_t>((valueIfConditionNonZero & testNz) | (valueIfConditionZero & testEqz));
+}
+SIMD_FORCE_INLINE float btSelect(unsigned condition, float valueIfConditionNonZero, float valueIfConditionZero) {
+#ifdef BT_HAVE_NATIVE_FSEL
+	return (float)btFsel((btScalar)condition - btScalar(1.0f), valueIfConditionNonZero, valueIfConditionZero);
+#else
+	return (condition != 0) ? valueIfConditionNonZero : valueIfConditionZero;
+#endif
+}
+
+template <typename T>
+SIMD_FORCE_INLINE void btSwap(T &a, T &b) {
+	T tmp = a;
+	a = b;
+	b = tmp;
+}
+
+//PCK: endian swapping functions
+SIMD_FORCE_INLINE unsigned btSwapEndian(unsigned val) {
+	return (((val & 0xff000000) >> 24) | ((val & 0x00ff0000) >> 8) | ((val & 0x0000ff00) << 8) | ((val & 0x000000ff) << 24));
+}
+
+SIMD_FORCE_INLINE unsigned short btSwapEndian(unsigned short val) {
+	return static_cast<unsigned short>(((val & 0xff00) >> 8) | ((val & 0x00ff) << 8));
+}
+
+SIMD_FORCE_INLINE unsigned btSwapEndian(int32_t val) {
+	return btSwapEndian((unsigned)val);
+}
+
+SIMD_FORCE_INLINE unsigned short btSwapEndian(short val) {
+	return btSwapEndian((unsigned short)val);
+}
+
+///btSwapFloat uses using char pointers to swap the endianness
+////btSwapFloat/btSwapDouble will NOT return a float, because the machine might 'correct' invalid floating point values
+///Not all values of sign/exponent/mantissa are valid floating point numbers according to IEEE 754.
+///When a floating point unit is faced with an invalid value, it may actually change the value, or worse, throw an exception.
+///In most systems, running user mode code, you wouldn't get an exception, but instead the hardware/os/runtime will 'fix' the number for you.
+///so instead of returning a float/double, we return integer/long long integer
+SIMD_FORCE_INLINE uint32_t btSwapEndianFloat(float d) {
+	uint32_t a = 0;
+	unsigned char *dst = (unsigned char *)&a;
+	unsigned char *src = (unsigned char *)&d;
+
+	dst[0] = src[3];
+	dst[1] = src[2];
+	dst[2] = src[1];
+	dst[3] = src[0];
+	return a;
+}
+
+// unswap using char pointers
+SIMD_FORCE_INLINE float btUnswapEndianFloat(uint32_t a) {
+	float d = 0.0f;
+	unsigned char *src = (unsigned char *)&a;
+	unsigned char *dst = (unsigned char *)&d;
+
+	dst[0] = src[3];
+	dst[1] = src[2];
+	dst[2] = src[1];
+	dst[3] = src[0];
+
+	return d;
+}
+
+// swap using char pointers
+SIMD_FORCE_INLINE void btSwapEndianDouble(double d, unsigned char *dst) {
+	unsigned char *src = (unsigned char *)&d;
+
+	dst[0] = src[7];
+	dst[1] = src[6];
+	dst[2] = src[5];
+	dst[3] = src[4];
+	dst[4] = src[3];
+	dst[5] = src[2];
+	dst[6] = src[1];
+	dst[7] = src[0];
+}
+
+// unswap using char pointers
+SIMD_FORCE_INLINE double btUnswapEndianDouble(const unsigned char *src) {
+	double d = 0.0;
+	unsigned char *dst = (unsigned char *)&d;
+
+	dst[0] = src[7];
+	dst[1] = src[6];
+	dst[2] = src[5];
+	dst[3] = src[4];
+	dst[4] = src[3];
+	dst[5] = src[2];
+	dst[6] = src[1];
+	dst[7] = src[0];
+
+	return d;
+}
+
+// returns normalized value in range [-SIMD_PI, SIMD_PI]
+SIMD_FORCE_INLINE btScalar btNormalizeAngle(btScalar angleInRadians) {
+	angleInRadians = btFmod(angleInRadians, SIMD_2_PI);
+	if (angleInRadians < -SIMD_PI) {
+		return angleInRadians + SIMD_2_PI;
+	} else if (angleInRadians > SIMD_PI) {
+		return angleInRadians - SIMD_2_PI;
+	} else {
+		return angleInRadians;
+	}
+}
+
+///rudimentary class to provide type info
+struct btTypedObject {
+	btTypedObject(int32_t objectType) :
+			m_objectType(objectType) {
+	}
+	int32_t m_objectType;
+	inline int32_t getObjectType() const {
+		return m_objectType;
+	}
+};
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_SCALAR_H
diff --git a/thirdparty/vhacd/inc/btVector3.h b/thirdparty/vhacd/inc/btVector3.h
new file mode 100644
index 0000000000..664728a419
--- /dev/null
+++ b/thirdparty/vhacd/inc/btVector3.h
@@ -0,0 +1,653 @@
+/*
+Copyright (c) 2003-2006 Gino van den Bergen / 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_VECTOR3_H
+#define BT_VECTOR3_H
+
+#include "btMinMax.h"
+#include "btScalar.h"
+
+#ifdef BT_USE_DOUBLE_PRECISION
+#define btVector3Data btVector3DoubleData
+#define btVector3DataName "btVector3DoubleData"
+#else
+#define btVector3Data btVector3FloatData
+#define btVector3DataName "btVector3FloatData"
+#endif //BT_USE_DOUBLE_PRECISION
+
+/**@brief btVector3 can be used to represent 3D points and vectors.
+ * It has an un-used w component to suit 16-byte alignment when btVector3 is stored in containers. This extra component can be used by derived classes (Quaternion?) or by user
+ * Ideally, this class should be replaced by a platform optimized SIMD version that keeps the data in registers
+ */
+//GODOT ADDITION
+namespace VHACD {
+//
+
+ATTRIBUTE_ALIGNED16(class)
+btVector3 {
+public:
+#if defined(__SPU__) && defined(__CELLOS_LV2__)
+	btScalar m_floats[4];
+
+public:
+	SIMD_FORCE_INLINE const vec_float4 &get128() const {
+		return *((const vec_float4 *)&m_floats[0]);
+	}
+
+public:
+#else //__CELLOS_LV2__ __SPU__
+#ifdef BT_USE_SSE // _WIN32
+	union {
+		__m128 mVec128;
+		btScalar m_floats[4];
+	};
+	SIMD_FORCE_INLINE __m128 get128() const {
+		return mVec128;
+	}
+	SIMD_FORCE_INLINE void set128(__m128 v128) {
+		mVec128 = v128;
+	}
+#else
+	btScalar m_floats[4];
+#endif
+#endif //__CELLOS_LV2__ __SPU__
+
+public:
+	/**@brief No initialization constructor */
+	SIMD_FORCE_INLINE btVector3() {}
+
+	/**@brief Constructor from scalars
+   * @param x X value
+   * @param y Y value 
+   * @param z Z value 
+   */
+	SIMD_FORCE_INLINE btVector3(const btScalar &x, const btScalar &y, const btScalar &z) {
+		m_floats[0] = x;
+		m_floats[1] = y;
+		m_floats[2] = z;
+		m_floats[3] = btScalar(0.);
+	}
+
+	/**@brief Add a vector to this one
+ * @param The vector to add to this one */
+	SIMD_FORCE_INLINE btVector3 &operator+=(const btVector3 &v) {
+
+		m_floats[0] += v.m_floats[0];
+		m_floats[1] += v.m_floats[1];
+		m_floats[2] += v.m_floats[2];
+		return *this;
+	}
+
+	/**@brief Subtract a vector from this one
+   * @param The vector to subtract */
+	SIMD_FORCE_INLINE btVector3 &operator-=(const btVector3 &v) {
+		m_floats[0] -= v.m_floats[0];
+		m_floats[1] -= v.m_floats[1];
+		m_floats[2] -= v.m_floats[2];
+		return *this;
+	}
+	/**@brief Scale the vector
+   * @param s Scale factor */
+	SIMD_FORCE_INLINE btVector3 &operator*=(const btScalar &s) {
+		m_floats[0] *= s;
+		m_floats[1] *= s;
+		m_floats[2] *= s;
+		return *this;
+	}
+
+	/**@brief Inversely scale the vector
+   * @param s Scale factor to divide by */
+	SIMD_FORCE_INLINE btVector3 &operator/=(const btScalar &s) {
+		btFullAssert(s != btScalar(0.0));
+		return *this *= btScalar(1.0) / s;
+	}
+
+	/**@brief Return the dot product
+   * @param v The other vector in the dot product */
+	SIMD_FORCE_INLINE btScalar dot(const btVector3 &v) const {
+		return m_floats[0] * v.m_floats[0] + m_floats[1] * v.m_floats[1] + m_floats[2] * v.m_floats[2];
+	}
+
+	/**@brief Return the length of the vector squared */
+	SIMD_FORCE_INLINE btScalar length2() const {
+		return dot(*this);
+	}
+
+	/**@brief Return the length of the vector */
+	SIMD_FORCE_INLINE btScalar length() const {
+		return btSqrt(length2());
+	}
+
+	/**@brief Return the distance squared between the ends of this and another vector
+   * This is symantically treating the vector like a point */
+	SIMD_FORCE_INLINE btScalar distance2(const btVector3 &v) const;
+
+	/**@brief Return the distance between the ends of this and another vector
+   * This is symantically treating the vector like a point */
+	SIMD_FORCE_INLINE btScalar distance(const btVector3 &v) const;
+
+	SIMD_FORCE_INLINE btVector3 &safeNormalize() {
+		btVector3 absVec = this->absolute();
+		int32_t maxIndex = absVec.maxAxis();
+		if (absVec[maxIndex] > 0) {
+			*this /= absVec[maxIndex];
+			return *this /= length();
+		}
+		setValue(1, 0, 0);
+		return *this;
+	}
+
+	/**@brief Normalize this vector
+   * x^2 + y^2 + z^2 = 1 */
+	SIMD_FORCE_INLINE btVector3 &normalize() {
+		return *this /= length();
+	}
+
+	/**@brief Return a normalized version of this vector */
+	SIMD_FORCE_INLINE btVector3 normalized() const;
+
+	/**@brief Return a rotated version of this vector
+   * @param wAxis The axis to rotate about 
+   * @param angle The angle to rotate by */
+	SIMD_FORCE_INLINE btVector3 rotate(const btVector3 &wAxis, const btScalar angle) const;
+
+	/**@brief Return the angle between this and another vector
+   * @param v The other vector */
+	SIMD_FORCE_INLINE btScalar angle(const btVector3 &v) const {
+		btScalar s = btSqrt(length2() * v.length2());
+		btFullAssert(s != btScalar(0.0));
+		return btAcos(dot(v) / s);
+	}
+	/**@brief Return a vector will the absolute values of each element */
+	SIMD_FORCE_INLINE btVector3 absolute() const {
+		return btVector3(
+				btFabs(m_floats[0]),
+				btFabs(m_floats[1]),
+				btFabs(m_floats[2]));
+	}
+	/**@brief Return the cross product between this and another vector
+   * @param v The other vector */
+	SIMD_FORCE_INLINE btVector3 cross(const btVector3 &v) const {
+		return btVector3(
+				m_floats[1] * v.m_floats[2] - m_floats[2] * v.m_floats[1],
+				m_floats[2] * v.m_floats[0] - m_floats[0] * v.m_floats[2],
+				m_floats[0] * v.m_floats[1] - m_floats[1] * v.m_floats[0]);
+	}
+
+	SIMD_FORCE_INLINE btScalar triple(const btVector3 &v1, const btVector3 &v2) const {
+		return m_floats[0] * (v1.m_floats[1] * v2.m_floats[2] - v1.m_floats[2] * v2.m_floats[1]) + m_floats[1] * (v1.m_floats[2] * v2.m_floats[0] - v1.m_floats[0] * v2.m_floats[2]) + m_floats[2] * (v1.m_floats[0] * v2.m_floats[1] - v1.m_floats[1] * v2.m_floats[0]);
+	}
+
+	/**@brief Return the axis with the smallest value
+   * Note return values are 0,1,2 for x, y, or z */
+	SIMD_FORCE_INLINE int32_t minAxis() const {
+		return m_floats[0] < m_floats[1] ? (m_floats[0] < m_floats[2] ? 0 : 2) : (m_floats[1] < m_floats[2] ? 1 : 2);
+	}
+
+	/**@brief Return the axis with the largest value
+   * Note return values are 0,1,2 for x, y, or z */
+	SIMD_FORCE_INLINE int32_t maxAxis() const {
+		return m_floats[0] < m_floats[1] ? (m_floats[1] < m_floats[2] ? 2 : 1) : (m_floats[0] < m_floats[2] ? 2 : 0);
+	}
+
+	SIMD_FORCE_INLINE int32_t furthestAxis() const {
+		return absolute().minAxis();
+	}
+
+	SIMD_FORCE_INLINE int32_t closestAxis() const {
+		return absolute().maxAxis();
+	}
+
+	SIMD_FORCE_INLINE void setInterpolate3(const btVector3 &v0, const btVector3 &v1, btScalar rt) {
+		btScalar s = btScalar(1.0) - rt;
+		m_floats[0] = s * v0.m_floats[0] + rt * v1.m_floats[0];
+		m_floats[1] = s * v0.m_floats[1] + rt * v1.m_floats[1];
+		m_floats[2] = s * v0.m_floats[2] + rt * v1.m_floats[2];
+		//don't do the unused w component
+		//		m_co[3] = s * v0[3] + rt * v1[3];
+	}
+
+	/**@brief Return the linear interpolation between this and another vector
+   * @param v The other vector 
+   * @param t The ration of this to v (t = 0 => return this, t=1 => return other) */
+	SIMD_FORCE_INLINE btVector3 lerp(const btVector3 &v, const btScalar &t) const {
+		return btVector3(m_floats[0] + (v.m_floats[0] - m_floats[0]) * t,
+				m_floats[1] + (v.m_floats[1] - m_floats[1]) * t,
+				m_floats[2] + (v.m_floats[2] - m_floats[2]) * t);
+	}
+
+	/**@brief Elementwise multiply this vector by the other
+   * @param v The other vector */
+	SIMD_FORCE_INLINE btVector3 &operator*=(const btVector3 &v) {
+		m_floats[0] *= v.m_floats[0];
+		m_floats[1] *= v.m_floats[1];
+		m_floats[2] *= v.m_floats[2];
+		return *this;
+	}
+
+	/**@brief Return the x value */
+	SIMD_FORCE_INLINE const btScalar &getX() const { return m_floats[0]; }
+	/**@brief Return the y value */
+	SIMD_FORCE_INLINE const btScalar &getY() const { return m_floats[1]; }
+	/**@brief Return the z value */
+	SIMD_FORCE_INLINE const btScalar &getZ() const { return m_floats[2]; }
+	/**@brief Set the x value */
+	SIMD_FORCE_INLINE void setX(btScalar x) { m_floats[0] = x; };
+	/**@brief Set the y value */
+	SIMD_FORCE_INLINE void setY(btScalar y) { m_floats[1] = y; };
+	/**@brief Set the z value */
+	SIMD_FORCE_INLINE void setZ(btScalar z) { m_floats[2] = z; };
+	/**@brief Set the w value */
+	SIMD_FORCE_INLINE void setW(btScalar w) { m_floats[3] = w; };
+	/**@brief Return the x value */
+	SIMD_FORCE_INLINE const btScalar &x() const { return m_floats[0]; }
+	/**@brief Return the y value */
+	SIMD_FORCE_INLINE const btScalar &y() const { return m_floats[1]; }
+	/**@brief Return the z value */
+	SIMD_FORCE_INLINE const btScalar &z() const { return m_floats[2]; }
+	/**@brief Return the w value */
+	SIMD_FORCE_INLINE const btScalar &w() const { return m_floats[3]; }
+
+	//SIMD_FORCE_INLINE btScalar&       operator[](int32_t i)       { return (&m_floats[0])[i];	}
+	//SIMD_FORCE_INLINE const btScalar& operator[](int32_t i) const { return (&m_floats[0])[i]; }
+	///operator btScalar*() replaces operator[], using implicit conversion. We added operator != and operator == to avoid pointer comparisons.
+	SIMD_FORCE_INLINE operator btScalar *() { return &m_floats[0]; }
+	SIMD_FORCE_INLINE operator const btScalar *() const { return &m_floats[0]; }
+
+	SIMD_FORCE_INLINE bool operator==(const btVector3 &other) const {
+		return ((m_floats[3] == other.m_floats[3]) && (m_floats[2] == other.m_floats[2]) && (m_floats[1] == other.m_floats[1]) && (m_floats[0] == other.m_floats[0]));
+	}
+
+	SIMD_FORCE_INLINE bool operator!=(const btVector3 &other) const {
+		return !(*this == other);
+	}
+
+	/**@brief Set each element to the max of the current values and the values of another btVector3
+   * @param other The other btVector3 to compare with 
+   */
+	SIMD_FORCE_INLINE void setMax(const btVector3 &other) {
+		btSetMax(m_floats[0], other.m_floats[0]);
+		btSetMax(m_floats[1], other.m_floats[1]);
+		btSetMax(m_floats[2], other.m_floats[2]);
+		btSetMax(m_floats[3], other.w());
+	}
+	/**@brief Set each element to the min of the current values and the values of another btVector3
+   * @param other The other btVector3 to compare with 
+   */
+	SIMD_FORCE_INLINE void setMin(const btVector3 &other) {
+		btSetMin(m_floats[0], other.m_floats[0]);
+		btSetMin(m_floats[1], other.m_floats[1]);
+		btSetMin(m_floats[2], other.m_floats[2]);
+		btSetMin(m_floats[3], other.w());
+	}
+
+	SIMD_FORCE_INLINE void setValue(const btScalar &x, const btScalar &y, const btScalar &z) {
+		m_floats[0] = x;
+		m_floats[1] = y;
+		m_floats[2] = z;
+		m_floats[3] = btScalar(0.);
+	}
+
+	void getSkewSymmetricMatrix(btVector3 * v0, btVector3 * v1, btVector3 * v2) const {
+		v0->setValue(0., -z(), y());
+		v1->setValue(z(), 0., -x());
+		v2->setValue(-y(), x(), 0.);
+	}
+
+	void setZero() {
+		setValue(btScalar(0.), btScalar(0.), btScalar(0.));
+	}
+
+	SIMD_FORCE_INLINE bool isZero() const {
+		return m_floats[0] == btScalar(0) && m_floats[1] == btScalar(0) && m_floats[2] == btScalar(0);
+	}
+
+	SIMD_FORCE_INLINE bool fuzzyZero() const {
+		return length2() < SIMD_EPSILON;
+	}
+
+	SIMD_FORCE_INLINE void serialize(struct btVector3Data & dataOut) const;
+
+	SIMD_FORCE_INLINE void deSerialize(const struct btVector3Data &dataIn);
+
+	SIMD_FORCE_INLINE void serializeFloat(struct btVector3FloatData & dataOut) const;
+
+	SIMD_FORCE_INLINE void deSerializeFloat(const struct btVector3FloatData &dataIn);
+
+	SIMD_FORCE_INLINE void serializeDouble(struct btVector3DoubleData & dataOut) const;
+
+	SIMD_FORCE_INLINE void deSerializeDouble(const struct btVector3DoubleData &dataIn);
+};
+
+/**@brief Return the sum of two vectors (Point symantics)*/
+SIMD_FORCE_INLINE btVector3
+operator+(const btVector3 &v1, const btVector3 &v2) {
+	return btVector3(v1.m_floats[0] + v2.m_floats[0], v1.m_floats[1] + v2.m_floats[1], v1.m_floats[2] + v2.m_floats[2]);
+}
+
+/**@brief Return the elementwise product of two vectors */
+SIMD_FORCE_INLINE btVector3
+operator*(const btVector3 &v1, const btVector3 &v2) {
+	return btVector3(v1.m_floats[0] * v2.m_floats[0], v1.m_floats[1] * v2.m_floats[1], v1.m_floats[2] * v2.m_floats[2]);
+}
+
+/**@brief Return the difference between two vectors */
+SIMD_FORCE_INLINE btVector3
+operator-(const btVector3 &v1, const btVector3 &v2) {
+	return btVector3(v1.m_floats[0] - v2.m_floats[0], v1.m_floats[1] - v2.m_floats[1], v1.m_floats[2] - v2.m_floats[2]);
+}
+/**@brief Return the negative of the vector */
+SIMD_FORCE_INLINE btVector3
+operator-(const btVector3 &v) {
+	return btVector3(-v.m_floats[0], -v.m_floats[1], -v.m_floats[2]);
+}
+
+/**@brief Return the vector scaled by s */
+SIMD_FORCE_INLINE btVector3
+operator*(const btVector3 &v, const btScalar &s) {
+	return btVector3(v.m_floats[0] * s, v.m_floats[1] * s, v.m_floats[2] * s);
+}
+
+/**@brief Return the vector scaled by s */
+SIMD_FORCE_INLINE btVector3
+operator*(const btScalar &s, const btVector3 &v) {
+	return v * s;
+}
+
+/**@brief Return the vector inversely scaled by s */
+SIMD_FORCE_INLINE btVector3
+operator/(const btVector3 &v, const btScalar &s) {
+	btFullAssert(s != btScalar(0.0));
+	return v * (btScalar(1.0) / s);
+}
+
+/**@brief Return the vector inversely scaled by s */
+SIMD_FORCE_INLINE btVector3
+operator/(const btVector3 &v1, const btVector3 &v2) {
+	return btVector3(v1.m_floats[0] / v2.m_floats[0], v1.m_floats[1] / v2.m_floats[1], v1.m_floats[2] / v2.m_floats[2]);
+}
+
+/**@brief Return the dot product between two vectors */
+SIMD_FORCE_INLINE btScalar
+btDot(const btVector3 &v1, const btVector3 &v2) {
+	return v1.dot(v2);
+}
+
+/**@brief Return the distance squared between two vectors */
+SIMD_FORCE_INLINE btScalar
+btDistance2(const btVector3 &v1, const btVector3 &v2) {
+	return v1.distance2(v2);
+}
+
+/**@brief Return the distance between two vectors */
+SIMD_FORCE_INLINE btScalar
+btDistance(const btVector3 &v1, const btVector3 &v2) {
+	return v1.distance(v2);
+}
+
+/**@brief Return the angle between two vectors */
+SIMD_FORCE_INLINE btScalar
+btAngle(const btVector3 &v1, const btVector3 &v2) {
+	return v1.angle(v2);
+}
+
+/**@brief Return the cross product of two vectors */
+SIMD_FORCE_INLINE btVector3
+btCross(const btVector3 &v1, const btVector3 &v2) {
+	return v1.cross(v2);
+}
+
+SIMD_FORCE_INLINE btScalar
+btTriple(const btVector3 &v1, const btVector3 &v2, const btVector3 &v3) {
+	return v1.triple(v2, v3);
+}
+
+/**@brief Return the linear interpolation between two vectors
+ * @param v1 One vector 
+ * @param v2 The other vector 
+ * @param t The ration of this to v (t = 0 => return v1, t=1 => return v2) */
+SIMD_FORCE_INLINE btVector3
+lerp(const btVector3 &v1, const btVector3 &v2, const btScalar &t) {
+	return v1.lerp(v2, t);
+}
+
+SIMD_FORCE_INLINE btScalar btVector3::distance2(const btVector3 &v) const {
+	return (v - *this).length2();
+}
+
+SIMD_FORCE_INLINE btScalar btVector3::distance(const btVector3 &v) const {
+	return (v - *this).length();
+}
+
+SIMD_FORCE_INLINE btVector3 btVector3::normalized() const {
+	return *this / length();
+}
+
+SIMD_FORCE_INLINE btVector3 btVector3::rotate(const btVector3 &wAxis, const btScalar angle) const {
+	// wAxis must be a unit lenght vector
+
+	btVector3 o = wAxis * wAxis.dot(*this);
+	btVector3 x = *this - o;
+	btVector3 y;
+
+	y = wAxis.cross(*this);
+
+	return (o + x * btCos(angle) + y * btSin(angle));
+}
+
+class btVector4 : public btVector3 {
+public:
+	SIMD_FORCE_INLINE btVector4() {}
+
+	SIMD_FORCE_INLINE btVector4(const btScalar &x, const btScalar &y, const btScalar &z, const btScalar &w) :
+			btVector3(x, y, z) {
+		m_floats[3] = w;
+	}
+
+	SIMD_FORCE_INLINE btVector4 absolute4() const {
+		return btVector4(
+				btFabs(m_floats[0]),
+				btFabs(m_floats[1]),
+				btFabs(m_floats[2]),
+				btFabs(m_floats[3]));
+	}
+
+	btScalar getW() const { return m_floats[3]; }
+
+	SIMD_FORCE_INLINE int32_t maxAxis4() const {
+		int32_t maxIndex = -1;
+		btScalar maxVal = btScalar(-BT_LARGE_FLOAT);
+		if (m_floats[0] > maxVal) {
+			maxIndex = 0;
+			maxVal = m_floats[0];
+		}
+		if (m_floats[1] > maxVal) {
+			maxIndex = 1;
+			maxVal = m_floats[1];
+		}
+		if (m_floats[2] > maxVal) {
+			maxIndex = 2;
+			maxVal = m_floats[2];
+		}
+		if (m_floats[3] > maxVal) {
+			maxIndex = 3;
+		}
+		return maxIndex;
+	}
+
+	SIMD_FORCE_INLINE int32_t minAxis4() const {
+		int32_t minIndex = -1;
+		btScalar minVal = btScalar(BT_LARGE_FLOAT);
+		if (m_floats[0] < minVal) {
+			minIndex = 0;
+			minVal = m_floats[0];
+		}
+		if (m_floats[1] < minVal) {
+			minIndex = 1;
+			minVal = m_floats[1];
+		}
+		if (m_floats[2] < minVal) {
+			minIndex = 2;
+			minVal = m_floats[2];
+		}
+		if (m_floats[3] < minVal) {
+			minIndex = 3;
+		}
+
+		return minIndex;
+	}
+
+	SIMD_FORCE_INLINE int32_t closestAxis4() const {
+		return absolute4().maxAxis4();
+	}
+
+	/**@brief Set x,y,z and zero w
+   * @param x Value of x
+   * @param y Value of y
+   * @param z Value of z
+   */
+
+	/*		void getValue(btScalar *m) const
+		{
+			m[0] = m_floats[0];
+			m[1] = m_floats[1];
+			m[2] =m_floats[2];
+		}
+*/
+	/**@brief Set the values
+   * @param x Value of x
+   * @param y Value of y
+   * @param z Value of z
+   * @param w Value of w
+   */
+	SIMD_FORCE_INLINE void setValue(const btScalar &x, const btScalar &y, const btScalar &z, const btScalar &w) {
+		m_floats[0] = x;
+		m_floats[1] = y;
+		m_floats[2] = z;
+		m_floats[3] = w;
+	}
+};
+
+///btSwapVector3Endian swaps vector endianness, useful for network and cross-platform serialization
+SIMD_FORCE_INLINE void btSwapScalarEndian(const btScalar &sourceVal, btScalar &destVal) {
+#ifdef BT_USE_DOUBLE_PRECISION
+	unsigned char *dest = (unsigned char *)&destVal;
+	unsigned char *src = (unsigned char *)&sourceVal;
+	dest[0] = src[7];
+	dest[1] = src[6];
+	dest[2] = src[5];
+	dest[3] = src[4];
+	dest[4] = src[3];
+	dest[5] = src[2];
+	dest[6] = src[1];
+	dest[7] = src[0];
+#else
+	unsigned char *dest = (unsigned char *)&destVal;
+	unsigned char *src = (unsigned char *)&sourceVal;
+	dest[0] = src[3];
+	dest[1] = src[2];
+	dest[2] = src[1];
+	dest[3] = src[0];
+#endif //BT_USE_DOUBLE_PRECISION
+}
+///btSwapVector3Endian swaps vector endianness, useful for network and cross-platform serialization
+SIMD_FORCE_INLINE void btSwapVector3Endian(const btVector3 &sourceVec, btVector3 &destVec) {
+	for (int32_t i = 0; i < 4; i++) {
+		btSwapScalarEndian(sourceVec[i], destVec[i]);
+	}
+}
+
+///btUnSwapVector3Endian swaps vector endianness, useful for network and cross-platform serialization
+SIMD_FORCE_INLINE void btUnSwapVector3Endian(btVector3 &vector) {
+
+	btVector3 swappedVec;
+	for (int32_t i = 0; i < 4; i++) {
+		btSwapScalarEndian(vector[i], swappedVec[i]);
+	}
+	vector = swappedVec;
+}
+
+template <class T>
+SIMD_FORCE_INLINE void btPlaneSpace1(const T &n, T &p, T &q) {
+	if (btFabs(n[2]) > SIMDSQRT12) {
+		// choose p in y-z plane
+		btScalar a = n[1] * n[1] + n[2] * n[2];
+		btScalar k = btRecipSqrt(a);
+		p[0] = 0;
+		p[1] = -n[2] * k;
+		p[2] = n[1] * k;
+		// set q = n x p
+		q[0] = a * k;
+		q[1] = -n[0] * p[2];
+		q[2] = n[0] * p[1];
+	} else {
+		// choose p in x-y plane
+		btScalar a = n[0] * n[0] + n[1] * n[1];
+		btScalar k = btRecipSqrt(a);
+		p[0] = -n[1] * k;
+		p[1] = n[0] * k;
+		p[2] = 0;
+		// set q = n x p
+		q[0] = -n[2] * p[1];
+		q[1] = n[2] * p[0];
+		q[2] = a * k;
+	}
+}
+
+struct btVector3FloatData {
+	float m_floats[4];
+};
+
+struct btVector3DoubleData {
+	double m_floats[4];
+};
+
+SIMD_FORCE_INLINE void btVector3::serializeFloat(struct btVector3FloatData &dataOut) const {
+	///could also do a memcpy, check if it is worth it
+	for (int32_t i = 0; i < 4; i++)
+		dataOut.m_floats[i] = float(m_floats[i]);
+}
+
+SIMD_FORCE_INLINE void btVector3::deSerializeFloat(const struct btVector3FloatData &dataIn) {
+	for (int32_t i = 0; i < 4; i++)
+		m_floats[i] = btScalar(dataIn.m_floats[i]);
+}
+
+SIMD_FORCE_INLINE void btVector3::serializeDouble(struct btVector3DoubleData &dataOut) const {
+	///could also do a memcpy, check if it is worth it
+	for (int32_t i = 0; i < 4; i++)
+		dataOut.m_floats[i] = double(m_floats[i]);
+}
+
+SIMD_FORCE_INLINE void btVector3::deSerializeDouble(const struct btVector3DoubleData &dataIn) {
+	for (int32_t i = 0; i < 4; i++)
+		m_floats[i] = btScalar(dataIn.m_floats[i]);
+}
+
+SIMD_FORCE_INLINE void btVector3::serialize(struct btVector3Data &dataOut) const {
+	///could also do a memcpy, check if it is worth it
+	for (int32_t i = 0; i < 4; i++)
+		dataOut.m_floats[i] = m_floats[i];
+}
+
+SIMD_FORCE_INLINE void btVector3::deSerialize(const struct btVector3Data &dataIn) {
+	for (int32_t i = 0; i < 4; i++)
+		m_floats[i] = dataIn.m_floats[i];
+}
+
+//GODOT ADDITION
+}; // namespace VHACD
+//
+
+#endif //BT_VECTOR3_H
diff --git a/thirdparty/vhacd/inc/vhacdCircularList.h b/thirdparty/vhacd/inc/vhacdCircularList.h
new file mode 100644
index 0000000000..0f5ddf9ecf
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdCircularList.h
@@ -0,0 +1,79 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_CIRCULAR_LIST_H
+#define VHACD_CIRCULAR_LIST_H
+#include <stdlib.h>
+namespace VHACD {
+//!    CircularListElement class.
+template <typename T>
+class CircularListElement {
+public:
+    T& GetData() { return m_data; }
+    const T& GetData() const { return m_data; }
+    CircularListElement<T>*& GetNext() { return m_next; }
+    CircularListElement<T>*& GetPrev() { return m_prev; }
+    const CircularListElement<T>*& GetNext() const { return m_next; }
+    const CircularListElement<T>*& GetPrev() const { return m_prev; }
+    //!    Constructor
+    CircularListElement(const T& data) { m_data = data; }
+    CircularListElement(void) {}
+    //! Destructor
+    ~CircularListElement(void) {}
+private:
+    T m_data;
+    CircularListElement<T>* m_next;
+    CircularListElement<T>* m_prev;
+
+    CircularListElement(const CircularListElement& rhs);
+};
+//!    CircularList class.
+template <typename T>
+class CircularList {
+public:
+    CircularListElement<T>*& GetHead() { return m_head; }
+    const CircularListElement<T>* GetHead() const { return m_head; }
+    bool IsEmpty() const { return (m_size == 0); }
+    size_t GetSize() const { return m_size; }
+    const T& GetData() const { return m_head->GetData(); }
+    T& GetData() { return m_head->GetData(); }
+    bool Delete();
+    bool Delete(CircularListElement<T>* element);
+    CircularListElement<T>* Add(const T* data = 0);
+    CircularListElement<T>* Add(const T& data);
+    bool Next();
+    bool Prev();
+    void Clear()
+    {
+        while (Delete())
+            ;
+    };
+    const CircularList& operator=(const CircularList& rhs);
+    //!    Constructor
+    CircularList()
+    {
+        m_head = 0;
+        m_size = 0;
+    }
+    CircularList(const CircularList& rhs);
+    //! Destructor
+    ~CircularList(void) { Clear(); };
+private:
+    CircularListElement<T>* m_head; //!< a pointer to the head of the circular list
+    size_t m_size; //!< number of element in the circular list
+};
+}
+#include "vhacdCircularList.inl"
+#endif // VHACD_CIRCULAR_LIST_H
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdCircularList.inl b/thirdparty/vhacd/inc/vhacdCircularList.inl
new file mode 100644
index 0000000000..2be5180524
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdCircularList.inl
@@ -0,0 +1,161 @@
+#pragma once
+#ifndef HACD_CIRCULAR_LIST_INL
+#define HACD_CIRCULAR_LIST_INL
+namespace VHACD
+{
+    template < typename T > 
+    inline bool CircularList<T>::Delete(CircularListElement<T> * element)
+    {
+        if (!element)
+        {
+            return false;
+        }
+        if (m_size > 1)
+        {
+            CircularListElement<T> * next = element->GetNext();
+            CircularListElement<T> * prev = element->GetPrev();
+            delete element;
+            m_size--;
+            if (element == m_head)
+            {
+                m_head = next;
+            }
+            next->GetPrev() = prev;
+            prev->GetNext() = next;
+            return true;
+        }
+        else if (m_size == 1)
+        {
+            delete m_head;
+            m_size--;
+            m_head = 0;
+            return true;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    
+    template < typename T > 
+    inline bool CircularList<T>::Delete()
+    {
+        if (m_size > 1)
+        {
+            CircularListElement<T> * next = m_head->GetNext();
+            CircularListElement<T> * prev = m_head->GetPrev();
+            delete m_head;
+            m_size--;
+            m_head = next;
+            next->GetPrev() = prev;
+            prev->GetNext() = next;
+            return true;
+        }
+        else if (m_size == 1)
+        {
+            delete m_head;
+            m_size--;
+            m_head = 0;
+            return true;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    template < typename T > 
+    inline CircularListElement<T> * CircularList<T>::Add(const T * data)
+    {
+        if (m_size == 0)
+        {
+            if (data)
+            {
+                m_head = new CircularListElement<T>(*data);
+            }
+            else
+            {
+                m_head = new CircularListElement<T>();
+            }
+            m_head->GetNext() = m_head->GetPrev() = m_head;
+        }
+        else
+        {
+            CircularListElement<T> * next = m_head->GetNext();
+            CircularListElement<T> * element = m_head;
+            if (data)
+            {
+                m_head = new CircularListElement<T>(*data);
+            }
+            else
+            {
+                m_head = new CircularListElement<T>();
+            }
+            m_head->GetNext() = next;
+            m_head->GetPrev() = element;
+            element->GetNext() = m_head;
+            next->GetPrev() = m_head;
+        }
+        m_size++;
+        return m_head;
+    }
+    template < typename T > 
+    inline CircularListElement<T> * CircularList<T>::Add(const T & data)
+    {
+        const T * pData = &data;
+        return Add(pData);
+    }
+    template < typename T > 
+    inline bool CircularList<T>::Next()
+    {
+        if (m_size == 0)
+        {
+            return false;
+        }
+        m_head = m_head->GetNext();
+        return true;
+    }
+    template < typename T > 
+    inline bool CircularList<T>::Prev()
+    {
+        if (m_size == 0)
+        {
+            return false;
+        }
+        m_head = m_head->GetPrev();
+        return true;
+    }
+    template < typename T > 
+    inline CircularList<T>::CircularList(const CircularList& rhs)
+    {
+        if (rhs.m_size > 0)
+        {
+            CircularListElement<T> * current = rhs.m_head;
+            do
+            {
+                current = current->GetNext();
+                Add(current->GetData());
+            }
+            while ( current != rhs.m_head );
+        }
+    }
+    template < typename T > 
+    inline const CircularList<T>& CircularList<T>::operator=(const CircularList& rhs)
+    {
+        if (&rhs != this)
+        {
+            Clear();
+            if (rhs.m_size > 0)
+            {
+                CircularListElement<T> * current = rhs.m_head;
+                do
+                {
+                    current = current->GetNext();
+                    Add(current->GetData());
+                }
+                while ( current != rhs.m_head );
+            }
+        }
+        return (*this);
+    }
+}
+#endif
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdICHull.h b/thirdparty/vhacd/inc/vhacdICHull.h
new file mode 100644
index 0000000000..132bdcfb3e
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdICHull.h
@@ -0,0 +1,98 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_ICHULL_H
+#define VHACD_ICHULL_H
+#include "vhacdManifoldMesh.h"
+#include "vhacdVector.h"
+
+namespace VHACD {
+//!    Incremental Convex Hull algorithm (cf. http://cs.smith.edu/~orourke/books/ftp.html ).
+enum ICHullError {
+    ICHullErrorOK = 0,
+    ICHullErrorCoplanarPoints,
+    ICHullErrorNoVolume,
+    ICHullErrorInconsistent,
+    ICHullErrorNotEnoughPoints
+};
+class ICHull {
+public:
+    static const double sc_eps;
+    //!
+    bool IsFlat() { return m_isFlat; }
+    //! Returns the computed mesh
+    TMMesh& GetMesh() { return m_mesh; }
+    //!    Add one point to the convex-hull
+    bool AddPoint(const Vec3<double>& point) { return AddPoints(&point, 1); }
+    //!    Add one point to the convex-hull
+    bool AddPoint(const Vec3<double>& point, int32_t id);
+    //!    Add points to the convex-hull
+    bool AddPoints(const Vec3<double>* points, size_t nPoints);
+    //!
+    ICHullError Process();
+    //!
+    ICHullError Process(const uint32_t nPointsCH, const double minVolume = 0.0);
+    //!
+    bool IsInside(const Vec3<double>& pt0, const double eps = 0.0);
+    //!
+    const ICHull& operator=(ICHull& rhs);
+
+    //!    Constructor
+    ICHull();
+    //! Destructor
+    ~ICHull(void){};
+
+private:
+    //!    DoubleTriangle builds the initial double triangle.  It first finds 3 noncollinear points and makes two faces out of them, in opposite order. It then finds a fourth point that is not coplanar with that face.  The vertices are stored in the face structure in counterclockwise order so that the volume between the face and the point is negative. Lastly, the 3 newfaces to the fourth point are constructed and the data structures are cleaned up.
+    ICHullError DoubleTriangle();
+    //!    MakeFace creates a new face structure from three vertices (in ccw order).  It returns a pointer to the face.
+    CircularListElement<TMMTriangle>* MakeFace(CircularListElement<TMMVertex>* v0,
+        CircularListElement<TMMVertex>* v1,
+        CircularListElement<TMMVertex>* v2,
+        CircularListElement<TMMTriangle>* fold);
+    //!
+    CircularListElement<TMMTriangle>* MakeConeFace(CircularListElement<TMMEdge>* e, CircularListElement<TMMVertex>* v);
+    //!
+    bool ProcessPoint();
+    //!
+    bool ComputePointVolume(double& totalVolume, bool markVisibleFaces);
+    //!
+    bool FindMaxVolumePoint(const double minVolume = 0.0);
+    //!
+    bool CleanEdges();
+    //!
+    bool CleanVertices(uint32_t& addedPoints);
+    //!
+    bool CleanTriangles();
+    //!
+    bool CleanUp(uint32_t& addedPoints);
+    //!
+    bool MakeCCW(CircularListElement<TMMTriangle>* f,
+        CircularListElement<TMMEdge>* e,
+        CircularListElement<TMMVertex>* v);
+    void Clear();
+
+private:
+    static const int32_t sc_dummyIndex;
+    TMMesh m_mesh;
+    SArray<CircularListElement<TMMEdge>*> m_edgesToDelete;
+    SArray<CircularListElement<TMMEdge>*> m_edgesToUpdate;
+    SArray<CircularListElement<TMMTriangle>*> m_trianglesToDelete;
+    Vec3<double> m_normal;
+    bool m_isFlat;
+    ICHull(const ICHull& rhs);
+};
+}
+#endif // VHACD_ICHULL_H
diff --git a/thirdparty/vhacd/inc/vhacdManifoldMesh.h b/thirdparty/vhacd/inc/vhacdManifoldMesh.h
new file mode 100644
index 0000000000..a48f53c5c5
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdManifoldMesh.h
@@ -0,0 +1,142 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+All rights reserved.
+
+
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*/
+#pragma once
+#ifndef VHACD_MANIFOLD_MESH_H
+#define VHACD_MANIFOLD_MESH_H
+#include "vhacdCircularList.h"
+#include "vhacdSArray.h"
+#include "vhacdVector.h"
+namespace VHACD {
+class TMMTriangle;
+class TMMEdge;
+class TMMesh;
+class ICHull;
+
+//!    Vertex data structure used in a triangular manifold mesh (TMM).
+class TMMVertex {
+public:
+    void Initialize();
+    TMMVertex(void);
+    ~TMMVertex(void);
+
+private:
+    Vec3<double> m_pos;
+    int32_t m_name;
+    size_t m_id;
+    CircularListElement<TMMEdge>* m_duplicate; // pointer to incident cone edge (or NULL)
+    bool m_onHull;
+    bool m_tag;
+    TMMVertex(const TMMVertex& rhs);
+    friend class ICHull;
+    friend class TMMesh;
+    friend class TMMTriangle;
+    friend class TMMEdge;
+};
+
+//!    Edge data structure used in a triangular manifold mesh (TMM).
+class TMMEdge {
+public:
+    void Initialize();
+    TMMEdge(void);
+    ~TMMEdge(void);
+
+private:
+    size_t m_id;
+    CircularListElement<TMMTriangle>* m_triangles[2];
+    CircularListElement<TMMVertex>* m_vertices[2];
+    CircularListElement<TMMTriangle>* m_newFace;
+    TMMEdge(const TMMEdge& rhs);
+    friend class ICHull;
+    friend class TMMTriangle;
+    friend class TMMVertex;
+    friend class TMMesh;
+};
+
+//!    Triangle data structure used in a triangular manifold mesh (TMM).
+class TMMTriangle {
+public:
+    void Initialize();
+    TMMTriangle(void);
+    ~TMMTriangle(void);
+
+private:
+    size_t m_id;
+    CircularListElement<TMMEdge>* m_edges[3];
+    CircularListElement<TMMVertex>* m_vertices[3];
+    bool m_visible;
+
+    TMMTriangle(const TMMTriangle& rhs);
+    friend class ICHull;
+    friend class TMMesh;
+    friend class TMMVertex;
+    friend class TMMEdge;
+};
+//!    triangular manifold mesh data structure.
+class TMMesh {
+public:
+    //! Returns the number of vertices>
+    inline size_t GetNVertices() const { return m_vertices.GetSize(); }
+    //! Returns the number of edges
+    inline size_t GetNEdges() const { return m_edges.GetSize(); }
+    //! Returns the number of triangles
+    inline size_t GetNTriangles() const { return m_triangles.GetSize(); }
+    //! Returns the vertices circular list
+    inline const CircularList<TMMVertex>& GetVertices() const { return m_vertices; }
+    //! Returns the edges circular list
+    inline const CircularList<TMMEdge>& GetEdges() const { return m_edges; }
+    //! Returns the triangles circular list
+    inline const CircularList<TMMTriangle>& GetTriangles() const { return m_triangles; }
+    //! Returns the vertices circular list
+    inline CircularList<TMMVertex>& GetVertices() { return m_vertices; }
+    //! Returns the edges circular list
+    inline CircularList<TMMEdge>& GetEdges() { return m_edges; }
+    //! Returns the triangles circular list
+    inline CircularList<TMMTriangle>& GetTriangles() { return m_triangles; }
+    //! Add vertex to the mesh
+    CircularListElement<TMMVertex>* AddVertex() { return m_vertices.Add(); }
+    //! Add vertex to the mesh
+    CircularListElement<TMMEdge>* AddEdge() { return m_edges.Add(); }
+    //! Add vertex to the mesh
+    CircularListElement<TMMTriangle>* AddTriangle() { return m_triangles.Add(); }
+    //! Print mesh information
+    void Print();
+    //!
+    void GetIFS(Vec3<double>* const points, Vec3<int32_t>* const triangles);
+    //!
+    void Clear();
+    //!
+    void Copy(TMMesh& mesh);
+    //!
+    bool CheckConsistancy();
+    //!
+    bool Normalize();
+    //!
+    bool Denormalize();
+    //!    Constructor
+    TMMesh();
+    //! Destructor
+    virtual ~TMMesh(void);
+
+private:
+    CircularList<TMMVertex> m_vertices;
+    CircularList<TMMEdge> m_edges;
+    CircularList<TMMTriangle> m_triangles;
+
+    // not defined
+    TMMesh(const TMMesh& rhs);
+    friend class ICHull;
+};
+}
+#endif // VHACD_MANIFOLD_MESH_H
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdMesh.h b/thirdparty/vhacd/inc/vhacdMesh.h
new file mode 100644
index 0000000000..a282e2b4fb
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdMesh.h
@@ -0,0 +1,130 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_MESH_H
+#define VHACD_MESH_H
+#include "vhacdSArray.h"
+#include "vhacdVector.h"
+
+#define VHACD_DEBUG_MESH
+
+namespace VHACD {
+enum AXIS {
+    AXIS_X = 0,
+    AXIS_Y = 1,
+    AXIS_Z = 2
+};
+struct Plane {
+    double m_a;
+    double m_b;
+    double m_c;
+    double m_d;
+    AXIS m_axis;
+    short m_index;
+};
+#ifdef VHACD_DEBUG_MESH
+struct Material {
+
+    Vec3<double> m_diffuseColor;
+    double m_ambientIntensity;
+    Vec3<double> m_specularColor;
+    Vec3<double> m_emissiveColor;
+    double m_shininess;
+    double m_transparency;
+    Material(void)
+    {
+        m_diffuseColor.X() = 0.5;
+        m_diffuseColor.Y() = 0.5;
+        m_diffuseColor.Z() = 0.5;
+        m_specularColor.X() = 0.5;
+        m_specularColor.Y() = 0.5;
+        m_specularColor.Z() = 0.5;
+        m_ambientIntensity = 0.4;
+        m_emissiveColor.X() = 0.0;
+        m_emissiveColor.Y() = 0.0;
+        m_emissiveColor.Z() = 0.0;
+        m_shininess = 0.4;
+        m_transparency = 0.0;
+    };
+};
+#endif // VHACD_DEBUG_MESH
+
+//! Triangular mesh data structure
+class Mesh {
+public:
+    void AddPoint(const Vec3<double>& pt) { m_points.PushBack(pt); };
+    void SetPoint(size_t index, const Vec3<double>& pt) { m_points[index] = pt; };
+    const Vec3<double>& GetPoint(size_t index) const { return m_points[index]; };
+    Vec3<double>& GetPoint(size_t index) { return m_points[index]; };
+    size_t GetNPoints() const { return m_points.Size(); };
+    double* GetPoints() { return (double*)m_points.Data(); } // ugly
+    const double* const GetPoints() const { return (double*)m_points.Data(); } // ugly
+    const Vec3<double>* const GetPointsBuffer() const { return m_points.Data(); } //
+    Vec3<double>* const GetPointsBuffer() { return m_points.Data(); } //
+    void AddTriangle(const Vec3<int32_t>& tri) { m_triangles.PushBack(tri); };
+    void SetTriangle(size_t index, const Vec3<int32_t>& tri) { m_triangles[index] = tri; };
+    const Vec3<int32_t>& GetTriangle(size_t index) const { return m_triangles[index]; };
+    Vec3<int32_t>& GetTriangle(size_t index) { return m_triangles[index]; };
+    size_t GetNTriangles() const { return m_triangles.Size(); };
+    int32_t* GetTriangles() { return (int32_t*)m_triangles.Data(); } // ugly
+    const int32_t* const GetTriangles() const { return (int32_t*)m_triangles.Data(); } // ugly
+    const Vec3<int32_t>* const GetTrianglesBuffer() const { return m_triangles.Data(); }
+    Vec3<int32_t>* const GetTrianglesBuffer() { return m_triangles.Data(); }
+    const Vec3<double>& GetCenter() const { return m_center; }
+    const Vec3<double>& GetMinBB() const { return m_minBB; }
+    const Vec3<double>& GetMaxBB() const { return m_maxBB; }
+    void ClearPoints() { m_points.Clear(); }
+    void ClearTriangles() { m_triangles.Clear(); }
+    void Clear()
+    {
+        ClearPoints();
+        ClearTriangles();
+    }
+    void ResizePoints(size_t nPts) { m_points.Resize(nPts); }
+    void ResizeTriangles(size_t nTri) { m_triangles.Resize(nTri); }
+    void CopyPoints(SArray<Vec3<double> >& points) const { points = m_points; }
+    double GetDiagBB() const { return m_diag; }
+    double ComputeVolume() const;
+    void ComputeConvexHull(const double* const pts,
+        const size_t nPts);
+    void Clip(const Plane& plane,
+        SArray<Vec3<double> >& positivePart,
+        SArray<Vec3<double> >& negativePart) const;
+    bool IsInside(const Vec3<double>& pt) const;
+    double ComputeDiagBB();
+	Vec3<double> &ComputeCenter(void);
+
+#ifdef VHACD_DEBUG_MESH
+    bool LoadOFF(const std::string& fileName, bool invert);
+    bool SaveVRML2(const std::string& fileName) const;
+    bool SaveVRML2(std::ofstream& fout, const Material& material) const;
+    bool SaveOFF(const std::string& fileName) const;
+#endif // VHACD_DEBUG_MESH
+
+    //! Constructor.
+    Mesh();
+    //! Destructor.
+    ~Mesh(void);
+
+private:
+    SArray<Vec3<double> > m_points;
+    SArray<Vec3<int32_t> > m_triangles;
+    Vec3<double> m_minBB;
+    Vec3<double> m_maxBB;
+    Vec3<double> m_center;
+    double m_diag;
+};
+}
+#endif
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdMutex.h b/thirdparty/vhacd/inc/vhacdMutex.h
new file mode 100644
index 0000000000..6b09259200
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdMutex.h
@@ -0,0 +1,148 @@
+/*!
+**
+** Copyright (c) 2009 by John W. Ratcliff mailto:jratcliffscarab@gmail.com
+**
+** Portions of this source has been released with the PhysXViewer application, as well as
+** Rocket, CreateDynamics, ODF, and as a number of sample code snippets.
+**
+** If you find this code useful or you are feeling particularily generous I would
+** ask that you please go to http://www.amillionpixels.us and make a donation
+** to Troy DeMolay.
+**
+** DeMolay is a youth group for young men between the ages of 12 and 21.
+** It teaches strong moral principles, as well as leadership skills and
+** public speaking.  The donations page uses the 'pay for pixels' paradigm
+** where, in this case, a pixel is only a single penny.  Donations can be
+** made for as small as $4 or as high as a $100 block.  Each person who donates
+** will get a link to their own site as well as acknowledgement on the
+** donations blog located here http://www.amillionpixels.blogspot.com/
+**
+** If you wish to contact me you can use the following methods:
+**
+** Skype ID: jratcliff63367
+** Yahoo: jratcliff63367
+** AOL: jratcliff1961
+** email: jratcliffscarab@gmail.com
+**
+**
+** The MIT license:
+**
+** Permission is hereby granted, free of charge, to any person obtaining a copy
+** of this software and associated documentation files (the "Software"), to deal
+** in the Software without restriction, including without limitation the rights
+** to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+** copies of the Software, and to permit persons to whom the Software is furnished
+** to do so, subject to the following conditions:
+**
+** The above copyright notice and this permission notice shall be included in all
+** copies or substantial portions of the Software.
+
+** THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+** IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+** FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+** AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+** WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+** CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+*/
+
+#pragma once
+#ifndef VHACD_MUTEX_H
+#define VHACD_MUTEX_H
+
+#if defined(WIN32)
+
+#ifndef _WIN32_WINNT
+#define _WIN32_WINNT 0x400
+#endif
+#include <windows.h>
+#pragma comment(lib, "winmm.lib")
+#endif
+
+#if defined(__linux__)
+//#include <sys/time.h>
+#include <errno.h>
+#include <time.h>
+#include <unistd.h>
+#define __stdcall
+#endif
+
+#if defined(__APPLE__) || defined(__linux__)
+#include <pthread.h>
+#endif
+
+#if defined(__APPLE__)
+#define PTHREAD_MUTEX_RECURSIVE_NP PTHREAD_MUTEX_RECURSIVE
+#endif
+
+#define VHACD_DEBUG
+
+//#define VHACD_NDEBUG
+#ifdef VHACD_NDEBUG
+#define VHACD_VERIFY(x) (x)
+#else
+#define VHACD_VERIFY(x) assert((x))
+#endif
+
+namespace VHACD {
+class Mutex {
+public:
+    Mutex(void)
+    {
+#if defined(WIN32) || defined(_XBOX)
+        InitializeCriticalSection(&m_mutex);
+#elif defined(__APPLE__) || defined(__linux__)
+        pthread_mutexattr_t mutexAttr; // Mutex Attribute
+        VHACD_VERIFY(pthread_mutexattr_init(&mutexAttr) == 0);
+        VHACD_VERIFY(pthread_mutexattr_settype(&mutexAttr, PTHREAD_MUTEX_RECURSIVE_NP) == 0);
+        VHACD_VERIFY(pthread_mutex_init(&m_mutex, &mutexAttr) == 0);
+        VHACD_VERIFY(pthread_mutexattr_destroy(&mutexAttr) == 0);
+#endif
+    }
+    ~Mutex(void)
+    {
+#if defined(WIN32) || defined(_XBOX)
+        DeleteCriticalSection(&m_mutex);
+#elif defined(__APPLE__) || defined(__linux__)
+        VHACD_VERIFY(pthread_mutex_destroy(&m_mutex) == 0);
+#endif
+    }
+    void Lock(void)
+    {
+#if defined(WIN32) || defined(_XBOX)
+        EnterCriticalSection(&m_mutex);
+#elif defined(__APPLE__) || defined(__linux__)
+        VHACD_VERIFY(pthread_mutex_lock(&m_mutex) == 0);
+#endif
+    }
+    bool TryLock(void)
+    {
+#if defined(WIN32) || defined(_XBOX)
+        bool bRet = false;
+        //assert(("TryEnterCriticalSection seems to not work on XP???", 0));
+        bRet = TryEnterCriticalSection(&m_mutex) ? true : false;
+        return bRet;
+#elif defined(__APPLE__) || defined(__linux__)
+        int32_t result = pthread_mutex_trylock(&m_mutex);
+        return (result == 0);
+#endif
+    }
+
+    void Unlock(void)
+    {
+#if defined(WIN32) || defined(_XBOX)
+        LeaveCriticalSection(&m_mutex);
+#elif defined(__APPLE__) || defined(__linux__)
+        VHACD_VERIFY(pthread_mutex_unlock(&m_mutex) == 0);
+#endif
+    }
+
+private:
+#if defined(WIN32) || defined(_XBOX)
+    CRITICAL_SECTION m_mutex;
+#elif defined(__APPLE__) || defined(__linux__)
+    pthread_mutex_t m_mutex;
+#endif
+};
+}
+#endif // VHACD_MUTEX_H
diff --git a/thirdparty/vhacd/inc/vhacdRaycastMesh.h b/thirdparty/vhacd/inc/vhacdRaycastMesh.h
new file mode 100644
index 0000000000..93bf299c07
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdRaycastMesh.h
@@ -0,0 +1,39 @@
+#ifndef RAYCAST_MESH_H
+
+#define RAYCAST_MESH_H
+
+#include <stdint.h>
+
+namespace VHACD
+{
+
+    // Very simple brute force raycast against a triangle mesh.  Tests every triangle; no hierachy.
+    // Does a deep copy, always does calculations with full double float precision
+    class RaycastMesh
+    {
+    public:
+        static RaycastMesh * createRaycastMesh(uint32_t vcount,		// The number of vertices in the source triangle mesh
+            const double *vertices,		// The array of vertex positions in the format x1,y1,z1..x2,y2,z2.. etc.
+            uint32_t tcount,		// The number of triangles in the source triangle mesh
+            const uint32_t *indices); // The triangle indices in the format of i1,i2,i3 ... i4,i5,i6, ...
+
+        static RaycastMesh * createRaycastMesh(uint32_t vcount,		// The number of vertices in the source triangle mesh
+            const float *vertices,		// The array of vertex positions in the format x1,y1,z1..x2,y2,z2.. etc.
+            uint32_t tcount,		// The number of triangles in the source triangle mesh
+            const uint32_t *indices); // The triangle indices in the format of i1,i2,i3 ... i4,i5,i6, ...
+
+
+        virtual bool raycast(const double *from,			// The starting point of the raycast
+            const double *to,				// The ending point of the raycast
+            const double *closestToPoint,	// The point to match the nearest hit location (can just be the 'from' location of no specific point)
+            double *hitLocation,			// The point where the ray hit nearest to the 'closestToPoint' location
+            double *hitDistance) = 0;		// The distance the ray traveled to the hit location
+
+        virtual void release(void) = 0;
+    protected:
+        virtual ~RaycastMesh(void) { };
+    };
+
+} // end of VHACD namespace
+
+#endif
diff --git a/thirdparty/vhacd/inc/vhacdSArray.h b/thirdparty/vhacd/inc/vhacdSArray.h
new file mode 100644
index 0000000000..4d84d1ae26
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdSArray.h
@@ -0,0 +1,158 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_SARRAY_H
+#define VHACD_SARRAY_H
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define SARRAY_DEFAULT_MIN_SIZE 16
+
+namespace VHACD {
+//!    SArray.
+template <typename T, size_t N = 64>
+class SArray {
+public:
+    T& operator[](size_t i)
+    {
+        T* const data = Data();
+        return data[i];
+    }
+    const T& operator[](size_t i) const
+    {
+        const T* const data = Data();
+        return data[i];
+    }
+    size_t Size() const
+    {
+        return m_size;
+    }
+    T* const Data()
+    {
+        return (m_maxSize == N) ? m_data0 : m_data;
+    }
+    const T* const Data() const
+    {
+        return (m_maxSize == N) ? m_data0 : m_data;
+    }
+    void Clear()
+    {
+        m_size = 0;
+        delete[] m_data;
+        m_data = 0;
+        m_maxSize = N;
+    }
+    void PopBack()
+    {
+        --m_size;
+    }
+    void Allocate(size_t size)
+    {
+        if (size > m_maxSize) {
+            T* temp = new T[size];
+            memcpy(temp, Data(), m_size * sizeof(T));
+            delete[] m_data;
+            m_data = temp;
+            m_maxSize = size;
+        }
+    }
+    void Resize(size_t size)
+    {
+        Allocate(size);
+        m_size = size;
+    }
+
+    void PushBack(const T& value)
+    {
+        if (m_size == m_maxSize) {
+            size_t maxSize = (m_maxSize << 1);
+            T* temp = new T[maxSize];
+            memcpy(temp, Data(), m_maxSize * sizeof(T));
+            delete[] m_data;
+            m_data = temp;
+            m_maxSize = maxSize;
+        }
+        T* const data = Data();
+        data[m_size++] = value;
+    }
+    bool Find(const T& value, size_t& pos)
+    {
+        T* const data = Data();
+        for (pos = 0; pos < m_size; ++pos)
+            if (value == data[pos])
+                return true;
+        return false;
+    }
+    bool Insert(const T& value)
+    {
+        size_t pos;
+        if (Find(value, pos))
+            return false;
+        PushBack(value);
+        return true;
+    }
+    bool Erase(const T& value)
+    {
+        size_t pos;
+        T* const data = Data();
+        if (Find(value, pos)) {
+            for (size_t j = pos + 1; j < m_size; ++j)
+                data[j - 1] = data[j];
+            --m_size;
+            return true;
+        }
+        return false;
+    }
+    void operator=(const SArray& rhs)
+    {
+        if (m_maxSize < rhs.m_size) {
+            delete[] m_data;
+            m_maxSize = rhs.m_maxSize;
+            m_data = new T[m_maxSize];
+        }
+        m_size = rhs.m_size;
+        memcpy(Data(), rhs.Data(), m_size * sizeof(T));
+    }
+    void Initialize()
+    {
+        m_data = 0;
+        m_size = 0;
+        m_maxSize = N;
+    }
+    SArray(const SArray& rhs)
+    {
+        m_data = 0;
+        m_size = 0;
+        m_maxSize = N;
+        *this = rhs;
+    }
+    SArray()
+    {
+        Initialize();
+    }
+    ~SArray()
+    {
+        delete[] m_data;
+    }
+
+private:
+    T m_data0[N];
+    T* m_data;
+    size_t m_size;
+    size_t m_maxSize;
+};
+}
+#endif
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdTimer.h b/thirdparty/vhacd/inc/vhacdTimer.h
new file mode 100644
index 0000000000..ba0e2c3364
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdTimer.h
@@ -0,0 +1,121 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_TIMER_H
+#define VHACD_TIMER_H
+
+#ifdef _WIN32
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN // Exclude rarely-used stuff from Windows headers
+#endif
+#include <windows.h>
+#elif __MACH__
+#include <mach/clock.h>
+#include <mach/mach.h>
+#else
+#include <sys/time.h>
+#include <time.h>
+#endif
+
+namespace VHACD {
+#ifdef _WIN32
+class Timer {
+public:
+    Timer(void)
+    {
+        m_start.QuadPart = 0;
+        m_stop.QuadPart = 0;
+        QueryPerformanceFrequency(&m_freq);
+    };
+    ~Timer(void){};
+    void Tic()
+    {
+        QueryPerformanceCounter(&m_start);
+    }
+    void Toc()
+    {
+        QueryPerformanceCounter(&m_stop);
+    }
+    double GetElapsedTime() // in ms
+    {
+        LARGE_INTEGER delta;
+        delta.QuadPart = m_stop.QuadPart - m_start.QuadPart;
+        return (1000.0 * delta.QuadPart) / (double)m_freq.QuadPart;
+    }
+
+private:
+    LARGE_INTEGER m_start;
+    LARGE_INTEGER m_stop;
+    LARGE_INTEGER m_freq;
+};
+
+#elif __MACH__
+class Timer {
+public:
+    Timer(void)
+    {
+        memset(this, 0, sizeof(Timer));
+        host_get_clock_service(mach_host_self(), CALENDAR_CLOCK, &m_cclock);
+    };
+    ~Timer(void)
+    {
+        mach_port_deallocate(mach_task_self(), m_cclock);
+    };
+    void Tic()
+    {
+        clock_get_time(m_cclock, &m_start);
+    }
+    void Toc()
+    {
+        clock_get_time(m_cclock, &m_stop);
+    }
+    double GetElapsedTime() // in ms
+    {
+        return 1000.0 * (m_stop.tv_sec - m_start.tv_sec + (1.0E-9) * (m_stop.tv_nsec - m_start.tv_nsec));
+    }
+
+private:
+    clock_serv_t m_cclock;
+    mach_timespec_t m_start;
+    mach_timespec_t m_stop;
+};
+#else
+class Timer {
+public:
+    Timer(void)
+    {
+        memset(this, 0, sizeof(Timer));
+    };
+    ~Timer(void){};
+    void Tic()
+    {
+        clock_gettime(CLOCK_REALTIME, &m_start);
+    }
+    void Toc()
+    {
+        clock_gettime(CLOCK_REALTIME, &m_stop);
+    }
+    double GetElapsedTime() // in ms
+    {
+        return 1000.0 * (m_stop.tv_sec - m_start.tv_sec + (1.0E-9) * (m_stop.tv_nsec - m_start.tv_nsec));
+    }
+
+private:
+    struct timespec m_start;
+    struct timespec m_stop;
+};
+#endif
+}
+#endif // VHACD_TIMER_H
diff --git a/thirdparty/vhacd/inc/vhacdVHACD.h b/thirdparty/vhacd/inc/vhacdVHACD.h
new file mode 100644
index 0000000000..d67b107642
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdVHACD.h
@@ -0,0 +1,371 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+All rights reserved.
+
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*/
+#pragma once
+#ifndef VHACD_VHACD_H
+#define VHACD_VHACD_H
+
+#ifdef OPENCL_FOUND
+#ifdef __MACH__
+#include <OpenCL/cl.h>
+#else
+#include <CL/cl.h>
+#endif
+#endif //OPENCL_FOUND
+
+#include "vhacdMutex.h"
+#include "vhacdVolume.h"
+#include "vhacdRaycastMesh.h"
+#include <vector>
+
+#define USE_THREAD 1
+#define OCL_MIN_NUM_PRIMITIVES 4096
+#define CH_APP_MIN_NUM_PRIMITIVES 64000
+namespace VHACD {
+class VHACD : public IVHACD {
+public:
+    //! Constructor.
+    VHACD()
+    {
+#if USE_THREAD == 1 && _OPENMP
+        m_ompNumProcessors = 2 * omp_get_num_procs();
+        omp_set_num_threads(m_ompNumProcessors);
+#else //USE_THREAD == 1 && _OPENMP
+        m_ompNumProcessors = 1;
+#endif //USE_THREAD == 1 && _OPENMP
+#ifdef CL_VERSION_1_1
+        m_oclWorkGroupSize = 0;
+        m_oclDevice = 0;
+        m_oclQueue = 0;
+        m_oclKernelComputePartialVolumes = 0;
+        m_oclKernelComputeSum = 0;
+#endif //CL_VERSION_1_1
+        Init();
+    }
+    //! Destructor.
+    ~VHACD(void) 
+    {
+    }
+    uint32_t GetNConvexHulls() const
+    {
+        return (uint32_t)m_convexHulls.Size();
+    }
+    void Cancel()
+    {
+        SetCancel(true);
+    }
+    void GetConvexHull(const uint32_t index, ConvexHull& ch) const
+    {
+        Mesh* mesh = m_convexHulls[index];
+        ch.m_nPoints = (uint32_t)mesh->GetNPoints();
+        ch.m_nTriangles = (uint32_t)mesh->GetNTriangles();
+        ch.m_points = mesh->GetPoints();
+        ch.m_triangles = (uint32_t *)mesh->GetTriangles();
+		ch.m_volume = mesh->ComputeVolume();
+		Vec3<double> &center = mesh->ComputeCenter();
+		ch.m_center[0] = center.X();
+		ch.m_center[1] = center.Y();
+		ch.m_center[2] = center.Z();
+    }
+    void Clean(void)
+    {
+        if (mRaycastMesh)
+        {
+            mRaycastMesh->release();
+            mRaycastMesh = nullptr;
+        }
+        delete m_volume;
+        delete m_pset;
+        size_t nCH = m_convexHulls.Size();
+        for (size_t p = 0; p < nCH; ++p) {
+            delete m_convexHulls[p];
+        }
+        m_convexHulls.Clear();
+        Init();
+    }
+    void Release(void)
+    {
+        delete this;
+    }
+    bool Compute(const float* const points,
+        const uint32_t nPoints,
+        const uint32_t* const triangles,
+        const uint32_t nTriangles,
+        const Parameters& params);
+    bool Compute(const double* const points,
+        const uint32_t nPoints,
+        const uint32_t* const triangles,
+        const uint32_t nTriangles,
+        const Parameters& params);
+    bool OCLInit(void* const oclDevice,
+        IUserLogger* const logger = 0);
+    bool OCLRelease(IUserLogger* const logger = 0);
+
+	virtual bool ComputeCenterOfMass(double centerOfMass[3]) const;
+
+private:
+    void SetCancel(bool cancel)
+    {
+        m_cancelMutex.Lock();
+        m_cancel = cancel;
+        m_cancelMutex.Unlock();
+    }
+    bool GetCancel()
+    {
+
+        m_cancelMutex.Lock();
+        bool cancel = m_cancel;
+        m_cancelMutex.Unlock();
+        return cancel;
+    }
+    void Update(const double stageProgress,
+        const double operationProgress,
+        const Parameters& params)
+    {
+        m_stageProgress = stageProgress;
+        m_operationProgress = operationProgress;
+        if (params.m_callback) {
+            params.m_callback->Update(m_overallProgress,
+                m_stageProgress,
+                m_operationProgress,
+                m_stage.c_str(),
+                m_operation.c_str());
+        }
+    }
+    void Init()
+    {
+		if (mRaycastMesh)
+		{
+			mRaycastMesh->release();
+			mRaycastMesh = nullptr;
+		}
+        memset(m_rot, 0, sizeof(double) * 9);
+        m_dim = 64;
+        m_volume = 0;
+        m_volumeCH0 = 0.0;
+        m_pset = 0;
+        m_overallProgress = 0.0;
+        m_stageProgress = 0.0;
+        m_operationProgress = 0.0;
+        m_stage = "";
+        m_operation = "";
+        m_barycenter[0] = m_barycenter[1] = m_barycenter[2] = 0.0;
+        m_rot[0][0] = m_rot[1][1] = m_rot[2][2] = 1.0;
+        SetCancel(false);
+    }
+    void ComputePrimitiveSet(const Parameters& params);
+    void ComputeACD(const Parameters& params);
+    void MergeConvexHulls(const Parameters& params);
+    void SimplifyConvexHull(Mesh* const ch, const size_t nvertices, const double minVolume);
+    void SimplifyConvexHulls(const Parameters& params);
+    void ComputeBestClippingPlane(const PrimitiveSet* inputPSet,
+        const double volume,
+        const SArray<Plane>& planes,
+        const Vec3<double>& preferredCuttingDirection,
+        const double w,
+        const double alpha,
+        const double beta,
+        const int32_t convexhullDownsampling,
+        const double progress0,
+        const double progress1,
+        Plane& bestPlane,
+        double& minConcavity,
+        const Parameters& params);
+    template <class T>
+    void AlignMesh(const T* const points,
+        const uint32_t stridePoints,
+        const uint32_t nPoints,
+        const int32_t* const triangles,
+        const uint32_t strideTriangles,
+        const uint32_t nTriangles,
+        const Parameters& params)
+    {
+        if (GetCancel() || !params.m_pca) {
+            return;
+        }
+        m_timer.Tic();
+
+        m_stage = "Align mesh";
+        m_operation = "Voxelization";
+
+        std::ostringstream msg;
+        if (params.m_logger) {
+            msg << "+ " << m_stage << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+
+        Update(0.0, 0.0, params);
+        if (GetCancel()) {
+            return;
+        }
+        m_dim = (size_t)(pow((double)params.m_resolution, 1.0 / 3.0) + 0.5);
+        Volume volume;
+        volume.Voxelize(points, stridePoints, nPoints,
+            triangles, strideTriangles, nTriangles,
+            m_dim, m_barycenter, m_rot);
+        size_t n = volume.GetNPrimitivesOnSurf() + volume.GetNPrimitivesInsideSurf();
+        Update(50.0, 100.0, params);
+
+        if (params.m_logger) {
+            msg.str("");
+            msg << "\t dim = " << m_dim << "\t-> " << n << " voxels" << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+        if (GetCancel()) {
+            return;
+        }
+        m_operation = "PCA";
+        Update(50.0, 0.0, params);
+        volume.AlignToPrincipalAxes(m_rot);
+        m_overallProgress = 1.0;
+        Update(100.0, 100.0, params);
+
+        m_timer.Toc();
+        if (params.m_logger) {
+            msg.str("");
+            msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+    }
+    template <class T>
+    void VoxelizeMesh(const T* const points,
+        const uint32_t stridePoints,
+        const uint32_t nPoints,
+        const int32_t* const triangles,
+        const uint32_t strideTriangles,
+        const uint32_t nTriangles,
+        const Parameters& params)
+    {
+        if (GetCancel()) {
+            return;
+        }
+
+        m_timer.Tic();
+        m_stage = "Voxelization";
+
+        std::ostringstream msg;
+        if (params.m_logger) {
+            msg << "+ " << m_stage << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+
+        delete m_volume;
+        m_volume = 0;
+        int32_t iteration = 0;
+        const int32_t maxIteration = 5;
+        double progress = 0.0;
+        while (iteration++ < maxIteration && !m_cancel) {
+            msg.str("");
+            msg << "Iteration " << iteration;
+            m_operation = msg.str();
+
+            progress = iteration * 100.0 / maxIteration;
+            Update(progress, 0.0, params);
+
+            m_volume = new Volume;
+            m_volume->Voxelize(points, stridePoints, nPoints,
+                triangles, strideTriangles, nTriangles,
+                m_dim, m_barycenter, m_rot);
+
+            Update(progress, 100.0, params);
+
+            size_t n = m_volume->GetNPrimitivesOnSurf() + m_volume->GetNPrimitivesInsideSurf();
+            if (params.m_logger) {
+                msg.str("");
+                msg << "\t dim = " << m_dim << "\t-> " << n << " voxels" << std::endl;
+                params.m_logger->Log(msg.str().c_str());
+            }
+
+            double a = pow((double)(params.m_resolution) / n, 0.33);
+            size_t dim_next = (size_t)(m_dim * a + 0.5);
+            if (n < params.m_resolution && iteration < maxIteration && m_volume->GetNPrimitivesOnSurf() < params.m_resolution / 8 && m_dim != dim_next) {
+                delete m_volume;
+                m_volume = 0;
+                m_dim = dim_next;
+            }
+            else {
+                break;
+            }
+        }
+        m_overallProgress = 10.0;
+        Update(100.0, 100.0, params);
+
+        m_timer.Toc();
+        if (params.m_logger) {
+            msg.str("");
+            msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+    }
+    template <class T>
+    bool ComputeACD(const T* const points,
+        const uint32_t nPoints,
+        const uint32_t* const triangles,
+        const uint32_t nTriangles,
+        const Parameters& params)
+    {
+        Init();
+        if (params.m_projectHullVertices)
+        {
+            mRaycastMesh = RaycastMesh::createRaycastMesh(nPoints, points, nTriangles, (const uint32_t *)triangles);
+        }
+        if (params.m_oclAcceleration) {
+            // build kernels
+        }
+        AlignMesh(points, 3, nPoints, (int32_t *)triangles, 3, nTriangles, params);
+        VoxelizeMesh(points, 3, nPoints, (int32_t *)triangles, 3, nTriangles, params);
+        ComputePrimitiveSet(params);
+        ComputeACD(params);
+        MergeConvexHulls(params);
+        SimplifyConvexHulls(params);
+        if (params.m_oclAcceleration) {
+            // Release kernels
+        }
+        if (GetCancel()) {
+            Clean();
+            return false;
+        }
+        return true;
+    }
+
+private:
+	RaycastMesh		*mRaycastMesh{ nullptr };
+    SArray<Mesh*> m_convexHulls;
+    std::string m_stage;
+    std::string m_operation;
+    double m_overallProgress;
+    double m_stageProgress;
+    double m_operationProgress;
+    double m_rot[3][3];
+    double m_volumeCH0;
+    Vec3<double> m_barycenter;
+    Timer m_timer;
+    size_t m_dim;
+    Volume* m_volume;
+    PrimitiveSet* m_pset;
+    Mutex m_cancelMutex;
+    bool m_cancel;
+    int32_t m_ompNumProcessors;
+#ifdef CL_VERSION_1_1
+    cl_device_id* m_oclDevice;
+    cl_context m_oclContext;
+    cl_program m_oclProgram;
+    cl_command_queue* m_oclQueue;
+    cl_kernel* m_oclKernelComputePartialVolumes;
+    cl_kernel* m_oclKernelComputeSum;
+    size_t m_oclWorkGroupSize;
+#endif //CL_VERSION_1_1
+};
+}
+#endif // VHACD_VHACD_H
diff --git a/thirdparty/vhacd/inc/vhacdVector.h b/thirdparty/vhacd/inc/vhacdVector.h
new file mode 100644
index 0000000000..efcfcf6ad7
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdVector.h
@@ -0,0 +1,168 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_VECTOR_H
+#define VHACD_VECTOR_H
+#include <iostream>
+#include <math.h>
+
+namespace VHACD {
+//!    Vector dim 3.
+template <typename T>
+class Vec3 {
+public:
+    T& operator[](size_t i) { return m_data[i]; }
+    const T& operator[](size_t i) const { return m_data[i]; }
+    T& X();
+    T& Y();
+    T& Z();
+    const T& X() const;
+    const T& Y() const;
+    const T& Z() const;
+    void Normalize();
+    T GetNorm() const;
+    void operator=(const Vec3& rhs);
+    void operator+=(const Vec3& rhs);
+    void operator-=(const Vec3& rhs);
+    void operator-=(T a);
+    void operator+=(T a);
+    void operator/=(T a);
+    void operator*=(T a);
+    Vec3 operator^(const Vec3& rhs) const;
+    T operator*(const Vec3& rhs) const;
+    Vec3 operator+(const Vec3& rhs) const;
+    Vec3 operator-(const Vec3& rhs) const;
+    Vec3 operator-() const;
+    Vec3 operator*(T rhs) const;
+    Vec3 operator/(T rhs) const;
+    bool operator<(const Vec3& rhs) const;
+    bool operator>(const Vec3& rhs) const;
+    Vec3();
+    Vec3(T a);
+    Vec3(T x, T y, T z);
+    Vec3(const Vec3& rhs);
+    /*virtual*/ ~Vec3(void);
+
+    // Compute the center of this bounding box and return the diagonal length
+    T GetCenter(const Vec3 &bmin, const Vec3 &bmax)
+    {
+        X() = (bmin.X() + bmax.X())*0.5;
+        Y() = (bmin.Y() + bmax.Y())*0.5;
+        Z() = (bmin.Z() + bmax.Z())*0.5;
+        T dx = bmax.X() - bmin.X();
+        T dy = bmax.Y() - bmin.Y();
+        T dz = bmax.Z() - bmin.Z();
+        T diagonal = T(sqrt(dx*dx + dy*dy + dz*dz));
+        return diagonal;
+    }
+
+    // Update the min/max values relative to this point
+    void UpdateMinMax(Vec3 &bmin,Vec3 &bmax) const
+    {
+        if (X() < bmin.X())
+        {
+            bmin.X() = X();
+        }
+        if (Y() < bmin.Y())
+        {
+            bmin.Y() = Y();
+        }
+        if (Z() < bmin.Z())
+        {
+            bmin.Z() = Z();
+        }
+        if (X() > bmax.X())
+        {
+            bmax.X() = X();
+        }
+        if (X() > bmax.X())
+        {
+            bmax.X() = X();
+        }
+        if (Y() > bmax.Y())
+        {
+            bmax.Y() = Y();
+        }
+        if (Z() > bmax.Z())
+        {
+            bmax.Z() = Z();
+        }
+    }
+
+    // Returns the squared distance between these two points
+    T GetDistanceSquared(const Vec3 &p) const
+    {
+        T dx = X() - p.X();
+        T dy = Y() - p.Y();
+        T dz = Z() - p.Z();
+        return dx*dx + dy*dy + dz*dz;
+    }
+
+    T GetDistance(const Vec3 &p) const
+    {
+        return sqrt(GetDistanceSquared(p));
+    }
+
+    // Returns the raw vector data as a pointer
+    T* GetData(void) 
+    {
+        return m_data;
+    }
+private:
+    T m_data[3];
+};
+//!    Vector dim 2.
+template <typename T>
+class Vec2 {
+public:
+    T& operator[](size_t i) { return m_data[i]; }
+    const T& operator[](size_t i) const { return m_data[i]; }
+    T& X();
+    T& Y();
+    const T& X() const;
+    const T& Y() const;
+    void Normalize();
+    T GetNorm() const;
+    void operator=(const Vec2& rhs);
+    void operator+=(const Vec2& rhs);
+    void operator-=(const Vec2& rhs);
+    void operator-=(T a);
+    void operator+=(T a);
+    void operator/=(T a);
+    void operator*=(T a);
+    T operator^(const Vec2& rhs) const;
+    T operator*(const Vec2& rhs) const;
+    Vec2 operator+(const Vec2& rhs) const;
+    Vec2 operator-(const Vec2& rhs) const;
+    Vec2 operator-() const;
+    Vec2 operator*(T rhs) const;
+    Vec2 operator/(T rhs) const;
+    Vec2();
+    Vec2(T a);
+    Vec2(T x, T y);
+    Vec2(const Vec2& rhs);
+    /*virtual*/ ~Vec2(void);
+
+private:
+    T m_data[2];
+};
+
+template <typename T>
+const bool Colinear(const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c);
+template <typename T>
+const T ComputeVolume4(const Vec3<T>& a, const Vec3<T>& b, const Vec3<T>& c, const Vec3<T>& d);
+}
+#include "vhacdVector.inl" // template implementation
+#endif
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdVector.inl b/thirdparty/vhacd/inc/vhacdVector.inl
new file mode 100644
index 0000000000..223c2ef173
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdVector.inl
@@ -0,0 +1,362 @@
+#pragma once
+#ifndef VHACD_VECTOR_INL
+#define VHACD_VECTOR_INL
+namespace VHACD
+{
+    template <typename T> 
+    inline Vec3<T> operator*(T lhs, const Vec3<T> & rhs)
+    {
+        return Vec3<T>(lhs * rhs.X(), lhs * rhs.Y(), lhs * rhs.Z());
+    }
+    template <typename T> 
+    inline T & Vec3<T>::X() 
+    {
+        return m_data[0];
+    }
+    template <typename T> 
+    inline  T &    Vec3<T>::Y() 
+    {
+        return m_data[1];
+    }
+    template <typename T>    
+    inline  T &    Vec3<T>::Z() 
+    {
+        return m_data[2];
+    }
+    template <typename T> 
+    inline  const T & Vec3<T>::X() const 
+    {
+        return m_data[0];
+    }
+    template <typename T> 
+    inline  const T & Vec3<T>::Y() const 
+    {
+        return m_data[1];
+    }
+    template <typename T> 
+    inline  const T & Vec3<T>::Z() const 
+    {
+        return m_data[2];
+    }
+    template <typename T> 
+    inline  void Vec3<T>::Normalize()
+    {
+        T n = sqrt(m_data[0]*m_data[0]+m_data[1]*m_data[1]+m_data[2]*m_data[2]);
+        if (n != 0.0) (*this) /= n;
+    }
+    template <typename T> 
+    inline  T Vec3<T>::GetNorm() const 
+    { 
+        return sqrt(m_data[0]*m_data[0]+m_data[1]*m_data[1]+m_data[2]*m_data[2]);
+    }
+    template <typename T> 
+    inline  void Vec3<T>::operator= (const Vec3 & rhs)
+    { 
+        this->m_data[0] = rhs.m_data[0]; 
+        this->m_data[1] = rhs.m_data[1]; 
+        this->m_data[2] = rhs.m_data[2]; 
+    }
+    template <typename T> 
+    inline  void Vec3<T>::operator+=(const Vec3 & rhs)
+    { 
+        this->m_data[0] += rhs.m_data[0]; 
+        this->m_data[1] += rhs.m_data[1]; 
+        this->m_data[2] += rhs.m_data[2]; 
+    }     
+    template <typename T>  
+    inline void Vec3<T>::operator-=(const Vec3 & rhs)
+    { 
+        this->m_data[0] -= rhs.m_data[0]; 
+        this->m_data[1] -= rhs.m_data[1]; 
+        this->m_data[2] -= rhs.m_data[2]; 
+    }
+    template <typename T> 
+    inline void Vec3<T>::operator-=(T a)
+    { 
+        this->m_data[0] -= a; 
+        this->m_data[1] -= a; 
+        this->m_data[2] -= a; 
+    }
+    template <typename T> 
+    inline void Vec3<T>::operator+=(T a)
+    { 
+        this->m_data[0] += a; 
+        this->m_data[1] += a; 
+        this->m_data[2] += a; 
+    }
+    template <typename T> 
+    inline void Vec3<T>::operator/=(T a)
+    { 
+        this->m_data[0] /= a; 
+        this->m_data[1] /= a; 
+        this->m_data[2] /= a; 
+    }
+    template <typename T>  
+    inline void Vec3<T>::operator*=(T a)
+    { 
+        this->m_data[0] *= a; 
+        this->m_data[1] *= a; 
+        this->m_data[2] *= a; 
+    }  
+    template <typename T> 
+    inline Vec3<T> Vec3<T>::operator^ (const Vec3<T> & rhs) const
+    {
+        return Vec3<T>(m_data[1] * rhs.m_data[2] - m_data[2] * rhs.m_data[1],
+                       m_data[2] * rhs.m_data[0] - m_data[0] * rhs.m_data[2],
+                       m_data[0] * rhs.m_data[1] - m_data[1] * rhs.m_data[0]);
+    }
+    template <typename T>
+    inline T Vec3<T>::operator*(const Vec3<T> & rhs) const
+    {
+        return (m_data[0] * rhs.m_data[0] + m_data[1] * rhs.m_data[1] + m_data[2] * rhs.m_data[2]);
+    }        
+    template <typename T>
+    inline Vec3<T> Vec3<T>::operator+(const Vec3<T> & rhs) const
+    {
+        return Vec3<T>(m_data[0] + rhs.m_data[0],m_data[1] + rhs.m_data[1],m_data[2] + rhs.m_data[2]);
+    }
+    template <typename T> 
+    inline  Vec3<T> Vec3<T>::operator-(const Vec3<T> & rhs) const
+    {
+        return Vec3<T>(m_data[0] - rhs.m_data[0],m_data[1] - rhs.m_data[1],m_data[2] - rhs.m_data[2]) ;
+    }     
+    template <typename T> 
+    inline  Vec3<T> Vec3<T>::operator-() const
+    {
+        return Vec3<T>(-m_data[0],-m_data[1],-m_data[2]) ;
+    }     
+
+    template <typename T> 
+    inline Vec3<T> Vec3<T>::operator*(T rhs) const
+    {
+        return Vec3<T>(rhs * this->m_data[0], rhs * this->m_data[1], rhs * this->m_data[2]);
+    }
+    template <typename T>
+    inline Vec3<T> Vec3<T>::operator/ (T rhs) const
+    {
+        return Vec3<T>(m_data[0] / rhs, m_data[1] / rhs, m_data[2] / rhs);
+    }
+    template <typename T>
+    inline Vec3<T>::Vec3(T a) 
+    { 
+        m_data[0] = m_data[1] = m_data[2] = a; 
+    }
+    template <typename T>
+    inline Vec3<T>::Vec3(T x, T y, T z)
+    {
+        m_data[0] = x;
+        m_data[1] = y;
+        m_data[2] = z;
+    }
+    template <typename T>
+    inline Vec3<T>::Vec3(const Vec3 & rhs)
+    {        
+        m_data[0] = rhs.m_data[0];
+        m_data[1] = rhs.m_data[1];
+        m_data[2] = rhs.m_data[2];
+    }
+    template <typename T>
+    inline Vec3<T>::~Vec3(void){};
+
+    template <typename T>
+    inline Vec3<T>::Vec3() {}
+    
+    template<typename T>
+    inline const bool Colinear(const Vec3<T> & a, const Vec3<T> & b, const Vec3<T> & c)
+    {
+        return  ((c.Z() - a.Z()) * (b.Y() - a.Y()) - (b.Z() - a.Z()) * (c.Y() - a.Y()) == 0.0 /*EPS*/) &&
+                ((b.Z() - a.Z()) * (c.X() - a.X()) - (b.X() - a.X()) * (c.Z() - a.Z()) == 0.0 /*EPS*/) &&
+                ((b.X() - a.X()) * (c.Y() - a.Y()) - (b.Y() - a.Y()) * (c.X() - a.X()) == 0.0 /*EPS*/);
+    }
+    
+    template<typename T>
+    inline const T ComputeVolume4(const Vec3<T> & a, const Vec3<T> & b, const Vec3<T> & c, const Vec3<T> & d)
+    {
+        return (a-d) * ((b-d) ^ (c-d));
+    }
+
+    template <typename T> 
+    inline bool Vec3<T>::operator<(const Vec3 & rhs) const
+    {
+        if (X() == rhs[0])
+        {
+            if (Y() == rhs[1])
+            {
+                return (Z()<rhs[2]);
+            }
+            return (Y()<rhs[1]);
+        }
+        return (X()<rhs[0]);
+    }
+    template <typename T> 
+    inline  bool Vec3<T>::operator>(const Vec3 & rhs) const
+    {
+        if (X() == rhs[0])
+        {
+            if (Y() == rhs[1])
+            {
+                return (Z()>rhs[2]);
+            }
+            return (Y()>rhs[1]);
+        }
+        return (X()>rhs[0]);
+    } 
+    template <typename T> 
+    inline Vec2<T> operator*(T lhs, const Vec2<T> & rhs)
+    {
+        return Vec2<T>(lhs * rhs.X(), lhs * rhs.Y());
+    }
+    template <typename T> 
+    inline T & Vec2<T>::X() 
+    {
+        return m_data[0];
+    }
+    template <typename T>    
+    inline  T &    Vec2<T>::Y() 
+    {
+        return m_data[1];
+    }
+    template <typename T>    
+    inline  const T & Vec2<T>::X() const 
+    {
+        return m_data[0];
+    }
+    template <typename T>    
+    inline  const T & Vec2<T>::Y() const 
+    {
+        return m_data[1];
+    }
+    template <typename T>    
+    inline  void Vec2<T>::Normalize()
+    {
+        T n = sqrt(m_data[0]*m_data[0]+m_data[1]*m_data[1]);
+        if (n != 0.0) (*this) /= n;
+    }
+    template <typename T>    
+    inline  T Vec2<T>::GetNorm() const 
+    { 
+        return sqrt(m_data[0]*m_data[0]+m_data[1]*m_data[1]);
+    }
+    template <typename T>    
+    inline  void Vec2<T>::operator= (const Vec2 & rhs)
+    { 
+        this->m_data[0] = rhs.m_data[0]; 
+        this->m_data[1] = rhs.m_data[1]; 
+    }
+    template <typename T>    
+    inline  void Vec2<T>::operator+=(const Vec2 & rhs)
+    { 
+        this->m_data[0] += rhs.m_data[0]; 
+        this->m_data[1] += rhs.m_data[1]; 
+    }     
+    template <typename T>  
+    inline void Vec2<T>::operator-=(const Vec2 & rhs)
+    { 
+        this->m_data[0] -= rhs.m_data[0]; 
+        this->m_data[1] -= rhs.m_data[1]; 
+    }
+    template <typename T>  
+    inline void Vec2<T>::operator-=(T a)
+    { 
+        this->m_data[0] -= a; 
+        this->m_data[1] -= a; 
+    }
+    template <typename T>  
+    inline void Vec2<T>::operator+=(T a)
+    { 
+        this->m_data[0] += a; 
+        this->m_data[1] += a; 
+    }
+    template <typename T>  
+    inline void Vec2<T>::operator/=(T a)
+    { 
+        this->m_data[0] /= a; 
+        this->m_data[1] /= a; 
+    }
+    template <typename T>  
+    inline void Vec2<T>::operator*=(T a)
+    { 
+        this->m_data[0] *= a; 
+        this->m_data[1] *= a; 
+    }  
+    template <typename T> 
+    inline T Vec2<T>::operator^ (const Vec2<T> & rhs) const
+    {
+        return m_data[0] * rhs.m_data[1] - m_data[1] * rhs.m_data[0];
+    }
+    template <typename T>
+    inline T Vec2<T>::operator*(const Vec2<T> & rhs) const
+    {
+        return (m_data[0] * rhs.m_data[0] + m_data[1] * rhs.m_data[1]);
+    }        
+    template <typename T>
+    inline Vec2<T> Vec2<T>::operator+(const Vec2<T> & rhs) const
+    {
+        return Vec2<T>(m_data[0] + rhs.m_data[0],m_data[1] + rhs.m_data[1]);
+    }
+    template <typename T> 
+    inline  Vec2<T> Vec2<T>::operator-(const Vec2<T> & rhs) const
+    {
+        return Vec2<T>(m_data[0] - rhs.m_data[0],m_data[1] - rhs.m_data[1]);
+    }     
+    template <typename T> 
+    inline  Vec2<T> Vec2<T>::operator-() const
+    {
+        return Vec2<T>(-m_data[0],-m_data[1]) ;
+    }     
+
+    template <typename T> 
+    inline Vec2<T> Vec2<T>::operator*(T rhs) const
+    {
+        return Vec2<T>(rhs * this->m_data[0], rhs * this->m_data[1]);
+    }
+    template <typename T>
+    inline Vec2<T> Vec2<T>::operator/ (T rhs) const
+    {
+        return Vec2<T>(m_data[0] / rhs, m_data[1] / rhs);
+    }
+    template <typename T>
+    inline Vec2<T>::Vec2(T a) 
+    { 
+        m_data[0] = m_data[1] = a; 
+    }
+    template <typename T>
+    inline Vec2<T>::Vec2(T x, T y)
+    {
+        m_data[0] = x;
+        m_data[1] = y;
+    }
+    template <typename T>
+    inline Vec2<T>::Vec2(const Vec2 & rhs)
+    {        
+        m_data[0] = rhs.m_data[0];
+        m_data[1] = rhs.m_data[1];
+    }
+    template <typename T>
+    inline Vec2<T>::~Vec2(void){};
+
+    template <typename T>
+    inline Vec2<T>::Vec2() {}
+
+   /*
+     InsideTriangle decides if a point P is Inside of the triangle
+     defined by A, B, C.
+   */
+    template<typename T>
+    inline const bool InsideTriangle(const Vec2<T> & a, const Vec2<T> & b, const Vec2<T> & c, const Vec2<T> & p)
+    {
+        T ax, ay, bx, by, cx, cy, apx, apy, bpx, bpy, cpx, cpy;
+        T cCROSSap, bCROSScp, aCROSSbp;
+        ax = c.X() - b.X();  ay = c.Y() - b.Y();
+        bx = a.X() - c.X();  by = a.Y() - c.Y();
+        cx = b.X() - a.X();  cy = b.Y() - a.Y();
+        apx= p.X() - a.X();  apy= p.Y() - a.Y();
+        bpx= p.X() - b.X();  bpy= p.Y() - b.Y();
+        cpx= p.X() - c.X();  cpy= p.Y() - c.Y();
+        aCROSSbp = ax*bpy - ay*bpx;
+        cCROSSap = cx*apy - cy*apx;
+        bCROSScp = bx*cpy - by*cpx;
+        return ((aCROSSbp >= 0.0) && (bCROSScp >= 0.0) && (cCROSSap >= 0.0));
+    }
+}
+#endif //VHACD_VECTOR_INL
\ No newline at end of file
diff --git a/thirdparty/vhacd/inc/vhacdVolume.h b/thirdparty/vhacd/inc/vhacdVolume.h
new file mode 100644
index 0000000000..8c47fa1e2c
--- /dev/null
+++ b/thirdparty/vhacd/inc/vhacdVolume.h
@@ -0,0 +1,430 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_VOLUME_H
+#define VHACD_VOLUME_H
+#include "vhacdMesh.h"
+#include "vhacdVector.h"
+#include <assert.h>
+
+#ifdef _MSC_VER
+#pragma warning(push)
+#pragma warning(disable:4456 4701)
+#endif
+
+namespace VHACD {
+
+enum VOXEL_VALUE {
+    PRIMITIVE_UNDEFINED = 0,
+    PRIMITIVE_OUTSIDE_SURFACE = 1,
+    PRIMITIVE_INSIDE_SURFACE = 2,
+    PRIMITIVE_ON_SURFACE = 3
+};
+
+struct Voxel {
+public:
+    short m_coord[3];
+    short m_data;
+};
+
+class PrimitiveSet {
+public:
+    virtual ~PrimitiveSet(){};
+    virtual PrimitiveSet* Create() const = 0;
+    virtual const size_t GetNPrimitives() const = 0;
+    virtual const size_t GetNPrimitivesOnSurf() const = 0;
+    virtual const size_t GetNPrimitivesInsideSurf() const = 0;
+    virtual const double GetEigenValue(AXIS axis) const = 0;
+    virtual const double ComputeMaxVolumeError() const = 0;
+    virtual const double ComputeVolume() const = 0;
+    virtual void Clip(const Plane& plane, PrimitiveSet* const positivePart,
+        PrimitiveSet* const negativePart) const = 0;
+    virtual void Intersect(const Plane& plane, SArray<Vec3<double> >* const positivePts,
+        SArray<Vec3<double> >* const negativePts, const size_t sampling) const = 0;
+    virtual void ComputeExteriorPoints(const Plane& plane, const Mesh& mesh,
+        SArray<Vec3<double> >* const exteriorPts) const = 0;
+    virtual void ComputeClippedVolumes(const Plane& plane, double& positiveVolume,
+        double& negativeVolume) const = 0;
+    virtual void SelectOnSurface(PrimitiveSet* const onSurfP) const = 0;
+    virtual void ComputeConvexHull(Mesh& meshCH, const size_t sampling = 1) const = 0;
+    virtual void ComputeBB() = 0;
+    virtual void ComputePrincipalAxes() = 0;
+    virtual void AlignToPrincipalAxes() = 0;
+    virtual void RevertAlignToPrincipalAxes() = 0;
+    virtual void Convert(Mesh& mesh, const VOXEL_VALUE value) const = 0;
+    const Mesh& GetConvexHull() const { return m_convexHull; };
+    Mesh& GetConvexHull() { return m_convexHull; };
+private:
+    Mesh m_convexHull;
+};
+
+//!
+class VoxelSet : public PrimitiveSet {
+    friend class Volume;
+
+public:
+    //! Destructor.
+    ~VoxelSet(void);
+    //! Constructor.
+    VoxelSet();
+
+    const size_t GetNPrimitives() const { return m_voxels.Size(); }
+    const size_t GetNPrimitivesOnSurf() const { return m_numVoxelsOnSurface; }
+    const size_t GetNPrimitivesInsideSurf() const { return m_numVoxelsInsideSurface; }
+    const double GetEigenValue(AXIS axis) const { return m_D[axis][axis]; }
+    const double ComputeVolume() const { return m_unitVolume * m_voxels.Size(); }
+    const double ComputeMaxVolumeError() const { return m_unitVolume * m_numVoxelsOnSurface; }
+    const Vec3<short>& GetMinBBVoxels() const { return m_minBBVoxels; }
+    const Vec3<short>& GetMaxBBVoxels() const { return m_maxBBVoxels; }
+    const Vec3<double>& GetMinBB() const { return m_minBB; }
+    const double& GetScale() const { return m_scale; }
+    const double& GetUnitVolume() const { return m_unitVolume; }
+    Vec3<double> GetPoint(Vec3<short> voxel) const
+    {
+        return Vec3<double>(voxel[0] * m_scale + m_minBB[0],
+            voxel[1] * m_scale + m_minBB[1],
+            voxel[2] * m_scale + m_minBB[2]);
+    }
+    Vec3<double> GetPoint(const Voxel& voxel) const
+    {
+        return Vec3<double>(voxel.m_coord[0] * m_scale + m_minBB[0],
+            voxel.m_coord[1] * m_scale + m_minBB[1],
+            voxel.m_coord[2] * m_scale + m_minBB[2]);
+    }
+    Vec3<double> GetPoint(Vec3<double> voxel) const
+    {
+        return Vec3<double>(voxel[0] * m_scale + m_minBB[0],
+            voxel[1] * m_scale + m_minBB[1],
+            voxel[2] * m_scale + m_minBB[2]);
+    }
+    void GetPoints(const Voxel& voxel, Vec3<double>* const pts) const;
+    void ComputeConvexHull(Mesh& meshCH, const size_t sampling = 1) const;
+    void Clip(const Plane& plane, PrimitiveSet* const positivePart, PrimitiveSet* const negativePart) const;
+    void Intersect(const Plane& plane, SArray<Vec3<double> >* const positivePts,
+        SArray<Vec3<double> >* const negativePts, const size_t sampling) const;
+    void ComputeExteriorPoints(const Plane& plane, const Mesh& mesh,
+        SArray<Vec3<double> >* const exteriorPts) const;
+    void ComputeClippedVolumes(const Plane& plane, double& positiveVolume, double& negativeVolume) const;
+    void SelectOnSurface(PrimitiveSet* const onSurfP) const;
+    void ComputeBB();
+    void Convert(Mesh& mesh, const VOXEL_VALUE value) const;
+    void ComputePrincipalAxes();
+    PrimitiveSet* Create() const
+    {
+        return new VoxelSet();
+    }
+    void AlignToPrincipalAxes(){};
+    void RevertAlignToPrincipalAxes(){};
+    Voxel* const GetVoxels() { return m_voxels.Data(); }
+    const Voxel* const GetVoxels() const { return m_voxels.Data(); }
+
+private:
+    size_t m_numVoxelsOnSurface;
+    size_t m_numVoxelsInsideSurface;
+    Vec3<double> m_minBB;
+    double m_scale;
+    SArray<Voxel, 8> m_voxels;
+    double m_unitVolume;
+    Vec3<double> m_minBBPts;
+    Vec3<double> m_maxBBPts;
+    Vec3<short> m_minBBVoxels;
+    Vec3<short> m_maxBBVoxels;
+    Vec3<short> m_barycenter;
+    double m_Q[3][3];
+    double m_D[3][3];
+    Vec3<double> m_barycenterPCA;
+};
+
+struct Tetrahedron {
+public:
+    Vec3<double> m_pts[4];
+    unsigned char m_data;
+};
+
+//!
+class TetrahedronSet : public PrimitiveSet {
+    friend class Volume;
+
+public:
+    //! Destructor.
+    ~TetrahedronSet(void);
+    //! Constructor.
+    TetrahedronSet();
+
+    const size_t GetNPrimitives() const { return m_tetrahedra.Size(); }
+    const size_t GetNPrimitivesOnSurf() const { return m_numTetrahedraOnSurface; }
+    const size_t GetNPrimitivesInsideSurf() const { return m_numTetrahedraInsideSurface; }
+    const Vec3<double>& GetMinBB() const { return m_minBB; }
+    const Vec3<double>& GetMaxBB() const { return m_maxBB; }
+    const Vec3<double>& GetBarycenter() const { return m_barycenter; }
+    const double GetEigenValue(AXIS axis) const { return m_D[axis][axis]; }
+    const double GetSacle() const { return m_scale; }
+    const double ComputeVolume() const;
+    const double ComputeMaxVolumeError() const;
+    void ComputeConvexHull(Mesh& meshCH, const size_t sampling = 1) const;
+    void ComputePrincipalAxes();
+    void AlignToPrincipalAxes();
+    void RevertAlignToPrincipalAxes();
+    void Clip(const Plane& plane, PrimitiveSet* const positivePart, PrimitiveSet* const negativePart) const;
+    void Intersect(const Plane& plane, SArray<Vec3<double> >* const positivePts,
+        SArray<Vec3<double> >* const negativePts, const size_t sampling) const;
+    void ComputeExteriorPoints(const Plane& plane, const Mesh& mesh,
+        SArray<Vec3<double> >* const exteriorPts) const;
+    void ComputeClippedVolumes(const Plane& plane, double& positiveVolume, double& negativeVolume) const;
+    void SelectOnSurface(PrimitiveSet* const onSurfP) const;
+    void ComputeBB();
+    void Convert(Mesh& mesh, const VOXEL_VALUE value) const;
+    inline bool Add(Tetrahedron& tetrahedron);
+    PrimitiveSet* Create() const
+    {
+        return new TetrahedronSet();
+    }
+    static const double EPS;
+
+private:
+    void AddClippedTetrahedra(const Vec3<double> (&pts)[10], const int32_t nPts);
+
+    size_t m_numTetrahedraOnSurface;
+    size_t m_numTetrahedraInsideSurface;
+    double m_scale;
+    Vec3<double> m_minBB;
+    Vec3<double> m_maxBB;
+    Vec3<double> m_barycenter;
+    SArray<Tetrahedron, 8> m_tetrahedra;
+    double m_Q[3][3];
+    double m_D[3][3];
+};
+
+//!
+class Volume {
+public:
+    //! Destructor.
+    ~Volume(void);
+
+    //! Constructor.
+    Volume();
+
+    //! Voxelize
+    template <class T>
+    void Voxelize(const T* const points, const uint32_t stridePoints, const uint32_t nPoints,
+        const int32_t* const triangles, const uint32_t strideTriangles, const uint32_t nTriangles,
+        const size_t dim, const Vec3<double>& barycenter, const double (&rot)[3][3]);
+    unsigned char& GetVoxel(const size_t i, const size_t j, const size_t k)
+    {
+        assert(i < m_dim[0] || i >= 0);
+        assert(j < m_dim[0] || j >= 0);
+        assert(k < m_dim[0] || k >= 0);
+        return m_data[i + j * m_dim[0] + k * m_dim[0] * m_dim[1]];
+    }
+    const unsigned char& GetVoxel(const size_t i, const size_t j, const size_t k) const
+    {
+        assert(i < m_dim[0] || i >= 0);
+        assert(j < m_dim[0] || j >= 0);
+        assert(k < m_dim[0] || k >= 0);
+        return m_data[i + j * m_dim[0] + k * m_dim[0] * m_dim[1]];
+    }
+    const size_t GetNPrimitivesOnSurf() const { return m_numVoxelsOnSurface; }
+    const size_t GetNPrimitivesInsideSurf() const { return m_numVoxelsInsideSurface; }
+    void Convert(Mesh& mesh, const VOXEL_VALUE value) const;
+    void Convert(VoxelSet& vset) const;
+    void Convert(TetrahedronSet& tset) const;
+    void AlignToPrincipalAxes(double (&rot)[3][3]) const;
+
+private:
+    void FillOutsideSurface(const size_t i0, const size_t j0, const size_t k0, const size_t i1,
+        const size_t j1, const size_t k1);
+    void FillInsideSurface();
+    template <class T>
+    void ComputeBB(const T* const points, const uint32_t stridePoints, const uint32_t nPoints,
+        const Vec3<double>& barycenter, const double (&rot)[3][3]);
+    void Allocate();
+    void Free();
+
+    Vec3<double> m_minBB;
+    Vec3<double> m_maxBB;
+    double m_scale;
+    size_t m_dim[3]; //>! dim
+    size_t m_numVoxelsOnSurface;
+    size_t m_numVoxelsInsideSurface;
+    size_t m_numVoxelsOutsideSurface;
+    unsigned char* m_data;
+};
+int32_t TriBoxOverlap(const Vec3<double>& boxcenter, const Vec3<double>& boxhalfsize, const Vec3<double>& triver0,
+    const Vec3<double>& triver1, const Vec3<double>& triver2);
+template <class T>
+inline void ComputeAlignedPoint(const T* const points, const uint32_t idx, const Vec3<double>& barycenter,
+    const double (&rot)[3][3], Vec3<double>& pt){};
+template <>
+inline void ComputeAlignedPoint<float>(const float* const points, const uint32_t idx, const Vec3<double>& barycenter, const double (&rot)[3][3], Vec3<double>& pt)
+{
+    double x = points[idx + 0] - barycenter[0];
+    double y = points[idx + 1] - barycenter[1];
+    double z = points[idx + 2] - barycenter[2];
+    pt[0] = rot[0][0] * x + rot[1][0] * y + rot[2][0] * z;
+    pt[1] = rot[0][1] * x + rot[1][1] * y + rot[2][1] * z;
+    pt[2] = rot[0][2] * x + rot[1][2] * y + rot[2][2] * z;
+}
+template <>
+inline void ComputeAlignedPoint<double>(const double* const points, const uint32_t idx, const Vec3<double>& barycenter, const double (&rot)[3][3], Vec3<double>& pt)
+{
+    double x = points[idx + 0] - barycenter[0];
+    double y = points[idx + 1] - barycenter[1];
+    double z = points[idx + 2] - barycenter[2];
+    pt[0] = rot[0][0] * x + rot[1][0] * y + rot[2][0] * z;
+    pt[1] = rot[0][1] * x + rot[1][1] * y + rot[2][1] * z;
+    pt[2] = rot[0][2] * x + rot[1][2] * y + rot[2][2] * z;
+}
+template <class T>
+void Volume::ComputeBB(const T* const points, const uint32_t stridePoints, const uint32_t nPoints,
+    const Vec3<double>& barycenter, const double (&rot)[3][3])
+{
+    Vec3<double> pt;
+    ComputeAlignedPoint(points, 0, barycenter, rot, pt);
+    m_maxBB = pt;
+    m_minBB = pt;
+    for (uint32_t v = 1; v < nPoints; ++v) {
+        ComputeAlignedPoint(points, v * stridePoints, barycenter, rot, pt);
+        for (int32_t i = 0; i < 3; ++i) {
+            if (pt[i] < m_minBB[i])
+                m_minBB[i] = pt[i];
+            else if (pt[i] > m_maxBB[i])
+                m_maxBB[i] = pt[i];
+        }
+    }
+}
+template <class T>
+void Volume::Voxelize(const T* const points, const uint32_t stridePoints, const uint32_t nPoints,
+    const int32_t* const triangles, const uint32_t strideTriangles, const uint32_t nTriangles,
+    const size_t dim, const Vec3<double>& barycenter, const double (&rot)[3][3])
+{
+    if (nPoints == 0) {
+        return;
+    }
+    ComputeBB(points, stridePoints, nPoints, barycenter, rot);
+
+    double d[3] = { m_maxBB[0] - m_minBB[0], m_maxBB[1] - m_minBB[1], m_maxBB[2] - m_minBB[2] };
+    double r;
+    if (d[0] > d[1] && d[0] > d[2]) {
+        r = d[0];
+        m_dim[0] = dim;
+        m_dim[1] = 2 + static_cast<size_t>(dim * d[1] / d[0]);
+        m_dim[2] = 2 + static_cast<size_t>(dim * d[2] / d[0]);
+    }
+    else if (d[1] > d[0] && d[1] > d[2]) {
+        r = d[1];
+        m_dim[1] = dim;
+        m_dim[0] = 2 + static_cast<size_t>(dim * d[0] / d[1]);
+        m_dim[2] = 2 + static_cast<size_t>(dim * d[2] / d[1]);
+    }
+    else {
+        r = d[2];
+        m_dim[2] = dim;
+        m_dim[0] = 2 + static_cast<size_t>(dim * d[0] / d[2]);
+        m_dim[1] = 2 + static_cast<size_t>(dim * d[1] / d[2]);
+    }
+
+    m_scale = r / (dim - 1);
+    double invScale = (dim - 1) / r;
+
+    Allocate();
+    m_numVoxelsOnSurface = 0;
+    m_numVoxelsInsideSurface = 0;
+    m_numVoxelsOutsideSurface = 0;
+
+    Vec3<double> p[3];
+    size_t i, j, k;
+    size_t i0, j0, k0;
+    size_t i1, j1, k1;
+    Vec3<double> boxcenter;
+    Vec3<double> pt;
+    const Vec3<double> boxhalfsize(0.5, 0.5, 0.5);
+    for (size_t t = 0, ti = 0; t < nTriangles; ++t, ti += strideTriangles) {
+        Vec3<int32_t> tri(triangles[ti + 0],
+            triangles[ti + 1],
+            triangles[ti + 2]);
+        for (int32_t c = 0; c < 3; ++c) {
+            ComputeAlignedPoint(points, tri[c] * stridePoints, barycenter, rot, pt);
+            p[c][0] = (pt[0] - m_minBB[0]) * invScale;
+            p[c][1] = (pt[1] - m_minBB[1]) * invScale;
+            p[c][2] = (pt[2] - m_minBB[2]) * invScale;
+            i = static_cast<size_t>(p[c][0] + 0.5);
+            j = static_cast<size_t>(p[c][1] + 0.5);
+            k = static_cast<size_t>(p[c][2] + 0.5);
+            assert(i < m_dim[0] && i >= 0 && j < m_dim[1] && j >= 0 && k < m_dim[2] && k >= 0);
+
+            if (c == 0) {
+                i0 = i1 = i;
+                j0 = j1 = j;
+                k0 = k1 = k;
+            }
+            else {
+                if (i < i0)
+                    i0 = i;
+                if (j < j0)
+                    j0 = j;
+                if (k < k0)
+                    k0 = k;
+                if (i > i1)
+                    i1 = i;
+                if (j > j1)
+                    j1 = j;
+                if (k > k1)
+                    k1 = k;
+            }
+        }
+        if (i0 > 0)
+            --i0;
+        if (j0 > 0)
+            --j0;
+        if (k0 > 0)
+            --k0;
+        if (i1 < m_dim[0])
+            ++i1;
+        if (j1 < m_dim[1])
+            ++j1;
+        if (k1 < m_dim[2])
+            ++k1;
+        for (size_t i = i0; i < i1; ++i) {
+            boxcenter[0] = (double)i;
+            for (size_t j = j0; j < j1; ++j) {
+                boxcenter[1] = (double)j;
+                for (size_t k = k0; k < k1; ++k) {
+                    boxcenter[2] = (double)k;
+                    int32_t res = TriBoxOverlap(boxcenter, boxhalfsize, p[0], p[1], p[2]);
+                    unsigned char& value = GetVoxel(i, j, k);
+                    if (res == 1 && value == PRIMITIVE_UNDEFINED) {
+                        value = PRIMITIVE_ON_SURFACE;
+                        ++m_numVoxelsOnSurface;
+                    }
+                }
+            }
+        }
+    }
+    FillOutsideSurface(0, 0, 0, m_dim[0], m_dim[1], 1);
+    FillOutsideSurface(0, 0, m_dim[2] - 1, m_dim[0], m_dim[1], m_dim[2]);
+    FillOutsideSurface(0, 0, 0, m_dim[0], 1, m_dim[2]);
+    FillOutsideSurface(0, m_dim[1] - 1, 0, m_dim[0], m_dim[1], m_dim[2]);
+    FillOutsideSurface(0, 0, 0, 1, m_dim[1], m_dim[2]);
+    FillOutsideSurface(m_dim[0] - 1, 0, 0, m_dim[0], m_dim[1], m_dim[2]);
+    FillInsideSurface();
+}
+}
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
+
+#endif // VHACD_VOLUME_H
diff --git a/thirdparty/vhacd/public/VHACD.h b/thirdparty/vhacd/public/VHACD.h
new file mode 100644
index 0000000000..863aeda9de
--- /dev/null
+++ b/thirdparty/vhacd/public/VHACD.h
@@ -0,0 +1,153 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#pragma once
+#ifndef VHACD_H
+#define VHACD_H
+
+#define VHACD_VERSION_MAJOR 2
+#define VHACD_VERSION_MINOR 3
+
+// Changes for version 2.3
+//
+// m_gamma : Has been removed.  This used to control the error metric to merge convex hulls.  Now it uses the 'm_maxConvexHulls' value instead.
+// m_maxConvexHulls : This is the maximum number of convex hulls to produce from the merge operation; replaces 'm_gamma'.
+//
+// Note that decomposition depth is no longer a user provided value.  It is now derived from the 
+// maximum number of hulls requested.
+//
+// As a convenience to the user, each convex hull produced now includes the volume of the hull as well as it's center.
+//
+// This version supports a convenience method to automatically make V-HACD run asynchronously in a background thread.
+// To get a fully asynchronous version, call 'CreateVHACD_ASYNC' instead of 'CreateVHACD'.  You get the same interface however,
+// now when computing convex hulls, it is no longer a blocking operation.  All callback messages are still returned
+// in the application's thread so you don't need to worry about mutex locks or anything in that case.
+// To tell if the operation is complete, the application should call 'IsReady'.  This will return true if
+// the last approximation operation is complete and will dispatch any pending messages.
+// If you call 'Compute' while a previous operation was still running, it will automatically cancel the last request
+// and begin a new one.  To cancel a currently running approximation just call 'Cancel'.
+#include <stdint.h>
+
+namespace VHACD {
+class IVHACD {
+public:
+    class IUserCallback {
+    public:
+        virtual ~IUserCallback(){};
+        virtual void Update(const double overallProgress,
+            const double stageProgress,
+            const double operationProgress,
+            const char* const stage,
+            const char* const operation)
+            = 0;
+    };
+
+    class IUserLogger {
+    public:
+        virtual ~IUserLogger(){};
+        virtual void Log(const char* const msg) = 0;
+    };
+
+    class ConvexHull {
+    public:
+        double* m_points;
+        uint32_t* m_triangles;
+        uint32_t m_nPoints;
+        uint32_t m_nTriangles;
+		double		m_volume;
+		double		m_center[3];
+    };
+
+    class Parameters {
+    public:
+        Parameters(void) { Init(); }
+        void Init(void)
+        {
+            m_resolution = 100000;
+            m_concavity = 0.001;
+            m_planeDownsampling = 4;
+            m_convexhullDownsampling = 4;
+            m_alpha = 0.05;
+            m_beta = 0.05;
+            m_pca = 0;
+            m_mode = 0; // 0: voxel-based (recommended), 1: tetrahedron-based
+            m_maxNumVerticesPerCH = 64;
+            m_minVolumePerCH = 0.0001;
+            m_callback = 0;
+            m_logger = 0;
+            m_convexhullApproximation = true;
+            m_oclAcceleration = true;
+            m_maxConvexHulls = 1024;
+			m_projectHullVertices = true; // This will project the output convex hull vertices onto the original source mesh to increase the floating point accuracy of the results
+        }
+        double m_concavity;
+        double m_alpha;
+        double m_beta;
+        double m_minVolumePerCH;
+        IUserCallback* m_callback;
+        IUserLogger* m_logger;
+        uint32_t m_resolution;
+        uint32_t m_maxNumVerticesPerCH;
+        uint32_t m_planeDownsampling;
+        uint32_t m_convexhullDownsampling;
+        uint32_t m_pca;
+        uint32_t m_mode;
+        uint32_t m_convexhullApproximation;
+        uint32_t m_oclAcceleration;
+        uint32_t	m_maxConvexHulls;
+		bool	m_projectHullVertices;
+    };
+
+    virtual void Cancel() = 0;
+    virtual bool Compute(const float* const points,
+        const uint32_t countPoints,
+        const uint32_t* const triangles,
+        const uint32_t countTriangles,
+        const Parameters& params)
+        = 0;
+    virtual bool Compute(const double* const points,
+        const uint32_t countPoints,
+        const uint32_t* const triangles,
+        const uint32_t countTriangles,
+        const Parameters& params)
+        = 0;
+    virtual uint32_t GetNConvexHulls() const = 0;
+    virtual void GetConvexHull(const uint32_t index, ConvexHull& ch) const = 0;
+    virtual void Clean(void) = 0; // release internally allocated memory
+    virtual void Release(void) = 0; // release IVHACD
+    virtual bool OCLInit(void* const oclDevice,
+        IUserLogger* const logger = 0)
+        = 0;
+    virtual bool OCLRelease(IUserLogger* const logger = 0) = 0;
+
+	// Will compute the center of mass of the convex hull decomposition results and return it
+	// in 'centerOfMass'.  Returns false if the center of mass could not be computed.
+	virtual bool ComputeCenterOfMass(double centerOfMass[3]) const = 0;
+
+	// In synchronous mode (non-multi-threaded) the state is always 'ready'
+	// In asynchronous mode, this returns true if the background thread is not still actively computing
+	// a new solution.  In an asynchronous config the 'IsReady' call will report any update or log
+	// messages in the caller's current thread.
+	virtual bool IsReady(void) const
+	{
+		return true;
+	}
+
+protected:
+    virtual ~IVHACD(void) {}
+};
+IVHACD* CreateVHACD(void);
+IVHACD* CreateVHACD_ASYNC(void);
+}
+#endif // VHACD_H
diff --git a/thirdparty/vhacd/src/FloatMath.cpp b/thirdparty/vhacd/src/FloatMath.cpp
new file mode 100644
index 0000000000..481e875104
--- /dev/null
+++ b/thirdparty/vhacd/src/FloatMath.cpp
@@ -0,0 +1,17 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#include <math.h>
+#include <float.h>
+#include "FloatMath.h"
+#include <vector>
+
+#define REAL float
+
+#include "FloatMath.inl"
+
+#undef REAL
+#define REAL double
+
+#include "FloatMath.inl"
diff --git a/thirdparty/vhacd/src/FloatMath.inl b/thirdparty/vhacd/src/FloatMath.inl
new file mode 100644
index 0000000000..ce529e6f71
--- /dev/null
+++ b/thirdparty/vhacd/src/FloatMath.inl
@@ -0,0 +1,5276 @@
+// a set of routines that let you do common 3d math
+// operations without any vector, matrix, or quaternion
+// classes or templates.
+//
+// a vector (or point) is a 'float *' to 3 floating point numbers.
+// a matrix is a 'float *' to an array of 16 floating point numbers representing a 4x4 transformation matrix compatible with D3D or OGL
+// a quaternion is a 'float *' to 4 floats representing a quaternion x,y,z,w
+//
+
+namespace FLOAT_MATH
+{
+
+void fm_inverseRT(const REAL matrix[16],const REAL pos[3],REAL t[3]) // inverse rotate translate the point.
+{
+
+	REAL _x = pos[0] - matrix[3*4+0];
+	REAL _y = pos[1] - matrix[3*4+1];
+	REAL _z = pos[2] - matrix[3*4+2];
+
+	// Multiply inverse-translated source vector by inverted rotation transform
+
+	t[0] = (matrix[0*4+0] * _x) + (matrix[0*4+1] * _y) + (matrix[0*4+2] * _z);
+	t[1] = (matrix[1*4+0] * _x) + (matrix[1*4+1] * _y) + (matrix[1*4+2] * _z);
+	t[2] = (matrix[2*4+0] * _x) + (matrix[2*4+1] * _y) + (matrix[2*4+2] * _z);
+
+}
+
+REAL fm_getDeterminant(const REAL matrix[16])
+{
+  REAL tempv[3];
+  REAL p0[3];
+  REAL p1[3];
+  REAL p2[3];
+
+
+	p0[0] = matrix[0*4+0];
+	p0[1] = matrix[0*4+1];
+	p0[2] = matrix[0*4+2];
+
+	p1[0] = matrix[1*4+0];
+	p1[1] = matrix[1*4+1];
+	p1[2] = matrix[1*4+2];
+
+	p2[0] = matrix[2*4+0];
+	p2[1] = matrix[2*4+1];
+	p2[2] = matrix[2*4+2];
+
+  fm_cross(tempv,p1,p2);
+
+  return fm_dot(p0,tempv);
+
+}
+
+REAL fm_squared(REAL x) { return x*x; };
+
+void fm_decomposeTransform(const REAL local_transform[16],REAL trans[3],REAL rot[4],REAL scale[3])
+{
+
+  trans[0] = local_transform[12];
+  trans[1] = local_transform[13];
+  trans[2] = local_transform[14];
+
+  scale[0] = (REAL)sqrt(fm_squared(local_transform[0*4+0]) + fm_squared(local_transform[0*4+1]) + fm_squared(local_transform[0*4+2]));
+  scale[1] = (REAL)sqrt(fm_squared(local_transform[1*4+0]) + fm_squared(local_transform[1*4+1]) + fm_squared(local_transform[1*4+2]));
+  scale[2] = (REAL)sqrt(fm_squared(local_transform[2*4+0]) + fm_squared(local_transform[2*4+1]) + fm_squared(local_transform[2*4+2]));
+
+  REAL m[16];
+  memcpy(m,local_transform,sizeof(REAL)*16);
+
+  REAL sx = 1.0f / scale[0];
+  REAL sy = 1.0f / scale[1];
+  REAL sz = 1.0f / scale[2];
+
+  m[0*4+0]*=sx;
+  m[0*4+1]*=sx;
+  m[0*4+2]*=sx;
+
+  m[1*4+0]*=sy;
+  m[1*4+1]*=sy;
+  m[1*4+2]*=sy;
+
+  m[2*4+0]*=sz;
+  m[2*4+1]*=sz;
+  m[2*4+2]*=sz;
+
+  fm_matrixToQuat(m,rot);
+
+}
+
+void fm_getSubMatrix(int32_t ki,int32_t kj,REAL pDst[16],const REAL matrix[16])
+{
+	int32_t row, col;
+	int32_t dstCol = 0, dstRow = 0;
+
+	for ( col = 0; col < 4; col++ )
+	{
+		if ( col == kj )
+		{
+			continue;
+		}
+		for ( dstRow = 0, row = 0; row < 4; row++ )
+		{
+			if ( row == ki )
+			{
+				continue;
+			}
+			pDst[dstCol*4+dstRow] = matrix[col*4+row];
+			dstRow++;
+		}
+		dstCol++;
+	}
+}
+
+void  fm_inverseTransform(const REAL matrix[16],REAL inverse_matrix[16])
+{
+	REAL determinant = fm_getDeterminant(matrix);
+	determinant = 1.0f / determinant;
+	for (int32_t i = 0; i < 4; i++ )
+	{
+		for (int32_t j = 0; j < 4; j++ )
+		{
+			int32_t sign = 1 - ( ( i + j ) % 2 ) * 2;
+			REAL subMat[16];
+			fm_identity(subMat);
+			fm_getSubMatrix( i, j, subMat, matrix );
+			REAL subDeterminant = fm_getDeterminant(subMat);
+			inverse_matrix[i*4+j] = ( subDeterminant * sign ) * determinant;
+		}
+	}
+}
+
+void fm_identity(REAL matrix[16]) // set 4x4 matrix to identity.
+{
+	matrix[0*4+0] = 1;
+	matrix[1*4+1] = 1;
+	matrix[2*4+2] = 1;
+	matrix[3*4+3] = 1;
+
+	matrix[1*4+0] = 0;
+	matrix[2*4+0] = 0;
+	matrix[3*4+0] = 0;
+
+	matrix[0*4+1] = 0;
+	matrix[2*4+1] = 0;
+	matrix[3*4+1] = 0;
+
+	matrix[0*4+2] = 0;
+	matrix[1*4+2] = 0;
+	matrix[3*4+2] = 0;
+
+	matrix[0*4+3] = 0;
+	matrix[1*4+3] = 0;
+	matrix[2*4+3] = 0;
+
+}
+
+void  fm_quatToEuler(const REAL quat[4],REAL &ax,REAL &ay,REAL &az)
+{
+	REAL x = quat[0];
+	REAL y = quat[1];
+	REAL z = quat[2];
+	REAL w = quat[3];
+
+	REAL sint	     = (2.0f * w * y) - (2.0f * x * z);
+	REAL cost_temp = 1.0f - (sint * sint);
+	REAL cost	   	 = 0;
+
+	if ( (REAL)fabs(cost_temp) > 0.001f )
+	{
+		cost = (REAL)sqrt( cost_temp );
+	}
+
+	REAL sinv, cosv, sinf, cosf;
+	if ( (REAL)fabs(cost) > 0.001f )
+	{
+	cost = 1.0f / cost;
+		sinv = ((2.0f * y * z) + (2.0f * w * x)) * cost;
+		cosv = (1.0f - (2.0f * x * x) - (2.0f * y * y)) * cost;
+		sinf = ((2.0f * x * y) + (2.0f * w * z)) * cost;
+		cosf = (1.0f - (2.0f * y * y) - (2.0f * z * z)) * cost;
+	}
+	else
+	{
+		sinv = (2.0f * w * x) - (2.0f * y * z);
+		cosv = 1.0f - (2.0f * x * x) - (2.0f * z * z);
+		sinf = 0;
+		cosf = 1.0f;
+	}
+
+	// compute output rotations
+	ax	= (REAL)atan2( sinv, cosv );
+	ay	= (REAL)atan2( sint, cost );
+	az	= (REAL)atan2( sinf, cosf );
+
+}
+
+void fm_eulerToMatrix(REAL ax,REAL ay,REAL az,REAL *matrix) // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+{
+  REAL quat[4];
+  fm_eulerToQuat(ax,ay,az,quat);
+  fm_quatToMatrix(quat,matrix);
+}
+
+void fm_getAABB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *bmin,REAL *bmax)
+{
+
+  const uint8_t *source = (const uint8_t *) points;
+
+	bmin[0] = points[0];
+	bmin[1] = points[1];
+	bmin[2] = points[2];
+
+	bmax[0] = points[0];
+	bmax[1] = points[1];
+	bmax[2] = points[2];
+
+
+  for (uint32_t i=1; i<vcount; i++)
+  {
+	source+=pstride;
+	const REAL *p = (const REAL *) source;
+
+	if ( p[0] < bmin[0] ) bmin[0] = p[0];
+	if ( p[1] < bmin[1] ) bmin[1] = p[1];
+	if ( p[2] < bmin[2] ) bmin[2] = p[2];
+
+		if ( p[0] > bmax[0] ) bmax[0] = p[0];
+		if ( p[1] > bmax[1] ) bmax[1] = p[1];
+		if ( p[2] > bmax[2] ) bmax[2] = p[2];
+
+  }
+}
+
+void  fm_eulerToQuat(const REAL *euler,REAL *quat) // convert euler angles to quaternion.
+{
+  fm_eulerToQuat(euler[0],euler[1],euler[2],quat);
+}
+
+void fm_eulerToQuat(REAL roll,REAL pitch,REAL yaw,REAL *quat) // convert euler angles to quaternion.
+{
+	roll  *= 0.5f;
+	pitch *= 0.5f;
+	yaw   *= 0.5f;
+
+	REAL cr = (REAL)cos(roll);
+	REAL cp = (REAL)cos(pitch);
+	REAL cy = (REAL)cos(yaw);
+
+	REAL sr = (REAL)sin(roll);
+	REAL sp = (REAL)sin(pitch);
+	REAL sy = (REAL)sin(yaw);
+
+	REAL cpcy = cp * cy;
+	REAL spsy = sp * sy;
+	REAL spcy = sp * cy;
+	REAL cpsy = cp * sy;
+
+	quat[0]   = ( sr * cpcy - cr * spsy);
+	quat[1]   = ( cr * spcy + sr * cpsy);
+	quat[2]   = ( cr * cpsy - sr * spcy);
+	quat[3]   = cr * cpcy + sr * spsy;
+}
+
+void fm_quatToMatrix(const REAL *quat,REAL *matrix) // convert quaterinion rotation to matrix, zeros out the translation component.
+{
+
+	REAL xx = quat[0]*quat[0];
+	REAL yy = quat[1]*quat[1];
+	REAL zz = quat[2]*quat[2];
+	REAL xy = quat[0]*quat[1];
+	REAL xz = quat[0]*quat[2];
+	REAL yz = quat[1]*quat[2];
+	REAL wx = quat[3]*quat[0];
+	REAL wy = quat[3]*quat[1];
+	REAL wz = quat[3]*quat[2];
+
+	matrix[0*4+0] = 1 - 2 * ( yy + zz );
+	matrix[1*4+0] =     2 * ( xy - wz );
+	matrix[2*4+0] =     2 * ( xz + wy );
+
+	matrix[0*4+1] =     2 * ( xy + wz );
+	matrix[1*4+1] = 1 - 2 * ( xx + zz );
+	matrix[2*4+1] =     2 * ( yz - wx );
+
+	matrix[0*4+2] =     2 * ( xz - wy );
+	matrix[1*4+2] =     2 * ( yz + wx );
+	matrix[2*4+2] = 1 - 2 * ( xx + yy );
+
+	matrix[3*4+0] = matrix[3*4+1] = matrix[3*4+2] = (REAL) 0.0f;
+	matrix[0*4+3] = matrix[1*4+3] = matrix[2*4+3] = (REAL) 0.0f;
+	matrix[3*4+3] =(REAL) 1.0f;
+
+}
+
+
+void fm_quatRotate(const REAL *quat,const REAL *v,REAL *r) // rotate a vector directly by a quaternion.
+{
+  REAL left[4];
+
+	left[0] =   quat[3]*v[0] + quat[1]*v[2] - v[1]*quat[2];
+	left[1] =   quat[3]*v[1] + quat[2]*v[0] - v[2]*quat[0];
+	left[2] =   quat[3]*v[2] + quat[0]*v[1] - v[0]*quat[1];
+	left[3] = - quat[0]*v[0] - quat[1]*v[1] - quat[2]*v[2];
+
+	r[0] = (left[3]*-quat[0]) + (quat[3]*left[0]) + (left[1]*-quat[2]) - (-quat[1]*left[2]);
+	r[1] = (left[3]*-quat[1]) + (quat[3]*left[1]) + (left[2]*-quat[0]) - (-quat[2]*left[0]);
+	r[2] = (left[3]*-quat[2]) + (quat[3]*left[2]) + (left[0]*-quat[1]) - (-quat[0]*left[1]);
+
+}
+
+
+void fm_getTranslation(const REAL *matrix,REAL *t)
+{
+	t[0] = matrix[3*4+0];
+	t[1] = matrix[3*4+1];
+	t[2] = matrix[3*4+2];
+}
+
+void fm_matrixToQuat(const REAL *matrix,REAL *quat) // convert the 3x3 portion of a 4x4 matrix into a quaterion as x,y,z,w
+{
+
+	REAL tr = matrix[0*4+0] + matrix[1*4+1] + matrix[2*4+2];
+
+	// check the diagonal
+
+	if (tr > 0.0f )
+	{
+		REAL s = (REAL) sqrt ( (double) (tr + 1.0f) );
+		quat[3] = s * 0.5f;
+		s = 0.5f / s;
+		quat[0] = (matrix[1*4+2] - matrix[2*4+1]) * s;
+		quat[1] = (matrix[2*4+0] - matrix[0*4+2]) * s;
+		quat[2] = (matrix[0*4+1] - matrix[1*4+0]) * s;
+
+	}
+	else
+	{
+		// diagonal is negative
+		int32_t nxt[3] = {1, 2, 0};
+		REAL  qa[4];
+
+		int32_t i = 0;
+
+		if (matrix[1*4+1] > matrix[0*4+0]) i = 1;
+		if (matrix[2*4+2] > matrix[i*4+i]) i = 2;
+
+		int32_t j = nxt[i];
+		int32_t k = nxt[j];
+
+		REAL s = (REAL)sqrt ( ((matrix[i*4+i] - (matrix[j*4+j] + matrix[k*4+k])) + 1.0f) );
+
+		qa[i] = s * 0.5f;
+
+		if (s != 0.0f ) s = 0.5f / s;
+
+		qa[3] = (matrix[j*4+k] - matrix[k*4+j]) * s;
+		qa[j] = (matrix[i*4+j] + matrix[j*4+i]) * s;
+		qa[k] = (matrix[i*4+k] + matrix[k*4+i]) * s;
+
+		quat[0] = qa[0];
+		quat[1] = qa[1];
+		quat[2] = qa[2];
+		quat[3] = qa[3];
+	}
+//	fm_normalizeQuat(quat);
+}
+
+
+REAL fm_sphereVolume(REAL radius) // return's the volume of a sphere of this radius (4/3 PI * R cubed )
+{
+	return (4.0f / 3.0f ) * FM_PI * radius * radius * radius;
+}
+
+
+REAL fm_cylinderVolume(REAL radius,REAL h)
+{
+	return FM_PI * radius * radius *h;
+}
+
+REAL fm_capsuleVolume(REAL radius,REAL h)
+{
+	REAL volume = fm_sphereVolume(radius); // volume of the sphere portion.
+	REAL ch = h-radius*2; // this is the cylinder length
+	if ( ch > 0 )
+	{
+		volume+=fm_cylinderVolume(radius,ch);
+	}
+	return volume;
+}
+
+void  fm_transform(const REAL matrix[16],const REAL v[3],REAL t[3]) // rotate and translate this point
+{
+  if ( matrix )
+  {
+	REAL tx = (matrix[0*4+0] * v[0]) +  (matrix[1*4+0] * v[1]) + (matrix[2*4+0] * v[2]) + matrix[3*4+0];
+	REAL ty = (matrix[0*4+1] * v[0]) +  (matrix[1*4+1] * v[1]) + (matrix[2*4+1] * v[2]) + matrix[3*4+1];
+	REAL tz = (matrix[0*4+2] * v[0]) +  (matrix[1*4+2] * v[1]) + (matrix[2*4+2] * v[2]) + matrix[3*4+2];
+	t[0] = tx;
+	t[1] = ty;
+	t[2] = tz;
+  }
+  else
+  {
+	t[0] = v[0];
+	t[1] = v[1];
+	t[2] = v[2];
+  }
+}
+
+void  fm_rotate(const REAL matrix[16],const REAL v[3],REAL t[3]) // rotate and translate this point
+{
+  if ( matrix )
+  {
+	REAL tx = (matrix[0*4+0] * v[0]) +  (matrix[1*4+0] * v[1]) + (matrix[2*4+0] * v[2]);
+	REAL ty = (matrix[0*4+1] * v[0]) +  (matrix[1*4+1] * v[1]) + (matrix[2*4+1] * v[2]);
+	REAL tz = (matrix[0*4+2] * v[0]) +  (matrix[1*4+2] * v[1]) + (matrix[2*4+2] * v[2]);
+	t[0] = tx;
+	t[1] = ty;
+	t[2] = tz;
+  }
+  else
+  {
+	t[0] = v[0];
+	t[1] = v[1];
+	t[2] = v[2];
+  }
+}
+
+
+REAL fm_distance(const REAL *p1,const REAL *p2)
+{
+	REAL dx = p1[0] - p2[0];
+	REAL dy = p1[1] - p2[1];
+	REAL dz = p1[2] - p2[2];
+
+	return (REAL)sqrt( dx*dx + dy*dy + dz *dz );
+}
+
+REAL fm_distanceSquared(const REAL *p1,const REAL *p2)
+{
+	REAL dx = p1[0] - p2[0];
+	REAL dy = p1[1] - p2[1];
+	REAL dz = p1[2] - p2[2];
+
+	return dx*dx + dy*dy + dz *dz;
+}
+
+
+REAL fm_distanceSquaredXZ(const REAL *p1,const REAL *p2)
+{
+	REAL dx = p1[0] - p2[0];
+	REAL dz = p1[2] - p2[2];
+
+	return dx*dx +  dz *dz;
+}
+
+
+REAL fm_computePlane(const REAL *A,const REAL *B,const REAL *C,REAL *n) // returns D
+{
+	REAL vx = (B[0] - C[0]);
+	REAL vy = (B[1] - C[1]);
+	REAL vz = (B[2] - C[2]);
+
+	REAL wx = (A[0] - B[0]);
+	REAL wy = (A[1] - B[1]);
+	REAL wz = (A[2] - B[2]);
+
+	REAL vw_x = vy * wz - vz * wy;
+	REAL vw_y = vz * wx - vx * wz;
+	REAL vw_z = vx * wy - vy * wx;
+
+	REAL mag = (REAL)sqrt((vw_x * vw_x) + (vw_y * vw_y) + (vw_z * vw_z));
+
+	if ( mag < 0.000001f )
+	{
+		mag = 0;
+	}
+	else
+	{
+		mag = 1.0f/mag;
+	}
+
+	REAL x = vw_x * mag;
+	REAL y = vw_y * mag;
+	REAL z = vw_z * mag;
+
+
+	REAL D = 0.0f - ((x*A[0])+(y*A[1])+(z*A[2]));
+
+  n[0] = x;
+  n[1] = y;
+  n[2] = z;
+
+	return D;
+}
+
+REAL fm_distToPlane(const REAL *plane,const REAL *p) // computes the distance of this point from the plane.
+{
+  return p[0]*plane[0]+p[1]*plane[1]+p[2]*plane[2]+plane[3];
+}
+
+REAL fm_dot(const REAL *p1,const REAL *p2)
+{
+  return p1[0]*p2[0]+p1[1]*p2[1]+p1[2]*p2[2];
+}
+
+void fm_cross(REAL *cross,const REAL *a,const REAL *b)
+{
+	cross[0] = a[1]*b[2] - a[2]*b[1];
+	cross[1] = a[2]*b[0] - a[0]*b[2];
+	cross[2] = a[0]*b[1] - a[1]*b[0];
+}
+
+REAL fm_computeNormalVector(REAL *n,const REAL *p1,const REAL *p2)
+{
+  n[0] = p2[0] - p1[0];
+  n[1] = p2[1] - p1[1];
+  n[2] = p2[2] - p1[2];
+  return fm_normalize(n);
+}
+
+bool  fm_computeWindingOrder(const REAL *p1,const REAL *p2,const REAL *p3) // returns true if the triangle is clockwise.
+{
+  bool ret = false;
+
+  REAL v1[3];
+  REAL v2[3];
+
+  fm_computeNormalVector(v1,p1,p2); // p2-p1 (as vector) and then normalized
+  fm_computeNormalVector(v2,p1,p3); // p3-p1 (as vector) and then normalized
+
+  REAL cross[3];
+
+  fm_cross(cross, v1, v2 );
+  REAL ref[3] = { 1, 0, 0 };
+
+  REAL d = fm_dot( cross, ref );
+
+
+  if ( d <= 0 )
+	ret = false;
+  else
+	ret = true;
+
+  return ret;
+}
+
+REAL fm_normalize(REAL *n) // normalize this vector
+{
+  REAL dist = (REAL)sqrt(n[0]*n[0] + n[1]*n[1] + n[2]*n[2]);
+  if ( dist > 0.0000001f )
+  {
+	REAL mag = 1.0f / dist;
+	n[0]*=mag;
+	n[1]*=mag;
+	n[2]*=mag;
+  }
+  else
+  {
+	n[0] = 1;
+	n[1] = 0;
+	n[2] = 0;
+  }
+
+  return dist;
+}
+
+
+void  fm_matrixMultiply(const REAL *pA,const REAL *pB,REAL *pM)
+{
+#if 1
+
+  REAL a = pA[0*4+0] * pB[0*4+0] + pA[0*4+1] * pB[1*4+0] + pA[0*4+2] * pB[2*4+0] + pA[0*4+3] * pB[3*4+0];
+  REAL b = pA[0*4+0] * pB[0*4+1] + pA[0*4+1] * pB[1*4+1] + pA[0*4+2] * pB[2*4+1] + pA[0*4+3] * pB[3*4+1];
+  REAL c = pA[0*4+0] * pB[0*4+2] + pA[0*4+1] * pB[1*4+2] + pA[0*4+2] * pB[2*4+2] + pA[0*4+3] * pB[3*4+2];
+  REAL d = pA[0*4+0] * pB[0*4+3] + pA[0*4+1] * pB[1*4+3] + pA[0*4+2] * pB[2*4+3] + pA[0*4+3] * pB[3*4+3];
+
+  REAL e = pA[1*4+0] * pB[0*4+0] + pA[1*4+1] * pB[1*4+0] + pA[1*4+2] * pB[2*4+0] + pA[1*4+3] * pB[3*4+0];
+  REAL f = pA[1*4+0] * pB[0*4+1] + pA[1*4+1] * pB[1*4+1] + pA[1*4+2] * pB[2*4+1] + pA[1*4+3] * pB[3*4+1];
+  REAL g = pA[1*4+0] * pB[0*4+2] + pA[1*4+1] * pB[1*4+2] + pA[1*4+2] * pB[2*4+2] + pA[1*4+3] * pB[3*4+2];
+  REAL h = pA[1*4+0] * pB[0*4+3] + pA[1*4+1] * pB[1*4+3] + pA[1*4+2] * pB[2*4+3] + pA[1*4+3] * pB[3*4+3];
+
+  REAL i = pA[2*4+0] * pB[0*4+0] + pA[2*4+1] * pB[1*4+0] + pA[2*4+2] * pB[2*4+0] + pA[2*4+3] * pB[3*4+0];
+  REAL j = pA[2*4+0] * pB[0*4+1] + pA[2*4+1] * pB[1*4+1] + pA[2*4+2] * pB[2*4+1] + pA[2*4+3] * pB[3*4+1];
+  REAL k = pA[2*4+0] * pB[0*4+2] + pA[2*4+1] * pB[1*4+2] + pA[2*4+2] * pB[2*4+2] + pA[2*4+3] * pB[3*4+2];
+  REAL l = pA[2*4+0] * pB[0*4+3] + pA[2*4+1] * pB[1*4+3] + pA[2*4+2] * pB[2*4+3] + pA[2*4+3] * pB[3*4+3];
+
+  REAL m = pA[3*4+0] * pB[0*4+0] + pA[3*4+1] * pB[1*4+0] + pA[3*4+2] * pB[2*4+0] + pA[3*4+3] * pB[3*4+0];
+  REAL n = pA[3*4+0] * pB[0*4+1] + pA[3*4+1] * pB[1*4+1] + pA[3*4+2] * pB[2*4+1] + pA[3*4+3] * pB[3*4+1];
+  REAL o = pA[3*4+0] * pB[0*4+2] + pA[3*4+1] * pB[1*4+2] + pA[3*4+2] * pB[2*4+2] + pA[3*4+3] * pB[3*4+2];
+  REAL p = pA[3*4+0] * pB[0*4+3] + pA[3*4+1] * pB[1*4+3] + pA[3*4+2] * pB[2*4+3] + pA[3*4+3] * pB[3*4+3];
+
+  pM[0] = a;
+  pM[1] = b;
+  pM[2] = c;
+  pM[3] = d;
+
+  pM[4] = e;
+  pM[5] = f;
+  pM[6] = g;
+  pM[7] = h;
+
+  pM[8] = i;
+  pM[9] = j;
+  pM[10] = k;
+  pM[11] = l;
+
+  pM[12] = m;
+  pM[13] = n;
+  pM[14] = o;
+  pM[15] = p;
+
+
+#else
+	memset(pM, 0, sizeof(REAL)*16);
+	for(int32_t i=0; i<4; i++ )
+		for(int32_t j=0; j<4; j++ )
+			for(int32_t k=0; k<4; k++ )
+				pM[4*i+j] +=  pA[4*i+k] * pB[4*k+j];
+#endif
+}
+
+
+void  fm_eulerToQuatDX(REAL x,REAL y,REAL z,REAL *quat) // convert euler angles to quaternion using the fucked up DirectX method
+{
+  REAL matrix[16];
+  fm_eulerToMatrix(x,y,z,matrix);
+  fm_matrixToQuat(matrix,quat);
+}
+
+// implementation copied from: http://blogs.msdn.com/mikepelton/archive/2004/10/29/249501.aspx
+void  fm_eulerToMatrixDX(REAL x,REAL y,REAL z,REAL *matrix) // convert euler angles to quaternion using the fucked up DirectX method.
+{
+  fm_identity(matrix);
+  matrix[0*4+0] = (REAL)(cos(z)*cos(y) + sin(z)*sin(x)*sin(y));
+  matrix[0*4+1] = (REAL)(sin(z)*cos(x));
+  matrix[0*4+2] = (REAL)(cos(z)*-sin(y) + sin(z)*sin(x)*cos(y));
+
+  matrix[1*4+0] = (REAL)(-sin(z)*cos(y)+cos(z)*sin(x)*sin(y));
+  matrix[1*4+1] = (REAL)(cos(z)*cos(x));
+  matrix[1*4+2] = (REAL)(sin(z)*sin(y) +cos(z)*sin(x)*cos(y));
+
+  matrix[2*4+0] = (REAL)(cos(x)*sin(y));
+  matrix[2*4+1] = (REAL)(-sin(x));
+  matrix[2*4+2] = (REAL)(cos(x)*cos(y));
+}
+
+
+void  fm_scale(REAL x,REAL y,REAL z,REAL *fscale) // apply scale to the matrix.
+{
+  fscale[0*4+0] = x;
+  fscale[1*4+1] = y;
+  fscale[2*4+2] = z;
+}
+
+
+void  fm_composeTransform(const REAL *position,const REAL *quat,const REAL *scale,REAL *matrix)
+{
+  fm_identity(matrix);
+  fm_quatToMatrix(quat,matrix);
+
+  if ( scale && ( scale[0] != 1 || scale[1] != 1 || scale[2] != 1 ) )
+  {
+	REAL work[16];
+	memcpy(work,matrix,sizeof(REAL)*16);
+	REAL mscale[16];
+	fm_identity(mscale);
+	fm_scale(scale[0],scale[1],scale[2],mscale);
+	fm_matrixMultiply(work,mscale,matrix);
+  }
+
+  matrix[12] = position[0];
+  matrix[13] = position[1];
+  matrix[14] = position[2];
+}
+
+
+void  fm_setTranslation(const REAL *translation,REAL *matrix)
+{
+  matrix[12] = translation[0];
+  matrix[13] = translation[1];
+  matrix[14] = translation[2];
+}
+
+static REAL enorm0_3d ( REAL x0, REAL y0, REAL z0, REAL x1, REAL y1, REAL z1 )
+
+/**********************************************************************/
+
+/*
+Purpose:
+
+ENORM0_3D computes the Euclidean norm of (P1-P0) in 3D.
+
+Modified:
+
+18 April 1999
+
+Author:
+
+John Burkardt
+
+Parameters:
+
+Input, REAL X0, Y0, Z0, X1, Y1, Z1, the coordinates of the points 
+P0 and P1.
+
+Output, REAL ENORM0_3D, the Euclidean norm of (P1-P0).
+*/
+{
+  REAL value;
+
+  value = (REAL)sqrt (
+	( x1 - x0 ) * ( x1 - x0 ) + 
+	( y1 - y0 ) * ( y1 - y0 ) + 
+	( z1 - z0 ) * ( z1 - z0 ) );
+
+  return value;
+}
+
+
+static REAL triangle_area_3d ( REAL x1, REAL y1, REAL z1, REAL x2,REAL y2, REAL z2, REAL x3, REAL y3, REAL z3 )
+
+						/**********************************************************************/
+
+						/*
+						Purpose:
+
+						TRIANGLE_AREA_3D computes the area of a triangle in 3D.
+
+						Modified:
+
+						22 April 1999
+
+						Author:
+
+						John Burkardt
+
+						Parameters:
+
+						Input, REAL X1, Y1, Z1, X2, Y2, Z2, X3, Y3, Z3, the (X,Y,Z)
+						coordinates of the corners of the triangle.
+
+						Output, REAL TRIANGLE_AREA_3D, the area of the triangle.
+						*/
+{
+  REAL a;
+  REAL alpha;
+  REAL area;
+  REAL b;
+  REAL base;
+  REAL c;
+  REAL dot;
+  REAL height;
+  /*
+  Find the projection of (P3-P1) onto (P2-P1).
+  */
+  dot = 
+	( x2 - x1 ) * ( x3 - x1 ) +
+	( y2 - y1 ) * ( y3 - y1 ) +
+	( z2 - z1 ) * ( z3 - z1 );
+
+  base = enorm0_3d ( x1, y1, z1, x2, y2, z2 );
+  /*
+  The height of the triangle is the length of (P3-P1) after its
+  projection onto (P2-P1) has been subtracted.
+  */
+  if ( base == 0.0 ) {
+
+	height = 0.0;
+
+  }
+  else {
+
+	alpha = dot / ( base * base );
+
+	a = x3 - x1 - alpha * ( x2 - x1 );
+	b = y3 - y1 - alpha * ( y2 - y1 );
+	c = z3 - z1 - alpha * ( z2 - z1 );
+
+	height = (REAL)sqrt ( a * a + b * b + c * c );
+
+  }
+
+  area = 0.5f * base * height;
+
+  return area;
+}
+
+
+REAL fm_computeArea(const REAL *p1,const REAL *p2,const REAL *p3)
+{
+  REAL ret = 0;
+
+  ret = triangle_area_3d(p1[0],p1[1],p1[2],p2[0],p2[1],p2[2],p3[0],p3[1],p3[2]);
+
+  return ret;
+}
+
+
+void  fm_lerp(const REAL *p1,const REAL *p2,REAL *dest,REAL lerpValue)
+{
+  dest[0] = ((p2[0] - p1[0])*lerpValue) + p1[0];
+  dest[1] = ((p2[1] - p1[1])*lerpValue) + p1[1];
+  dest[2] = ((p2[2] - p1[2])*lerpValue) + p1[2];
+}
+
+bool fm_pointTestXZ(const REAL *p,const REAL *i,const REAL *j)
+{
+  bool ret = false;
+
+  if (((( i[2] <= p[2] ) && ( p[2]  < j[2] )) || (( j[2] <= p[2] ) && ( p[2]  < i[2] ))) && ( p[0] < (j[0] - i[0]) * (p[2] - i[2]) / (j[2] - i[2]) + i[0]))
+	ret = true;
+
+  return ret;
+};
+
+
+bool  fm_insideTriangleXZ(const REAL *p,const REAL *p1,const REAL *p2,const REAL *p3)
+{
+  bool ret = false;
+
+  int32_t c = 0;
+  if ( fm_pointTestXZ(p,p1,p2) ) c = !c;
+  if ( fm_pointTestXZ(p,p2,p3) ) c = !c;
+  if ( fm_pointTestXZ(p,p3,p1) ) c = !c;
+  if ( c ) ret = true;
+
+  return ret;
+}
+
+bool  fm_insideAABB(const REAL *pos,const REAL *bmin,const REAL *bmax)
+{
+  bool ret = false;
+
+  if ( pos[0] >= bmin[0] && pos[0] <= bmax[0] &&
+	   pos[1] >= bmin[1] && pos[1] <= bmax[1] &&
+	   pos[2] >= bmin[2] && pos[2] <= bmax[2] )
+	ret = true;
+
+  return ret;
+}
+
+
+uint32_t fm_clipTestPoint(const REAL *bmin,const REAL *bmax,const REAL *pos)
+{
+  uint32_t ret = 0;
+
+  if ( pos[0] < bmin[0] )
+	ret|=FMCS_XMIN;
+  else if ( pos[0] > bmax[0] )
+	ret|=FMCS_XMAX;
+
+  if ( pos[1] < bmin[1] )
+	ret|=FMCS_YMIN;
+  else if ( pos[1] > bmax[1] )
+	ret|=FMCS_YMAX;
+
+  if ( pos[2] < bmin[2] )
+	ret|=FMCS_ZMIN;
+  else if ( pos[2] > bmax[2] )
+	ret|=FMCS_ZMAX;
+
+  return ret;
+}
+
+uint32_t fm_clipTestPointXZ(const REAL *bmin,const REAL *bmax,const REAL *pos) // only tests X and Z, not Y
+{
+  uint32_t ret = 0;
+
+  if ( pos[0] < bmin[0] )
+	ret|=FMCS_XMIN;
+  else if ( pos[0] > bmax[0] )
+	ret|=FMCS_XMAX;
+
+  if ( pos[2] < bmin[2] )
+	ret|=FMCS_ZMIN;
+  else if ( pos[2] > bmax[2] )
+	ret|=FMCS_ZMAX;
+
+  return ret;
+}
+
+uint32_t fm_clipTestAABB(const REAL *bmin,const REAL *bmax,const REAL *p1,const REAL *p2,const REAL *p3,uint32_t &andCode)
+{
+  uint32_t orCode = 0;
+
+  andCode = FMCS_XMIN | FMCS_XMAX | FMCS_YMIN | FMCS_YMAX | FMCS_ZMIN | FMCS_ZMAX;
+
+  uint32_t c = fm_clipTestPoint(bmin,bmax,p1);
+  orCode|=c;
+  andCode&=c;
+
+  c = fm_clipTestPoint(bmin,bmax,p2);
+  orCode|=c;
+  andCode&=c;
+
+  c = fm_clipTestPoint(bmin,bmax,p3);
+  orCode|=c;
+  andCode&=c;
+
+  return orCode;
+}
+
+bool intersect(const REAL *si,const REAL *ei,const REAL *bmin,const REAL *bmax,REAL *time)
+{
+  REAL st,et,fst = 0,fet = 1;
+
+  for (int32_t i = 0; i < 3; i++)
+  {
+	if (*si < *ei)
+	{
+	  if (*si > *bmax || *ei < *bmin)
+		return false;
+	  REAL di = *ei - *si;
+	  st = (*si < *bmin)? (*bmin - *si) / di: 0;
+	  et = (*ei > *bmax)? (*bmax - *si) / di: 1;
+	}
+	else
+	{
+	  if (*ei > *bmax || *si < *bmin)
+		return false;
+	  REAL di = *ei - *si;
+	  st = (*si > *bmax)? (*bmax - *si) / di: 0;
+	  et = (*ei < *bmin)? (*bmin - *si) / di: 1;
+	}
+
+	if (st > fst) fst = st;
+	if (et < fet) fet = et;
+	if (fet < fst)
+	  return false;
+	bmin++; bmax++;
+	si++; ei++;
+  }
+
+  *time = fst;
+  return true;
+}
+
+
+
+bool fm_lineTestAABB(const REAL *p1,const REAL *p2,const REAL *bmin,const REAL *bmax,REAL &time)
+{
+  bool sect = intersect(p1,p2,bmin,bmax,&time);
+  return sect;
+}
+
+
+bool fm_lineTestAABBXZ(const REAL *p1,const REAL *p2,const REAL *bmin,const REAL *bmax,REAL &time)
+{
+  REAL _bmin[3];
+  REAL _bmax[3];
+
+  _bmin[0] = bmin[0];
+  _bmin[1] = -1e9;
+  _bmin[2] = bmin[2];
+
+  _bmax[0] = bmax[0];
+  _bmax[1] = 1e9;
+  _bmax[2] = bmax[2];
+
+  bool sect = intersect(p1,p2,_bmin,_bmax,&time);
+
+  return sect;
+}
+
+void  fm_minmax(const REAL *p,REAL *bmin,REAL *bmax) // accmulate to a min-max value
+{
+
+  if ( p[0] < bmin[0] ) bmin[0] = p[0];
+  if ( p[1] < bmin[1] ) bmin[1] = p[1];
+  if ( p[2] < bmin[2] ) bmin[2] = p[2];
+
+  if ( p[0] > bmax[0] ) bmax[0] = p[0];
+  if ( p[1] > bmax[1] ) bmax[1] = p[1];
+  if ( p[2] > bmax[2] ) bmax[2] = p[2];
+
+}
+
+REAL fm_solveX(const REAL *plane,REAL y,REAL z) // solve for X given this plane equation and the other two components.
+{
+  REAL x = (y*plane[1]+z*plane[2]+plane[3]) / -plane[0];
+  return x;
+}
+
+REAL fm_solveY(const REAL *plane,REAL x,REAL z) // solve for Y given this plane equation and the other two components.
+{
+  REAL y = (x*plane[0]+z*plane[2]+plane[3]) / -plane[1];
+  return y;
+}
+
+
+REAL fm_solveZ(const REAL *plane,REAL x,REAL y) // solve for Y given this plane equation and the other two components.
+{
+  REAL z = (x*plane[0]+y*plane[1]+plane[3]) / -plane[2];
+  return z;
+}
+
+
+void  fm_getAABBCenter(const REAL *bmin,const REAL *bmax,REAL *center)
+{
+  center[0] = (bmax[0]-bmin[0])*0.5f+bmin[0];
+  center[1] = (bmax[1]-bmin[1])*0.5f+bmin[1];
+  center[2] = (bmax[2]-bmin[2])*0.5f+bmin[2];
+}
+
+FM_Axis fm_getDominantAxis(const REAL normal[3])
+{
+  FM_Axis ret = FM_XAXIS;
+
+  REAL x = (REAL)fabs(normal[0]);
+  REAL y = (REAL)fabs(normal[1]);
+  REAL z = (REAL)fabs(normal[2]);
+
+  if ( y > x && y > z )
+	ret = FM_YAXIS;
+  else if ( z > x && z > y )
+	ret = FM_ZAXIS;
+
+  return ret;
+}
+
+
+bool fm_lineSphereIntersect(const REAL *center,REAL radius,const REAL *p1,const REAL *p2,REAL *intersect)
+{
+  bool ret = false;
+
+  REAL dir[3];
+
+  dir[0] = p2[0]-p1[0];
+  dir[1] = p2[1]-p1[1];
+  dir[2] = p2[2]-p1[2];
+
+  REAL distance = (REAL)sqrt( dir[0]*dir[0]+dir[1]*dir[1]+dir[2]*dir[2]);
+
+  if ( distance > 0 )
+  {
+	REAL recip = 1.0f / distance;
+	dir[0]*=recip;
+	dir[1]*=recip;
+	dir[2]*=recip;
+	ret = fm_raySphereIntersect(center,radius,p1,dir,distance,intersect);
+  }
+  else
+  {
+	dir[0] = center[0]-p1[0];
+	dir[1] = center[1]-p1[1];
+	dir[2] = center[2]-p1[2];
+	REAL d2 = dir[0]*dir[0]+dir[1]*dir[1]+dir[2]*dir[2];
+	REAL r2 = radius*radius;
+	if ( d2 < r2 )
+	{
+	  ret = true;
+	  if ( intersect )
+	  {
+		intersect[0] = p1[0];
+		intersect[1] = p1[1];
+		intersect[2] = p1[2];
+	  }
+	}
+  }
+  return ret;
+}
+
+#define DOT(p1,p2) (p1[0]*p2[0]+p1[1]*p2[1]+p1[2]*p2[2])
+
+bool fm_raySphereIntersect(const REAL *center,REAL radius,const REAL *pos,const REAL *dir,REAL distance,REAL *intersect)
+{
+  bool ret = false;
+
+  REAL E0[3];
+
+  E0[0] = center[0] - pos[0];
+  E0[1] = center[1] - pos[1];
+  E0[2] = center[2] - pos[2];
+
+  REAL V[3];
+
+  V[0]  = dir[0];
+  V[1]  = dir[1];
+  V[2]  = dir[2];
+
+
+  REAL dist2   = E0[0]*E0[0] + E0[1]*E0[1] + E0[2] * E0[2];
+  REAL radius2 = radius*radius; // radius squared..
+
+  // Bug Fix For Gem, if origin is *inside* the sphere, invert the
+  // direction vector so that we get a valid intersection location.
+  if ( dist2 < radius2 )
+  {
+	V[0]*=-1;
+	V[1]*=-1;
+	V[2]*=-1;
+  }
+
+
+	REAL v = DOT(E0,V);
+
+	REAL disc = radius2 - (dist2 - v*v);
+
+	if (disc > 0.0f)
+	{
+		if ( intersect )
+		{
+		  REAL d = (REAL)sqrt(disc);
+	  REAL diff = v-d;
+	  if ( diff < distance )
+	  {
+		intersect[0] = pos[0]+V[0]*diff;
+		intersect[1] = pos[1]+V[1]*diff;
+		intersect[2] = pos[2]+V[2]*diff;
+		ret = true;
+	  }
+	}
+	}
+
+	return ret;
+}
+
+
+void fm_catmullRom(REAL *out_vector,const REAL *p1,const REAL *p2,const REAL *p3,const REAL *p4, const REAL s)
+{
+  REAL s_squared = s * s;
+  REAL s_cubed = s_squared * s;
+
+  REAL coefficient_p1 = -s_cubed + 2*s_squared - s;
+  REAL coefficient_p2 = 3 * s_cubed - 5 * s_squared + 2;
+  REAL coefficient_p3 = -3 * s_cubed +4 * s_squared + s;
+  REAL coefficient_p4 = s_cubed - s_squared;
+
+  out_vector[0] = (coefficient_p1 * p1[0] + coefficient_p2 * p2[0] + coefficient_p3 * p3[0] + coefficient_p4 * p4[0])*0.5f;
+  out_vector[1] = (coefficient_p1 * p1[1] + coefficient_p2 * p2[1] + coefficient_p3 * p3[1] + coefficient_p4 * p4[1])*0.5f;
+  out_vector[2] = (coefficient_p1 * p1[2] + coefficient_p2 * p2[2] + coefficient_p3 * p3[2] + coefficient_p4 * p4[2])*0.5f;
+}
+
+bool fm_intersectAABB(const REAL *bmin1,const REAL *bmax1,const REAL *bmin2,const REAL *bmax2)
+{
+  if ((bmin1[0] > bmax2[0]) || (bmin2[0] > bmax1[0])) return false;
+  if ((bmin1[1] > bmax2[1]) || (bmin2[1] > bmax1[1])) return false;
+  if ((bmin1[2] > bmax2[2]) || (bmin2[2] > bmax1[2])) return false;
+  return true;
+
+}
+
+bool  fm_insideAABB(const REAL *obmin,const REAL *obmax,const REAL *tbmin,const REAL *tbmax) // test if bounding box tbmin/tmbax is fully inside obmin/obmax
+{
+  bool ret = false;
+
+  if ( tbmax[0] <= obmax[0] &&
+	   tbmax[1] <= obmax[1] &&
+	   tbmax[2] <= obmax[2] &&
+	   tbmin[0] >= obmin[0] &&
+	   tbmin[1] >= obmin[1] &&
+	   tbmin[2] >= obmin[2] ) ret = true;
+
+  return ret;
+}
+
+
+// Reference, from Stan Melax in Game Gems I
+//  Quaternion q;
+//  vector3 c = CrossProduct(v0,v1);
+//  REAL   d = DotProduct(v0,v1);
+//  REAL   s = (REAL)sqrt((1+d)*2);
+//  q.x = c.x / s;
+//  q.y = c.y / s;
+//  q.z = c.z / s;
+//  q.w = s /2.0f;
+//  return q;
+void fm_rotationArc(const REAL *v0,const REAL *v1,REAL *quat)
+{
+  REAL cross[3];
+
+  fm_cross(cross,v0,v1);
+  REAL d = fm_dot(v0,v1);
+
+  if( d<= -0.99999f ) // 180 about x axis
+  {
+	  if ( fabsf((float)v0[0]) < 0.1f )
+	  {
+		  quat[0] = 0;
+		  quat[1] = v0[2];
+		  quat[2] = -v0[1];
+		  quat[3] = 0;
+	  }
+	  else
+	  {
+		  quat[0] = v0[1];
+		  quat[1] = -v0[0];
+		  quat[2] = 0;
+		  quat[3] = 0;
+	  }
+	  REAL magnitudeSquared = quat[0]*quat[0] + quat[1]*quat[1] + quat[2]*quat[2] + quat[3]*quat[3];
+	  REAL magnitude = sqrtf((float)magnitudeSquared);
+	  REAL recip = 1.0f / magnitude;
+	  quat[0]*=recip;
+	  quat[1]*=recip;
+	  quat[2]*=recip;
+	  quat[3]*=recip;
+  }
+  else
+  {
+	  REAL s = (REAL)sqrt((1+d)*2);
+	  REAL recip = 1.0f / s;
+
+	  quat[0] = cross[0] * recip;
+	  quat[1] = cross[1] * recip;
+	  quat[2] = cross[2] * recip;
+	  quat[3] = s * 0.5f;
+  }
+}
+
+
+REAL fm_distancePointLineSegment(const REAL *Point,const REAL *LineStart,const REAL *LineEnd,REAL *intersection,LineSegmentType &type,REAL epsilon)
+{
+  REAL ret;
+
+  REAL LineMag = fm_distance( LineEnd, LineStart );
+
+  if ( LineMag > 0 )
+  {
+	REAL U = ( ( ( Point[0] - LineStart[0] ) * ( LineEnd[0] - LineStart[0] ) ) + ( ( Point[1] - LineStart[1] ) * ( LineEnd[1] - LineStart[1] ) ) + ( ( Point[2] - LineStart[2] ) * ( LineEnd[2] - LineStart[2] ) ) ) / ( LineMag * LineMag );
+	if( U < 0.0f || U > 1.0f )
+	{
+	  REAL d1 = fm_distanceSquared(Point,LineStart);
+	  REAL d2 = fm_distanceSquared(Point,LineEnd);
+	  if ( d1 <= d2 )
+	  {
+		ret = (REAL)sqrt(d1);
+		intersection[0] = LineStart[0];
+		intersection[1] = LineStart[1];
+		intersection[2] = LineStart[2];
+		type = LS_START;
+	  }
+	  else
+	  {
+		ret = (REAL)sqrt(d2);
+		intersection[0] = LineEnd[0];
+		intersection[1] = LineEnd[1];
+		intersection[2] = LineEnd[2];
+		type = LS_END;
+	  }
+	}
+	else
+	{
+	  intersection[0] = LineStart[0] + U * ( LineEnd[0] - LineStart[0] );
+	  intersection[1] = LineStart[1] + U * ( LineEnd[1] - LineStart[1] );
+	  intersection[2] = LineStart[2] + U * ( LineEnd[2] - LineStart[2] );
+
+	  ret = fm_distance(Point,intersection);
+
+	  REAL d1 = fm_distanceSquared(intersection,LineStart);
+	  REAL d2 = fm_distanceSquared(intersection,LineEnd);
+	  REAL mag = (epsilon*2)*(epsilon*2);
+
+	  if ( d1 < mag ) // if less than 1/100th the total distance, treat is as the 'start'
+	  {
+		type = LS_START;
+	  }
+	  else if ( d2 < mag )
+	  {
+		type = LS_END;
+	  }
+	  else
+	  {
+		type = LS_MIDDLE;
+	  }
+
+	}
+  }
+  else
+  {
+	ret = LineMag;
+	intersection[0] = LineEnd[0];
+	intersection[1] = LineEnd[1];
+	intersection[2] = LineEnd[2];
+	type = LS_END;
+  }
+
+  return ret;
+}
+
+
+#ifndef BEST_FIT_PLANE_H
+
+#define BEST_FIT_PLANE_H
+
+template <class Type> class Eigen
+{
+public:
+
+
+  void DecrSortEigenStuff(void)
+  {
+	Tridiagonal(); //diagonalize the matrix.
+	QLAlgorithm(); //
+	DecreasingSort();
+	GuaranteeRotation();
+  }
+
+  void Tridiagonal(void)
+  {
+	Type fM00 = mElement[0][0];
+	Type fM01 = mElement[0][1];
+	Type fM02 = mElement[0][2];
+	Type fM11 = mElement[1][1];
+	Type fM12 = mElement[1][2];
+	Type fM22 = mElement[2][2];
+
+	m_afDiag[0] = fM00;
+	m_afSubd[2] = 0;
+	if (fM02 != (Type)0.0)
+	{
+	  Type fLength = (REAL)sqrt(fM01*fM01+fM02*fM02);
+	  Type fInvLength = ((Type)1.0)/fLength;
+	  fM01 *= fInvLength;
+	  fM02 *= fInvLength;
+	  Type fQ = ((Type)2.0)*fM01*fM12+fM02*(fM22-fM11);
+	  m_afDiag[1] = fM11+fM02*fQ;
+	  m_afDiag[2] = fM22-fM02*fQ;
+	  m_afSubd[0] = fLength;
+	  m_afSubd[1] = fM12-fM01*fQ;
+	  mElement[0][0] = (Type)1.0;
+	  mElement[0][1] = (Type)0.0;
+	  mElement[0][2] = (Type)0.0;
+	  mElement[1][0] = (Type)0.0;
+	  mElement[1][1] = fM01;
+	  mElement[1][2] = fM02;
+	  mElement[2][0] = (Type)0.0;
+	  mElement[2][1] = fM02;
+	  mElement[2][2] = -fM01;
+	  m_bIsRotation = false;
+	}
+	else
+	{
+	  m_afDiag[1] = fM11;
+	  m_afDiag[2] = fM22;
+	  m_afSubd[0] = fM01;
+	  m_afSubd[1] = fM12;
+	  mElement[0][0] = (Type)1.0;
+	  mElement[0][1] = (Type)0.0;
+	  mElement[0][2] = (Type)0.0;
+	  mElement[1][0] = (Type)0.0;
+	  mElement[1][1] = (Type)1.0;
+	  mElement[1][2] = (Type)0.0;
+	  mElement[2][0] = (Type)0.0;
+	  mElement[2][1] = (Type)0.0;
+	  mElement[2][2] = (Type)1.0;
+	  m_bIsRotation = true;
+	}
+  }
+
+  bool QLAlgorithm(void)
+  {
+	const int32_t iMaxIter = 32;
+
+	for (int32_t i0 = 0; i0 <3; i0++)
+	{
+	  int32_t i1;
+	  for (i1 = 0; i1 < iMaxIter; i1++)
+	  {
+		int32_t i2;
+		for (i2 = i0; i2 <= (3-2); i2++)
+		{
+		  Type fTmp = fabs(m_afDiag[i2]) + fabs(m_afDiag[i2+1]);
+		  if ( fabs(m_afSubd[i2]) + fTmp == fTmp )
+			break;
+		}
+		if (i2 == i0)
+		{
+		  break;
+		}
+
+		Type fG = (m_afDiag[i0+1] - m_afDiag[i0])/(((Type)2.0) * m_afSubd[i0]);
+		Type fR = (REAL)sqrt(fG*fG+(Type)1.0);
+		if (fG < (Type)0.0)
+		{
+		  fG = m_afDiag[i2]-m_afDiag[i0]+m_afSubd[i0]/(fG-fR);
+		}
+		else
+		{
+		  fG = m_afDiag[i2]-m_afDiag[i0]+m_afSubd[i0]/(fG+fR);
+		}
+		Type fSin = (Type)1.0, fCos = (Type)1.0, fP = (Type)0.0;
+		for (int32_t i3 = i2-1; i3 >= i0; i3--)
+		{
+		  Type fF = fSin*m_afSubd[i3];
+		  Type fB = fCos*m_afSubd[i3];
+		  if (fabs(fF) >= fabs(fG))
+		  {
+			fCos = fG/fF;
+			fR = (REAL)sqrt(fCos*fCos+(Type)1.0);
+			m_afSubd[i3+1] = fF*fR;
+			fSin = ((Type)1.0)/fR;
+			fCos *= fSin;
+		  }
+		  else
+		  {
+			fSin = fF/fG;
+			fR = (REAL)sqrt(fSin*fSin+(Type)1.0);
+			m_afSubd[i3+1] = fG*fR;
+			fCos = ((Type)1.0)/fR;
+			fSin *= fCos;
+		  }
+		  fG = m_afDiag[i3+1]-fP;
+		  fR = (m_afDiag[i3]-fG)*fSin+((Type)2.0)*fB*fCos;
+		  fP = fSin*fR;
+		  m_afDiag[i3+1] = fG+fP;
+		  fG = fCos*fR-fB;
+		  for (int32_t i4 = 0; i4 < 3; i4++)
+		  {
+			fF = mElement[i4][i3+1];
+			mElement[i4][i3+1] = fSin*mElement[i4][i3]+fCos*fF;
+			mElement[i4][i3] = fCos*mElement[i4][i3]-fSin*fF;
+		  }
+		}
+		m_afDiag[i0] -= fP;
+		m_afSubd[i0] = fG;
+		m_afSubd[i2] = (Type)0.0;
+	  }
+	  if (i1 == iMaxIter)
+	  {
+		return false;
+	  }
+	}
+	return true;
+  }
+
+  void DecreasingSort(void)
+  {
+	//sort eigenvalues in decreasing order, e[0] >= ... >= e[iSize-1]
+	for (int32_t i0 = 0, i1; i0 <= 3-2; i0++)
+	{
+	  // locate maximum eigenvalue
+	  i1 = i0;
+	  Type fMax = m_afDiag[i1];
+	  int32_t i2;
+	  for (i2 = i0+1; i2 < 3; i2++)
+	  {
+		if (m_afDiag[i2] > fMax)
+		{
+		  i1 = i2;
+		  fMax = m_afDiag[i1];
+		}
+	  }
+
+	  if (i1 != i0)
+	  {
+		// swap eigenvalues
+		m_afDiag[i1] = m_afDiag[i0];
+		m_afDiag[i0] = fMax;
+		// swap eigenvectors
+		for (i2 = 0; i2 < 3; i2++)
+		{
+		  Type fTmp = mElement[i2][i0];
+		  mElement[i2][i0] = mElement[i2][i1];
+		  mElement[i2][i1] = fTmp;
+		  m_bIsRotation = !m_bIsRotation;
+		}
+	  }
+	}
+  }
+
+
+  void GuaranteeRotation(void)
+  {
+	if (!m_bIsRotation)
+	{
+	  // change sign on the first column
+	  for (int32_t iRow = 0; iRow <3; iRow++)
+	  {
+		mElement[iRow][0] = -mElement[iRow][0];
+	  }
+	}
+  }
+
+  Type mElement[3][3];
+  Type m_afDiag[3];
+  Type m_afSubd[3];
+  bool m_bIsRotation;
+};
+
+#endif
+
+bool fm_computeBestFitPlane(uint32_t vcount,
+					 const REAL *points,
+					 uint32_t vstride,
+					 const REAL *weights,
+					 uint32_t wstride,
+					 REAL *plane,
+					REAL *center)
+{
+  bool ret = false;
+
+  REAL kOrigin[3] = { 0, 0, 0 };
+
+  REAL wtotal = 0;
+
+  {
+	const char *source  = (const char *) points;
+	const char *wsource = (const char *) weights;
+
+	for (uint32_t i=0; i<vcount; i++)
+	{
+
+	  const REAL *p = (const REAL *) source;
+
+	  REAL w = 1;
+
+	  if ( wsource )
+	  {
+		const REAL *ws = (const REAL *) wsource;
+		w = *ws; //
+		wsource+=wstride;
+	  }
+
+	  kOrigin[0]+=p[0]*w;
+	  kOrigin[1]+=p[1]*w;
+	  kOrigin[2]+=p[2]*w;
+
+	  wtotal+=w;
+
+	  source+=vstride;
+	}
+  }
+
+  REAL recip = 1.0f / wtotal; // reciprocol of total weighting
+
+  kOrigin[0]*=recip;
+  kOrigin[1]*=recip;
+  kOrigin[2]*=recip;
+
+  center[0] = kOrigin[0];
+  center[1] = kOrigin[1];
+  center[2] = kOrigin[2];
+
+
+  REAL fSumXX=0;
+  REAL fSumXY=0;
+  REAL fSumXZ=0;
+
+  REAL fSumYY=0;
+  REAL fSumYZ=0;
+  REAL fSumZZ=0;
+
+
+  {
+	const char *source  = (const char *) points;
+	const char *wsource = (const char *) weights;
+
+	for (uint32_t i=0; i<vcount; i++)
+	{
+
+	  const REAL *p = (const REAL *) source;
+
+	  REAL w = 1;
+
+	  if ( wsource )
+	  {
+		const REAL *ws = (const REAL *) wsource;
+		w = *ws; //
+		wsource+=wstride;
+	  }
+
+	  REAL kDiff[3];
+
+	  kDiff[0] = w*(p[0] - kOrigin[0]); // apply vertex weighting!
+	  kDiff[1] = w*(p[1] - kOrigin[1]);
+	  kDiff[2] = w*(p[2] - kOrigin[2]);
+
+	  fSumXX+= kDiff[0] * kDiff[0]; // sume of the squares of the differences.
+	  fSumXY+= kDiff[0] * kDiff[1]; // sume of the squares of the differences.
+	  fSumXZ+= kDiff[0] * kDiff[2]; // sume of the squares of the differences.
+
+	  fSumYY+= kDiff[1] * kDiff[1];
+	  fSumYZ+= kDiff[1] * kDiff[2];
+	  fSumZZ+= kDiff[2] * kDiff[2];
+
+
+	  source+=vstride;
+	}
+  }
+
+  fSumXX *= recip;
+  fSumXY *= recip;
+  fSumXZ *= recip;
+  fSumYY *= recip;
+  fSumYZ *= recip;
+  fSumZZ *= recip;
+
+  // setup the eigensolver
+  Eigen<REAL> kES;
+
+  kES.mElement[0][0] = fSumXX;
+  kES.mElement[0][1] = fSumXY;
+  kES.mElement[0][2] = fSumXZ;
+
+  kES.mElement[1][0] = fSumXY;
+  kES.mElement[1][1] = fSumYY;
+  kES.mElement[1][2] = fSumYZ;
+
+  kES.mElement[2][0] = fSumXZ;
+  kES.mElement[2][1] = fSumYZ;
+  kES.mElement[2][2] = fSumZZ;
+
+  // compute eigenstuff, smallest eigenvalue is in last position
+  kES.DecrSortEigenStuff();
+
+  REAL kNormal[3];
+
+  kNormal[0] = kES.mElement[0][2];
+  kNormal[1] = kES.mElement[1][2];
+  kNormal[2] = kES.mElement[2][2];
+
+  // the minimum energy
+  plane[0] = kNormal[0];
+  plane[1] = kNormal[1];
+  plane[2] = kNormal[2];
+
+  plane[3] = 0 - fm_dot(kNormal,kOrigin);
+
+  ret = true;
+
+  return ret;
+}
+
+
+bool fm_colinear(const REAL a1[3],const REAL a2[3],const REAL b1[3],const REAL b2[3],REAL epsilon)  // true if these two line segments are co-linear.
+{
+  bool ret = false;
+
+  REAL dir1[3];
+  REAL dir2[3];
+
+  dir1[0] = (a2[0] - a1[0]);
+  dir1[1] = (a2[1] - a1[1]);
+  dir1[2] = (a2[2] - a1[2]);
+
+  dir2[0] = (b2[0]-a1[0]) - (b1[0]-a1[0]);
+  dir2[1] = (b2[1]-a1[1]) - (b1[1]-a1[1]);
+  dir2[2] = (b2[2]-a2[2]) - (b1[2]-a2[2]);
+
+  fm_normalize(dir1);
+  fm_normalize(dir2);
+
+  REAL dot = fm_dot(dir1,dir2);
+
+  if ( dot >= epsilon )
+  {
+	ret = true;
+  }
+
+
+  return ret;
+}
+
+bool fm_colinear(const REAL *p1,const REAL *p2,const REAL *p3,REAL epsilon)
+{
+  bool ret = false;
+
+  REAL dir1[3];
+  REAL dir2[3];
+
+  dir1[0] = p2[0] - p1[0];
+  dir1[1] = p2[1] - p1[1];
+  dir1[2] = p2[2] - p1[2];
+
+  dir2[0] = p3[0] - p2[0];
+  dir2[1] = p3[1] - p2[1];
+  dir2[2] = p3[2] - p2[2];
+
+  fm_normalize(dir1);
+  fm_normalize(dir2);
+
+  REAL dot = fm_dot(dir1,dir2);
+
+  if ( dot >= epsilon )
+  {
+	ret = true;
+  }
+
+
+  return ret;
+}
+
+void  fm_initMinMax(const REAL *p,REAL *bmin,REAL *bmax)
+{
+  bmax[0] = bmin[0] = p[0];
+  bmax[1] = bmin[1] = p[1];
+  bmax[2] = bmin[2] = p[2];
+}
+
+IntersectResult fm_intersectLineSegments2d(const REAL *a1,const REAL *a2,const REAL *b1,const REAL *b2,REAL *intersection)
+{
+  IntersectResult ret;
+
+  REAL denom  = ((b2[1] - b1[1])*(a2[0] - a1[0])) - ((b2[0] - b1[0])*(a2[1] - a1[1]));
+  REAL nume_a = ((b2[0] - b1[0])*(a1[1] - b1[1])) - ((b2[1] - b1[1])*(a1[0] - b1[0]));
+  REAL nume_b = ((a2[0] - a1[0])*(a1[1] - b1[1])) - ((a2[1] - a1[1])*(a1[0] - b1[0]));
+  if (denom == 0 )
+  {
+	if(nume_a == 0 && nume_b == 0)
+	{
+	  ret = IR_COINCIDENT;
+	}
+	else
+	{
+	  ret = IR_PARALLEL;
+	}
+  }
+  else
+  {
+
+	REAL recip = 1 / denom;
+	REAL ua = nume_a * recip;
+	REAL ub = nume_b * recip;
+
+	if(ua >= 0 && ua <= 1 && ub >= 0 && ub <= 1 )
+	{
+	  // Get the intersection point.
+	  intersection[0] = a1[0] + ua*(a2[0] - a1[0]);
+	  intersection[1] = a1[1] + ua*(a2[1] - a1[1]);
+	  ret = IR_DO_INTERSECT;
+	}
+	else
+	{
+	  ret = IR_DONT_INTERSECT;
+	}
+  }
+  return ret;
+}
+
+IntersectResult fm_intersectLineSegments2dTime(const REAL *a1,const REAL *a2,const REAL *b1,const REAL *b2,REAL &t1,REAL &t2)
+{
+  IntersectResult ret;
+
+  REAL denom  = ((b2[1] - b1[1])*(a2[0] - a1[0])) - ((b2[0] - b1[0])*(a2[1] - a1[1]));
+  REAL nume_a = ((b2[0] - b1[0])*(a1[1] - b1[1])) - ((b2[1] - b1[1])*(a1[0] - b1[0]));
+  REAL nume_b = ((a2[0] - a1[0])*(a1[1] - b1[1])) - ((a2[1] - a1[1])*(a1[0] - b1[0]));
+  if (denom == 0 )
+  {
+	if(nume_a == 0 && nume_b == 0)
+	{
+	  ret = IR_COINCIDENT;
+	}
+	else
+	{
+	  ret = IR_PARALLEL;
+	}
+  }
+  else
+  {
+
+	REAL recip = 1 / denom;
+	REAL ua = nume_a * recip;
+	REAL ub = nume_b * recip;
+
+	if(ua >= 0 && ua <= 1 && ub >= 0 && ub <= 1 )
+	{
+	  t1 = ua;
+	  t2 = ub;
+	  ret = IR_DO_INTERSECT;
+	}
+	else
+	{
+	  ret = IR_DONT_INTERSECT;
+	}
+  }
+  return ret;
+}
+
+//**** Plane Triangle Intersection
+
+
+
+
+
+// assumes that the points are on opposite sides of the plane!
+bool fm_intersectPointPlane(const REAL *p1,const REAL *p2,REAL *split,const REAL *plane)
+{
+
+  REAL dp1 = fm_distToPlane(plane,p1);
+  REAL dp2 = fm_distToPlane(plane, p2);
+  if (dp1 <= 0 && dp2 <= 0)
+  {
+	  return false;
+  }
+  if (dp1 >= 0 && dp2 >= 0)
+  {
+	  return false;
+  }
+
+  REAL dir[3];
+
+  dir[0] = p2[0] - p1[0];
+  dir[1] = p2[1] - p1[1];
+  dir[2] = p2[2] - p1[2];
+
+  REAL dot1 = dir[0]*plane[0] + dir[1]*plane[1] + dir[2]*plane[2];
+  REAL dot2 = dp1 - plane[3];
+
+  REAL    t = -(plane[3] + dot2 ) / dot1;
+
+  split[0] = (dir[0]*t)+p1[0];
+  split[1] = (dir[1]*t)+p1[1];
+  split[2] = (dir[2]*t)+p1[2];
+
+  return true;
+}
+
+PlaneTriResult fm_getSidePlane(const REAL *p,const REAL *plane,REAL epsilon)
+{
+  PlaneTriResult ret = PTR_ON_PLANE;
+
+  REAL d = fm_distToPlane(plane,p);
+
+  if ( d < -epsilon || d > epsilon )
+  {
+	if ( d > 0 )
+		ret =  PTR_FRONT; // it is 'in front' within the provided epsilon value.
+	else
+	  ret = PTR_BACK;
+  }
+
+  return ret;
+}
+
+
+
+#ifndef PLANE_TRIANGLE_INTERSECTION_H
+
+#define PLANE_TRIANGLE_INTERSECTION_H
+
+#define MAXPTS 256
+
+template <class Type> class point
+{
+public:
+
+  void set(const Type *p)
+  {
+	x = p[0];
+	y = p[1];
+	z = p[2];
+  }
+
+  Type x;
+  Type y;
+  Type z;
+};
+
+template <class Type> class plane
+{
+public:
+  plane(const Type *p)
+  {
+	normal.x = p[0];
+	normal.y = p[1];
+	normal.z = p[2];
+	D        = p[3];
+  }
+
+  Type Classify_Point(const point<Type> &p)
+  {
+	return p.x*normal.x + p.y*normal.y + p.z*normal.z + D;
+  }
+
+  point<Type> normal;
+  Type  D;
+};
+
+template <class Type> class polygon
+{
+public:
+  polygon(void)
+  {
+	mVcount = 0;
+  }
+
+  polygon(const Type *p1,const Type *p2,const Type *p3)
+  {
+	mVcount = 3;
+	mVertices[0].set(p1);
+	mVertices[1].set(p2);
+	mVertices[2].set(p3);
+  }
+
+
+  int32_t NumVertices(void) const { return mVcount; };
+
+  const point<Type>& Vertex(int32_t index)
+  {
+	if ( index < 0 ) index+=mVcount;
+	return mVertices[index];
+  };
+
+
+  void set(const point<Type> *pts,int32_t count)
+  {
+	for (int32_t i=0; i<count; i++)
+	{
+	  mVertices[i] = pts[i];
+	}
+	mVcount = count;
+  }
+
+
+  void Split_Polygon(polygon<Type> *poly,plane<Type> *part, polygon<Type> &front, polygon<Type> &back)
+  {
+	int32_t   count = poly->NumVertices ();
+	int32_t   out_c = 0, in_c = 0;
+	point<Type> ptA, ptB,outpts[MAXPTS],inpts[MAXPTS];
+	Type sideA, sideB;
+	ptA = poly->Vertex (count - 1);
+	sideA = part->Classify_Point (ptA);
+	for (int32_t i = -1; ++i < count;)
+	{
+	  ptB = poly->Vertex(i);
+	  sideB = part->Classify_Point(ptB);
+	  if (sideB > 0)
+	  {
+		if (sideA < 0)
+		{
+			  point<Type> v;
+		  fm_intersectPointPlane(&ptB.x, &ptA.x, &v.x, &part->normal.x );
+		  outpts[out_c++] = inpts[in_c++] = v;
+		}
+		outpts[out_c++] = ptB;
+	  }
+	  else if (sideB < 0)
+	  {
+		if (sideA > 0)
+		{
+		  point<Type> v;
+		  fm_intersectPointPlane(&ptB.x, &ptA.x, &v.x, &part->normal.x );
+		  outpts[out_c++] = inpts[in_c++] = v;
+		}
+		inpts[in_c++] = ptB;
+	  }
+	  else
+		 outpts[out_c++] = inpts[in_c++] = ptB;
+	  ptA = ptB;
+	  sideA = sideB;
+	}
+
+	front.set(&outpts[0], out_c);
+	back.set(&inpts[0], in_c);
+  }
+
+  int32_t           mVcount;
+  point<Type>   mVertices[MAXPTS];
+};
+
+
+
+#endif
+
+static inline void add(const REAL *p,REAL *dest,uint32_t tstride,uint32_t &pcount)
+{
+  char *d = (char *) dest;
+  d = d + pcount*tstride;
+  dest = (REAL *) d;
+  dest[0] = p[0];
+  dest[1] = p[1];
+  dest[2] = p[2];
+  pcount++;
+	assert( pcount <= 4 );
+}
+
+
+PlaneTriResult fm_planeTriIntersection(const REAL *_plane,    // the plane equation in Ax+By+Cz+D format
+									const REAL *triangle, // the source triangle.
+									uint32_t tstride,  // stride in bytes of the input and output *vertices*
+									REAL        epsilon,  // the co-planar epsilon value.
+									REAL       *front,    // the triangle in front of the
+									uint32_t &fcount,  // number of vertices in the 'front' triangle
+									REAL       *back,     // the triangle in back of the plane
+									uint32_t &bcount) // the number of vertices in the 'back' triangle.
+{
+
+  fcount = 0;
+  bcount = 0;
+
+  const char *tsource = (const char *) triangle;
+
+  // get the three vertices of the triangle.
+  const REAL *p1     = (const REAL *) (tsource);
+  const REAL *p2     = (const REAL *) (tsource+tstride);
+  const REAL *p3     = (const REAL *) (tsource+tstride*2);
+
+
+  PlaneTriResult r1   = fm_getSidePlane(p1,_plane,epsilon); // compute the side of the plane each vertex is on
+  PlaneTriResult r2   = fm_getSidePlane(p2,_plane,epsilon);
+  PlaneTriResult r3   = fm_getSidePlane(p3,_plane,epsilon);
+
+  // If any of the points lay right *on* the plane....
+  if ( r1 == PTR_ON_PLANE || r2 == PTR_ON_PLANE || r3 == PTR_ON_PLANE )
+  {
+	// If the triangle is completely co-planar, then just treat it as 'front' and return!
+	if ( r1 == PTR_ON_PLANE && r2 == PTR_ON_PLANE && r3 == PTR_ON_PLANE )
+	{
+	  add(p1,front,tstride,fcount);
+	  add(p2,front,tstride,fcount);
+	  add(p3,front,tstride,fcount);
+	  return PTR_FRONT;
+	}
+	// Decide to place the co-planar points on the same side as the co-planar point.
+	PlaneTriResult r= PTR_ON_PLANE;
+	if ( r1 != PTR_ON_PLANE )
+	  r = r1;
+	else if ( r2 != PTR_ON_PLANE )
+	  r = r2;
+	else if ( r3 != PTR_ON_PLANE )
+	  r = r3;
+
+	if ( r1 == PTR_ON_PLANE ) r1 = r;
+	if ( r2 == PTR_ON_PLANE ) r2 = r;
+	if ( r3 == PTR_ON_PLANE ) r3 = r;
+
+  }
+
+  if ( r1 == r2 && r1 == r3 ) // if all three vertices are on the same side of the plane.
+  {
+	if ( r1 == PTR_FRONT ) // if all three are in front of the plane, then copy to the 'front' output triangle.
+	{
+	  add(p1,front,tstride,fcount);
+	  add(p2,front,tstride,fcount);
+	  add(p3,front,tstride,fcount);
+	}
+	else
+	{
+	  add(p1,back,tstride,bcount); // if all three are in 'back' then copy to the 'back' output triangle.
+	  add(p2,back,tstride,bcount);
+	  add(p3,back,tstride,bcount);
+	}
+	return r1; // if all three points are on the same side of the plane return result
+  }
+
+
+  polygon<REAL> pi(p1,p2,p3);
+  polygon<REAL>  pfront,pback;
+
+  plane<REAL>    part(_plane);
+
+  pi.Split_Polygon(&pi,&part,pfront,pback);
+
+  for (int32_t i=0; i<pfront.mVcount; i++)
+  {
+	add( &pfront.mVertices[i].x, front, tstride, fcount );
+  }
+
+  for (int32_t i=0; i<pback.mVcount; i++)
+  {
+	add( &pback.mVertices[i].x, back, tstride, bcount );
+  }
+
+  PlaneTriResult ret = PTR_SPLIT;
+
+  if ( fcount < 3 ) fcount = 0;
+  if ( bcount < 3 ) bcount = 0;
+
+  if ( fcount == 0 && bcount )
+	ret = PTR_BACK;
+
+  if ( bcount == 0 && fcount )
+	ret = PTR_FRONT;
+
+
+  return ret;
+}
+
+// computes the OBB for this set of points relative to this transform matrix.
+void computeOBB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *sides,REAL *matrix)
+{
+  const char *src = (const char *) points;
+
+  REAL bmin[3] = { 1e9, 1e9, 1e9 };
+  REAL bmax[3] = { -1e9, -1e9, -1e9 };
+
+  for (uint32_t i=0; i<vcount; i++)
+  {
+	const REAL *p = (const REAL *) src;
+	REAL t[3];
+
+	fm_inverseRT(matrix, p, t ); // inverse rotate translate
+
+	if ( t[0] < bmin[0] ) bmin[0] = t[0];
+	if ( t[1] < bmin[1] ) bmin[1] = t[1];
+	if ( t[2] < bmin[2] ) bmin[2] = t[2];
+
+	if ( t[0] > bmax[0] ) bmax[0] = t[0];
+	if ( t[1] > bmax[1] ) bmax[1] = t[1];
+	if ( t[2] > bmax[2] ) bmax[2] = t[2];
+
+	src+=pstride;
+  }
+
+  REAL center[3];
+
+  sides[0] = bmax[0]-bmin[0];
+  sides[1] = bmax[1]-bmin[1];
+  sides[2] = bmax[2]-bmin[2];
+
+  center[0] = sides[0]*0.5f+bmin[0];
+  center[1] = sides[1]*0.5f+bmin[1];
+  center[2] = sides[2]*0.5f+bmin[2];
+
+  REAL ocenter[3];
+
+  fm_rotate(matrix,center,ocenter);
+
+  matrix[12]+=ocenter[0];
+  matrix[13]+=ocenter[1];
+  matrix[14]+=ocenter[2];
+
+}
+
+void fm_computeBestFitOBB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *sides,REAL *matrix,bool bruteForce)
+{
+  REAL plane[4];
+  REAL center[3];
+  fm_computeBestFitPlane(vcount,points,pstride,0,0,plane,center);
+  fm_planeToMatrix(plane,matrix);
+  computeOBB( vcount, points, pstride, sides, matrix );
+
+  REAL refmatrix[16];
+  memcpy(refmatrix,matrix,16*sizeof(REAL));
+
+  REAL volume = sides[0]*sides[1]*sides[2];
+  if ( bruteForce )
+  {
+	for (REAL a=10; a<180; a+=10)
+	{
+	  REAL quat[4];
+	  fm_eulerToQuat(0,a*FM_DEG_TO_RAD,0,quat);
+	  REAL temp[16];
+	  REAL pmatrix[16];
+	  fm_quatToMatrix(quat,temp);
+	  fm_matrixMultiply(temp,refmatrix,pmatrix);
+	  REAL psides[3];
+	  computeOBB( vcount, points, pstride, psides, pmatrix );
+	  REAL v = psides[0]*psides[1]*psides[2];
+	  if ( v < volume )
+	  {
+		volume = v;
+		memcpy(matrix,pmatrix,sizeof(REAL)*16);
+		sides[0] = psides[0];
+		sides[1] = psides[1];
+		sides[2] = psides[2];
+	  }
+	}
+  }
+}
+
+void fm_computeBestFitOBB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *sides,REAL *pos,REAL *quat,bool bruteForce)
+{
+  REAL matrix[16];
+  fm_computeBestFitOBB(vcount,points,pstride,sides,matrix,bruteForce);
+  fm_getTranslation(matrix,pos);
+  fm_matrixToQuat(matrix,quat);
+}
+
+void fm_computeBestFitABB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *sides,REAL *pos)
+{
+	REAL bmin[3];
+	REAL bmax[3];
+
+  bmin[0] = points[0];
+  bmin[1] = points[1];
+  bmin[2] = points[2];
+
+  bmax[0] = points[0];
+  bmax[1] = points[1];
+  bmax[2] = points[2];
+
+	const char *cp = (const char *) points;
+	for (uint32_t i=0; i<vcount; i++)
+	{
+		const REAL *p = (const REAL *) cp;
+
+		if ( p[0] < bmin[0] ) bmin[0] = p[0];
+		if ( p[1] < bmin[1] ) bmin[1] = p[1];
+		if ( p[2] < bmin[2] ) bmin[2] = p[2];
+
+	if ( p[0] > bmax[0] ) bmax[0] = p[0];
+	if ( p[1] > bmax[1] ) bmax[1] = p[1];
+	if ( p[2] > bmax[2] ) bmax[2] = p[2];
+
+	cp+=pstride;
+	}
+
+
+	sides[0] = bmax[0] - bmin[0];
+	sides[1] = bmax[1] - bmin[1];
+	sides[2] = bmax[2] - bmin[2];
+
+	pos[0] = bmin[0]+sides[0]*0.5f;
+	pos[1] = bmin[1]+sides[1]*0.5f;
+	pos[2] = bmin[2]+sides[2]*0.5f;
+
+}
+
+
+void fm_planeToMatrix(const REAL *plane,REAL *matrix) // convert a plane equation to a 4x4 rotation matrix
+{
+  REAL ref[3] = { 0, 1, 0 };
+  REAL quat[4];
+  fm_rotationArc(ref,plane,quat);
+  fm_quatToMatrix(quat,matrix);
+  REAL origin[3] = { 0, -plane[3], 0 };
+  REAL center[3];
+  fm_transform(matrix,origin,center);
+  fm_setTranslation(center,matrix);
+}
+
+void fm_planeToQuat(const REAL *plane,REAL *quat,REAL *pos) // convert a plane equation to a quaternion and translation
+{
+  REAL ref[3] = { 0, 1, 0 };
+  REAL matrix[16];
+  fm_rotationArc(ref,plane,quat);
+  fm_quatToMatrix(quat,matrix);
+  REAL origin[3] = { 0, plane[3], 0 };
+  fm_transform(matrix,origin,pos);
+}
+
+void fm_eulerMatrix(REAL ax,REAL ay,REAL az,REAL *matrix) // convert euler (in radians) to a dest 4x4 matrix (translation set to zero)
+{
+  REAL quat[4];
+  fm_eulerToQuat(ax,ay,az,quat);
+  fm_quatToMatrix(quat,matrix);
+}
+
+
+//**********************************************************
+//**********************************************************
+//**** Vertex Welding
+//**********************************************************
+//**********************************************************
+
+#ifndef VERTEX_INDEX_H
+
+#define VERTEX_INDEX_H
+
+namespace VERTEX_INDEX
+{
+
+class KdTreeNode;
+
+typedef std::vector< KdTreeNode * > KdTreeNodeVector;
+
+enum Axes
+{
+  X_AXIS = 0,
+  Y_AXIS = 1,
+  Z_AXIS = 2
+};
+
+class KdTreeFindNode
+{
+public:
+  KdTreeFindNode(void)
+  {
+	mNode = 0;
+	mDistance = 0;
+  }
+  KdTreeNode  *mNode;
+  double        mDistance;
+};
+
+class KdTreeInterface
+{
+public:
+  virtual const double * getPositionDouble(uint32_t index) const = 0;
+  virtual const float  * getPositionFloat(uint32_t index) const = 0;
+};
+
+class KdTreeNode
+{
+public:
+  KdTreeNode(void)
+  {
+	mIndex = 0;
+	mLeft = 0;
+	mRight = 0;
+  }
+
+  KdTreeNode(uint32_t index)
+  {
+	mIndex = index;
+	mLeft = 0;
+	mRight = 0;
+  };
+
+	~KdTreeNode(void)
+  {
+  }
+
+
+  void addDouble(KdTreeNode *node,Axes dim,const KdTreeInterface *iface)
+  {
+	const double *nodePosition = iface->getPositionDouble( node->mIndex );
+	const double *position     = iface->getPositionDouble( mIndex );
+	switch ( dim )
+	{
+	  case X_AXIS:
+		if ( nodePosition[0] <= position[0] )
+		{
+		  if ( mLeft )
+			mLeft->addDouble(node,Y_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addDouble(node,Y_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	  case Y_AXIS:
+		if ( nodePosition[1] <= position[1] )
+		{
+		  if ( mLeft )
+			mLeft->addDouble(node,Z_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addDouble(node,Z_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	  case Z_AXIS:
+		if ( nodePosition[2] <= position[2] )
+		{
+		  if ( mLeft )
+			mLeft->addDouble(node,X_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addDouble(node,X_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	}
+
+  }
+
+
+  void addFloat(KdTreeNode *node,Axes dim,const KdTreeInterface *iface)
+  {
+	const float *nodePosition = iface->getPositionFloat( node->mIndex );
+	const float *position     = iface->getPositionFloat( mIndex );
+	switch ( dim )
+	{
+	  case X_AXIS:
+		if ( nodePosition[0] <= position[0] )
+		{
+		  if ( mLeft )
+			mLeft->addFloat(node,Y_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addFloat(node,Y_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	  case Y_AXIS:
+		if ( nodePosition[1] <= position[1] )
+		{
+		  if ( mLeft )
+			mLeft->addFloat(node,Z_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addFloat(node,Z_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	  case Z_AXIS:
+		if ( nodePosition[2] <= position[2] )
+		{
+		  if ( mLeft )
+			mLeft->addFloat(node,X_AXIS,iface);
+		  else
+			mLeft = node;
+		}
+		else
+		{
+		  if ( mRight )
+			mRight->addFloat(node,X_AXIS,iface);
+		  else
+			mRight = node;
+		}
+		break;
+	}
+
+  }
+
+
+  uint32_t getIndex(void) const { return mIndex; };
+
+  void search(Axes axis,const double *pos,double radius,uint32_t &count,uint32_t maxObjects,KdTreeFindNode *found,const KdTreeInterface *iface)
+  {
+
+	const double *position = iface->getPositionDouble(mIndex);
+
+	double dx = pos[0] - position[0];
+	double dy = pos[1] - position[1];
+	double dz = pos[2] - position[2];
+
+	KdTreeNode *search1 = 0;
+	KdTreeNode *search2 = 0;
+
+	switch ( axis )
+	{
+	  case X_AXIS:
+	   if ( dx <= 0 )     // JWR  if we are to the left
+	   {
+		search1 = mLeft; // JWR  then search to the left
+		if ( -dx < radius )  // JWR  if distance to the right is less than our search radius, continue on the right as well.
+		  search2 = mRight;
+	   }
+	   else
+	   {
+		 search1 = mRight; // JWR  ok, we go down the left tree
+		 if ( dx < radius ) // JWR  if the distance from the right is less than our search radius
+				search2 = mLeft;
+		}
+		axis = Y_AXIS;
+		break;
+	  case Y_AXIS:
+		if ( dy <= 0 )
+		{
+		  search1 = mLeft;
+		  if ( -dy < radius )
+					search2 = mRight;
+		}
+		else
+		{
+		  search1 = mRight;
+		  if ( dy < radius )
+					search2 = mLeft;
+		}
+		axis = Z_AXIS;
+		break;
+	  case Z_AXIS:
+		if ( dz <= 0 )
+		{
+		  search1 = mLeft;
+		  if ( -dz < radius )
+					search2 = mRight;
+		}
+		else
+		{
+		  search1 = mRight;
+		  if ( dz < radius )
+					search2 = mLeft;
+		}
+		axis = X_AXIS;
+		break;
+	}
+
+	double r2 = radius*radius;
+	double m  = dx*dx+dy*dy+dz*dz;
+
+	if ( m < r2 )
+	{
+	  switch ( count )
+	  {
+		case 0:
+		  found[count].mNode = this;
+		  found[count].mDistance = m;
+		  break;
+		case 1:
+		  if ( m < found[0].mDistance )
+		  {
+			if ( maxObjects == 1 )
+			{
+			  found[0].mNode = this;
+			  found[0].mDistance = m;
+			}
+			else
+			{
+			  found[1] = found[0];
+			  found[0].mNode = this;
+			  found[0].mDistance = m;
+			}
+		  }
+		  else if ( maxObjects > 1)
+		  {
+			found[1].mNode = this;
+			found[1].mDistance = m;
+		  }
+		  break;
+		default:
+		  {
+			bool inserted = false;
+
+			for (uint32_t i=0; i<count; i++)
+			{
+			  if ( m < found[i].mDistance ) // if this one is closer than a pre-existing one...
+			  {
+				// insertion sort...
+				uint32_t scan = count;
+				if ( scan >= maxObjects ) scan=maxObjects-1;
+				for (uint32_t j=scan; j>i; j--)
+				{
+				  found[j] = found[j-1];
+				}
+				found[i].mNode = this;
+				found[i].mDistance = m;
+				inserted = true;
+				break;
+			  }
+			}
+
+			if ( !inserted && count < maxObjects )
+			{
+			  found[count].mNode = this;
+			  found[count].mDistance = m;
+			}
+		  }
+		  break;
+	  }
+	  count++;
+	  if ( count > maxObjects )
+	  {
+		count = maxObjects;
+	  }
+	}
+
+
+	if ( search1 )
+		search1->search( axis, pos,radius, count, maxObjects, found, iface);
+
+	if ( search2 )
+		search2->search( axis, pos,radius, count, maxObjects, found, iface);
+
+  }
+
+  void search(Axes axis,const float *pos,float radius,uint32_t &count,uint32_t maxObjects,KdTreeFindNode *found,const KdTreeInterface *iface)
+  {
+
+	const float *position = iface->getPositionFloat(mIndex);
+
+	float dx = pos[0] - position[0];
+	float dy = pos[1] - position[1];
+	float dz = pos[2] - position[2];
+
+	KdTreeNode *search1 = 0;
+	KdTreeNode *search2 = 0;
+
+	switch ( axis )
+	{
+	  case X_AXIS:
+	   if ( dx <= 0 )     // JWR  if we are to the left
+	   {
+		search1 = mLeft; // JWR  then search to the left
+		if ( -dx < radius )  // JWR  if distance to the right is less than our search radius, continue on the right as well.
+		  search2 = mRight;
+	   }
+	   else
+	   {
+		 search1 = mRight; // JWR  ok, we go down the left tree
+		 if ( dx < radius ) // JWR  if the distance from the right is less than our search radius
+				search2 = mLeft;
+		}
+		axis = Y_AXIS;
+		break;
+	  case Y_AXIS:
+		if ( dy <= 0 )
+		{
+		  search1 = mLeft;
+		  if ( -dy < radius )
+					search2 = mRight;
+		}
+		else
+		{
+		  search1 = mRight;
+		  if ( dy < radius )
+					search2 = mLeft;
+		}
+		axis = Z_AXIS;
+		break;
+	  case Z_AXIS:
+		if ( dz <= 0 )
+		{
+		  search1 = mLeft;
+		  if ( -dz < radius )
+					search2 = mRight;
+		}
+		else
+		{
+		  search1 = mRight;
+		  if ( dz < radius )
+					search2 = mLeft;
+		}
+		axis = X_AXIS;
+		break;
+	}
+
+	float r2 = radius*radius;
+	float m  = dx*dx+dy*dy+dz*dz;
+
+	if ( m < r2 )
+	{
+	  switch ( count )
+	  {
+		case 0:
+		  found[count].mNode = this;
+		  found[count].mDistance = m;
+		  break;
+		case 1:
+		  if ( m < found[0].mDistance )
+		  {
+			if ( maxObjects == 1 )
+			{
+			  found[0].mNode = this;
+			  found[0].mDistance = m;
+			}
+			else
+			{
+			  found[1] = found[0];
+			  found[0].mNode = this;
+			  found[0].mDistance = m;
+			}
+		  }
+		  else if ( maxObjects > 1)
+		  {
+			found[1].mNode = this;
+			found[1].mDistance = m;
+		  }
+		  break;
+		default:
+		  {
+			bool inserted = false;
+
+			for (uint32_t i=0; i<count; i++)
+			{
+			  if ( m < found[i].mDistance ) // if this one is closer than a pre-existing one...
+			  {
+				// insertion sort...
+				uint32_t scan = count;
+				if ( scan >= maxObjects ) scan=maxObjects-1;
+				for (uint32_t j=scan; j>i; j--)
+				{
+				  found[j] = found[j-1];
+				}
+				found[i].mNode = this;
+				found[i].mDistance = m;
+				inserted = true;
+				break;
+			  }
+			}
+
+			if ( !inserted && count < maxObjects )
+			{
+			  found[count].mNode = this;
+			  found[count].mDistance = m;
+			}
+		  }
+		  break;
+	  }
+	  count++;
+	  if ( count > maxObjects )
+	  {
+		count = maxObjects;
+	  }
+	}
+
+
+	if ( search1 )
+		search1->search( axis, pos,radius, count, maxObjects, found, iface);
+
+	if ( search2 )
+		search2->search( axis, pos,radius, count, maxObjects, found, iface);
+
+  }
+
+private:
+
+  void setLeft(KdTreeNode *left) { mLeft = left; };
+  void setRight(KdTreeNode *right) { mRight = right; };
+
+	KdTreeNode *getLeft(void)         { return mLeft; }
+	KdTreeNode *getRight(void)        { return mRight; }
+
+  uint32_t          mIndex;
+  KdTreeNode     *mLeft;
+  KdTreeNode     *mRight;
+};
+
+
+#define MAX_BUNDLE_SIZE 1024  // 1024 nodes at a time, to minimize memory allocation and guarantee that pointers are persistent.
+
+class KdTreeNodeBundle 
+{
+public:
+
+  KdTreeNodeBundle(void)
+  {
+	mNext = 0;
+	mIndex = 0;
+  }
+
+  bool isFull(void) const
+  {
+	return (bool)( mIndex == MAX_BUNDLE_SIZE );
+  }
+
+  KdTreeNode * getNextNode(void)
+  {
+	assert(mIndex<MAX_BUNDLE_SIZE);
+	KdTreeNode *ret = &mNodes[mIndex];
+	mIndex++;
+	return ret;
+  }
+
+  KdTreeNodeBundle  *mNext;
+  uint32_t             mIndex;
+  KdTreeNode         mNodes[MAX_BUNDLE_SIZE];
+};
+
+
+typedef std::vector< double > DoubleVector;
+typedef std::vector< float >  FloatVector;
+
+class KdTree : public KdTreeInterface
+{
+public:
+  KdTree(void)
+  {
+	mRoot = 0;
+	mBundle = 0;
+	mVcount = 0;
+	mUseDouble = false;
+  }
+
+  virtual ~KdTree(void)
+  {
+	reset();
+  }
+
+  const double * getPositionDouble(uint32_t index) const
+  {
+	assert( mUseDouble );
+	assert ( index < mVcount );
+	return  &mVerticesDouble[index*3];
+  }
+
+  const float * getPositionFloat(uint32_t index) const
+  {
+	assert( !mUseDouble );
+	assert ( index < mVcount );
+	return  &mVerticesFloat[index*3];
+  }
+
+  uint32_t search(const double *pos,double radius,uint32_t maxObjects,KdTreeFindNode *found) const
+  {
+	assert( mUseDouble );
+	if ( !mRoot )	return 0;
+	uint32_t count = 0;
+	mRoot->search(X_AXIS,pos,radius,count,maxObjects,found,this);
+	return count;
+  }
+
+  uint32_t search(const float *pos,float radius,uint32_t maxObjects,KdTreeFindNode *found) const
+  {
+	assert( !mUseDouble );
+	if ( !mRoot )	return 0;
+	uint32_t count = 0;
+	mRoot->search(X_AXIS,pos,radius,count,maxObjects,found,this);
+	return count;
+  }
+
+  void reset(void)
+  {
+	mRoot = 0;
+	mVerticesDouble.clear();
+	mVerticesFloat.clear();
+	KdTreeNodeBundle *bundle = mBundle;
+	while ( bundle )
+	{
+	  KdTreeNodeBundle *next = bundle->mNext;
+	  delete bundle;
+	  bundle = next;
+	}
+	mBundle = 0;
+	mVcount = 0;
+  }
+
+  uint32_t add(double x,double y,double z)
+  {
+	assert(mUseDouble);
+	uint32_t ret = mVcount;
+	mVerticesDouble.push_back(x);
+	mVerticesDouble.push_back(y);
+	mVerticesDouble.push_back(z);
+	mVcount++;
+	KdTreeNode *node = getNewNode(ret);
+	if ( mRoot )
+	{
+	  mRoot->addDouble(node,X_AXIS,this);
+	}
+	else
+	{
+	  mRoot = node;
+	}
+	return ret;
+  }
+
+  uint32_t add(float x,float y,float z)
+  {
+	assert(!mUseDouble);
+	uint32_t ret = mVcount;
+	mVerticesFloat.push_back(x);
+	mVerticesFloat.push_back(y);
+	mVerticesFloat.push_back(z);
+	mVcount++;
+	KdTreeNode *node = getNewNode(ret);
+	if ( mRoot )
+	{
+	  mRoot->addFloat(node,X_AXIS,this);
+	}
+	else
+	{
+	  mRoot = node;
+	}
+	return ret;
+  }
+
+  KdTreeNode * getNewNode(uint32_t index)
+  {
+	if ( mBundle == 0 )
+	{
+	  mBundle = new KdTreeNodeBundle;
+	}
+	if ( mBundle->isFull() )
+	{
+	  KdTreeNodeBundle *bundle = new KdTreeNodeBundle;
+	  mBundle->mNext = bundle;
+	  mBundle = bundle;
+	}
+	KdTreeNode *node = mBundle->getNextNode();
+	new ( node ) KdTreeNode(index);
+	return node;
+  }
+
+  uint32_t getNearest(const double *pos,double radius,bool &_found) const // returns the nearest possible neighbor's index.
+  {
+	assert( mUseDouble );
+	uint32_t ret = 0;
+
+	_found = false;
+	KdTreeFindNode found[1];
+	uint32_t count = search(pos,radius,1,found);
+	if ( count )
+	{
+	  KdTreeNode *node = found[0].mNode;
+	  ret = node->getIndex();
+	  _found = true;
+	}
+	return ret;
+  }
+
+  uint32_t getNearest(const float *pos,float radius,bool &_found) const // returns the nearest possible neighbor's index.
+  {
+	assert( !mUseDouble );
+	uint32_t ret = 0;
+
+	_found = false;
+	KdTreeFindNode found[1];
+	uint32_t count = search(pos,radius,1,found);
+	if ( count )
+	{
+	  KdTreeNode *node = found[0].mNode;
+	  ret = node->getIndex();
+	  _found = true;
+	}
+	return ret;
+  }
+
+  const double * getVerticesDouble(void) const
+  {
+	assert( mUseDouble );
+	const double *ret = 0;
+	if ( !mVerticesDouble.empty() )
+	{
+	  ret = &mVerticesDouble[0];
+	}
+	return ret;
+  }
+
+  const float * getVerticesFloat(void) const
+  {
+	assert( !mUseDouble );
+	const float * ret = 0;
+	if ( !mVerticesFloat.empty() )
+	{
+	  ret = &mVerticesFloat[0];
+	}
+	return ret;
+  }
+
+  uint32_t getVcount(void) const { return mVcount; };
+
+  void setUseDouble(bool useDouble)
+  {
+	mUseDouble = useDouble;
+  }
+
+private:
+  bool                    mUseDouble;
+  KdTreeNode             *mRoot;
+  KdTreeNodeBundle       *mBundle;
+  uint32_t                  mVcount;
+  DoubleVector            mVerticesDouble;
+  FloatVector             mVerticesFloat;
+};
+
+}; // end of namespace VERTEX_INDEX
+
+class MyVertexIndex : public fm_VertexIndex
+{
+public:
+  MyVertexIndex(double granularity,bool snapToGrid)
+  {
+	mDoubleGranularity = granularity;
+	mFloatGranularity  = (float)granularity;
+	mSnapToGrid        = snapToGrid;
+	mUseDouble         = true;
+	mKdTree.setUseDouble(true);
+  }
+
+  MyVertexIndex(float granularity,bool snapToGrid)
+  {
+	mDoubleGranularity = granularity;
+	mFloatGranularity  = (float)granularity;
+	mSnapToGrid        = snapToGrid;
+	mUseDouble         = false;
+	mKdTree.setUseDouble(false);
+  }
+
+  virtual ~MyVertexIndex(void)
+  {
+
+  }
+
+
+  double snapToGrid(double p)
+  {
+	double m = fmod(p,mDoubleGranularity);
+	p-=m;
+	return p;
+  }
+
+  float snapToGrid(float p)
+  {
+	float m = fmodf(p,mFloatGranularity);
+	p-=m;
+	return p;
+  }
+
+  uint32_t    getIndex(const float *_p,bool &newPos)  // get index for a vector float
+  {
+	uint32_t ret;
+
+	if ( mUseDouble )
+	{
+	  double p[3];
+	  p[0] = _p[0];
+	  p[1] = _p[1];
+	  p[2] = _p[2];
+	  return getIndex(p,newPos);
+	}
+
+	newPos = false;
+
+	float p[3];
+
+	if ( mSnapToGrid )
+	{
+	  p[0] = snapToGrid(_p[0]);
+	  p[1] = snapToGrid(_p[1]);
+	  p[2] = snapToGrid(_p[2]);
+	}
+	else
+	{
+	  p[0] = _p[0];
+	  p[1] = _p[1];
+	  p[2] = _p[2];
+	}
+
+	bool found;
+	ret = mKdTree.getNearest(p,mFloatGranularity,found);
+	if ( !found )
+	{
+	  newPos = true;
+	  ret = mKdTree.add(p[0],p[1],p[2]);
+	}
+
+
+	return ret;
+  }
+
+  uint32_t    getIndex(const double *_p,bool &newPos)  // get index for a vector double
+  {
+	uint32_t ret;
+
+	if ( !mUseDouble )
+	{
+	  float p[3];
+	  p[0] = (float)_p[0];
+	  p[1] = (float)_p[1];
+	  p[2] = (float)_p[2];
+	  return getIndex(p,newPos);
+	}
+
+	newPos = false;
+
+	double p[3];
+
+	if ( mSnapToGrid )
+	{
+	  p[0] = snapToGrid(_p[0]);
+	  p[1] = snapToGrid(_p[1]);
+	  p[2] = snapToGrid(_p[2]);
+	}
+	else
+	{
+	  p[0] = _p[0];
+	  p[1] = _p[1];
+	  p[2] = _p[2];
+	}
+
+	bool found;
+	ret = mKdTree.getNearest(p,mDoubleGranularity,found);
+	if ( !found )
+	{
+	  newPos = true;
+	  ret = mKdTree.add(p[0],p[1],p[2]);
+	}
+
+
+	return ret;
+  }
+
+  const float *   getVerticesFloat(void) const
+  {
+	const float * ret = 0;
+
+	assert( !mUseDouble );
+
+	ret = mKdTree.getVerticesFloat();
+
+	return ret;
+  }
+
+  const double *  getVerticesDouble(void) const
+  {
+	const double * ret = 0;
+
+	assert( mUseDouble );
+
+	ret = mKdTree.getVerticesDouble();
+
+	return ret;
+  }
+
+  const float *   getVertexFloat(uint32_t index) const
+  {
+	const float * ret  = 0;
+	assert( !mUseDouble );
+#ifdef _DEBUG
+	uint32_t vcount = mKdTree.getVcount();
+	assert( index < vcount );
+#endif
+	ret =  mKdTree.getVerticesFloat();
+	ret = &ret[index*3];
+	return ret;
+  }
+
+  const double *   getVertexDouble(uint32_t index) const
+  {
+	const double * ret = 0;
+	assert( mUseDouble );
+#ifdef _DEBUG
+	uint32_t vcount = mKdTree.getVcount();
+	assert( index < vcount );
+#endif
+	ret =  mKdTree.getVerticesDouble();
+	ret = &ret[index*3];
+
+	return ret;
+  }
+
+  uint32_t    getVcount(void) const
+  {
+	return mKdTree.getVcount();
+  }
+
+  bool isDouble(void) const
+  {
+	return mUseDouble;
+  }
+
+
+  bool            saveAsObj(const char *fname,uint32_t tcount,uint32_t *indices)
+  {
+	bool ret = false;
+
+
+	FILE *fph = fopen(fname,"wb");
+	if ( fph )
+	{
+	  ret = true;
+
+	  uint32_t vcount    = getVcount();
+	  if ( mUseDouble )
+	  {
+		const double *v  = getVerticesDouble();
+		for (uint32_t i=0; i<vcount; i++)
+		{
+		  fprintf(fph,"v %0.9f %0.9f %0.9f\r\n", (float)v[0], (float)v[1], (float)v[2] );
+		  v+=3;
+		}
+	  }
+	  else
+	  {
+		const float *v  = getVerticesFloat();
+		for (uint32_t i=0; i<vcount; i++)
+		{
+		  fprintf(fph,"v %0.9f %0.9f %0.9f\r\n", v[0], v[1], v[2] );
+		  v+=3;
+		}
+	  }
+
+	  for (uint32_t i=0; i<tcount; i++)
+	  {
+		uint32_t i1 = *indices++;
+		uint32_t i2 = *indices++;
+		uint32_t i3 = *indices++;
+		fprintf(fph,"f %d %d %d\r\n", i1+1, i2+1, i3+1 );
+	  }
+	  fclose(fph);
+	}
+
+	return ret;
+  }
+
+private:
+  bool    mUseDouble:1;
+  bool    mSnapToGrid:1;
+  double  mDoubleGranularity;
+  float   mFloatGranularity;
+  VERTEX_INDEX::KdTree  mKdTree;
+};
+
+fm_VertexIndex * fm_createVertexIndex(double granularity,bool snapToGrid) // create an indexed vertex system for doubles
+{
+  MyVertexIndex *ret = new MyVertexIndex(granularity,snapToGrid);
+  return static_cast< fm_VertexIndex *>(ret);
+}
+
+fm_VertexIndex * fm_createVertexIndex(float granularity,bool snapToGrid)  // create an indexed vertext system for floats
+{
+  MyVertexIndex *ret = new MyVertexIndex(granularity,snapToGrid);
+  return static_cast< fm_VertexIndex *>(ret);
+}
+
+void          fm_releaseVertexIndex(fm_VertexIndex *vindex)
+{
+  MyVertexIndex *m = static_cast< MyVertexIndex *>(vindex);
+  delete m;
+}
+
+#endif   // END OF VERTEX WELDING CODE
+
+
+REAL fm_computeBestFitAABB(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *bmin,REAL *bmax) // returns the diagonal distance
+{
+
+  const uint8_t *source = (const uint8_t *) points;
+
+	bmin[0] = points[0];
+	bmin[1] = points[1];
+	bmin[2] = points[2];
+
+	bmax[0] = points[0];
+	bmax[1] = points[1];
+	bmax[2] = points[2];
+
+
+  for (uint32_t i=1; i<vcount; i++)
+  {
+	source+=pstride;
+	const REAL *p = (const REAL *) source;
+
+	if ( p[0] < bmin[0] ) bmin[0] = p[0];
+	if ( p[1] < bmin[1] ) bmin[1] = p[1];
+	if ( p[2] < bmin[2] ) bmin[2] = p[2];
+
+		if ( p[0] > bmax[0] ) bmax[0] = p[0];
+		if ( p[1] > bmax[1] ) bmax[1] = p[1];
+		if ( p[2] > bmax[2] ) bmax[2] = p[2];
+
+  }
+
+  REAL dx = bmax[0] - bmin[0];
+  REAL dy = bmax[1] - bmin[1];
+  REAL dz = bmax[2] - bmin[2];
+
+	return (REAL) sqrt( dx*dx + dy*dy + dz*dz );
+
+}
+
+
+
+/* a = b - c */
+#define vector(a,b,c) \
+	(a)[0] = (b)[0] - (c)[0];	\
+	(a)[1] = (b)[1] - (c)[1];	\
+	(a)[2] = (b)[2] - (c)[2];
+
+
+
+#define innerProduct(v,q) \
+		((v)[0] * (q)[0] + \
+		(v)[1] * (q)[1] + \
+		(v)[2] * (q)[2])
+
+#define crossProduct(a,b,c) \
+	(a)[0] = (b)[1] * (c)[2] - (c)[1] * (b)[2]; \
+	(a)[1] = (b)[2] * (c)[0] - (c)[2] * (b)[0]; \
+	(a)[2] = (b)[0] * (c)[1] - (c)[0] * (b)[1];
+
+
+bool fm_lineIntersectsTriangle(const REAL *rayStart,const REAL *rayEnd,const REAL *p1,const REAL *p2,const REAL *p3,REAL *sect)
+{
+	REAL dir[3];
+
+  dir[0] = rayEnd[0] - rayStart[0];
+  dir[1] = rayEnd[1] - rayStart[1];
+  dir[2] = rayEnd[2] - rayStart[2];
+
+  REAL d = (REAL)sqrt(dir[0]*dir[0] + dir[1]*dir[1] + dir[2]*dir[2]);
+  REAL r = 1.0f / d;
+
+  dir[0]*=r;
+  dir[1]*=r;
+  dir[2]*=r;
+
+
+  REAL t;
+
+	bool ret = fm_rayIntersectsTriangle(rayStart, dir, p1, p2, p3, t );
+
+	if ( ret )
+	{
+		if ( t > d )
+		{
+			sect[0] = rayStart[0] + dir[0]*t;
+			sect[1] = rayStart[1] + dir[1]*t;
+			sect[2] = rayStart[2] + dir[2]*t;
+		}
+		else
+		{
+			ret = false;
+		}
+	}
+
+  return ret;
+}
+
+
+
+bool fm_rayIntersectsTriangle(const REAL *p,const REAL *d,const REAL *v0,const REAL *v1,const REAL *v2,REAL &t)
+{
+	REAL e1[3],e2[3],h[3],s[3],q[3];
+	REAL a,f,u,v;
+
+	vector(e1,v1,v0);
+	vector(e2,v2,v0);
+	crossProduct(h,d,e2);
+	a = innerProduct(e1,h);
+
+	if (a > -0.00001 && a < 0.00001)
+		return(false);
+
+	f = 1/a;
+	vector(s,p,v0);
+	u = f * (innerProduct(s,h));
+
+	if (u < 0.0 || u > 1.0)
+		return(false);
+
+	crossProduct(q,s,e1);
+	v = f * innerProduct(d,q);
+	if (v < 0.0 || u + v > 1.0)
+		return(false);
+	// at this stage we can compute t to find out where
+	// the intersection point is on the line
+	t = f * innerProduct(e2,q);
+	if (t > 0) // ray intersection
+		return(true);
+	else // this means that there is a line intersection
+		 // but not a ray intersection
+		 return (false);
+}
+
+
+inline REAL det(const REAL *p1,const REAL *p2,const REAL *p3)
+{
+  return  p1[0]*p2[1]*p3[2] + p2[0]*p3[1]*p1[2] + p3[0]*p1[1]*p2[2] -p1[0]*p3[1]*p2[2] - p2[0]*p1[1]*p3[2] - p3[0]*p2[1]*p1[2];
+}
+
+
+REAL  fm_computeMeshVolume(const REAL *vertices,uint32_t tcount,const uint32_t *indices)
+{
+	REAL volume = 0;
+
+	for (uint32_t i=0; i<tcount; i++,indices+=3)
+	{
+	const REAL *p1 = &vertices[ indices[0]*3 ];
+		const REAL *p2 = &vertices[ indices[1]*3 ];
+		const REAL *p3 = &vertices[ indices[2]*3 ];
+		volume+=det(p1,p2,p3); // compute the volume of the tetrahedran relative to the origin.
+	}
+
+	volume*=(1.0f/6.0f);
+	if ( volume < 0 )
+		volume*=-1;
+	return volume;
+}
+
+
+const REAL * fm_getPoint(const REAL *points,uint32_t pstride,uint32_t index)
+{
+  const uint8_t *scan = (const uint8_t *)points;
+  scan+=(index*pstride);
+  return (REAL *)scan;
+}
+
+
+bool fm_insideTriangle(REAL Ax, REAL Ay,
+					  REAL Bx, REAL By,
+					  REAL Cx, REAL Cy,
+					  REAL Px, REAL Py)
+
+{
+  REAL ax, ay, bx, by, cx, cy, apx, apy, bpx, bpy, cpx, cpy;
+  REAL cCROSSap, bCROSScp, aCROSSbp;
+
+  ax = Cx - Bx;  ay = Cy - By;
+  bx = Ax - Cx;  by = Ay - Cy;
+  cx = Bx - Ax;  cy = By - Ay;
+  apx= Px - Ax;  apy= Py - Ay;
+  bpx= Px - Bx;  bpy= Py - By;
+  cpx= Px - Cx;  cpy= Py - Cy;
+
+  aCROSSbp = ax*bpy - ay*bpx;
+  cCROSSap = cx*apy - cy*apx;
+  bCROSScp = bx*cpy - by*cpx;
+
+  return ((aCROSSbp >= 0.0f) && (bCROSScp >= 0.0f) && (cCROSSap >= 0.0f));
+}
+
+
+REAL fm_areaPolygon2d(uint32_t pcount,const REAL *points,uint32_t pstride)
+{
+  int32_t n = (int32_t)pcount;
+
+  REAL A=0.0f;
+  for(int32_t p=n-1,q=0; q<n; p=q++)
+  {
+	const REAL *p1 = fm_getPoint(points,pstride,p);
+	const REAL *p2 = fm_getPoint(points,pstride,q);
+	A+= p1[0]*p2[1] - p2[0]*p1[1];
+  }
+  return A*0.5f;
+}
+
+
+bool  fm_pointInsidePolygon2d(uint32_t pcount,const REAL *points,uint32_t pstride,const REAL *point,uint32_t xindex,uint32_t yindex)
+{
+  uint32_t j = pcount-1;
+  int32_t oddNodes = 0;
+
+  REAL x = point[xindex];
+  REAL y = point[yindex];
+
+  for (uint32_t i=0; i<pcount; i++)
+  {
+	const REAL *p1 = fm_getPoint(points,pstride,i);
+	const REAL *p2 = fm_getPoint(points,pstride,j);
+
+	REAL x1 = p1[xindex];
+	REAL y1 = p1[yindex];
+
+	REAL x2 = p2[xindex];
+	REAL y2 = p2[yindex];
+
+	if ( (y1 < y && y2 >= y) ||  (y2 < y && y1 >= y) )
+	{
+	  if (x1+(y-y1)/(y2-y1)*(x2-x1)<x)
+	  {
+		oddNodes = 1-oddNodes;
+	  }
+	}
+	j = i;
+  }
+
+  return oddNodes ? true : false;
+}
+
+
+uint32_t fm_consolidatePolygon(uint32_t pcount,const REAL *points,uint32_t pstride,REAL *_dest,REAL epsilon) // collapses co-linear edges.
+{
+  uint32_t ret = 0;
+
+
+  if ( pcount >= 3 )
+  {
+	const REAL *prev = fm_getPoint(points,pstride,pcount-1);
+	const REAL *current = points;
+	const REAL *next    = fm_getPoint(points,pstride,1);
+	REAL *dest = _dest;
+
+	for (uint32_t i=0; i<pcount; i++)
+	{
+
+	  next = (i+1)==pcount ? points : next;
+
+	  if ( !fm_colinear(prev,current,next,epsilon) )
+	  {
+		dest[0] = current[0];
+		dest[1] = current[1];
+		dest[2] = current[2];
+
+		dest+=3;
+		ret++;
+	  }
+
+	  prev = current;
+	  current+=3;
+	  next+=3;
+
+	}
+  }
+
+  return ret;
+}
+
+
+#ifndef RECT3D_TEMPLATE
+
+#define RECT3D_TEMPLATE
+
+template <class T> class Rect3d
+{
+public:
+  Rect3d(void) { };
+
+  Rect3d(const T *bmin,const T *bmax)
+  {
+
+	mMin[0] = bmin[0];
+	mMin[1] = bmin[1];
+	mMin[2] = bmin[2];
+
+	mMax[0] = bmax[0];
+	mMax[1] = bmax[1];
+	mMax[2] = bmax[2];
+
+  }
+
+  void SetMin(const T *bmin)
+  {
+	mMin[0] = bmin[0];
+	mMin[1] = bmin[1];
+	mMin[2] = bmin[2];
+  }
+
+  void SetMax(const T *bmax)
+  {
+	mMax[0] = bmax[0];
+	mMax[1] = bmax[1];
+	mMax[2] = bmax[2];
+  }
+
+	void SetMin(T x,T y,T z)
+	{
+		mMin[0] = x;
+		mMin[1] = y;
+		mMin[2] = z;
+	}
+
+	void SetMax(T x,T y,T z)
+	{
+		mMax[0] = x;
+		mMax[1] = y;
+		mMax[2] = z;
+	}
+
+  T mMin[3];
+  T mMax[3];
+};
+
+#endif
+
+void splitRect(uint32_t axis,
+						   const Rect3d<REAL> &source,
+							 Rect3d<REAL> &b1,
+							 Rect3d<REAL> &b2,
+							 const REAL *midpoint)
+{
+	switch ( axis )
+	{
+		case 0:
+			b1.SetMin(source.mMin);
+			b1.SetMax( midpoint[0], source.mMax[1], source.mMax[2] );
+
+			b2.SetMin( midpoint[0], source.mMin[1], source.mMin[2] );
+			b2.SetMax(source.mMax);
+
+			break;
+		case 1:
+			b1.SetMin(source.mMin);
+			b1.SetMax( source.mMax[0], midpoint[1], source.mMax[2] );
+
+			b2.SetMin( source.mMin[0], midpoint[1], source.mMin[2] );
+			b2.SetMax(source.mMax);
+
+			break;
+		case 2:
+			b1.SetMin(source.mMin);
+			b1.SetMax( source.mMax[0], source.mMax[1], midpoint[2] );
+
+			b2.SetMin( source.mMin[0], source.mMin[1], midpoint[2] );
+			b2.SetMax(source.mMax);
+
+			break;
+	}
+}
+
+bool fm_computeSplitPlane(uint32_t vcount,
+						  const REAL *vertices,
+						  uint32_t /* tcount */,
+						  const uint32_t * /* indices */,
+						  REAL *plane)
+{
+
+  REAL sides[3];
+  REAL matrix[16];
+
+  fm_computeBestFitOBB( vcount, vertices, sizeof(REAL)*3, sides, matrix );
+
+  REAL bmax[3];
+  REAL bmin[3];
+
+  bmax[0] = sides[0]*0.5f;
+  bmax[1] = sides[1]*0.5f;
+  bmax[2] = sides[2]*0.5f;
+
+  bmin[0] = -bmax[0];
+  bmin[1] = -bmax[1];
+  bmin[2] = -bmax[2];
+
+
+  REAL dx = sides[0];
+  REAL dy = sides[1];
+  REAL dz = sides[2];
+
+
+	uint32_t axis = 0;
+
+	if ( dy > dx )
+	{
+		axis = 1;
+	}
+
+	if ( dz > dx && dz > dy )
+	{
+		axis = 2;
+	}
+
+  REAL p1[3];
+  REAL p2[3];
+  REAL p3[3];
+
+  p3[0] = p2[0] = p1[0] = bmin[0] + dx*0.5f;
+  p3[1] = p2[1] = p1[1] = bmin[1] + dy*0.5f;
+  p3[2] = p2[2] = p1[2] = bmin[2] + dz*0.5f;
+
+  Rect3d<REAL> b(bmin,bmax);
+
+  Rect3d<REAL> b1,b2;
+
+  splitRect(axis,b,b1,b2,p1);
+
+
+  switch ( axis )
+  {
+	case 0:
+	  p2[1] = bmin[1];
+	  p2[2] = bmin[2];
+
+	  if ( dz > dy )
+	  {
+		p3[1] = bmax[1];
+		p3[2] = bmin[2];
+	  }
+	  else
+	  {
+		p3[1] = bmin[1];
+		p3[2] = bmax[2];
+	  }
+
+	  break;
+	case 1:
+	  p2[0] = bmin[0];
+	  p2[2] = bmin[2];
+
+	  if ( dx > dz )
+	  {
+		p3[0] = bmax[0];
+		p3[2] = bmin[2];
+	  }
+	  else
+	  {
+		p3[0] = bmin[0];
+		p3[2] = bmax[2];
+	  }
+
+	  break;
+	case 2:
+	  p2[0] = bmin[0];
+	  p2[1] = bmin[1];
+
+	  if ( dx > dy )
+	  {
+		p3[0] = bmax[0];
+		p3[1] = bmin[1];
+	  }
+	  else
+	  {
+		p3[0] = bmin[0];
+		p3[1] = bmax[1];
+	  }
+
+	  break;
+  }
+
+  REAL tp1[3];
+  REAL tp2[3];
+  REAL tp3[3];
+
+  fm_transform(matrix,p1,tp1);
+  fm_transform(matrix,p2,tp2);
+  fm_transform(matrix,p3,tp3);
+
+	plane[3] = fm_computePlane(tp1,tp2,tp3,plane);
+
+  return true;
+
+}
+
+#pragma warning(disable:4100)
+
+void fm_nearestPointInTriangle(const REAL * /*nearestPoint*/,const REAL * /*p1*/,const REAL * /*p2*/,const REAL * /*p3*/,REAL * /*nearest*/)
+{
+
+}
+
+static REAL Partial(const REAL *a,const REAL *p) 
+{
+	return (a[0]*p[1]) - (p[0]*a[1]);
+}
+
+REAL  fm_areaTriangle(const REAL *p0,const REAL *p1,const REAL *p2)
+{
+  REAL A = Partial(p0,p1);
+	A+= Partial(p1,p2);
+	A+= Partial(p2,p0);
+	return A*0.5f;
+}
+
+void fm_subtract(const REAL *A,const REAL *B,REAL *diff) // compute A-B and store the result in 'diff'
+{
+  diff[0] = A[0]-B[0];
+  diff[1] = A[1]-B[1];
+  diff[2] = A[2]-B[2];
+}
+
+
+void  fm_multiplyTransform(const REAL *pA,const REAL *pB,REAL *pM)
+{
+
+  REAL a = pA[0*4+0] * pB[0*4+0] + pA[0*4+1] * pB[1*4+0] + pA[0*4+2] * pB[2*4+0] + pA[0*4+3] * pB[3*4+0];
+  REAL b = pA[0*4+0] * pB[0*4+1] + pA[0*4+1] * pB[1*4+1] + pA[0*4+2] * pB[2*4+1] + pA[0*4+3] * pB[3*4+1];
+  REAL c = pA[0*4+0] * pB[0*4+2] + pA[0*4+1] * pB[1*4+2] + pA[0*4+2] * pB[2*4+2] + pA[0*4+3] * pB[3*4+2];
+  REAL d = pA[0*4+0] * pB[0*4+3] + pA[0*4+1] * pB[1*4+3] + pA[0*4+2] * pB[2*4+3] + pA[0*4+3] * pB[3*4+3];
+
+  REAL e = pA[1*4+0] * pB[0*4+0] + pA[1*4+1] * pB[1*4+0] + pA[1*4+2] * pB[2*4+0] + pA[1*4+3] * pB[3*4+0];
+  REAL f = pA[1*4+0] * pB[0*4+1] + pA[1*4+1] * pB[1*4+1] + pA[1*4+2] * pB[2*4+1] + pA[1*4+3] * pB[3*4+1];
+  REAL g = pA[1*4+0] * pB[0*4+2] + pA[1*4+1] * pB[1*4+2] + pA[1*4+2] * pB[2*4+2] + pA[1*4+3] * pB[3*4+2];
+  REAL h = pA[1*4+0] * pB[0*4+3] + pA[1*4+1] * pB[1*4+3] + pA[1*4+2] * pB[2*4+3] + pA[1*4+3] * pB[3*4+3];
+
+  REAL i = pA[2*4+0] * pB[0*4+0] + pA[2*4+1] * pB[1*4+0] + pA[2*4+2] * pB[2*4+0] + pA[2*4+3] * pB[3*4+0];
+  REAL j = pA[2*4+0] * pB[0*4+1] + pA[2*4+1] * pB[1*4+1] + pA[2*4+2] * pB[2*4+1] + pA[2*4+3] * pB[3*4+1];
+  REAL k = pA[2*4+0] * pB[0*4+2] + pA[2*4+1] * pB[1*4+2] + pA[2*4+2] * pB[2*4+2] + pA[2*4+3] * pB[3*4+2];
+  REAL l = pA[2*4+0] * pB[0*4+3] + pA[2*4+1] * pB[1*4+3] + pA[2*4+2] * pB[2*4+3] + pA[2*4+3] * pB[3*4+3];
+
+  REAL m = pA[3*4+0] * pB[0*4+0] + pA[3*4+1] * pB[1*4+0] + pA[3*4+2] * pB[2*4+0] + pA[3*4+3] * pB[3*4+0];
+  REAL n = pA[3*4+0] * pB[0*4+1] + pA[3*4+1] * pB[1*4+1] + pA[3*4+2] * pB[2*4+1] + pA[3*4+3] * pB[3*4+1];
+  REAL o = pA[3*4+0] * pB[0*4+2] + pA[3*4+1] * pB[1*4+2] + pA[3*4+2] * pB[2*4+2] + pA[3*4+3] * pB[3*4+2];
+  REAL p = pA[3*4+0] * pB[0*4+3] + pA[3*4+1] * pB[1*4+3] + pA[3*4+2] * pB[2*4+3] + pA[3*4+3] * pB[3*4+3];
+
+  pM[0] = a;  pM[1] = b;  pM[2] = c;  pM[3] = d;
+
+  pM[4] = e;  pM[5] = f;  pM[6] = g;  pM[7] = h;
+
+  pM[8] = i;  pM[9] = j;  pM[10] = k;  pM[11] = l;
+
+  pM[12] = m;  pM[13] = n;  pM[14] = o;  pM[15] = p;
+}
+
+void fm_multiply(REAL *A,REAL scaler)
+{
+  A[0]*=scaler;
+  A[1]*=scaler;
+  A[2]*=scaler;
+}
+
+void fm_add(const REAL *A,const REAL *B,REAL *sum)
+{
+  sum[0] = A[0]+B[0];
+  sum[1] = A[1]+B[1];
+  sum[2] = A[2]+B[2];
+}
+
+void fm_copy3(const REAL *source,REAL *dest)
+{
+  dest[0] = source[0];
+  dest[1] = source[1];
+  dest[2] = source[2];
+}
+
+
+uint32_t  fm_copyUniqueVertices(uint32_t vcount,const REAL *input_vertices,REAL *output_vertices,uint32_t tcount,const uint32_t *input_indices,uint32_t *output_indices)
+{
+  uint32_t ret = 0;
+
+  REAL *vertices = (REAL *)malloc(sizeof(REAL)*vcount*3);
+  memcpy(vertices,input_vertices,sizeof(REAL)*vcount*3);
+  REAL *dest = output_vertices;
+
+  uint32_t *reindex = (uint32_t *)malloc(sizeof(uint32_t)*vcount);
+  memset(reindex,0xFF,sizeof(uint32_t)*vcount);
+
+  uint32_t icount = tcount*3;
+
+  for (uint32_t i=0; i<icount; i++)
+  {
+	uint32_t index = *input_indices++;
+
+	assert( index < vcount );
+
+	if ( reindex[index] == 0xFFFFFFFF )
+	{
+	  *output_indices++ = ret;
+	  reindex[index] = ret;
+	  const REAL *pos = &vertices[index*3];
+	  dest[0] = pos[0];
+	  dest[1] = pos[1];
+	  dest[2] = pos[2];
+	  dest+=3;
+	  ret++;
+	}
+	else
+	{
+	  *output_indices++ = reindex[index];
+	}
+  }
+  free(vertices);
+  free(reindex);
+  return ret;
+}
+
+bool    fm_isMeshCoplanar(uint32_t tcount,const uint32_t *indices,const REAL *vertices,bool doubleSided) // returns true if this collection of indexed triangles are co-planar!
+{
+  bool ret = true;
+
+  if ( tcount > 0 )
+  {
+	uint32_t i1 = indices[0];
+	uint32_t i2 = indices[1];
+	uint32_t i3 = indices[2];
+	const REAL *p1 = &vertices[i1*3];
+	const REAL *p2 = &vertices[i2*3];
+	const REAL *p3 = &vertices[i3*3];
+	REAL plane[4];
+	plane[3] = fm_computePlane(p1,p2,p3,plane);
+	const uint32_t *scan = &indices[3];
+	for (uint32_t i=1; i<tcount; i++)
+	{
+	  i1 = *scan++;
+	  i2 = *scan++;
+	  i3 = *scan++;
+	  p1 = &vertices[i1*3];
+	  p2 = &vertices[i2*3];
+	  p3 = &vertices[i3*3];
+	  REAL _plane[4];
+	  _plane[3] = fm_computePlane(p1,p2,p3,_plane);
+	  if ( !fm_samePlane(plane,_plane,0.01f,0.001f,doubleSided) )
+	  {
+		ret = false;
+		break;
+	  }
+	}
+  }
+  return ret;
+}
+
+
+bool fm_samePlane(const REAL p1[4],const REAL p2[4],REAL normalEpsilon,REAL dEpsilon,bool doubleSided)
+{
+  bool ret = false;
+
+#if 0
+  if (p1[0] == p2[0] &&
+	  p1[1] == p2[1] &&
+	  p1[2] == p2[2] &&
+	  p1[3] == p2[3])
+  {
+	  ret = true;
+  }
+#else
+  REAL diff = (REAL) fabs(p1[3]-p2[3]);
+  if ( diff < dEpsilon ) // if the plane -d  co-efficient is within our epsilon
+  {
+	REAL dot = fm_dot(p1,p2); // compute the dot-product of the vector normals.
+	if ( doubleSided ) dot = (REAL)fabs(dot);
+	REAL dmin = 1 - normalEpsilon;
+	REAL dmax = 1 + normalEpsilon;
+	if ( dot >= dmin && dot <= dmax )
+	{
+	  ret = true; // then the plane equation is for practical purposes identical.
+	}
+  }
+#endif
+  return ret;
+}
+
+
+void  fm_initMinMax(REAL bmin[3],REAL bmax[3])
+{
+  bmin[0] = FLT_MAX;
+  bmin[1] = FLT_MAX;
+  bmin[2] = FLT_MAX;
+
+  bmax[0] = -FLT_MAX;
+  bmax[1] = -FLT_MAX;
+  bmax[2] = -FLT_MAX;
+}
+
+void fm_inflateMinMax(REAL bmin[3], REAL bmax[3], REAL ratio)
+{
+	REAL inflate = fm_distance(bmin, bmax)*0.5f*ratio;
+
+	bmin[0] -= inflate;
+	bmin[1] -= inflate;
+	bmin[2] -= inflate;
+
+	bmax[0] += inflate;
+	bmax[1] += inflate;
+	bmax[2] += inflate;
+}
+
+#ifndef TESSELATE_H
+
+#define TESSELATE_H
+
+typedef std::vector< uint32_t > UintVector;
+
+class Myfm_Tesselate : public fm_Tesselate
+{
+public:
+  virtual ~Myfm_Tesselate(void)
+  {
+
+  }
+
+  const uint32_t * tesselate(fm_VertexIndex *vindex,uint32_t tcount,const uint32_t *indices,float longEdge,uint32_t maxDepth,uint32_t &outcount)
+  {
+	const uint32_t *ret = 0;
+
+	mMaxDepth = maxDepth;
+	mLongEdge  = longEdge*longEdge;
+	mLongEdgeD = mLongEdge;
+	mVertices = vindex;
+
+	if ( mVertices->isDouble() )
+	{
+	  uint32_t vcount = mVertices->getVcount();
+	  double *vertices = (double *)malloc(sizeof(double)*vcount*3);
+	  memcpy(vertices,mVertices->getVerticesDouble(),sizeof(double)*vcount*3);
+
+	  for (uint32_t i=0; i<tcount; i++)
+	  {
+		uint32_t i1 = *indices++;
+		uint32_t i2 = *indices++;
+		uint32_t i3 = *indices++;
+
+		const double *p1 = &vertices[i1*3];
+		const double *p2 = &vertices[i2*3];
+		const double *p3 = &vertices[i3*3];
+
+		tesselate(p1,p2,p3,0);
+
+	  }
+	  free(vertices);
+	}
+	else
+	{
+	  uint32_t vcount = mVertices->getVcount();
+	  float *vertices = (float *)malloc(sizeof(float)*vcount*3);
+	  memcpy(vertices,mVertices->getVerticesFloat(),sizeof(float)*vcount*3);
+
+
+	  for (uint32_t i=0; i<tcount; i++)
+	  {
+		uint32_t i1 = *indices++;
+		uint32_t i2 = *indices++;
+		uint32_t i3 = *indices++;
+
+		const float *p1 = &vertices[i1*3];
+		const float *p2 = &vertices[i2*3];
+		const float *p3 = &vertices[i3*3];
+
+		tesselate(p1,p2,p3,0);
+
+	  }
+	  free(vertices);
+	}
+
+	outcount = (uint32_t)(mIndices.size()/3);
+	ret = &mIndices[0];
+
+
+	return ret;
+  }
+
+  void tesselate(const float *p1,const float *p2,const float *p3,uint32_t recurse)
+  {
+	bool split = false;
+	float l1,l2,l3;
+
+	l1 = l2 = l3 = 0;
+
+	if ( recurse < mMaxDepth )
+	{
+	  l1 = fm_distanceSquared(p1,p2);
+		l2 = fm_distanceSquared(p2,p3);
+		l3 = fm_distanceSquared(p3,p1);
+
+	  if (  l1 > mLongEdge || l2 > mLongEdge || l3 > mLongEdge )
+		split = true;
+
+	}
+
+	if ( split )
+	{
+		uint32_t edge;
+
+		if ( l1 >= l2 && l1 >= l3 )
+			edge = 0;
+		else if ( l2 >= l1 && l2 >= l3 )
+			edge = 1;
+		else
+			edge = 2;
+
+			float splits[3];
+
+		switch ( edge )
+		{
+			case 0:
+				{
+			fm_lerp(p1,p2,splits,0.5f);
+			tesselate(p1,splits,p3, recurse+1 );
+			tesselate(splits,p2,p3, recurse+1 );
+				}
+				break;
+			case 1:
+				{
+			fm_lerp(p2,p3,splits,0.5f);
+			tesselate(p1,p2,splits, recurse+1 );
+			tesselate(p1,splits,p3, recurse+1 );
+				}
+				break;
+			case 2:
+				{
+					fm_lerp(p3,p1,splits,0.5f);
+			tesselate(p1,p2,splits, recurse+1 );
+			tesselate(splits,p2,p3, recurse+1 );
+				}
+				break;
+		}
+	}
+	else
+	{
+	  bool newp;
+
+	  uint32_t i1 = mVertices->getIndex(p1,newp);
+	  uint32_t i2 = mVertices->getIndex(p2,newp);
+	  uint32_t i3 = mVertices->getIndex(p3,newp);
+
+	  mIndices.push_back(i1);
+	  mIndices.push_back(i2);
+	  mIndices.push_back(i3);
+	}
+
+  }
+
+  void tesselate(const double *p1,const double *p2,const double *p3,uint32_t recurse)
+  {
+	bool split = false;
+	double l1,l2,l3;
+
+	l1 = l2 = l3 = 0;
+
+	if ( recurse < mMaxDepth )
+	{
+	  l1 = fm_distanceSquared(p1,p2);
+		l2 = fm_distanceSquared(p2,p3);
+		l3 = fm_distanceSquared(p3,p1);
+
+	  if (  l1 > mLongEdgeD || l2 > mLongEdgeD || l3 > mLongEdgeD )
+		split = true;
+
+	}
+
+	if ( split )
+	{
+		uint32_t edge;
+
+		if ( l1 >= l2 && l1 >= l3 )
+			edge = 0;
+		else if ( l2 >= l1 && l2 >= l3 )
+			edge = 1;
+		else
+			edge = 2;
+
+			double splits[3];
+
+		switch ( edge )
+		{
+			case 0:
+				{
+			fm_lerp(p1,p2,splits,0.5);
+			tesselate(p1,splits,p3, recurse+1 );
+			tesselate(splits,p2,p3, recurse+1 );
+				}
+				break;
+			case 1:
+				{
+			fm_lerp(p2,p3,splits,0.5);
+			tesselate(p1,p2,splits, recurse+1 );
+			tesselate(p1,splits,p3, recurse+1 );
+				}
+				break;
+			case 2:
+				{
+					fm_lerp(p3,p1,splits,0.5);
+			tesselate(p1,p2,splits, recurse+1 );
+			tesselate(splits,p2,p3, recurse+1 );
+				}
+				break;
+		}
+	}
+	else
+	{
+	  bool newp;
+
+	  uint32_t i1 = mVertices->getIndex(p1,newp);
+	  uint32_t i2 = mVertices->getIndex(p2,newp);
+	  uint32_t i3 = mVertices->getIndex(p3,newp);
+
+	  mIndices.push_back(i1);
+	  mIndices.push_back(i2);
+	  mIndices.push_back(i3);
+	}
+
+  }
+
+private:
+  float           mLongEdge;
+  double          mLongEdgeD;
+  fm_VertexIndex *mVertices;
+  UintVector    mIndices;
+  uint32_t          mMaxDepth;
+};
+
+fm_Tesselate * fm_createTesselate(void)
+{
+  Myfm_Tesselate *m = new Myfm_Tesselate;
+  return static_cast< fm_Tesselate * >(m);
+}
+
+void           fm_releaseTesselate(fm_Tesselate *t)
+{
+  Myfm_Tesselate *m = static_cast< Myfm_Tesselate *>(t);
+  delete m;
+}
+
+#endif
+
+
+#ifndef RAY_ABB_INTERSECT
+
+#define RAY_ABB_INTERSECT
+
+//! Integer representation of a floating-point value.
+#define IR(x)	((uint32_t&)x)
+
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+/**
+*	A method to compute a ray-AABB intersection.
+*	Original code by Andrew Woo, from "Graphics Gems", Academic Press, 1990
+*	Optimized code by Pierre Terdiman, 2000 (~20-30% faster on my Celeron 500)
+*	Epsilon value added by Klaus Hartmann. (discarding it saves a few cycles only)
+*
+*	Hence this version is faster as well as more robust than the original one.
+*
+*	Should work provided:
+*	1) the integer representation of 0.0f is 0x00000000
+*	2) the sign bit of the float is the most significant one
+*
+*	Report bugs: p.terdiman@codercorner.com
+*
+*	\param		aabb		[in] the axis-aligned bounding box
+*	\param		origin		[in] ray origin
+*	\param		dir			[in] ray direction
+*	\param		coord		[out] impact coordinates
+*	\return		true if ray intersects AABB
+*/
+///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define RAYAABB_EPSILON 0.00001f
+bool fm_intersectRayAABB(const float MinB[3],const float MaxB[3],const float origin[3],const float dir[3],float coord[3])
+{
+  bool Inside = true;
+  float MaxT[3];
+  MaxT[0]=MaxT[1]=MaxT[2]=-1.0f;
+
+  // Find candidate planes.
+  for(uint32_t i=0;i<3;i++)
+  {
+	if(origin[i] < MinB[i])
+	{
+	  coord[i]	= MinB[i];
+	  Inside		= false;
+
+	  // Calculate T distances to candidate planes
+	  if(IR(dir[i]))	MaxT[i] = (MinB[i] - origin[i]) / dir[i];
+	}
+	else if(origin[i] > MaxB[i])
+	{
+	  coord[i]	= MaxB[i];
+	  Inside		= false;
+
+	  // Calculate T distances to candidate planes
+	  if(IR(dir[i]))	MaxT[i] = (MaxB[i] - origin[i]) / dir[i];
+	}
+  }
+
+  // Ray origin inside bounding box
+  if(Inside)
+  {
+	coord[0] = origin[0];
+	coord[1] = origin[1];
+	coord[2] = origin[2];
+	return true;
+  }
+
+  // Get largest of the maxT's for final choice of intersection
+  uint32_t WhichPlane = 0;
+  if(MaxT[1] > MaxT[WhichPlane])	WhichPlane = 1;
+  if(MaxT[2] > MaxT[WhichPlane])	WhichPlane = 2;
+
+  // Check final candidate actually inside box
+  if(IR(MaxT[WhichPlane])&0x80000000) return false;
+
+  for(uint32_t i=0;i<3;i++)
+  {
+	if(i!=WhichPlane)
+	{
+	  coord[i] = origin[i] + MaxT[WhichPlane] * dir[i];
+#ifdef RAYAABB_EPSILON
+	  if(coord[i] < MinB[i] - RAYAABB_EPSILON || coord[i] > MaxB[i] + RAYAABB_EPSILON)	return false;
+#else
+	  if(coord[i] < MinB[i] || coord[i] > MaxB[i])	return false;
+#endif
+	}
+  }
+  return true;	// ray hits box
+}
+
+bool fm_intersectLineSegmentAABB(const float bmin[3],const float bmax[3],const float p1[3],const float p2[3],float intersect[3])
+{
+  bool ret = false;
+
+  float dir[3];
+  dir[0] = p2[0] - p1[0];
+  dir[1] = p2[1] - p1[1];
+  dir[2] = p2[2] - p1[2];
+  float dist = fm_normalize(dir);
+  if ( dist > RAYAABB_EPSILON )
+  {
+	ret = fm_intersectRayAABB(bmin,bmax,p1,dir,intersect);
+	if ( ret )
+	{
+	  float d = fm_distanceSquared(p1,intersect);
+	  if ( d  > (dist*dist) )
+	  {
+		ret = false;
+	  }
+	}
+  }
+  return ret;
+}
+
+#endif
+
+#ifndef OBB_TO_AABB
+
+#define OBB_TO_AABB
+
+#pragma warning(disable:4100)
+
+void    fm_OBBtoAABB(const float /*obmin*/[3],const float /*obmax*/[3],const float /*matrix*/[16],float /*abmin*/[3],float /*abmax*/[3])
+{
+  assert(0); // not yet implemented.
+}
+
+
+const REAL * computePos(uint32_t index,const REAL *vertices,uint32_t vstride)
+{
+  const char *tmp = (const char *)vertices;
+  tmp+=(index*vstride);
+  return (const REAL*)tmp;
+}
+
+void computeNormal(uint32_t index,REAL *normals,uint32_t nstride,const REAL *normal)
+{
+  char *tmp = (char *)normals;
+  tmp+=(index*nstride);
+  REAL *dest = (REAL *)tmp;
+  dest[0]+=normal[0];
+  dest[1]+=normal[1];
+  dest[2]+=normal[2];
+}
+
+void fm_computeMeanNormals(uint32_t vcount,       // the number of vertices
+						   const REAL *vertices,     // the base address of the vertex position data.
+						   uint32_t vstride,      // the stride between position data.
+						   REAL *normals,            // the base address  of the destination for mean vector normals
+						   uint32_t nstride,      // the stride between normals
+						   uint32_t tcount,       // the number of triangles
+						   const uint32_t *indices)     // the triangle indices
+{
+
+  // Step #1 : Zero out the vertex normals
+  char *dest = (char *)normals;
+  for (uint32_t i=0; i<vcount; i++)
+  {
+	REAL *n = (REAL *)dest;
+	n[0] = 0;
+	n[1] = 0;
+	n[2] = 0;
+	dest+=nstride;
+  }
+
+  // Step #2 : Compute the face normals and accumulate them
+  const uint32_t *scan = indices;
+  for (uint32_t i=0; i<tcount; i++)
+  {
+
+	uint32_t i1 = *scan++;
+	uint32_t i2 = *scan++;
+	uint32_t i3 = *scan++;
+
+	const REAL *p1 = computePos(i1,vertices,vstride);
+	const REAL *p2 = computePos(i2,vertices,vstride);
+	const REAL *p3 = computePos(i3,vertices,vstride);
+
+	REAL normal[3];
+	fm_computePlane(p3,p2,p1,normal);
+
+	computeNormal(i1,normals,nstride,normal);
+	computeNormal(i2,normals,nstride,normal);
+	computeNormal(i3,normals,nstride,normal);
+  }
+
+
+  // Normalize the accumulated normals
+  dest = (char *)normals;
+  for (uint32_t i=0; i<vcount; i++)
+  {
+	REAL *n = (REAL *)dest;
+	fm_normalize(n);
+	dest+=nstride;
+  }
+
+}
+
+#endif
+
+
+#define BIGNUMBER 100000000.0  		/* hundred million */
+
+static inline void Set(REAL *n,REAL x,REAL y,REAL z)
+{
+	n[0] = x;
+	n[1] = y;
+	n[2] = z;
+};
+
+static inline void Copy(REAL *dest,const REAL *source)
+{
+	dest[0] = source[0];
+	dest[1] = source[1];
+	dest[2] = source[2];
+}
+
+
+REAL  fm_computeBestFitSphere(uint32_t vcount,const REAL *points,uint32_t pstride,REAL *center)
+{
+	REAL radius;
+	REAL radius2;
+
+	REAL xmin[3];
+	REAL xmax[3];
+	REAL ymin[3];
+	REAL ymax[3];
+	REAL zmin[3];
+	REAL zmax[3];
+	REAL dia1[3];
+	REAL dia2[3];
+
+	/* FIRST PASS: find 6 minima/maxima points */
+	Set(xmin,BIGNUMBER,BIGNUMBER,BIGNUMBER);
+	Set(xmax,-BIGNUMBER,-BIGNUMBER,-BIGNUMBER);
+	Set(ymin,BIGNUMBER,BIGNUMBER,BIGNUMBER);
+	Set(ymax,-BIGNUMBER,-BIGNUMBER,-BIGNUMBER);
+	Set(zmin,BIGNUMBER,BIGNUMBER,BIGNUMBER);
+	Set(zmax,-BIGNUMBER,-BIGNUMBER,-BIGNUMBER);
+
+	{
+		const char *scan = (const char *)points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+			const REAL *caller_p = (const REAL *)scan;
+			if (caller_p[0]<xmin[0])
+				Copy(xmin,caller_p); /* New xminimum point */
+			if (caller_p[0]>xmax[0])
+				Copy(xmax,caller_p);
+			if (caller_p[1]<ymin[1])
+				Copy(ymin,caller_p);
+			if (caller_p[1]>ymax[1])
+				Copy(ymax,caller_p);
+			if (caller_p[2]<zmin[2])
+				Copy(zmin,caller_p);
+			if (caller_p[2]>zmax[2])
+				Copy(zmax,caller_p);
+			scan+=pstride;
+		}
+	}
+
+	/* Set xspan = distance between the 2 points xmin & xmax (squared) */
+	REAL dx = xmax[0] - xmin[0];
+	REAL dy = xmax[1] - xmin[1];
+	REAL dz = xmax[2] - xmin[2];
+	REAL xspan = dx*dx + dy*dy + dz*dz;
+
+/* Same for y & z spans */
+	dx = ymax[0] - ymin[0];
+	dy = ymax[1] - ymin[1];
+	dz = ymax[2] - ymin[2];
+	REAL yspan = dx*dx + dy*dy + dz*dz;
+
+	dx = zmax[0] - zmin[0];
+	dy = zmax[1] - zmin[1];
+	dz = zmax[2] - zmin[2];
+	REAL zspan = dx*dx + dy*dy + dz*dz;
+
+	/* Set points dia1 & dia2 to the maximally separated pair */
+	Copy(dia1,xmin);
+	Copy(dia2,xmax); /* assume xspan biggest */
+	REAL maxspan = xspan;
+
+	if (yspan>maxspan)
+	{
+		maxspan = yspan;
+		Copy(dia1,ymin);
+		Copy(dia2,ymax);
+	}
+
+	if (zspan>maxspan)
+	{
+		maxspan = zspan;
+		Copy(dia1,zmin);
+		Copy(dia2,zmax);
+	}
+
+
+	/* dia1,dia2 is a diameter of initial sphere */
+	/* calc initial center */
+	center[0] = (dia1[0]+dia2[0])*0.5f;
+	center[1] = (dia1[1]+dia2[1])*0.5f;
+	center[2] = (dia1[2]+dia2[2])*0.5f;
+
+	/* calculate initial radius**2 and radius */
+
+	dx = dia2[0]-center[0]; /* x component of radius vector */
+	dy = dia2[1]-center[1]; /* y component of radius vector */
+	dz = dia2[2]-center[2]; /* z component of radius vector */
+
+	radius2 = dx*dx + dy*dy + dz*dz;
+	radius = REAL(sqrt(radius2));
+
+	/* SECOND PASS: increment current sphere */
+	{
+		const char *scan = (const char *)points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+			const REAL *caller_p = (const REAL *)scan;
+			dx = caller_p[0]-center[0];
+			dy = caller_p[1]-center[1];
+			dz = caller_p[2]-center[2];
+			REAL old_to_p_sq = dx*dx + dy*dy + dz*dz;
+			if (old_to_p_sq > radius2) 	/* do r**2 test first */
+			{ 	/* this point is outside of current sphere */
+				REAL old_to_p = REAL(sqrt(old_to_p_sq));
+				/* calc radius of new sphere */
+				radius = (radius + old_to_p) * 0.5f;
+				radius2 = radius*radius; 	/* for next r**2 compare */
+				REAL old_to_new = old_to_p - radius;
+				/* calc center of new sphere */
+				REAL recip = 1.0f /old_to_p;
+				REAL cx = (radius*center[0] + old_to_new*caller_p[0]) * recip;
+				REAL cy = (radius*center[1] + old_to_new*caller_p[1]) * recip;
+				REAL cz = (radius*center[2] + old_to_new*caller_p[2]) * recip;
+				Set(center,cx,cy,cz);
+				scan+=pstride;
+			}
+		}
+	}
+	return radius;
+}
+
+
+void fm_computeBestFitCapsule(uint32_t vcount,const REAL *points,uint32_t pstride,REAL &radius,REAL &height,REAL matrix[16],bool bruteForce)
+{
+  REAL sides[3];
+  REAL omatrix[16];
+  fm_computeBestFitOBB(vcount,points,pstride,sides,omatrix,bruteForce);
+
+  int32_t axis = 0;
+  if ( sides[0] > sides[1] && sides[0] > sides[2] )
+	axis = 0;
+  else if ( sides[1] > sides[0] && sides[1] > sides[2] )
+	axis = 1;
+  else 
+	axis = 2;
+
+  REAL localTransform[16];
+
+  REAL maxDist = 0;
+  REAL maxLen = 0;
+
+  switch ( axis )
+  {
+	case 0:
+	  {
+		fm_eulerMatrix(0,0,FM_PI/2,localTransform);
+		fm_matrixMultiply(localTransform,omatrix,matrix);
+
+		const uint8_t *scan = (const uint8_t *)points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+		  const REAL *p = (const REAL *)scan;
+		  REAL t[3];
+		  fm_inverseRT(omatrix,p,t);
+		  REAL dist = t[1]*t[1]+t[2]*t[2];
+		  if ( dist > maxDist )
+		  {
+			maxDist = dist;
+		  }
+		  REAL l = (REAL) fabs(t[0]);
+		  if ( l > maxLen )
+		  {
+			maxLen = l;
+		  }
+		  scan+=pstride;
+		}
+	  }
+	  height = sides[0];
+	  break;
+	case 1:
+	  {
+		fm_eulerMatrix(0,FM_PI/2,0,localTransform);
+		fm_matrixMultiply(localTransform,omatrix,matrix);
+
+		const uint8_t *scan = (const uint8_t *)points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+		  const REAL *p = (const REAL *)scan;
+		  REAL t[3];
+		  fm_inverseRT(omatrix,p,t);
+		  REAL dist = t[0]*t[0]+t[2]*t[2];
+		  if ( dist > maxDist )
+		  {
+			maxDist = dist;
+		  }
+		  REAL l = (REAL) fabs(t[1]);
+		  if ( l > maxLen )
+		  {
+			maxLen = l;
+		  }
+		  scan+=pstride;
+		}
+	  }
+	  height = sides[1];
+	  break;
+	case 2:
+	  {
+		fm_eulerMatrix(FM_PI/2,0,0,localTransform);
+		fm_matrixMultiply(localTransform,omatrix,matrix);
+
+		const uint8_t *scan = (const uint8_t *)points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+		  const REAL *p = (const REAL *)scan;
+		  REAL t[3];
+		  fm_inverseRT(omatrix,p,t);
+		  REAL dist = t[0]*t[0]+t[1]*t[1];
+		  if ( dist > maxDist )
+		  {
+			maxDist = dist;
+		  }
+		  REAL l = (REAL) fabs(t[2]);
+		  if ( l > maxLen )
+		  {
+			maxLen = l;
+		  }
+		  scan+=pstride;
+		}
+	  }
+	  height = sides[2];
+	  break;
+  }
+  radius = (REAL)sqrt(maxDist);
+  height = (maxLen*2)-(radius*2);
+}
+
+
+//************* Triangulation
+
+#ifndef TRIANGULATE_H
+
+#define TRIANGULATE_H
+
+typedef uint32_t TU32;
+
+class TVec
+{
+public:
+	TVec(double _x,double _y,double _z) { x = _x; y = _y; z = _z; };
+	TVec(void) { };
+
+  double x;
+  double y;
+  double z;
+};
+
+typedef std::vector< TVec >  TVecVector;
+typedef std::vector< TU32 >  TU32Vector;
+
+class CTriangulator
+{
+public:
+	///     Default constructor
+	CTriangulator();
+
+	///     Default destructor
+	virtual ~CTriangulator();
+
+	///     Triangulates the contour
+	void triangulate(TU32Vector &indices);
+
+	///     Returns the given point in the triangulator array
+	inline TVec get(const TU32 id) { return mPoints[id]; }
+
+	virtual void reset(void)
+	{
+		mInputPoints.clear();
+		mPoints.clear();
+		mIndices.clear();
+	}
+
+	virtual void addPoint(double x,double y,double z)
+	{
+		TVec v(x,y,z);
+		// update bounding box...
+		if ( mInputPoints.empty() )
+		{
+			mMin = v;
+			mMax = v;
+		}
+		else
+		{
+			if ( x < mMin.x ) mMin.x = x;
+			if ( y < mMin.y ) mMin.y = y;
+			if ( z < mMin.z ) mMin.z = z;
+
+			if ( x > mMax.x ) mMax.x = x;
+			if ( y > mMax.y ) mMax.y = y;
+			if ( z > mMax.z ) mMax.z = z;
+		}
+		mInputPoints.push_back(v);
+	}
+
+	// Triangulation happens in 2d.  We could inverse transform the polygon around the normal direction, or we just use the two most signficant axes
+	// Here we find the two longest axes and use them to triangulate.  Inverse transforming them would introduce more doubleing point error and isn't worth it.
+	virtual uint32_t * triangulate(uint32_t &tcount,double epsilon)
+	{
+		uint32_t *ret = 0;
+		tcount = 0;
+		mEpsilon = epsilon;
+
+		if ( !mInputPoints.empty() )
+		{
+			mPoints.clear();
+
+		  double dx = mMax.x - mMin.x; // locate the first, second and third longest edges and store them in i1, i2, i3
+		  double dy = mMax.y - mMin.y;
+		  double dz = mMax.z - mMin.z;
+
+		  uint32_t i1,i2,i3;
+
+		  if ( dx > dy && dx > dz )
+		  {
+			  i1 = 0;
+			  if ( dy > dz )
+			  {
+				  i2 = 1;
+				  i3 = 2;
+			  }
+			  else
+			  {
+				  i2 = 2;
+				  i3 = 1;
+			  }
+		  }
+		  else if ( dy > dx && dy > dz )
+		  {
+			  i1 = 1;
+			  if ( dx > dz )
+			  {
+				  i2 = 0;
+				  i3 = 2;
+			  }
+			  else
+			  {
+				  i2 = 2;
+				  i3 = 0;
+			  }
+		  }
+		  else
+		  {
+			  i1 = 2;
+			  if ( dx > dy )
+			  {
+				  i2 = 0;
+				  i3 = 1;
+			  }
+			  else
+			  {
+				  i2 = 1;
+				  i3 = 0;
+			  }
+		  }
+
+		  uint32_t pcount = (uint32_t)mInputPoints.size();
+		  const double *points = &mInputPoints[0].x;
+		  for (uint32_t i=0; i<pcount; i++)
+		  {
+			TVec v( points[i1], points[i2], points[i3] );
+			mPoints.push_back(v);
+			points+=3;
+		  }
+
+		  mIndices.clear();
+		  triangulate(mIndices);
+		  tcount = (uint32_t)mIndices.size()/3;
+		  if ( tcount )
+		  {
+			  ret = &mIndices[0];
+		  }
+		}
+		return ret;
+	}
+
+	virtual const double * getPoint(uint32_t index)
+	{
+		return &mInputPoints[index].x;
+	}
+
+
+private:
+	double                  mEpsilon;
+	TVec                   mMin;
+	TVec                   mMax;
+	TVecVector             mInputPoints;
+	TVecVector             mPoints;
+	TU32Vector             mIndices;
+
+	///     Tests if a point is inside the given triangle
+	bool _insideTriangle(const TVec& A, const TVec& B, const TVec& C,const TVec& P);
+
+	///     Returns the area of the contour
+	double _area();
+
+	bool _snip(int32_t u, int32_t v, int32_t w, int32_t n, int32_t *V);
+
+	///     Processes the triangulation
+	void _process(TU32Vector &indices);
+
+};
+
+///     Default constructor
+CTriangulator::CTriangulator(void)
+{
+}
+
+///     Default destructor
+CTriangulator::~CTriangulator()
+{
+}
+
+///     Triangulates the contour
+void CTriangulator::triangulate(TU32Vector &indices)
+{
+	_process(indices);
+}
+
+///     Processes the triangulation
+void CTriangulator::_process(TU32Vector &indices)
+{
+	const int32_t n = (const int32_t)mPoints.size();
+	if (n < 3)
+		return;
+	int32_t *V = (int32_t *)malloc(sizeof(int32_t)*n);
+
+	bool flipped = false;
+
+	if (0.0f < _area())
+	{
+		for (int32_t v = 0; v < n; v++)
+			V[v] = v;
+	}
+	else
+	{
+		flipped = true;
+		for (int32_t v = 0; v < n; v++)
+			V[v] = (n - 1) - v;
+	}
+
+	int32_t nv = n;
+	int32_t count = 2 * nv;
+	for (int32_t m = 0, v = nv - 1; nv > 2;)
+	{
+		if (0 >= (count--))
+			return;
+
+		int32_t u = v;
+		if (nv <= u)
+			u = 0;
+		v = u + 1;
+		if (nv <= v)
+			v = 0;
+		int32_t w = v + 1;
+		if (nv <= w)
+			w = 0;
+
+		if (_snip(u, v, w, nv, V))
+		{
+			int32_t a, b, c, s, t;
+			a = V[u];
+			b = V[v];
+			c = V[w];
+			if ( flipped )
+			{
+				indices.push_back(a);
+				indices.push_back(b);
+				indices.push_back(c);
+			}
+			else
+			{
+				indices.push_back(c);
+				indices.push_back(b);
+				indices.push_back(a);
+			}
+			m++;
+			for (s = v, t = v + 1; t < nv; s++, t++)
+				V[s] = V[t];
+			nv--;
+			count = 2 * nv;
+		}
+	}
+
+	free(V);
+}
+
+///     Returns the area of the contour
+double CTriangulator::_area()
+{
+	int32_t n = (uint32_t)mPoints.size();
+	double A = 0.0f;
+	for (int32_t p = n - 1, q = 0; q < n; p = q++)
+	{
+		const TVec &pval = mPoints[p];
+		const TVec &qval = mPoints[q];
+		A += pval.x * qval.y - qval.x * pval.y;
+	}
+	A*=0.5f;
+	return A;
+}
+
+bool CTriangulator::_snip(int32_t u, int32_t v, int32_t w, int32_t n, int32_t *V)
+{
+	int32_t p;
+
+	const TVec &A = mPoints[ V[u] ];
+	const TVec &B = mPoints[ V[v] ];
+	const TVec &C = mPoints[ V[w] ];
+
+	if (mEpsilon > (((B.x - A.x) * (C.y - A.y)) - ((B.y - A.y) * (C.x - A.x))) )
+		return false;
+
+	for (p = 0; p < n; p++)
+	{
+		if ((p == u) || (p == v) || (p == w))
+			continue;
+		const TVec &P = mPoints[ V[p] ];
+		if (_insideTriangle(A, B, C, P))
+			return false;
+	}
+	return true;
+}
+
+///     Tests if a point is inside the given triangle
+bool CTriangulator::_insideTriangle(const TVec& A, const TVec& B, const TVec& C,const TVec& P)
+{
+	double ax, ay, bx, by, cx, cy, apx, apy, bpx, bpy, cpx, cpy;
+	double cCROSSap, bCROSScp, aCROSSbp;
+
+	ax = C.x - B.x;  ay = C.y - B.y;
+	bx = A.x - C.x;  by = A.y - C.y;
+	cx = B.x - A.x;  cy = B.y - A.y;
+	apx = P.x - A.x;  apy = P.y - A.y;
+	bpx = P.x - B.x;  bpy = P.y - B.y;
+	cpx = P.x - C.x;  cpy = P.y - C.y;
+
+	aCROSSbp = ax * bpy - ay * bpx;
+	cCROSSap = cx * apy - cy * apx;
+	bCROSScp = bx * cpy - by * cpx;
+
+	return ((aCROSSbp >= 0.0f) && (bCROSScp >= 0.0f) && (cCROSSap >= 0.0f));
+}
+
+class Triangulate : public fm_Triangulate
+{
+public:
+  Triangulate(void)
+  {
+	mPointsFloat = 0;
+	mPointsDouble = 0;
+  }
+
+  virtual ~Triangulate(void)
+  {
+	reset();
+  }
+  void reset(void)
+  {
+	free(mPointsFloat);
+	free(mPointsDouble);
+	mPointsFloat = 0;
+	mPointsDouble = 0;
+  }
+
+  virtual const double *       triangulate3d(uint32_t pcount,
+											 const double *_points,
+											 uint32_t vstride,
+											 uint32_t &tcount,
+											 bool consolidate,
+											 double epsilon)
+  {
+	reset();
+
+	double *points = (double *)malloc(sizeof(double)*pcount*3);
+	if ( consolidate )
+	{
+	  pcount = fm_consolidatePolygon(pcount,_points,vstride,points,1-epsilon);
+	}
+	else
+	{
+	  double *dest = points;
+	  for (uint32_t i=0; i<pcount; i++)
+	  {
+		const double *src = fm_getPoint(_points,vstride,i);
+		dest[0] = src[0];
+		dest[1] = src[1];
+		dest[2] = src[2];
+		dest+=3;
+	  }
+	  vstride = sizeof(double)*3;
+	}
+
+	if ( pcount >= 3 )
+	{
+	  CTriangulator ct;
+	  for (uint32_t i=0; i<pcount; i++)
+	  {
+		const double *src = fm_getPoint(points,vstride,i);
+		ct.addPoint( src[0], src[1], src[2] );
+	  }
+	  uint32_t _tcount;
+	  uint32_t *indices = ct.triangulate(_tcount,epsilon);
+	  if ( indices )
+	  {
+		tcount = _tcount;
+		mPointsDouble = (double *)malloc(sizeof(double)*tcount*3*3);
+		double *dest = mPointsDouble;
+		for (uint32_t i=0; i<tcount; i++)
+		{
+		  uint32_t i1 = indices[i*3+0];
+		  uint32_t i2 = indices[i*3+1];
+		  uint32_t i3 = indices[i*3+2];
+		  const double *p1 = ct.getPoint(i1);
+		  const double *p2 = ct.getPoint(i2);
+		  const double *p3 = ct.getPoint(i3);
+
+		  dest[0] = p1[0];
+		  dest[1] = p1[1];
+		  dest[2] = p1[2];
+
+		  dest[3] = p2[0];
+		  dest[4] = p2[1];
+		  dest[5] = p2[2];
+
+		  dest[6] = p3[0];
+		  dest[7] = p3[1];
+		  dest[8] = p3[2];
+		  dest+=9;
+		}
+	  }
+	}
+	free(points);
+
+	return mPointsDouble;
+  }
+
+  virtual const float  *       triangulate3d(uint32_t pcount,
+											 const float  *points,
+											 uint32_t vstride,
+											 uint32_t &tcount,
+											 bool consolidate,
+											 float epsilon)
+  {
+	reset();
+
+	double *temp = (double *)malloc(sizeof(double)*pcount*3);
+	double *dest = temp;
+	for (uint32_t i=0; i<pcount; i++)
+	{
+	  const float *p = fm_getPoint(points,vstride,i);
+	  dest[0] = p[0];
+	  dest[1] = p[1];
+	  dest[2] = p[2];
+	  dest+=3;
+	}
+	const double *results = triangulate3d(pcount,temp,sizeof(double)*3,tcount,consolidate,epsilon);
+	if ( results )
+	{
+	  uint32_t fcount = tcount*3*3;
+	  mPointsFloat = (float *)malloc(sizeof(float)*tcount*3*3);
+	  for (uint32_t i=0; i<fcount; i++)
+	  {
+		mPointsFloat[i] = (float) results[i];
+	  }
+	  free(mPointsDouble);
+	  mPointsDouble = 0;
+	}
+	free(temp);
+
+	return mPointsFloat;
+  }
+
+private:
+  float *mPointsFloat;
+  double *mPointsDouble;
+};
+
+fm_Triangulate * fm_createTriangulate(void)
+{
+  Triangulate *t = new Triangulate;
+  return static_cast< fm_Triangulate *>(t);
+}
+
+void             fm_releaseTriangulate(fm_Triangulate *t)
+{
+  Triangulate *tt = static_cast< Triangulate *>(t);
+  delete tt;
+}
+
+#endif
+
+bool validDistance(const REAL *p1,const REAL *p2,REAL epsilon)
+{
+	bool ret = true;
+
+	REAL dx = p1[0] - p2[0];
+	REAL dy = p1[1] - p2[1];
+	REAL dz = p1[2] - p2[2];
+	REAL dist = dx*dx+dy*dy+dz*dz;
+	if ( dist < (epsilon*epsilon) )
+	{
+		ret = false;
+	}
+	return ret;
+}
+
+bool fm_isValidTriangle(const REAL *p1,const REAL *p2,const REAL *p3,REAL epsilon)
+{
+  bool ret = false;
+
+  if ( validDistance(p1,p2,epsilon) &&
+	   validDistance(p1,p3,epsilon) &&
+	   validDistance(p2,p3,epsilon) )
+  {
+
+	  REAL area = fm_computeArea(p1,p2,p3);
+	  if ( area > epsilon )
+	  {
+		REAL _vertices[3*3],vertices[64*3];
+
+		_vertices[0] = p1[0];
+		_vertices[1] = p1[1];
+		_vertices[2] = p1[2];
+
+		_vertices[3] = p2[0];
+		_vertices[4] = p2[1];
+		_vertices[5] = p2[2];
+
+		_vertices[6] = p3[0];
+		_vertices[7] = p3[1];
+		_vertices[8] = p3[2];
+
+		uint32_t pcount = fm_consolidatePolygon(3,_vertices,sizeof(REAL)*3,vertices,1-epsilon);
+		if ( pcount == 3 )
+		{
+		  ret = true;
+		}
+	  }
+  }
+  return ret;
+}
+
+
+void  fm_multiplyQuat(const REAL *left,const REAL *right,REAL *quat)
+{
+	REAL a,b,c,d;
+
+	a = left[3]*right[3] - left[0]*right[0] - left[1]*right[1] - left[2]*right[2];
+	b = left[3]*right[0] + right[3]*left[0] + left[1]*right[2] - right[1]*left[2];
+	c = left[3]*right[1] + right[3]*left[1] + left[2]*right[0] - right[2]*left[0];
+	d = left[3]*right[2] + right[3]*left[2] + left[0]*right[1] - right[0]*left[1];
+
+	quat[3] = a;
+	quat[0] = b;
+	quat[1] = c;
+	quat[2] = d;
+}
+
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+						 const REAL *points,     // starting address of points array.
+						 REAL *center)
+
+{
+	bool ret = false;
+	if ( vcount )
+	{
+		center[0] = 0;
+		center[1] = 0;
+		center[2] = 0;
+		const REAL *p = points;
+		for (uint32_t i=0; i<vcount; i++)
+		{
+			center[0]+=p[0];
+			center[1]+=p[1];
+			center[2]+=p[2];
+			p += 3;
+		}
+		REAL recip = 1.0f / (REAL)vcount;
+		center[0]*=recip;
+		center[1]*=recip;
+		center[2]*=recip;
+		ret = true;
+	}
+	return ret;
+}
+
+bool  fm_computeCentroid(uint32_t vcount,     // number of input data points
+	const REAL *points,     // starting address of points array.
+	uint32_t triCount,
+	const uint32_t *indices,
+	REAL *center)
+
+{
+	bool ret = false;
+	if (vcount)
+	{
+		center[0] = 0;
+		center[1] = 0;
+		center[2] = 0;
+
+		REAL numerator[3] = { 0, 0, 0 };
+		REAL denomintaor = 0;
+
+		for (uint32_t i = 0; i < triCount; i++)
+		{
+			uint32_t i1 = indices[i * 3 + 0];
+			uint32_t i2 = indices[i * 3 + 1];
+			uint32_t i3 = indices[i * 3 + 2];
+
+			const REAL *p1 = &points[i1 * 3];
+			const REAL *p2 = &points[i2 * 3];
+			const REAL *p3 = &points[i3 * 3];
+
+			// Compute the sum of the three positions
+			REAL sum[3];
+			sum[0] = p1[0] + p2[0] + p3[0];
+			sum[1] = p1[1] + p2[1] + p3[1];
+			sum[2] = p1[2] + p2[2] + p3[2];
+
+			// Compute the average of the three positions
+			sum[0] = sum[0] / 3;
+			sum[1] = sum[1] / 3;
+			sum[2] = sum[2] / 3;
+
+			// Compute the area of this triangle
+			REAL area = fm_computeArea(p1, p2, p3);
+
+			numerator[0]+= (sum[0] * area);
+			numerator[1]+= (sum[1] * area);
+			numerator[2]+= (sum[2] * area);
+
+			denomintaor += area;
+
+		}
+		REAL recip = 1 / denomintaor;
+		center[0] = numerator[0] * recip;
+		center[1] = numerator[1] * recip;
+		center[2] = numerator[2] * recip;
+		ret = true;
+	}
+	return ret;
+}
+
+
+#ifndef TEMPLATE_VEC3
+#define TEMPLATE_VEC3
+template <class Type> class Vec3
+{
+public:
+	Vec3(void)
+	{
+
+	}
+	Vec3(Type _x,Type _y,Type _z)
+	{
+		x = _x;
+		y = _y;
+		z = _z;
+	}
+	Type x;
+	Type y;
+	Type z;
+};
+#endif
+
+void fm_transformAABB(const REAL bmin[3],const REAL bmax[3],const REAL matrix[16],REAL tbmin[3],REAL tbmax[3])
+{
+	Vec3<REAL> box[8];
+	box[0] = Vec3< REAL >( bmin[0], bmin[1], bmin[2] );
+	box[1] = Vec3< REAL >( bmax[0], bmin[1], bmin[2] );
+	box[2] = Vec3< REAL >( bmax[0], bmax[1], bmin[2] );
+	box[3] = Vec3< REAL >( bmin[0], bmax[1], bmin[2] );
+	box[4] = Vec3< REAL >( bmin[0], bmin[1], bmax[2] );
+	box[5] = Vec3< REAL >( bmax[0], bmin[1], bmax[2] );
+	box[6] = Vec3< REAL >( bmax[0], bmax[1], bmax[2] );
+	box[7] = Vec3< REAL >( bmin[0], bmax[1], bmax[2] );
+	// transform all 8 corners of the box and then recompute a new AABB
+	for (unsigned int i=0; i<8; i++)
+	{
+		Vec3< REAL > &p = box[i];
+		fm_transform(matrix,&p.x,&p.x);
+		if ( i == 0 )
+		{
+			tbmin[0] = tbmax[0] = p.x;
+			tbmin[1] = tbmax[1] = p.y;
+			tbmin[2] = tbmax[2] = p.z;
+		}
+		else
+		{
+			if ( p.x < tbmin[0] ) tbmin[0] = p.x;
+			if ( p.y < tbmin[1] ) tbmin[1] = p.y;
+			if ( p.z < tbmin[2] ) tbmin[2] = p.z;
+			if ( p.x > tbmax[0] ) tbmax[0] = p.x;
+			if ( p.y > tbmax[1] ) tbmax[1] = p.y;
+			if ( p.z > tbmax[2] ) tbmax[2] = p.z;
+		}
+	}
+}
+
+REAL  fm_normalizeQuat(REAL n[4]) // normalize this quat
+{
+	REAL dx = n[0]*n[0];
+	REAL dy = n[1]*n[1];
+	REAL dz = n[2]*n[2];
+	REAL dw = n[3]*n[3];
+
+	REAL dist = dx*dx+dy*dy+dz*dz+dw*dw;
+
+	dist = (REAL)sqrt(dist);
+
+	REAL recip = 1.0f / dist;
+
+	n[0]*=recip;
+	n[1]*=recip;
+	n[2]*=recip;
+	n[3]*=recip;
+
+	return dist;
+}
+
+
+}; // end of namespace
diff --git a/thirdparty/vhacd/src/VHACD-ASYNC.cpp b/thirdparty/vhacd/src/VHACD-ASYNC.cpp
new file mode 100644
index 0000000000..baf309d6a1
--- /dev/null
+++ b/thirdparty/vhacd/src/VHACD-ASYNC.cpp
@@ -0,0 +1,334 @@
+#include "../public/VHACD.h"
+#include <stdlib.h>
+#include <string.h>
+#include <stdarg.h>
+#include <thread>
+#include <atomic>
+#include <mutex>
+#include <string>
+#include <float.h>
+
+#define ENABLE_ASYNC 1
+
+#define HACD_ALLOC(x) malloc(x)
+#define HACD_FREE(x) free(x)
+#define HACD_ASSERT(x) assert(x)
+
+namespace VHACD
+{
+
+class MyHACD_API : public VHACD::IVHACD, public VHACD::IVHACD::IUserCallback, VHACD::IVHACD::IUserLogger
+{
+public:
+	MyHACD_API(void)
+	{
+		mVHACD = VHACD::CreateVHACD();
+	}
+
+	virtual ~MyHACD_API(void)
+	{
+		releaseHACD();
+		Cancel();
+		mVHACD->Release();
+	}
+
+	
+	virtual bool Compute(const double* const _points,
+		const uint32_t countPoints,
+		const uint32_t* const _triangles,
+		const uint32_t countTriangles,
+		const Parameters& _desc) final
+	{
+#if ENABLE_ASYNC
+		Cancel(); // if we previously had a solution running; cancel it.
+		releaseHACD();
+
+		// We need to copy the input vertices and triangles into our own buffers so we can operate
+		// on them safely from the background thread.
+		mVertices = (double *)HACD_ALLOC(sizeof(double)*countPoints * 3);
+		mIndices = (uint32_t *)HACD_ALLOC(sizeof(uint32_t)*countTriangles * 3);
+		memcpy(mVertices, _points, sizeof(double)*countPoints * 3);
+		memcpy(mIndices, _triangles, sizeof(uint32_t)*countTriangles * 3);
+		mRunning = true;
+		mThread = new std::thread([this, countPoints, countTriangles, _desc]()
+		{
+			ComputeNow(mVertices, countPoints, mIndices, countTriangles, _desc);
+			mRunning = false;
+		});
+#else
+		releaseHACD();
+		ComputeNow(_points, countPoints, _triangles, countTriangles, _desc);
+#endif
+		return true;
+	}
+
+	bool ComputeNow(const double* const points,
+		const uint32_t countPoints,
+		const uint32_t* const triangles,
+		const uint32_t countTriangles,
+		const Parameters& _desc) 
+	{
+		uint32_t ret = 0;
+
+		mHullCount	= 0;
+		mCallback	= _desc.m_callback;
+		mLogger		= _desc.m_logger;
+
+		IVHACD::Parameters desc = _desc;
+		// Set our intercepting callback interfaces if non-null
+		desc.m_callback = desc.m_callback ? this : nullptr;
+		desc.m_logger = desc.m_logger ? this : nullptr;
+
+		if ( countPoints )
+		{
+			bool ok = mVHACD->Compute(points, countPoints, triangles, countTriangles, desc);
+			if (ok)
+			{
+				ret = mVHACD->GetNConvexHulls();
+				mHulls = new IVHACD::ConvexHull[ret];
+				for (uint32_t i = 0; i < ret; i++)
+				{
+					VHACD::IVHACD::ConvexHull vhull;
+					mVHACD->GetConvexHull(i, vhull);
+					VHACD::IVHACD::ConvexHull h;
+					h.m_nPoints = vhull.m_nPoints;
+					h.m_points = (double *)HACD_ALLOC(sizeof(double) * 3 * h.m_nPoints);
+					memcpy(h.m_points, vhull.m_points, sizeof(double) * 3 * h.m_nPoints);
+					h.m_nTriangles = vhull.m_nTriangles;
+					h.m_triangles = (uint32_t *)HACD_ALLOC(sizeof(uint32_t) * 3 * h.m_nTriangles);
+					memcpy(h.m_triangles, vhull.m_triangles, sizeof(uint32_t) * 3 * h.m_nTriangles);
+					h.m_volume = vhull.m_volume;
+					h.m_center[0] = vhull.m_center[0];
+					h.m_center[1] = vhull.m_center[1];
+					h.m_center[2] = vhull.m_center[2];
+					mHulls[i] = h;
+					if (mCancel)
+					{
+						ret = 0;
+						break;
+					}
+				}
+			}
+		}
+
+		mHullCount = ret;
+		return ret ? true : false;
+	}
+
+	void releaseHull(VHACD::IVHACD::ConvexHull &h)
+	{
+		HACD_FREE((void *)h.m_triangles);
+		HACD_FREE((void *)h.m_points);
+		h.m_triangles = nullptr;
+		h.m_points = nullptr;
+	}
+
+	virtual void GetConvexHull(const uint32_t index, VHACD::IVHACD::ConvexHull& ch) const final
+	{
+		if ( index < mHullCount )
+		{
+			ch = mHulls[index];
+		}
+	}
+
+	void	releaseHACD(void) // release memory associated with the last HACD request
+	{
+		for (uint32_t i=0; i<mHullCount; i++)
+		{
+			releaseHull(mHulls[i]);
+		}
+		delete[]mHulls;
+		mHulls = nullptr;
+		mHullCount = 0;
+		HACD_FREE(mVertices);
+		mVertices = nullptr;
+		HACD_FREE(mIndices);
+		mIndices = nullptr;
+	}
+
+
+	virtual void release(void) // release the HACD_API interface
+	{
+		delete this;
+	}
+
+	virtual uint32_t	getHullCount(void)
+	{
+		return mHullCount;
+	}
+
+	virtual void Cancel() final
+	{
+		if (mRunning)
+		{
+			mVHACD->Cancel();	// Set the cancel signal to the base VHACD
+		}
+		if (mThread)
+		{
+			mThread->join();	// Wait for the thread to fully exit before we delete the instance
+			delete mThread;
+			mThread = nullptr;
+			Log("Convex Decomposition thread canceled\n");
+		}
+		mCancel = false; // clear the cancel semaphore
+	}
+
+	virtual bool Compute(const float* const points,
+		const uint32_t countPoints,
+		const uint32_t* const triangles,
+		const uint32_t countTriangles,
+		const Parameters& params) final
+	{
+
+		double *vertices = (double *)HACD_ALLOC(sizeof(double)*countPoints * 3);
+		const float *source = points;
+		double *dest = vertices;
+		for (uint32_t i = 0; i < countPoints; i++)
+		{
+			dest[0] = source[0];
+			dest[1] = source[1];
+			dest[2] = source[2];
+			dest += 3;
+			source += 3;
+		}
+
+		bool ret =  Compute(vertices, countPoints, triangles, countTriangles, params);
+		HACD_FREE(vertices);
+		return ret;
+	}
+
+	virtual uint32_t GetNConvexHulls() const final
+	{
+		processPendingMessages();
+		return mHullCount;
+	}
+
+	virtual void Clean(void) final // release internally allocated memory
+	{
+		Cancel();
+		releaseHACD();
+		mVHACD->Clean();
+	}
+
+	virtual void Release(void) final  // release IVHACD
+	{
+		delete this;
+	}
+
+	virtual bool OCLInit(void* const oclDevice,
+		IVHACD::IUserLogger* const logger = 0) final
+	{
+		return mVHACD->OCLInit(oclDevice, logger);
+	}
+		
+	virtual bool OCLRelease(IVHACD::IUserLogger* const logger = 0) final
+	{
+		return mVHACD->OCLRelease(logger);
+	}
+
+	virtual void Update(const double overallProgress,
+		const double stageProgress,
+		const double operationProgress,
+		const char* const stage,
+		const char* const operation) final
+	{
+		mMessageMutex.lock();
+		mHaveUpdateMessage = true;
+		mOverallProgress = overallProgress;
+		mStageProgress = stageProgress;
+		mOperationProgress = operationProgress;
+		mStage = std::string(stage);
+		mOperation = std::string(operation);
+		mMessageMutex.unlock();
+	}
+
+	virtual void Log(const char* const msg) final
+	{
+		mMessageMutex.lock();
+		mHaveLogMessage = true;
+		mMessage = std::string(msg);
+		mMessageMutex.unlock();
+	}
+
+	virtual bool IsReady(void) const final
+	{
+		processPendingMessages();
+		return !mRunning; 
+	}
+
+	// As a convenience for the calling application we only send it update and log messages from it's own main
+	// thread.  This reduces the complexity burden on the caller by making sure it only has to deal with log
+	// messages in it's main application thread.
+	void processPendingMessages(void) const
+	{
+		// If we have a new update message and the user has specified a callback we send the message and clear the semaphore
+		if (mHaveUpdateMessage && mCallback)
+		{
+			mMessageMutex.lock();
+			mCallback->Update(mOverallProgress, mStageProgress, mOperationProgress, mStage.c_str(), mOperation.c_str());
+			mHaveUpdateMessage = false;
+			mMessageMutex.unlock();
+		}
+		// If we have a new log message and the user has specified a callback we send the message and clear the semaphore
+		if (mHaveLogMessage && mLogger)
+		{
+			mMessageMutex.lock();
+			mLogger->Log(mMessage.c_str());
+			mHaveLogMessage = false;
+			mMessageMutex.unlock();
+		}
+	}
+
+	// Will compute the center of mass of the convex hull decomposition results and return it
+	// in 'centerOfMass'.  Returns false if the center of mass could not be computed.
+	virtual bool ComputeCenterOfMass(double centerOfMass[3]) const
+	{
+		bool ret = false;
+
+		centerOfMass[0] = 0;
+		centerOfMass[1] = 0;
+		centerOfMass[2] = 0;
+
+		if (mVHACD && IsReady() )
+		{
+			ret = mVHACD->ComputeCenterOfMass(centerOfMass);
+		}
+		return ret;
+	}
+
+private:
+	double							*mVertices{ nullptr };
+	uint32_t						*mIndices{ nullptr };
+	std::atomic< uint32_t>			mHullCount{ 0 };
+	VHACD::IVHACD::ConvexHull		*mHulls{ nullptr };
+	VHACD::IVHACD::IUserCallback	*mCallback{ nullptr };
+	VHACD::IVHACD::IUserLogger		*mLogger{ nullptr };
+	VHACD::IVHACD					*mVHACD{ nullptr };
+	std::thread						*mThread{ nullptr };
+	std::atomic< bool >				mRunning{ false };
+	std::atomic<bool>				mCancel{ false };
+
+	// Thread safe caching mechanism for messages and update status.
+	// This is so that caller always gets messages in his own thread
+	// Member variables are marked as 'mutable' since the message dispatch function
+	// is called from const query methods.
+	mutable std::mutex						mMessageMutex;
+	mutable std::atomic< bool >				mHaveUpdateMessage{ false };
+	mutable std::atomic< bool >				mHaveLogMessage{ false };
+	mutable double							mOverallProgress{ 0 };
+	mutable double							mStageProgress{ 0 };
+	mutable double							mOperationProgress{ 0 };
+	mutable std::string						mStage;
+	mutable std::string						mOperation;
+	mutable std::string						mMessage;
+};
+
+IVHACD* CreateVHACD_ASYNC(void)
+{
+	MyHACD_API *m = new MyHACD_API;
+	return static_cast<IVHACD *>(m);
+}
+
+
+}; // end of VHACD namespace
+
diff --git a/thirdparty/vhacd/src/VHACD.cpp b/thirdparty/vhacd/src/VHACD.cpp
new file mode 100644
index 0000000000..54cc6e3761
--- /dev/null
+++ b/thirdparty/vhacd/src/VHACD.cpp
@@ -0,0 +1,1589 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _CRT_SECURE_NO_WARNINGS
+#define _CRT_SECURE_NO_WARNINGS
+#endif
+
+#include <algorithm>
+#include <fstream>
+#include <iomanip>
+#include <limits>
+#include <sstream>
+#if _OPENMP
+#include <omp.h>
+#endif // _OPENMP
+
+#include "../public/VHACD.h"
+#include "btConvexHullComputer.h"
+#include "vhacdICHull.h"
+#include "vhacdMesh.h"
+#include "vhacdSArray.h"
+#include "vhacdTimer.h"
+#include "vhacdVHACD.h"
+#include "vhacdVector.h"
+#include "vhacdVolume.h"
+#include "FloatMath.h"
+
+#define MAX(a, b) (((a) > (b)) ? (a) : (b))
+#define MIN(a, b) (((a) < (b)) ? (a) : (b))
+#define ABS(a) (((a) < 0) ? -(a) : (a))
+#define ZSGN(a) (((a) < 0) ? -1 : (a) > 0 ? 1 : 0)
+#define MAX_DOUBLE (1.79769e+308)
+
+#ifdef _MSC_VER
+#pragma warning(disable:4267 4100 4244 4456)
+#endif
+
+#ifdef USE_SSE
+#include <immintrin.h>
+
+const int32_t SIMD_WIDTH = 4;
+inline int32_t FindMinimumElement(const float* const d, float* const _, const int32_t n)
+{
+    // Min within vectors
+    __m128 min_i = _mm_set1_ps(-1.0f);
+    __m128 min_v = _mm_set1_ps(std::numeric_limits<float>::max());
+    for (int32_t i = 0; i <= n - SIMD_WIDTH; i += SIMD_WIDTH) {
+        const __m128 data = _mm_load_ps(&d[i]);
+        const __m128 pred = _mm_cmplt_ps(data, min_v);
+
+        min_i = _mm_blendv_ps(min_i, _mm_set1_ps(i), pred);
+        min_v = _mm_min_ps(data, min_v);
+    }
+
+    /* Min within vector */
+    const __m128 min1 = _mm_shuffle_ps(min_v, min_v, _MM_SHUFFLE(1, 0, 3, 2));
+    const __m128 min2 = _mm_min_ps(min_v, min1);
+    const __m128 min3 = _mm_shuffle_ps(min2, min2, _MM_SHUFFLE(0, 1, 0, 1));
+    const __m128 min4 = _mm_min_ps(min2, min3);
+    float min_d = _mm_cvtss_f32(min4);
+
+    // Min index
+    const int32_t min_idx = __builtin_ctz(_mm_movemask_ps(_mm_cmpeq_ps(min_v, min4)));
+    int32_t ret = min_i[min_idx] + min_idx;
+
+    // Trailing elements
+    for (int32_t i = (n & ~(SIMD_WIDTH - 1)); i < n; ++i) {
+        if (d[i] < min_d) {
+            min_d = d[i];
+            ret = i;
+        }
+    }
+
+    *m = min_d;
+    return ret;
+}
+
+inline int32_t FindMinimumElement(const float* const d, float* const m, const int32_t begin, const int32_t end)
+{
+    // Leading elements
+    int32_t min_i = -1;
+    float min_d = std::numeric_limits<float>::max();
+    const int32_t aligned = (begin & ~(SIMD_WIDTH - 1)) + ((begin & (SIMD_WIDTH - 1)) ? SIMD_WIDTH : 0);
+    for (int32_t i = begin; i < std::min(end, aligned); ++i) {
+        if (d[i] < min_d) {
+            min_d = d[i];
+            min_i = i;
+        }
+    }
+
+    // Middle and trailing elements
+    float r_m = std::numeric_limits<float>::max();
+    const int32_t n = end - aligned;
+    const int32_t r_i = (n > 0) ? FindMinimumElement(&d[aligned], &r_m, n) : 0;
+
+    // Pick the lowest
+    if (r_m < min_d) {
+        *m = r_m;
+        return r_i + aligned;
+    }
+    else {
+        *m = min_d;
+        return min_i;
+    }
+}
+#else
+inline int32_t FindMinimumElement(const float* const d, float* const m, const int32_t begin, const int32_t end)
+{
+    int32_t idx = -1;
+    float min = (std::numeric_limits<float>::max)();
+    for (size_t i = begin; i < size_t(end); ++i) {
+        if (d[i] < min) {
+            idx = i;
+            min = d[i];
+        }
+    }
+
+    *m = min;
+    return idx;
+}
+#endif
+
+//#define OCL_SOURCE_FROM_FILE
+#ifndef OCL_SOURCE_FROM_FILE
+const char* oclProgramSource = "\
+__kernel void ComputePartialVolumes(__global short4 * voxels,                    \
+                                    const    int      numVoxels,                 \
+                                    const    float4   plane,                     \
+                                    const    float4   minBB,                     \
+                                    const    float4   scale,                     \
+                                    __local  uint4 *  localPartialVolumes,       \
+                                    __global uint4 *  partialVolumes)            \
+{                                                                                \
+    int localId = get_local_id(0);                                               \
+    int groupSize = get_local_size(0);                                           \
+    int i0 = get_global_id(0) << 2;                                              \
+    float4 voxel;                                                                \
+    uint4  v;                                                                    \
+    voxel = convert_float4(voxels[i0]);                                          \
+    v.s0 = (dot(plane, mad(scale, voxel, minBB)) >= 0.0f) * (i0     < numVoxels);\
+    voxel = convert_float4(voxels[i0 + 1]);                                      \
+    v.s1 = (dot(plane, mad(scale, voxel, minBB)) >= 0.0f) * (i0 + 1 < numVoxels);\
+    voxel = convert_float4(voxels[i0 + 2]);                                      \
+    v.s2 = (dot(plane, mad(scale, voxel, minBB)) >= 0.0f) * (i0 + 2 < numVoxels);\
+    voxel = convert_float4(voxels[i0 + 3]);                                      \
+    v.s3 = (dot(plane, mad(scale, voxel, minBB)) >= 0.0f) * (i0 + 3 < numVoxels);\
+    localPartialVolumes[localId] = v;                                            \
+    barrier(CLK_LOCAL_MEM_FENCE);                                                \
+    for (int i = groupSize >> 1; i > 0; i >>= 1)                                 \
+    {                                                                            \
+        if (localId < i)                                                         \
+        {                                                                        \
+            localPartialVolumes[localId] += localPartialVolumes[localId + i];    \
+        }                                                                        \
+        barrier(CLK_LOCAL_MEM_FENCE);                                            \
+    }                                                                            \
+    if (localId == 0)                                                            \
+    {                                                                            \
+        partialVolumes[get_group_id(0)] = localPartialVolumes[0];                \
+    }                                                                            \
+}                                                                                \
+__kernel void ComputePartialSums(__global uint4 * data,                          \
+                                 const    int     dataSize,                      \
+                                 __local  uint4 * partialSums)                   \
+{                                                                                \
+    int globalId  = get_global_id(0);                                            \
+    int localId   = get_local_id(0);                                             \
+    int groupSize = get_local_size(0);                                           \
+    int i;                                                                       \
+    if (globalId < dataSize)                                                     \
+    {                                                                            \
+        partialSums[localId] = data[globalId];                                   \
+    }                                                                            \
+    else                                                                         \
+    {                                                                            \
+        partialSums[localId] = (0, 0, 0, 0);                                     \
+    }                                                                            \
+    barrier(CLK_LOCAL_MEM_FENCE);                                                \
+    for (i = groupSize >> 1; i > 0; i >>= 1)                                     \
+    {                                                                            \
+        if (localId < i)                                                         \
+        {                                                                        \
+            partialSums[localId] += partialSums[localId + i];                    \
+        }                                                                        \
+        barrier(CLK_LOCAL_MEM_FENCE);                                            \
+    }                                                                            \
+    if (localId == 0)                                                            \
+    {                                                                            \
+        data[get_group_id(0)] = partialSums[0];                                  \
+    }                                                                            \
+}";
+#endif //OCL_SOURCE_FROM_FILE
+
+namespace VHACD {
+IVHACD* CreateVHACD(void)
+{
+    return new VHACD();
+}
+bool VHACD::OCLInit(void* const oclDevice, IUserLogger* const logger)
+{
+#ifdef CL_VERSION_1_1
+    m_oclDevice = (cl_device_id*)oclDevice;
+    cl_int error;
+    m_oclContext = clCreateContext(NULL, 1, m_oclDevice, NULL, NULL, &error);
+    if (error != CL_SUCCESS) {
+        if (logger) {
+            logger->Log("Couldn't create context\n");
+        }
+        return false;
+    }
+
+#ifdef OCL_SOURCE_FROM_FILE
+    std::string cl_files = OPENCL_CL_FILES;
+// read kernal from file
+#ifdef _WIN32
+    std::replace(cl_files.begin(), cl_files.end(), '/', '\\');
+#endif // _WIN32
+
+    FILE* program_handle = fopen(cl_files.c_str(), "rb");
+    fseek(program_handle, 0, SEEK_END);
+    size_t program_size = ftell(program_handle);
+    rewind(program_handle);
+    char* program_buffer = new char[program_size + 1];
+    program_buffer[program_size] = '\0';
+    fread(program_buffer, sizeof(char), program_size, program_handle);
+    fclose(program_handle);
+    // create program
+    m_oclProgram = clCreateProgramWithSource(m_oclContext, 1, (const char**)&program_buffer, &program_size, &error);
+    delete[] program_buffer;
+#else
+    size_t program_size = strlen(oclProgramSource);
+    m_oclProgram = clCreateProgramWithSource(m_oclContext, 1, (const char**)&oclProgramSource, &program_size, &error);
+#endif
+    if (error != CL_SUCCESS) {
+        if (logger) {
+            logger->Log("Couldn't create program\n");
+        }
+        return false;
+    }
+
+    /* Build program */
+    error = clBuildProgram(m_oclProgram, 1, m_oclDevice, "-cl-denorms-are-zero", NULL, NULL);
+    if (error != CL_SUCCESS) {
+        size_t log_size;
+        /* Find Size of log and print to std output */
+        clGetProgramBuildInfo(m_oclProgram, *m_oclDevice, CL_PROGRAM_BUILD_LOG, 0, NULL, &log_size);
+        char* program_log = new char[log_size + 2];
+        program_log[log_size] = '\n';
+        program_log[log_size + 1] = '\0';
+        clGetProgramBuildInfo(m_oclProgram, *m_oclDevice, CL_PROGRAM_BUILD_LOG, log_size + 1, program_log, NULL);
+        if (logger) {
+            logger->Log("Couldn't build program\n");
+            logger->Log(program_log);
+        }
+        delete[] program_log;
+        return false;
+    }
+
+    delete[] m_oclQueue;
+    delete[] m_oclKernelComputePartialVolumes;
+    delete[] m_oclKernelComputeSum;
+    m_oclQueue = new cl_command_queue[m_ompNumProcessors];
+    m_oclKernelComputePartialVolumes = new cl_kernel[m_ompNumProcessors];
+    m_oclKernelComputeSum = new cl_kernel[m_ompNumProcessors];
+
+    const char nameKernelComputePartialVolumes[] = "ComputePartialVolumes";
+    const char nameKernelComputeSum[] = "ComputePartialSums";
+    for (int32_t k = 0; k < m_ompNumProcessors; ++k) {
+        m_oclKernelComputePartialVolumes[k] = clCreateKernel(m_oclProgram, nameKernelComputePartialVolumes, &error);
+        if (error != CL_SUCCESS) {
+            if (logger) {
+                logger->Log("Couldn't create kernel\n");
+            }
+            return false;
+        }
+        m_oclKernelComputeSum[k] = clCreateKernel(m_oclProgram, nameKernelComputeSum, &error);
+        if (error != CL_SUCCESS) {
+            if (logger) {
+                logger->Log("Couldn't create kernel\n");
+            }
+            return false;
+        }
+    }
+
+    error = clGetKernelWorkGroupInfo(m_oclKernelComputePartialVolumes[0],
+        *m_oclDevice,
+        CL_KERNEL_WORK_GROUP_SIZE,
+        sizeof(size_t),
+        &m_oclWorkGroupSize,
+        NULL);
+    size_t workGroupSize = 0;
+    error = clGetKernelWorkGroupInfo(m_oclKernelComputeSum[0],
+        *m_oclDevice,
+        CL_KERNEL_WORK_GROUP_SIZE,
+        sizeof(size_t),
+        &workGroupSize,
+        NULL);
+    if (error != CL_SUCCESS) {
+        if (logger) {
+            logger->Log("Couldn't query work group info\n");
+        }
+        return false;
+    }
+
+    if (workGroupSize < m_oclWorkGroupSize) {
+        m_oclWorkGroupSize = workGroupSize;
+    }
+
+    for (int32_t k = 0; k < m_ompNumProcessors; ++k) {
+        m_oclQueue[k] = clCreateCommandQueue(m_oclContext, *m_oclDevice, 0 /*CL_QUEUE_PROFILING_ENABLE*/, &error);
+        if (error != CL_SUCCESS) {
+            if (logger) {
+                logger->Log("Couldn't create queue\n");
+            }
+            return false;
+        }
+    }
+    return true;
+#else //CL_VERSION_1_1
+    return false;
+#endif //CL_VERSION_1_1
+}
+bool VHACD::OCLRelease(IUserLogger* const logger)
+{
+#ifdef CL_VERSION_1_1
+    cl_int error;
+    if (m_oclKernelComputePartialVolumes) {
+        for (int32_t k = 0; k < m_ompNumProcessors; ++k) {
+            error = clReleaseKernel(m_oclKernelComputePartialVolumes[k]);
+            if (error != CL_SUCCESS) {
+                if (logger) {
+                    logger->Log("Couldn't release kernal\n");
+                }
+                return false;
+            }
+        }
+        delete[] m_oclKernelComputePartialVolumes;
+    }
+    if (m_oclKernelComputeSum) {
+        for (int32_t k = 0; k < m_ompNumProcessors; ++k) {
+            error = clReleaseKernel(m_oclKernelComputeSum[k]);
+            if (error != CL_SUCCESS) {
+                if (logger) {
+                    logger->Log("Couldn't release kernal\n");
+                }
+                return false;
+            }
+        }
+        delete[] m_oclKernelComputeSum;
+    }
+    if (m_oclQueue) {
+        for (int32_t k = 0; k < m_ompNumProcessors; ++k) {
+            error = clReleaseCommandQueue(m_oclQueue[k]);
+            if (error != CL_SUCCESS) {
+                if (logger) {
+                    logger->Log("Couldn't release queue\n");
+                }
+                return false;
+            }
+        }
+        delete[] m_oclQueue;
+    }
+    error = clReleaseProgram(m_oclProgram);
+    if (error != CL_SUCCESS) {
+        if (logger) {
+            logger->Log("Couldn't release program\n");
+        }
+        return false;
+    }
+    error = clReleaseContext(m_oclContext);
+    if (error != CL_SUCCESS) {
+        if (logger) {
+            logger->Log("Couldn't release context\n");
+        }
+        return false;
+    }
+
+    return true;
+#else //CL_VERSION_1_1
+    return false;
+#endif //CL_VERSION_1_1
+}
+void VHACD::ComputePrimitiveSet(const Parameters& params)
+{
+    if (GetCancel()) {
+        return;
+    }
+    m_timer.Tic();
+
+    m_stage = "Compute primitive set";
+    m_operation = "Convert volume to pset";
+
+    std::ostringstream msg;
+    if (params.m_logger) {
+        msg << "+ " << m_stage << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+    Update(0.0, 0.0, params);
+    if (params.m_mode == 0) {
+        VoxelSet* vset = new VoxelSet;
+        m_volume->Convert(*vset);
+        m_pset = vset;
+    }
+    else {
+        TetrahedronSet* tset = new TetrahedronSet;
+        m_volume->Convert(*tset);
+        m_pset = tset;
+    }
+
+    delete m_volume;
+    m_volume = 0;
+
+    if (params.m_logger) {
+        msg.str("");
+        msg << "\t # primitives               " << m_pset->GetNPrimitives() << std::endl;
+        msg << "\t # inside surface           " << m_pset->GetNPrimitivesInsideSurf() << std::endl;
+        msg << "\t # on surface               " << m_pset->GetNPrimitivesOnSurf() << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+    m_overallProgress = 15.0;
+    Update(100.0, 100.0, params);
+    m_timer.Toc();
+    if (params.m_logger) {
+        msg.str("");
+        msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+}
+bool VHACD::Compute(const double* const points, const uint32_t nPoints,
+    const uint32_t* const triangles,const uint32_t nTriangles, const Parameters& params)
+{
+    return ComputeACD(points, nPoints, triangles, nTriangles, params);
+}
+bool VHACD::Compute(const float* const points,const uint32_t nPoints,
+    const uint32_t* const triangles,const uint32_t nTriangles, const Parameters& params)
+{
+    return ComputeACD(points, nPoints, triangles, nTriangles, params);
+}
+double ComputePreferredCuttingDirection(const PrimitiveSet* const tset, Vec3<double>& dir)
+{
+    double ex = tset->GetEigenValue(AXIS_X);
+    double ey = tset->GetEigenValue(AXIS_Y);
+    double ez = tset->GetEigenValue(AXIS_Z);
+    double vx = (ey - ez) * (ey - ez);
+    double vy = (ex - ez) * (ex - ez);
+    double vz = (ex - ey) * (ex - ey);
+    if (vx < vy && vx < vz) {
+        double e = ey * ey + ez * ez;
+        dir[0] = 1.0;
+        dir[1] = 0.0;
+        dir[2] = 0.0;
+        return (e == 0.0) ? 0.0 : 1.0 - vx / e;
+    }
+    else if (vy < vx && vy < vz) {
+        double e = ex * ex + ez * ez;
+        dir[0] = 0.0;
+        dir[1] = 1.0;
+        dir[2] = 0.0;
+        return (e == 0.0) ? 0.0 : 1.0 - vy / e;
+    }
+    else {
+        double e = ex * ex + ey * ey;
+        dir[0] = 0.0;
+        dir[1] = 0.0;
+        dir[2] = 1.0;
+        return (e == 0.0) ? 0.0 : 1.0 - vz / e;
+    }
+}
+void ComputeAxesAlignedClippingPlanes(const VoxelSet& vset, const short downsampling, SArray<Plane>& planes)
+{
+    const Vec3<short> minV = vset.GetMinBBVoxels();
+    const Vec3<short> maxV = vset.GetMaxBBVoxels();
+    Vec3<double> pt;
+    Plane plane;
+    const short i0 = minV[0];
+    const short i1 = maxV[0];
+    plane.m_a = 1.0;
+    plane.m_b = 0.0;
+    plane.m_c = 0.0;
+    plane.m_axis = AXIS_X;
+    for (short i = i0; i <= i1; i += downsampling) {
+        pt = vset.GetPoint(Vec3<double>(i + 0.5, 0.0, 0.0));
+        plane.m_d = -pt[0];
+        plane.m_index = i;
+        planes.PushBack(plane);
+    }
+    const short j0 = minV[1];
+    const short j1 = maxV[1];
+    plane.m_a = 0.0;
+    plane.m_b = 1.0;
+    plane.m_c = 0.0;
+    plane.m_axis = AXIS_Y;
+    for (short j = j0; j <= j1; j += downsampling) {
+        pt = vset.GetPoint(Vec3<double>(0.0, j + 0.5, 0.0));
+        plane.m_d = -pt[1];
+        plane.m_index = j;
+        planes.PushBack(plane);
+    }
+    const short k0 = minV[2];
+    const short k1 = maxV[2];
+    plane.m_a = 0.0;
+    plane.m_b = 0.0;
+    plane.m_c = 1.0;
+    plane.m_axis = AXIS_Z;
+    for (short k = k0; k <= k1; k += downsampling) {
+        pt = vset.GetPoint(Vec3<double>(0.0, 0.0, k + 0.5));
+        plane.m_d = -pt[2];
+        plane.m_index = k;
+        planes.PushBack(plane);
+    }
+}
+void ComputeAxesAlignedClippingPlanes(const TetrahedronSet& tset, const short downsampling, SArray<Plane>& planes)
+{
+    const Vec3<double> minV = tset.GetMinBB();
+    const Vec3<double> maxV = tset.GetMaxBB();
+    const double scale = tset.GetSacle();
+    const short i0 = 0;
+    const short j0 = 0;
+    const short k0 = 0;
+    const short i1 = static_cast<short>((maxV[0] - minV[0]) / scale + 0.5);
+    const short j1 = static_cast<short>((maxV[1] - minV[1]) / scale + 0.5);
+    const short k1 = static_cast<short>((maxV[2] - minV[2]) / scale + 0.5);
+
+    Plane plane;
+    plane.m_a = 1.0;
+    plane.m_b = 0.0;
+    plane.m_c = 0.0;
+    plane.m_axis = AXIS_X;
+    for (short i = i0; i <= i1; i += downsampling) {
+        double x = minV[0] + scale * i;
+        plane.m_d = -x;
+        plane.m_index = i;
+        planes.PushBack(plane);
+    }
+    plane.m_a = 0.0;
+    plane.m_b = 1.0;
+    plane.m_c = 0.0;
+    plane.m_axis = AXIS_Y;
+    for (short j = j0; j <= j1; j += downsampling) {
+        double y = minV[1] + scale * j;
+        plane.m_d = -y;
+        plane.m_index = j;
+        planes.PushBack(plane);
+    }
+    plane.m_a = 0.0;
+    plane.m_b = 0.0;
+    plane.m_c = 1.0;
+    plane.m_axis = AXIS_Z;
+    for (short k = k0; k <= k1; k += downsampling) {
+        double z = minV[2] + scale * k;
+        plane.m_d = -z;
+        plane.m_index = k;
+        planes.PushBack(plane);
+    }
+}
+void RefineAxesAlignedClippingPlanes(const VoxelSet& vset, const Plane& bestPlane, const short downsampling,
+    SArray<Plane>& planes)
+{
+    const Vec3<short> minV = vset.GetMinBBVoxels();
+    const Vec3<short> maxV = vset.GetMaxBBVoxels();
+    Vec3<double> pt;
+    Plane plane;
+
+    if (bestPlane.m_axis == AXIS_X) {
+        const short i0 = MAX(minV[0], bestPlane.m_index - downsampling);
+        const short i1 = MIN(maxV[0], bestPlane.m_index + downsampling);
+        plane.m_a = 1.0;
+        plane.m_b = 0.0;
+        plane.m_c = 0.0;
+        plane.m_axis = AXIS_X;
+        for (short i = i0; i <= i1; ++i) {
+            pt = vset.GetPoint(Vec3<double>(i + 0.5, 0.0, 0.0));
+            plane.m_d = -pt[0];
+            plane.m_index = i;
+            planes.PushBack(plane);
+        }
+    }
+    else if (bestPlane.m_axis == AXIS_Y) {
+        const short j0 = MAX(minV[1], bestPlane.m_index - downsampling);
+        const short j1 = MIN(maxV[1], bestPlane.m_index + downsampling);
+        plane.m_a = 0.0;
+        plane.m_b = 1.0;
+        plane.m_c = 0.0;
+        plane.m_axis = AXIS_Y;
+        for (short j = j0; j <= j1; ++j) {
+            pt = vset.GetPoint(Vec3<double>(0.0, j + 0.5, 0.0));
+            plane.m_d = -pt[1];
+            plane.m_index = j;
+            planes.PushBack(plane);
+        }
+    }
+    else {
+        const short k0 = MAX(minV[2], bestPlane.m_index - downsampling);
+        const short k1 = MIN(maxV[2], bestPlane.m_index + downsampling);
+        plane.m_a = 0.0;
+        plane.m_b = 0.0;
+        plane.m_c = 1.0;
+        plane.m_axis = AXIS_Z;
+        for (short k = k0; k <= k1; ++k) {
+            pt = vset.GetPoint(Vec3<double>(0.0, 0.0, k + 0.5));
+            plane.m_d = -pt[2];
+            plane.m_index = k;
+            planes.PushBack(plane);
+        }
+    }
+}
+void RefineAxesAlignedClippingPlanes(const TetrahedronSet& tset, const Plane& bestPlane, const short downsampling,
+    SArray<Plane>& planes)
+{
+    const Vec3<double> minV = tset.GetMinBB();
+    const Vec3<double> maxV = tset.GetMaxBB();
+    const double scale = tset.GetSacle();
+    Plane plane;
+
+    if (bestPlane.m_axis == AXIS_X) {
+        const short i0 = MAX(0, bestPlane.m_index - downsampling);
+        const short i1 = static_cast<short>(MIN((maxV[0] - minV[0]) / scale + 0.5, bestPlane.m_index + downsampling));
+        plane.m_a = 1.0;
+        plane.m_b = 0.0;
+        plane.m_c = 0.0;
+        plane.m_axis = AXIS_X;
+        for (short i = i0; i <= i1; ++i) {
+            double x = minV[0] + scale * i;
+            plane.m_d = -x;
+            plane.m_index = i;
+            planes.PushBack(plane);
+        }
+    }
+    else if (bestPlane.m_axis == AXIS_Y) {
+        const short j0 = MAX(0, bestPlane.m_index - downsampling);
+        const short j1 = static_cast<short>(MIN((maxV[1] - minV[1]) / scale + 0.5, bestPlane.m_index + downsampling));
+        plane.m_a = 0.0;
+        plane.m_b = 1.0;
+        plane.m_c = 0.0;
+        plane.m_axis = AXIS_Y;
+        for (short j = j0; j <= j1; ++j) {
+            double y = minV[1] + scale * j;
+            plane.m_d = -y;
+            plane.m_index = j;
+            planes.PushBack(plane);
+        }
+    }
+    else {
+        const short k0 = MAX(0, bestPlane.m_index - downsampling);
+        const short k1 = static_cast<short>(MIN((maxV[2] - minV[2]) / scale + 0.5, bestPlane.m_index + downsampling));
+        plane.m_a = 0.0;
+        plane.m_b = 0.0;
+        plane.m_c = 1.0;
+        plane.m_axis = AXIS_Z;
+        for (short k = k0; k <= k1; ++k) {
+            double z = minV[2] + scale * k;
+            plane.m_d = -z;
+            plane.m_index = k;
+            planes.PushBack(plane);
+        }
+    }
+}
+inline double ComputeLocalConcavity(const double volume, const double volumeCH)
+{
+    return fabs(volumeCH - volume) / volumeCH;
+}
+inline double ComputeConcavity(const double volume, const double volumeCH, const double volume0)
+{
+    return fabs(volumeCH - volume) / volume0;
+}
+
+//#define DEBUG_TEMP
+void VHACD::ComputeBestClippingPlane(const PrimitiveSet* inputPSet, const double volume, const SArray<Plane>& planes,
+    const Vec3<double>& preferredCuttingDirection, const double w, const double alpha, const double beta,
+    const int32_t convexhullDownsampling, const double progress0, const double progress1, Plane& bestPlane,
+    double& minConcavity, const Parameters& params)
+{
+    if (GetCancel()) {
+        return;
+    }
+    char msg[256];
+    size_t nPrimitives = inputPSet->GetNPrimitives();
+    bool oclAcceleration = (nPrimitives > OCL_MIN_NUM_PRIMITIVES && params.m_oclAcceleration && params.m_mode == 0) ? true : false;
+    int32_t iBest = -1;
+    int32_t nPlanes = static_cast<int32_t>(planes.Size());
+    bool cancel = false;
+    int32_t done = 0;
+    double minTotal = MAX_DOUBLE;
+    double minBalance = MAX_DOUBLE;
+    double minSymmetry = MAX_DOUBLE;
+    minConcavity = MAX_DOUBLE;
+
+    SArray<Vec3<double> >* chPts = new SArray<Vec3<double> >[2 * m_ompNumProcessors];
+    Mesh* chs = new Mesh[2 * m_ompNumProcessors];
+    PrimitiveSet* onSurfacePSet = inputPSet->Create();
+    inputPSet->SelectOnSurface(onSurfacePSet);
+
+    PrimitiveSet** psets = 0;
+    if (!params.m_convexhullApproximation) {
+        psets = new PrimitiveSet*[2 * m_ompNumProcessors];
+        for (int32_t i = 0; i < 2 * m_ompNumProcessors; ++i) {
+            psets[i] = inputPSet->Create();
+        }
+    }
+
+#ifdef CL_VERSION_1_1
+    // allocate OpenCL data structures
+    cl_mem voxels;
+    cl_mem* partialVolumes = 0;
+    size_t globalSize = 0;
+    size_t nWorkGroups = 0;
+    double unitVolume = 0.0;
+    if (oclAcceleration) {
+        VoxelSet* vset = (VoxelSet*)inputPSet;
+        const Vec3<double> minBB = vset->GetMinBB();
+        const float fMinBB[4] = { (float)minBB[0], (float)minBB[1], (float)minBB[2], 1.0f };
+        const float fSclae[4] = { (float)vset->GetScale(), (float)vset->GetScale(), (float)vset->GetScale(), 0.0f };
+        const int32_t nVoxels = (int32_t)nPrimitives;
+        unitVolume = vset->GetUnitVolume();
+        nWorkGroups = (nPrimitives + 4 * m_oclWorkGroupSize - 1) / (4 * m_oclWorkGroupSize);
+        globalSize = nWorkGroups * m_oclWorkGroupSize;
+        cl_int error;
+        voxels = clCreateBuffer(m_oclContext,
+            CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR,
+            sizeof(Voxel) * nPrimitives,
+            vset->GetVoxels(),
+            &error);
+        if (error != CL_SUCCESS) {
+            if (params.m_logger) {
+                params.m_logger->Log("Couldn't create buffer\n");
+            }
+            SetCancel(true);
+        }
+
+        partialVolumes = new cl_mem[m_ompNumProcessors];
+        for (int32_t i = 0; i < m_ompNumProcessors; ++i) {
+            partialVolumes[i] = clCreateBuffer(m_oclContext,
+                CL_MEM_WRITE_ONLY,
+                sizeof(uint32_t) * 4 * nWorkGroups,
+                NULL,
+                &error);
+            if (error != CL_SUCCESS) {
+                if (params.m_logger) {
+                    params.m_logger->Log("Couldn't create buffer\n");
+                }
+                SetCancel(true);
+                break;
+            }
+            error = clSetKernelArg(m_oclKernelComputePartialVolumes[i], 0, sizeof(cl_mem), &voxels);
+            error |= clSetKernelArg(m_oclKernelComputePartialVolumes[i], 1, sizeof(uint32_t), &nVoxels);
+            error |= clSetKernelArg(m_oclKernelComputePartialVolumes[i], 3, sizeof(float) * 4, fMinBB);
+            error |= clSetKernelArg(m_oclKernelComputePartialVolumes[i], 4, sizeof(float) * 4, &fSclae);
+            error |= clSetKernelArg(m_oclKernelComputePartialVolumes[i], 5, sizeof(uint32_t) * 4 * m_oclWorkGroupSize, NULL);
+            error |= clSetKernelArg(m_oclKernelComputePartialVolumes[i], 6, sizeof(cl_mem), &(partialVolumes[i]));
+            error |= clSetKernelArg(m_oclKernelComputeSum[i], 0, sizeof(cl_mem), &(partialVolumes[i]));
+            error |= clSetKernelArg(m_oclKernelComputeSum[i], 2, sizeof(uint32_t) * 4 * m_oclWorkGroupSize, NULL);
+            if (error != CL_SUCCESS) {
+                if (params.m_logger) {
+                    params.m_logger->Log("Couldn't kernel arguments \n");
+                }
+                SetCancel(true);
+            }
+        }
+    }
+#else // CL_VERSION_1_1
+    oclAcceleration = false;
+#endif // CL_VERSION_1_1
+
+#ifdef DEBUG_TEMP
+    Timer timerComputeCost;
+    timerComputeCost.Tic();
+#endif // DEBUG_TEMP
+
+#if USE_THREAD == 1 && _OPENMP
+#pragma omp parallel for
+#endif
+    for (int32_t x = 0; x < nPlanes; ++x) {
+        int32_t threadID = 0;
+#if USE_THREAD == 1 && _OPENMP
+        threadID = omp_get_thread_num();
+#pragma omp flush(cancel)
+#endif
+        if (!cancel) {
+            //Update progress
+            if (GetCancel()) {
+                cancel = true;
+#if USE_THREAD == 1 && _OPENMP
+#pragma omp flush(cancel)
+#endif
+            }
+            Plane plane = planes[x];
+
+            if (oclAcceleration) {
+#ifdef CL_VERSION_1_1
+                const float fPlane[4] = { (float)plane.m_a, (float)plane.m_b, (float)plane.m_c, (float)plane.m_d };
+                cl_int error = clSetKernelArg(m_oclKernelComputePartialVolumes[threadID], 2, sizeof(float) * 4, fPlane);
+                if (error != CL_SUCCESS) {
+                    if (params.m_logger) {
+                        params.m_logger->Log("Couldn't kernel atguments \n");
+                    }
+                    SetCancel(true);
+                }
+
+                error = clEnqueueNDRangeKernel(m_oclQueue[threadID], m_oclKernelComputePartialVolumes[threadID],
+                    1, NULL, &globalSize, &m_oclWorkGroupSize, 0, NULL, NULL);
+                if (error != CL_SUCCESS) {
+                    if (params.m_logger) {
+                        params.m_logger->Log("Couldn't run kernel \n");
+                    }
+                    SetCancel(true);
+                }
+                int32_t nValues = (int32_t)nWorkGroups;
+                while (nValues > 1) {
+                    error = clSetKernelArg(m_oclKernelComputeSum[threadID], 1, sizeof(int32_t), &nValues);
+                    if (error != CL_SUCCESS) {
+                        if (params.m_logger) {
+                            params.m_logger->Log("Couldn't kernel atguments \n");
+                        }
+                        SetCancel(true);
+                    }
+                    size_t nWorkGroups = (nValues + m_oclWorkGroupSize - 1) / m_oclWorkGroupSize;
+                    size_t globalSize = nWorkGroups * m_oclWorkGroupSize;
+                    error = clEnqueueNDRangeKernel(m_oclQueue[threadID], m_oclKernelComputeSum[threadID],
+                        1, NULL, &globalSize, &m_oclWorkGroupSize, 0, NULL, NULL);
+                    if (error != CL_SUCCESS) {
+                        if (params.m_logger) {
+                            params.m_logger->Log("Couldn't run kernel \n");
+                        }
+                        SetCancel(true);
+                    }
+                    nValues = (int32_t)nWorkGroups;
+                }
+#endif // CL_VERSION_1_1
+            }
+
+            Mesh& leftCH = chs[threadID];
+            Mesh& rightCH = chs[threadID + m_ompNumProcessors];
+            rightCH.ResizePoints(0);
+            leftCH.ResizePoints(0);
+            rightCH.ResizeTriangles(0);
+            leftCH.ResizeTriangles(0);
+
+// compute convex-hulls
+#ifdef TEST_APPROX_CH
+            double volumeLeftCH1;
+            double volumeRightCH1;
+#endif //TEST_APPROX_CH
+            if (params.m_convexhullApproximation) {
+                SArray<Vec3<double> >& leftCHPts = chPts[threadID];
+                SArray<Vec3<double> >& rightCHPts = chPts[threadID + m_ompNumProcessors];
+                rightCHPts.Resize(0);
+                leftCHPts.Resize(0);
+                onSurfacePSet->Intersect(plane, &rightCHPts, &leftCHPts, convexhullDownsampling * 32);
+                inputPSet->GetConvexHull().Clip(plane, rightCHPts, leftCHPts);
+                rightCH.ComputeConvexHull((double*)rightCHPts.Data(), rightCHPts.Size());
+                leftCH.ComputeConvexHull((double*)leftCHPts.Data(), leftCHPts.Size());
+#ifdef TEST_APPROX_CH
+                Mesh leftCH1;
+                Mesh rightCH1;
+                VoxelSet right;
+                VoxelSet left;
+                onSurfacePSet->Clip(plane, &right, &left);
+                right.ComputeConvexHull(rightCH1, convexhullDownsampling);
+                left.ComputeConvexHull(leftCH1, convexhullDownsampling);
+
+                volumeLeftCH1 = leftCH1.ComputeVolume();
+                volumeRightCH1 = rightCH1.ComputeVolume();
+#endif //TEST_APPROX_CH
+            }
+            else {
+                PrimitiveSet* const right = psets[threadID];
+                PrimitiveSet* const left = psets[threadID + m_ompNumProcessors];
+                onSurfacePSet->Clip(plane, right, left);
+                right->ComputeConvexHull(rightCH, convexhullDownsampling);
+                left->ComputeConvexHull(leftCH, convexhullDownsampling);
+            }
+            double volumeLeftCH = leftCH.ComputeVolume();
+            double volumeRightCH = rightCH.ComputeVolume();
+
+            // compute clipped volumes
+            double volumeLeft = 0.0;
+            double volumeRight = 0.0;
+            if (oclAcceleration) {
+#ifdef CL_VERSION_1_1
+                uint32_t volumes[4];
+                cl_int error = clEnqueueReadBuffer(m_oclQueue[threadID], partialVolumes[threadID], CL_TRUE,
+                    0, sizeof(uint32_t) * 4, volumes, 0, NULL, NULL);
+                size_t nPrimitivesRight = volumes[0] + volumes[1] + volumes[2] + volumes[3];
+                size_t nPrimitivesLeft = nPrimitives - nPrimitivesRight;
+                volumeRight = nPrimitivesRight * unitVolume;
+                volumeLeft = nPrimitivesLeft * unitVolume;
+                if (error != CL_SUCCESS) {
+                    if (params.m_logger) {
+                        params.m_logger->Log("Couldn't read buffer \n");
+                    }
+                    SetCancel(true);
+                }
+#endif // CL_VERSION_1_1
+            }
+            else {
+                inputPSet->ComputeClippedVolumes(plane, volumeRight, volumeLeft);
+            }
+            double concavityLeft = ComputeConcavity(volumeLeft, volumeLeftCH, m_volumeCH0);
+            double concavityRight = ComputeConcavity(volumeRight, volumeRightCH, m_volumeCH0);
+            double concavity = (concavityLeft + concavityRight);
+
+            // compute cost
+            double balance = alpha * fabs(volumeLeft - volumeRight) / m_volumeCH0;
+            double d = w * (preferredCuttingDirection[0] * plane.m_a + preferredCuttingDirection[1] * plane.m_b + preferredCuttingDirection[2] * plane.m_c);
+            double symmetry = beta * d;
+            double total = concavity + balance + symmetry;
+
+#if USE_THREAD == 1 && _OPENMP
+#pragma omp critical
+#endif
+            {
+                if (total < minTotal || (total == minTotal && x < iBest)) {
+                    minConcavity = concavity;
+                    minBalance = balance;
+                    minSymmetry = symmetry;
+                    bestPlane = plane;
+                    minTotal = total;
+                    iBest = x;
+                }
+                ++done;
+                if (!(done & 127)) // reduce update frequency
+                {
+                    double progress = done * (progress1 - progress0) / nPlanes + progress0;
+                    Update(m_stageProgress, progress, params);
+                }
+            }
+        }
+    }
+
+#ifdef DEBUG_TEMP
+    timerComputeCost.Toc();
+    printf_s("Cost[%i] = %f\n", nPlanes, timerComputeCost.GetElapsedTime());
+#endif // DEBUG_TEMP
+
+#ifdef CL_VERSION_1_1
+    if (oclAcceleration) {
+        clReleaseMemObject(voxels);
+        for (int32_t i = 0; i < m_ompNumProcessors; ++i) {
+            clReleaseMemObject(partialVolumes[i]);
+        }
+        delete[] partialVolumes;
+    }
+#endif // CL_VERSION_1_1
+
+    if (psets) {
+        for (int32_t i = 0; i < 2 * m_ompNumProcessors; ++i) {
+            delete psets[i];
+        }
+        delete[] psets;
+    }
+    delete onSurfacePSet;
+    delete[] chPts;
+    delete[] chs;
+    if (params.m_logger) {
+        sprintf(msg, "\n\t\t\t Best  %04i T=%2.6f C=%2.6f B=%2.6f S=%2.6f (%1.1f, %1.1f, %1.1f, %3.3f)\n\n", iBest, minTotal, minConcavity, minBalance, minSymmetry, bestPlane.m_a, bestPlane.m_b, bestPlane.m_c, bestPlane.m_d);
+        params.m_logger->Log(msg);
+    }
+}
+void VHACD::ComputeACD(const Parameters& params)
+{
+    if (GetCancel()) {
+        return;
+    }
+    m_timer.Tic();
+
+    m_stage = "Approximate Convex Decomposition";
+    m_stageProgress = 0.0;
+    std::ostringstream msg;
+    if (params.m_logger) {
+        msg << "+ " << m_stage << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+    SArray<PrimitiveSet*> parts;
+    SArray<PrimitiveSet*> inputParts;
+    SArray<PrimitiveSet*> temp;
+    inputParts.PushBack(m_pset);
+    m_pset = 0;
+    SArray<Plane> planes;
+    SArray<Plane> planesRef;
+    uint32_t sub = 0;
+    bool firstIteration = true;
+    m_volumeCH0 = 1.0;
+
+	// Compute the decomposition depth based on the number of convex hulls being requested..
+	uint32_t hullCount = 2;
+	uint32_t depth = 1;
+	while (params.m_maxConvexHulls > hullCount)
+	{
+		depth++;
+		hullCount *= 2;
+	}
+	// We must always increment the decomposition depth one higher than the maximum number of hulls requested.
+	// The reason for this is as follows.
+	// Say, for example, the user requests 32 convex hulls exactly.  This would be a decomposition depth of 5.
+	// However, when we do that, we do *not* necessarily get 32 hulls as a result.  This is because, during
+	// the recursive descent of the binary tree, one or more of the leaf nodes may have no concavity and
+	// will not be split.  So, in this way, even with a decomposition depth of 5, you can produce fewer than
+	// 32 hulls.  So, in this case, we would set the decomposition depth to 6 (producing up to as high as 64 convex hulls).
+	// Then, the merge step which combines over-described hulls down to the user requested amount, we will end up
+	// getting exactly 32 convex hulls as a result.
+	// We could just allow the artist to directly control the decomposition depth directly, but this would be a bit
+	// too complex and the preference is simply to let them specify how many hulls they want and derive the solution
+	// from that.
+	depth++;
+
+
+    while (sub++ < depth && inputParts.Size() > 0 && !m_cancel) {
+        msg.str("");
+        msg << "Subdivision level " << sub;
+        m_operation = msg.str();
+
+        if (params.m_logger) {
+            msg.str("");
+            msg << "\t Subdivision level " << sub << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+
+        double maxConcavity = 0.0;
+        const size_t nInputParts = inputParts.Size();
+        Update(m_stageProgress, 0.0, params);
+        for (size_t p = 0; p < nInputParts && !m_cancel; ++p) {
+            const double progress0 = p * 100.0 / nInputParts;
+            const double progress1 = (p + 0.75) * 100.0 / nInputParts;
+            const double progress2 = (p + 1.00) * 100.0 / nInputParts;
+
+            Update(m_stageProgress, progress0, params);
+
+            PrimitiveSet* pset = inputParts[p];
+            inputParts[p] = 0;
+            double volume = pset->ComputeVolume();
+            pset->ComputeBB();
+            pset->ComputePrincipalAxes();
+            if (params.m_pca) {
+                pset->AlignToPrincipalAxes();
+            }
+
+            pset->ComputeConvexHull(pset->GetConvexHull());
+            double volumeCH = fabs(pset->GetConvexHull().ComputeVolume());
+            if (firstIteration) {
+                m_volumeCH0 = volumeCH;
+            }
+
+            double concavity = ComputeConcavity(volume, volumeCH, m_volumeCH0);
+            double error = 1.01 * pset->ComputeMaxVolumeError() / m_volumeCH0;
+
+            if (firstIteration) {
+                firstIteration = false;
+            }
+
+            if (params.m_logger) {
+                msg.str("");
+                msg << "\t -> Part[" << p
+                    << "] C  = " << concavity
+                    << ", E  = " << error
+                    << ", VS = " << pset->GetNPrimitivesOnSurf()
+                    << ", VI = " << pset->GetNPrimitivesInsideSurf()
+                    << std::endl;
+                params.m_logger->Log(msg.str().c_str());
+            }
+
+            if (concavity > params.m_concavity && concavity > error) {
+                Vec3<double> preferredCuttingDirection;
+                double w = ComputePreferredCuttingDirection(pset, preferredCuttingDirection);
+                planes.Resize(0);
+                if (params.m_mode == 0) {
+                    VoxelSet* vset = (VoxelSet*)pset;
+                    ComputeAxesAlignedClippingPlanes(*vset, params.m_planeDownsampling, planes);
+                }
+                else {
+                    TetrahedronSet* tset = (TetrahedronSet*)pset;
+                    ComputeAxesAlignedClippingPlanes(*tset, params.m_planeDownsampling, planes);
+                }
+
+                if (params.m_logger) {
+                    msg.str("");
+                    msg << "\t\t [Regular sampling] Number of clipping planes " << planes.Size() << std::endl;
+                    params.m_logger->Log(msg.str().c_str());
+                }
+
+                Plane bestPlane;
+                double minConcavity = MAX_DOUBLE;
+                ComputeBestClippingPlane(pset,
+                    volume,
+                    planes,
+                    preferredCuttingDirection,
+                    w,
+                    concavity * params.m_alpha,
+                    concavity * params.m_beta,
+                    params.m_convexhullDownsampling,
+                    progress0,
+                    progress1,
+                    bestPlane,
+                    minConcavity,
+                    params);
+                if (!m_cancel && (params.m_planeDownsampling > 1 || params.m_convexhullDownsampling > 1)) {
+                    planesRef.Resize(0);
+
+                    if (params.m_mode == 0) {
+                        VoxelSet* vset = (VoxelSet*)pset;
+                        RefineAxesAlignedClippingPlanes(*vset, bestPlane, params.m_planeDownsampling, planesRef);
+                    }
+                    else {
+                        TetrahedronSet* tset = (TetrahedronSet*)pset;
+                        RefineAxesAlignedClippingPlanes(*tset, bestPlane, params.m_planeDownsampling, planesRef);
+                    }
+
+                    if (params.m_logger) {
+                        msg.str("");
+                        msg << "\t\t [Refining] Number of clipping planes " << planesRef.Size() << std::endl;
+                        params.m_logger->Log(msg.str().c_str());
+                    }
+                    ComputeBestClippingPlane(pset,
+                        volume,
+                        planesRef,
+                        preferredCuttingDirection,
+                        w,
+                        concavity * params.m_alpha,
+                        concavity * params.m_beta,
+                        1, // convexhullDownsampling = 1
+                        progress1,
+                        progress2,
+                        bestPlane,
+                        minConcavity,
+                        params);
+                }
+                if (GetCancel()) {
+                    delete pset; // clean up
+                    break;
+                }
+                else {
+                    if (maxConcavity < minConcavity) {
+                        maxConcavity = minConcavity;
+                    }
+                    PrimitiveSet* bestLeft = pset->Create();
+                    PrimitiveSet* bestRight = pset->Create();
+                    temp.PushBack(bestLeft);
+                    temp.PushBack(bestRight);
+                    pset->Clip(bestPlane, bestRight, bestLeft);
+                    if (params.m_pca) {
+                        bestRight->RevertAlignToPrincipalAxes();
+                        bestLeft->RevertAlignToPrincipalAxes();
+                    }
+                    delete pset;
+                }
+            }
+            else {
+                if (params.m_pca) {
+                    pset->RevertAlignToPrincipalAxes();
+                }
+                parts.PushBack(pset);
+            }
+        }
+
+        Update(95.0 * (1.0 - maxConcavity) / (1.0 - params.m_concavity), 100.0, params);
+        if (GetCancel()) {
+            const size_t nTempParts = temp.Size();
+            for (size_t p = 0; p < nTempParts; ++p) {
+                delete temp[p];
+            }
+            temp.Resize(0);
+        }
+        else {
+            inputParts = temp;
+            temp.Resize(0);
+        }
+    }
+    const size_t nInputParts = inputParts.Size();
+    for (size_t p = 0; p < nInputParts; ++p) {
+        parts.PushBack(inputParts[p]);
+    }
+
+    if (GetCancel()) {
+        const size_t nParts = parts.Size();
+        for (size_t p = 0; p < nParts; ++p) {
+            delete parts[p];
+        }
+        return;
+    }
+
+    m_overallProgress = 90.0;
+    Update(m_stageProgress, 100.0, params);
+
+    msg.str("");
+    msg << "Generate convex-hulls";
+    m_operation = msg.str();
+    size_t nConvexHulls = parts.Size();
+    if (params.m_logger) {
+        msg.str("");
+        msg << "+ Generate " << nConvexHulls << " convex-hulls " << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+    Update(m_stageProgress, 0.0, params);
+    m_convexHulls.Resize(0);
+    for (size_t p = 0; p < nConvexHulls && !m_cancel; ++p) {
+        Update(m_stageProgress, p * 100.0 / nConvexHulls, params);
+        m_convexHulls.PushBack(new Mesh);
+        parts[p]->ComputeConvexHull(*m_convexHulls[p]);
+        size_t nv = m_convexHulls[p]->GetNPoints();
+        double x, y, z;
+        for (size_t i = 0; i < nv; ++i) {
+            Vec3<double>& pt = m_convexHulls[p]->GetPoint(i);
+            x = pt[0];
+            y = pt[1];
+            z = pt[2];
+            pt[0] = m_rot[0][0] * x + m_rot[0][1] * y + m_rot[0][2] * z + m_barycenter[0];
+            pt[1] = m_rot[1][0] * x + m_rot[1][1] * y + m_rot[1][2] * z + m_barycenter[1];
+            pt[2] = m_rot[2][0] * x + m_rot[2][1] * y + m_rot[2][2] * z + m_barycenter[2];
+        }
+    }
+
+    const size_t nParts = parts.Size();
+    for (size_t p = 0; p < nParts; ++p) {
+        delete parts[p];
+        parts[p] = 0;
+    }
+    parts.Resize(0);
+
+    if (GetCancel()) {
+        const size_t nConvexHulls = m_convexHulls.Size();
+        for (size_t p = 0; p < nConvexHulls; ++p) {
+            delete m_convexHulls[p];
+        }
+        m_convexHulls.Clear();
+        return;
+    }
+
+    m_overallProgress = 95.0;
+    Update(100.0, 100.0, params);
+    m_timer.Toc();
+    if (params.m_logger) {
+        msg.str("");
+        msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+}
+void AddPoints(const Mesh* const mesh, SArray<Vec3<double> >& pts)
+{
+    const int32_t n = (int32_t)mesh->GetNPoints();
+    for (int32_t i = 0; i < n; ++i) {
+        pts.PushBack(mesh->GetPoint(i));
+    }
+}
+void ComputeConvexHull(const Mesh* const ch1, const Mesh* const ch2, SArray<Vec3<double> >& pts, Mesh* const combinedCH)
+{
+    pts.Resize(0);
+    AddPoints(ch1, pts);
+    AddPoints(ch2, pts);
+
+    btConvexHullComputer ch;
+    ch.compute((double*)pts.Data(), 3 * sizeof(double), (int32_t)pts.Size(), -1.0, -1.0);
+    combinedCH->ResizePoints(0);
+    combinedCH->ResizeTriangles(0);
+    for (int32_t v = 0; v < ch.vertices.size(); v++) {
+        combinedCH->AddPoint(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+    }
+    const int32_t nt = ch.faces.size();
+    for (int32_t t = 0; t < nt; ++t) {
+        const btConvexHullComputer::Edge* sourceEdge = &(ch.edges[ch.faces[t]]);
+        int32_t a = sourceEdge->getSourceVertex();
+        int32_t b = sourceEdge->getTargetVertex();
+        const btConvexHullComputer::Edge* edge = sourceEdge->getNextEdgeOfFace();
+        int32_t c = edge->getTargetVertex();
+        while (c != a) {
+            combinedCH->AddTriangle(Vec3<int32_t>(a, b, c));
+            edge = edge->getNextEdgeOfFace();
+            b = c;
+            c = edge->getTargetVertex();
+        }
+    }
+}
+void VHACD::MergeConvexHulls(const Parameters& params)
+{
+    if (GetCancel()) {
+        return;
+    }
+    m_timer.Tic();
+
+    m_stage = "Merge Convex Hulls";
+
+    std::ostringstream msg;
+    if (params.m_logger) {
+        msg << "+ " << m_stage << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+	// Get the current number of convex hulls
+    size_t nConvexHulls = m_convexHulls.Size();
+	// Iteration counter
+    int32_t iteration = 0;
+	// While we have more than at least one convex hull and the user has not asked us to cancel the operation
+    if (nConvexHulls > 1 && !m_cancel) 
+	{
+		// Get the gamma error threshold for when to exit
+        SArray<Vec3<double> > pts;
+        Mesh combinedCH;
+
+        // Populate the cost matrix
+        size_t idx = 0;
+        SArray<float> costMatrix;
+        costMatrix.Resize(((nConvexHulls * nConvexHulls) - nConvexHulls) >> 1);
+        for (size_t p1 = 1; p1 < nConvexHulls; ++p1) 
+		{
+            const float volume1 = m_convexHulls[p1]->ComputeVolume();
+            for (size_t p2 = 0; p2 < p1; ++p2) 
+			{
+                ComputeConvexHull(m_convexHulls[p1], m_convexHulls[p2], pts, &combinedCH);
+                costMatrix[idx++] = ComputeConcavity(volume1 + m_convexHulls[p2]->ComputeVolume(), combinedCH.ComputeVolume(), m_volumeCH0);
+            }
+        }
+
+        // Until we cant merge below the maximum cost
+        size_t costSize = m_convexHulls.Size();
+        while (!m_cancel) 
+		{
+            msg.str("");
+            msg << "Iteration " << iteration++;
+            m_operation = msg.str();
+
+            // Search for lowest cost
+            float bestCost = (std::numeric_limits<float>::max)();
+            const size_t addr = FindMinimumElement(costMatrix.Data(), &bestCost, 0, costMatrix.Size());
+			if ( (costSize-1) < params.m_maxConvexHulls)
+			{
+				break;
+			}
+            const size_t addrI = (static_cast<int32_t>(sqrt(1 + (8 * addr))) - 1) >> 1;
+            const size_t p1 = addrI + 1;
+            const size_t p2 = addr - ((addrI * (addrI + 1)) >> 1);
+            assert(p1 >= 0);
+            assert(p2 >= 0);
+            assert(p1 < costSize);
+            assert(p2 < costSize);
+
+            if (params.m_logger) 
+			{
+                msg.str("");
+                msg << "\t\t Merging (" << p1 << ", " << p2 << ") " << bestCost << std::endl
+                    << std::endl;
+                params.m_logger->Log(msg.str().c_str());
+            }
+
+            // Make the lowest cost row and column into a new hull
+            Mesh* cch = new Mesh;
+            ComputeConvexHull(m_convexHulls[p1], m_convexHulls[p2], pts, cch);
+            delete m_convexHulls[p2];
+            m_convexHulls[p2] = cch;
+
+            delete m_convexHulls[p1];
+            std::swap(m_convexHulls[p1], m_convexHulls[m_convexHulls.Size() - 1]);
+            m_convexHulls.PopBack();
+
+            costSize = costSize - 1;
+
+            // Calculate costs versus the new hull
+            size_t rowIdx = ((p2 - 1) * p2) >> 1;
+            const float volume1 = m_convexHulls[p2]->ComputeVolume();
+            for (size_t i = 0; (i < p2) && (!m_cancel); ++i) 
+			{
+                ComputeConvexHull(m_convexHulls[p2], m_convexHulls[i], pts, &combinedCH);
+                costMatrix[rowIdx++] = ComputeConcavity(volume1 + m_convexHulls[i]->ComputeVolume(), combinedCH.ComputeVolume(), m_volumeCH0);
+            }
+
+            rowIdx += p2;
+            for (size_t i = p2 + 1; (i < costSize) && (!m_cancel); ++i) 
+			{
+                ComputeConvexHull(m_convexHulls[p2], m_convexHulls[i], pts, &combinedCH);
+                costMatrix[rowIdx] = ComputeConcavity(volume1 + m_convexHulls[i]->ComputeVolume(), combinedCH.ComputeVolume(), m_volumeCH0);
+                rowIdx += i;
+                assert(rowIdx >= 0);
+            }
+
+            // Move the top column in to replace its space
+            const size_t erase_idx = ((costSize - 1) * costSize) >> 1;
+            if (p1 < costSize) {
+                rowIdx = (addrI * p1) >> 1;
+                size_t top_row = erase_idx;
+                for (size_t i = 0; i < p1; ++i) {
+                    if (i != p2) {
+                        costMatrix[rowIdx] = costMatrix[top_row];
+                    }
+                    ++rowIdx;
+                    ++top_row;
+                }
+
+                ++top_row;
+                rowIdx += p1;
+                for (size_t i = p1 + 1; i < (costSize + 1); ++i) {
+                    costMatrix[rowIdx] = costMatrix[top_row++];
+                    rowIdx += i;
+                    assert(rowIdx >= 0);
+                }
+            }
+            costMatrix.Resize(erase_idx);
+        }
+    }
+    m_overallProgress = 99.0;
+    Update(100.0, 100.0, params);
+    m_timer.Toc();
+    if (params.m_logger) {
+        msg.str("");
+        msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+}
+void VHACD::SimplifyConvexHull(Mesh* const ch, const size_t nvertices, const double minVolume)
+{
+    if (nvertices <= 4) {
+        return;
+    }
+    ICHull icHull;
+    if (mRaycastMesh)
+    {
+        // We project these points onto the original source mesh to increase precision
+        // The voxelization process drops floating point precision so returned data points are not exactly lying on the 
+        // surface of the original source mesh.
+        // The first step is we need to compute the bounding box of the mesh we are trying to build a convex hull for.
+        // From this bounding box, we compute the length of the diagonal to get a relative size and center for point projection
+        uint32_t nPoints = ch->GetNPoints();
+        Vec3<double> *inputPoints = ch->GetPointsBuffer();
+        Vec3<double> bmin(inputPoints[0]);
+        Vec3<double> bmax(inputPoints[1]);
+        for (uint32_t i = 1; i < nPoints; i++)
+        {
+            const Vec3<double> &p = inputPoints[i];
+            p.UpdateMinMax(bmin, bmax);
+        }
+        Vec3<double> center;
+        double diagonalLength = center.GetCenter(bmin, bmax);   // Get the center of the bounding box
+        // This is the error threshold for determining if we should use the raycast result data point vs. the voxelized result.
+        double pointDistanceThreshold = diagonalLength * 0.05;
+        // If a new point is within 1/100th the diagonal length of the bounding volume we do not add it.  To do so would create a
+        // thin sliver in the resulting convex hull
+        double snapDistanceThreshold = diagonalLength * 0.01;
+        double snapDistanceThresholdSquared = snapDistanceThreshold*snapDistanceThreshold;
+
+        // Allocate buffer for projected vertices
+        Vec3<double> *outputPoints = new Vec3<double>[nPoints];
+        uint32_t outCount = 0;
+        for (uint32_t i = 0; i < nPoints; i++)
+        {
+            Vec3<double> &inputPoint = inputPoints[i];
+            Vec3<double> &outputPoint = outputPoints[outCount];
+            // Compute the direction vector from the center of this mesh to the vertex
+            Vec3<double> dir = inputPoint - center;
+            // Normalize the direction vector.
+            dir.Normalize();
+            // Multiply times the diagonal length of the mesh
+            dir *= diagonalLength;
+            // Add the center back in again to get the destination point
+            dir += center;
+            // By default the output point is equal to the input point
+            outputPoint = inputPoint;
+            double pointDistance;
+            if (mRaycastMesh->raycast(center.GetData(), dir.GetData(), inputPoint.GetData(), outputPoint.GetData(),&pointDistance) )
+            {
+                // If the nearest intersection point is too far away, we keep the original source data point.
+                // Not all points lie directly on the original mesh surface
+                if (pointDistance > pointDistanceThreshold)
+                {
+                    outputPoint = inputPoint;
+                }
+            }
+            // Ok, before we add this point, we do not want to create points which are extremely close to each other.
+            // This will result in tiny sliver triangles which are really bad for collision detection.
+            bool foundNearbyPoint = false;
+            for (uint32_t j = 0; j < outCount; j++)
+            {
+                // If this new point is extremely close to an existing point, we do not add it!
+                double squaredDistance = outputPoints[j].GetDistanceSquared(outputPoint);
+                if (squaredDistance < snapDistanceThresholdSquared )
+                {
+                    foundNearbyPoint = true;
+                    break;
+                }
+            }
+            if (!foundNearbyPoint)
+            {
+                outCount++;
+            }
+        }
+        icHull.AddPoints(outputPoints, outCount);
+        delete[]outputPoints;
+    }
+    else
+    {
+        icHull.AddPoints(ch->GetPointsBuffer(), ch->GetNPoints());
+    }
+    icHull.Process((uint32_t)nvertices, minVolume);
+    TMMesh& mesh = icHull.GetMesh();
+    const size_t nT = mesh.GetNTriangles();
+    const size_t nV = mesh.GetNVertices();
+    ch->ResizePoints(nV);
+    ch->ResizeTriangles(nT);
+    mesh.GetIFS(ch->GetPointsBuffer(), ch->GetTrianglesBuffer());
+}
+void VHACD::SimplifyConvexHulls(const Parameters& params)
+{
+    if (m_cancel || params.m_maxNumVerticesPerCH < 4) {
+        return;
+    }
+    m_timer.Tic();
+
+    m_stage = "Simplify convex-hulls";
+    m_operation = "Simplify convex-hulls";
+
+    std::ostringstream msg;
+    const size_t nConvexHulls = m_convexHulls.Size();
+    if (params.m_logger) {
+        msg << "+ Simplify " << nConvexHulls << " convex-hulls " << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+
+    Update(0.0, 0.0, params);
+    for (size_t i = 0; i < nConvexHulls && !m_cancel; ++i) {
+        if (params.m_logger) {
+            msg.str("");
+            msg << "\t\t Simplify CH[" << std::setfill('0') << std::setw(5) << i << "] " << m_convexHulls[i]->GetNPoints() << " V, " << m_convexHulls[i]->GetNTriangles() << " T" << std::endl;
+            params.m_logger->Log(msg.str().c_str());
+        }
+        SimplifyConvexHull(m_convexHulls[i], params.m_maxNumVerticesPerCH, m_volumeCH0 * params.m_minVolumePerCH);
+    }
+
+    m_overallProgress = 100.0;
+    Update(100.0, 100.0, params);
+    m_timer.Toc();
+    if (params.m_logger) {
+        msg.str("");
+        msg << "\t time " << m_timer.GetElapsedTime() / 1000.0 << "s" << std::endl;
+        params.m_logger->Log(msg.str().c_str());
+    }
+}
+
+bool VHACD::ComputeCenterOfMass(double centerOfMass[3]) const
+{
+	bool ret = false;
+
+	centerOfMass[0] = 0;
+	centerOfMass[1] = 0;
+	centerOfMass[2] = 0;
+	// Get number of convex hulls in the result
+	uint32_t hullCount = GetNConvexHulls();
+	if (hullCount) // if we have results
+	{
+		ret = true;
+		double totalVolume = 0;
+		// Initialize the center of mass to zero
+		centerOfMass[0] = 0;
+		centerOfMass[1] = 0;
+		centerOfMass[2] = 0;
+		// Compute the total volume of all convex hulls
+		for (uint32_t i = 0; i < hullCount; i++)
+		{
+			ConvexHull ch;
+			GetConvexHull(i, ch);
+			totalVolume += ch.m_volume;
+		}
+		// compute the reciprocal of the total volume
+		double recipVolume = 1.0 / totalVolume;
+		// Add in the weighted by volume average of the center point of each convex hull
+		for (uint32_t i = 0; i < hullCount; i++)
+		{
+			ConvexHull ch;
+			GetConvexHull(i, ch);
+			double ratio = ch.m_volume*recipVolume;
+			centerOfMass[0] += ch.m_center[0] * ratio;
+			centerOfMass[1] += ch.m_center[1] * ratio;
+			centerOfMass[2] += ch.m_center[2] * ratio;
+		}
+	}
+	return ret;
+}
+
+} // end of VHACD namespace
diff --git a/thirdparty/vhacd/src/btAlignedAllocator.cpp b/thirdparty/vhacd/src/btAlignedAllocator.cpp
new file mode 100644
index 0000000000..e137032463
--- /dev/null
+++ b/thirdparty/vhacd/src/btAlignedAllocator.cpp
@@ -0,0 +1,171 @@
+/*
+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 "btAlignedAllocator.h"
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+#ifdef _MSC_VER
+#pragma warning(disable : 4311 4302)
+#endif
+
+int32_t gNumAlignedAllocs = 0;
+int32_t gNumAlignedFree = 0;
+int32_t gTotalBytesAlignedAllocs = 0; //detect memory leaks
+
+static void *btAllocDefault(size_t size) {
+	return malloc(size);
+}
+
+static void btFreeDefault(void *ptr) {
+	free(ptr);
+}
+
+static btAllocFunc *sAllocFunc = btAllocDefault;
+static btFreeFunc *sFreeFunc = btFreeDefault;
+
+#if defined(BT_HAS_ALIGNED_ALLOCATOR)
+#include <malloc.h>
+static void *btAlignedAllocDefault(size_t size, int32_t alignment) {
+	return _aligned_malloc(size, (size_t)alignment);
+}
+
+static void btAlignedFreeDefault(void *ptr) {
+	_aligned_free(ptr);
+}
+#elif defined(__CELLOS_LV2__)
+#include <stdlib.h>
+
+static inline void *btAlignedAllocDefault(size_t size, int32_t alignment) {
+	return memalign(alignment, size);
+}
+
+static inline void btAlignedFreeDefault(void *ptr) {
+	free(ptr);
+}
+#else
+static inline void *btAlignedAllocDefault(size_t size, int32_t alignment) {
+	void *ret;
+	char *real;
+	unsigned long offset;
+
+	real = (char *)sAllocFunc(size + sizeof(void *) + (alignment - 1));
+	if (real) {
+		offset = (alignment - (unsigned long)(real + sizeof(void *))) & (alignment - 1);
+		ret = (void *)((real + sizeof(void *)) + offset);
+		*((void **)(ret)-1) = (void *)(real);
+	} else {
+		ret = (void *)(real);
+	}
+	return (ret);
+}
+
+static inline void btAlignedFreeDefault(void *ptr) {
+	void *real;
+
+	if (ptr) {
+		real = *((void **)(ptr)-1);
+		sFreeFunc(real);
+	}
+}
+#endif
+
+static btAlignedAllocFunc *sAlignedAllocFunc = btAlignedAllocDefault;
+static btAlignedFreeFunc *sAlignedFreeFunc = btAlignedFreeDefault;
+
+void btAlignedAllocSetCustomAligned(btAlignedAllocFunc *allocFunc, btAlignedFreeFunc *freeFunc) {
+	sAlignedAllocFunc = allocFunc ? allocFunc : btAlignedAllocDefault;
+	sAlignedFreeFunc = freeFunc ? freeFunc : btAlignedFreeDefault;
+}
+
+void btAlignedAllocSetCustom(btAllocFunc *allocFunc, btFreeFunc *freeFunc) {
+	sAllocFunc = allocFunc ? allocFunc : btAllocDefault;
+	sFreeFunc = freeFunc ? freeFunc : btFreeDefault;
+}
+
+#ifdef BT_DEBUG_MEMORY_ALLOCATIONS
+//this generic allocator provides the total allocated number of bytes
+#include <stdio.h>
+
+void *btAlignedAllocInternal(size_t size, int32_t alignment, int32_t line, char *filename) {
+	void *ret;
+	char *real;
+	unsigned long offset;
+
+	gTotalBytesAlignedAllocs += size;
+	gNumAlignedAllocs++;
+
+	real = (char *)sAllocFunc(size + 2 * sizeof(void *) + (alignment - 1));
+	if (real) {
+		offset = (alignment - (unsigned long)(real + 2 * sizeof(void *))) & (alignment - 1);
+		ret = (void *)((real + 2 * sizeof(void *)) + offset);
+		*((void **)(ret)-1) = (void *)(real);
+		*((int32_t *)(ret)-2) = size;
+	} else {
+		ret = (void *)(real); //??
+	}
+
+	printf("allocation#%d at address %x, from %s,line %d, size %d\n", gNumAlignedAllocs, real, filename, line, size);
+
+	int32_t *ptr = (int32_t *)ret;
+	*ptr = 12;
+	return (ret);
+}
+
+void btAlignedFreeInternal(void *ptr, int32_t line, char *filename) {
+
+	void *real;
+	gNumAlignedFree++;
+
+	if (ptr) {
+		real = *((void **)(ptr)-1);
+		int32_t size = *((int32_t *)(ptr)-2);
+		gTotalBytesAlignedAllocs -= size;
+
+		printf("free #%d at address %x, from %s,line %d, size %d\n", gNumAlignedFree, real, filename, line, size);
+
+		sFreeFunc(real);
+	} else {
+		printf("NULL ptr\n");
+	}
+}
+
+#else //BT_DEBUG_MEMORY_ALLOCATIONS
+
+void *btAlignedAllocInternal(size_t size, int32_t alignment) {
+	gNumAlignedAllocs++;
+	void *ptr;
+	ptr = sAlignedAllocFunc(size, alignment);
+	//	printf("btAlignedAllocInternal %d, %x\n",size,ptr);
+	return ptr;
+}
+
+void btAlignedFreeInternal(void *ptr) {
+	if (!ptr) {
+		return;
+	}
+
+	gNumAlignedFree++;
+	//	printf("btAlignedFreeInternal %x\n",ptr);
+	sAlignedFreeFunc(ptr);
+}
+
+//GODOT ADDITION
+};
+//
+
+#endif //BT_DEBUG_MEMORY_ALLOCATIONS
diff --git a/thirdparty/vhacd/src/btConvexHullComputer.cpp b/thirdparty/vhacd/src/btConvexHullComputer.cpp
new file mode 100644
index 0000000000..5ff3ab917e
--- /dev/null
+++ b/thirdparty/vhacd/src/btConvexHullComputer.cpp
@@ -0,0 +1,2336 @@
+/*
+Copyright (c) 2011 Ole Kniemeyer, MAXON, www.maxon.net
+
+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 <string.h>
+
+#include "btAlignedObjectArray.h"
+#include "btConvexHullComputer.h"
+#include "btMinMax.h"
+#include "btVector3.h"
+
+#ifdef __GNUC__
+#include <stdint.h>
+
+//GODOT ADDITION
+namespace VHACD {
+//
+
+#elif defined(_MSC_VER)
+typedef __int32 int32_t;
+typedef __int64 int64_t;
+typedef unsigned __int32 uint32_t;
+typedef unsigned __int64 uint64_t;
+#else
+typedef int32_t int32_t;
+typedef long long int32_t int64_t;
+typedef uint32_t uint32_t;
+typedef unsigned long long int32_t uint64_t;
+#endif
+
+#ifdef _MSC_VER
+#pragma warning(disable : 4458)
+#endif
+
+//The definition of USE_X86_64_ASM is moved into the build system. You can enable it manually by commenting out the following lines
+//#if (defined(__GNUC__) && defined(__x86_64__) && !defined(__ICL))  // || (defined(__ICL) && defined(_M_X64))   bug in Intel compiler, disable inline assembly
+//	#define USE_X86_64_ASM
+//#endif
+
+//#define DEBUG_CONVEX_HULL
+//#define SHOW_ITERATIONS
+
+#if defined(DEBUG_CONVEX_HULL) || defined(SHOW_ITERATIONS)
+#include <stdio.h>
+#endif
+
+// Convex hull implementation based on Preparata and Hong
+// Ole Kniemeyer, MAXON Computer GmbH
+class btConvexHullInternal {
+public:
+	class Point64 {
+	public:
+		int64_t x;
+		int64_t y;
+		int64_t z;
+
+		Point64(int64_t x, int64_t y, int64_t z) :
+				x(x),
+				y(y),
+				z(z) {
+		}
+
+		bool isZero() {
+			return (x == 0) && (y == 0) && (z == 0);
+		}
+
+		int64_t dot(const Point64 &b) const {
+			return x * b.x + y * b.y + z * b.z;
+		}
+	};
+
+	class Point32 {
+	public:
+		int32_t x;
+		int32_t y;
+		int32_t z;
+		int32_t index;
+
+		Point32() {
+		}
+
+		Point32(int32_t x, int32_t y, int32_t z) :
+				x(x),
+				y(y),
+				z(z),
+				index(-1) {
+		}
+
+		bool operator==(const Point32 &b) const {
+			return (x == b.x) && (y == b.y) && (z == b.z);
+		}
+
+		bool operator!=(const Point32 &b) const {
+			return (x != b.x) || (y != b.y) || (z != b.z);
+		}
+
+		bool isZero() {
+			return (x == 0) && (y == 0) && (z == 0);
+		}
+
+		Point64 cross(const Point32 &b) const {
+			return Point64(y * b.z - z * b.y, z * b.x - x * b.z, x * b.y - y * b.x);
+		}
+
+		Point64 cross(const Point64 &b) const {
+			return Point64(y * b.z - z * b.y, z * b.x - x * b.z, x * b.y - y * b.x);
+		}
+
+		int64_t dot(const Point32 &b) const {
+			return x * b.x + y * b.y + z * b.z;
+		}
+
+		int64_t dot(const Point64 &b) const {
+			return x * b.x + y * b.y + z * b.z;
+		}
+
+		Point32 operator+(const Point32 &b) const {
+			return Point32(x + b.x, y + b.y, z + b.z);
+		}
+
+		Point32 operator-(const Point32 &b) const {
+			return Point32(x - b.x, y - b.y, z - b.z);
+		}
+	};
+
+	class Int128 {
+	public:
+		uint64_t low;
+		uint64_t high;
+
+		Int128() {
+		}
+
+		Int128(uint64_t low, uint64_t high) :
+				low(low),
+				high(high) {
+		}
+
+		Int128(uint64_t low) :
+				low(low),
+				high(0) {
+		}
+
+		Int128(int64_t value) :
+				low(value),
+				high((value >= 0) ? 0 : (uint64_t)-1LL) {
+		}
+
+		static Int128 mul(int64_t a, int64_t b);
+
+		static Int128 mul(uint64_t a, uint64_t b);
+
+		Int128 operator-() const {
+			return Int128((uint64_t) - (int64_t)low, ~high + (low == 0));
+		}
+
+		Int128 operator+(const Int128 &b) const {
+#ifdef USE_X86_64_ASM
+			Int128 result;
+			__asm__("addq %[bl], %[rl]\n\t"
+					"adcq %[bh], %[rh]\n\t"
+					: [rl] "=r"(result.low), [rh] "=r"(result.high)
+					: "0"(low), "1"(high), [bl] "g"(b.low), [bh] "g"(b.high)
+					: "cc");
+			return result;
+#else
+			uint64_t lo = low + b.low;
+			return Int128(lo, high + b.high + (lo < low));
+#endif
+		}
+
+		Int128 operator-(const Int128 &b) const {
+#ifdef USE_X86_64_ASM
+			Int128 result;
+			__asm__("subq %[bl], %[rl]\n\t"
+					"sbbq %[bh], %[rh]\n\t"
+					: [rl] "=r"(result.low), [rh] "=r"(result.high)
+					: "0"(low), "1"(high), [bl] "g"(b.low), [bh] "g"(b.high)
+					: "cc");
+			return result;
+#else
+			return *this + -b;
+#endif
+		}
+
+		Int128 &operator+=(const Int128 &b) {
+#ifdef USE_X86_64_ASM
+			__asm__("addq %[bl], %[rl]\n\t"
+					"adcq %[bh], %[rh]\n\t"
+					: [rl] "=r"(low), [rh] "=r"(high)
+					: "0"(low), "1"(high), [bl] "g"(b.low), [bh] "g"(b.high)
+					: "cc");
+#else
+			uint64_t lo = low + b.low;
+			if (lo < low) {
+				++high;
+			}
+			low = lo;
+			high += b.high;
+#endif
+			return *this;
+		}
+
+		Int128 &operator++() {
+			if (++low == 0) {
+				++high;
+			}
+			return *this;
+		}
+
+		Int128 operator*(int64_t b) const;
+
+		btScalar toScalar() const {
+			return ((int64_t)high >= 0) ? btScalar(high) * (btScalar(0x100000000LL) * btScalar(0x100000000LL)) + btScalar(low) : -(-*this).toScalar();
+		}
+
+		int32_t getSign() const {
+			return ((int64_t)high < 0) ? -1 : (high || low) ? 1 : 0;
+		}
+
+		bool operator<(const Int128 &b) const {
+			return (high < b.high) || ((high == b.high) && (low < b.low));
+		}
+
+		int32_t ucmp(const Int128 &b) const {
+			if (high < b.high) {
+				return -1;
+			}
+			if (high > b.high) {
+				return 1;
+			}
+			if (low < b.low) {
+				return -1;
+			}
+			if (low > b.low) {
+				return 1;
+			}
+			return 0;
+		}
+	};
+
+	class Rational64 {
+	private:
+		uint64_t m_numerator;
+		uint64_t m_denominator;
+		int32_t sign;
+
+	public:
+		Rational64(int64_t numerator, int64_t denominator) {
+			if (numerator > 0) {
+				sign = 1;
+				m_numerator = (uint64_t)numerator;
+			} else if (numerator < 0) {
+				sign = -1;
+				m_numerator = (uint64_t)-numerator;
+			} else {
+				sign = 0;
+				m_numerator = 0;
+			}
+			if (denominator > 0) {
+				m_denominator = (uint64_t)denominator;
+			} else if (denominator < 0) {
+				sign = -sign;
+				m_denominator = (uint64_t)-denominator;
+			} else {
+				m_denominator = 0;
+			}
+		}
+
+		bool isNegativeInfinity() const {
+			return (sign < 0) && (m_denominator == 0);
+		}
+
+		bool isNaN() const {
+			return (sign == 0) && (m_denominator == 0);
+		}
+
+		int32_t compare(const Rational64 &b) const;
+
+		btScalar toScalar() const {
+			return sign * ((m_denominator == 0) ? SIMD_INFINITY : (btScalar)m_numerator / m_denominator);
+		}
+	};
+
+	class Rational128 {
+	private:
+		Int128 numerator;
+		Int128 denominator;
+		int32_t sign;
+		bool isInt64;
+
+	public:
+		Rational128(int64_t value) {
+			if (value > 0) {
+				sign = 1;
+				this->numerator = value;
+			} else if (value < 0) {
+				sign = -1;
+				this->numerator = -value;
+			} else {
+				sign = 0;
+				this->numerator = (uint64_t)0;
+			}
+			this->denominator = (uint64_t)1;
+			isInt64 = true;
+		}
+
+		Rational128(const Int128 &numerator, const Int128 &denominator) {
+			sign = numerator.getSign();
+			if (sign >= 0) {
+				this->numerator = numerator;
+			} else {
+				this->numerator = -numerator;
+			}
+			int32_t dsign = denominator.getSign();
+			if (dsign >= 0) {
+				this->denominator = denominator;
+			} else {
+				sign = -sign;
+				this->denominator = -denominator;
+			}
+			isInt64 = false;
+		}
+
+		int32_t compare(const Rational128 &b) const;
+
+		int32_t compare(int64_t b) const;
+
+		btScalar toScalar() const {
+			return sign * ((denominator.getSign() == 0) ? SIMD_INFINITY : numerator.toScalar() / denominator.toScalar());
+		}
+	};
+
+	class PointR128 {
+	public:
+		Int128 x;
+		Int128 y;
+		Int128 z;
+		Int128 denominator;
+
+		PointR128() {
+		}
+
+		PointR128(Int128 x, Int128 y, Int128 z, Int128 denominator) :
+				x(x),
+				y(y),
+				z(z),
+				denominator(denominator) {
+		}
+
+		btScalar xvalue() const {
+			return x.toScalar() / denominator.toScalar();
+		}
+
+		btScalar yvalue() const {
+			return y.toScalar() / denominator.toScalar();
+		}
+
+		btScalar zvalue() const {
+			return z.toScalar() / denominator.toScalar();
+		}
+	};
+
+	class Edge;
+	class Face;
+
+	class Vertex {
+	public:
+		Vertex *next;
+		Vertex *prev;
+		Edge *edges;
+		Face *firstNearbyFace;
+		Face *lastNearbyFace;
+		PointR128 point128;
+		Point32 point;
+		int32_t copy;
+
+		Vertex() :
+				next(NULL),
+				prev(NULL),
+				edges(NULL),
+				firstNearbyFace(NULL),
+				lastNearbyFace(NULL),
+				copy(-1) {
+		}
+
+#ifdef DEBUG_CONVEX_HULL
+		void print() {
+			printf("V%d (%d, %d, %d)", point.index, point.x, point.y, point.z);
+		}
+
+		void printGraph();
+#endif
+
+		Point32 operator-(const Vertex &b) const {
+			return point - b.point;
+		}
+
+		Rational128 dot(const Point64 &b) const {
+			return (point.index >= 0) ? Rational128(point.dot(b)) : Rational128(point128.x * b.x + point128.y * b.y + point128.z * b.z, point128.denominator);
+		}
+
+		btScalar xvalue() const {
+			return (point.index >= 0) ? btScalar(point.x) : point128.xvalue();
+		}
+
+		btScalar yvalue() const {
+			return (point.index >= 0) ? btScalar(point.y) : point128.yvalue();
+		}
+
+		btScalar zvalue() const {
+			return (point.index >= 0) ? btScalar(point.z) : point128.zvalue();
+		}
+
+		void receiveNearbyFaces(Vertex *src) {
+			if (lastNearbyFace) {
+				lastNearbyFace->nextWithSameNearbyVertex = src->firstNearbyFace;
+			} else {
+				firstNearbyFace = src->firstNearbyFace;
+			}
+			if (src->lastNearbyFace) {
+				lastNearbyFace = src->lastNearbyFace;
+			}
+			for (Face *f = src->firstNearbyFace; f; f = f->nextWithSameNearbyVertex) {
+				btAssert(f->nearbyVertex == src);
+				f->nearbyVertex = this;
+			}
+			src->firstNearbyFace = NULL;
+			src->lastNearbyFace = NULL;
+		}
+	};
+
+	class Edge {
+	public:
+		Edge *next;
+		Edge *prev;
+		Edge *reverse;
+		Vertex *target;
+		Face *face;
+		int32_t copy;
+
+		~Edge() {
+			next = NULL;
+			prev = NULL;
+			reverse = NULL;
+			target = NULL;
+			face = NULL;
+		}
+
+		void link(Edge *n) {
+			btAssert(reverse->target == n->reverse->target);
+			next = n;
+			n->prev = this;
+		}
+
+#ifdef DEBUG_CONVEX_HULL
+		void print() {
+			printf("E%p : %d -> %d,  n=%p p=%p   (0 %d\t%d\t%d) -> (%d %d %d)", this, reverse->target->point.index, target->point.index, next, prev,
+					reverse->target->point.x, reverse->target->point.y, reverse->target->point.z, target->point.x, target->point.y, target->point.z);
+		}
+#endif
+	};
+
+	class Face {
+	public:
+		Face *next;
+		Vertex *nearbyVertex;
+		Face *nextWithSameNearbyVertex;
+		Point32 origin;
+		Point32 dir0;
+		Point32 dir1;
+
+		Face() :
+				next(NULL),
+				nearbyVertex(NULL),
+				nextWithSameNearbyVertex(NULL) {
+		}
+
+		void init(Vertex *a, Vertex *b, Vertex *c) {
+			nearbyVertex = a;
+			origin = a->point;
+			dir0 = *b - *a;
+			dir1 = *c - *a;
+			if (a->lastNearbyFace) {
+				a->lastNearbyFace->nextWithSameNearbyVertex = this;
+			} else {
+				a->firstNearbyFace = this;
+			}
+			a->lastNearbyFace = this;
+		}
+
+		Point64 getNormal() {
+			return dir0.cross(dir1);
+		}
+	};
+
+	template <typename UWord, typename UHWord>
+	class DMul {
+	private:
+		static uint32_t high(uint64_t value) {
+			return (uint32_t)(value >> 32);
+		}
+
+		static uint32_t low(uint64_t value) {
+			return (uint32_t)value;
+		}
+
+		static uint64_t mul(uint32_t a, uint32_t b) {
+			return (uint64_t)a * (uint64_t)b;
+		}
+
+		static void shlHalf(uint64_t &value) {
+			value <<= 32;
+		}
+
+		static uint64_t high(Int128 value) {
+			return value.high;
+		}
+
+		static uint64_t low(Int128 value) {
+			return value.low;
+		}
+
+		static Int128 mul(uint64_t a, uint64_t b) {
+			return Int128::mul(a, b);
+		}
+
+		static void shlHalf(Int128 &value) {
+			value.high = value.low;
+			value.low = 0;
+		}
+
+	public:
+		static void mul(UWord a, UWord b, UWord &resLow, UWord &resHigh) {
+			UWord p00 = mul(low(a), low(b));
+			UWord p01 = mul(low(a), high(b));
+			UWord p10 = mul(high(a), low(b));
+			UWord p11 = mul(high(a), high(b));
+			UWord p0110 = UWord(low(p01)) + UWord(low(p10));
+			p11 += high(p01);
+			p11 += high(p10);
+			p11 += high(p0110);
+			shlHalf(p0110);
+			p00 += p0110;
+			if (p00 < p0110) {
+				++p11;
+			}
+			resLow = p00;
+			resHigh = p11;
+		}
+	};
+
+private:
+	class IntermediateHull {
+	public:
+		Vertex *minXy;
+		Vertex *maxXy;
+		Vertex *minYx;
+		Vertex *maxYx;
+
+		IntermediateHull() :
+				minXy(NULL),
+				maxXy(NULL),
+				minYx(NULL),
+				maxYx(NULL) {
+		}
+
+		void print();
+	};
+
+	enum Orientation { NONE,
+		CLOCKWISE,
+		COUNTER_CLOCKWISE };
+
+	template <typename T>
+	class PoolArray {
+	private:
+		T *array;
+		int32_t size;
+
+	public:
+		PoolArray<T> *next;
+
+		PoolArray(int32_t size) :
+				size(size),
+				next(NULL) {
+			array = (T *)btAlignedAlloc(sizeof(T) * size, 16);
+		}
+
+		~PoolArray() {
+			btAlignedFree(array);
+		}
+
+		T *init() {
+			T *o = array;
+			for (int32_t i = 0; i < size; i++, o++) {
+				o->next = (i + 1 < size) ? o + 1 : NULL;
+			}
+			return array;
+		}
+	};
+
+	template <typename T>
+	class Pool {
+	private:
+		PoolArray<T> *arrays;
+		PoolArray<T> *nextArray;
+		T *freeObjects;
+		int32_t arraySize;
+
+	public:
+		Pool() :
+				arrays(NULL),
+				nextArray(NULL),
+				freeObjects(NULL),
+				arraySize(256) {
+		}
+
+		~Pool() {
+			while (arrays) {
+				PoolArray<T> *p = arrays;
+				arrays = p->next;
+				p->~PoolArray<T>();
+				btAlignedFree(p);
+			}
+		}
+
+		void reset() {
+			nextArray = arrays;
+			freeObjects = NULL;
+		}
+
+		void setArraySize(int32_t arraySize) {
+			this->arraySize = arraySize;
+		}
+
+		T *newObject() {
+			T *o = freeObjects;
+			if (!o) {
+				PoolArray<T> *p = nextArray;
+				if (p) {
+					nextArray = p->next;
+				} else {
+					p = new (btAlignedAlloc(sizeof(PoolArray<T>), 16)) PoolArray<T>(arraySize);
+					p->next = arrays;
+					arrays = p;
+				}
+				o = p->init();
+			}
+			freeObjects = o->next;
+			return new (o) T();
+		};
+
+		void freeObject(T *object) {
+			object->~T();
+			object->next = freeObjects;
+			freeObjects = object;
+		}
+	};
+
+	btVector3 scaling;
+	btVector3 center;
+	Pool<Vertex> vertexPool;
+	Pool<Edge> edgePool;
+	Pool<Face> facePool;
+	btAlignedObjectArray<Vertex *> originalVertices;
+	int32_t mergeStamp;
+	int32_t minAxis;
+	int32_t medAxis;
+	int32_t maxAxis;
+	int32_t usedEdgePairs;
+	int32_t maxUsedEdgePairs;
+
+	static Orientation getOrientation(const Edge *prev, const Edge *next, const Point32 &s, const Point32 &t);
+	Edge *findMaxAngle(bool ccw, const Vertex *start, const Point32 &s, const Point64 &rxs, const Point64 &sxrxs, Rational64 &minCot);
+	void findEdgeForCoplanarFaces(Vertex *c0, Vertex *c1, Edge *&e0, Edge *&e1, Vertex *stop0, Vertex *stop1);
+
+	Edge *newEdgePair(Vertex *from, Vertex *to);
+
+	void removeEdgePair(Edge *edge) {
+		Edge *n = edge->next;
+		Edge *r = edge->reverse;
+
+		btAssert(edge->target && r->target);
+
+		if (n != edge) {
+			n->prev = edge->prev;
+			edge->prev->next = n;
+			r->target->edges = n;
+		} else {
+			r->target->edges = NULL;
+		}
+
+		n = r->next;
+
+		if (n != r) {
+			n->prev = r->prev;
+			r->prev->next = n;
+			edge->target->edges = n;
+		} else {
+			edge->target->edges = NULL;
+		}
+
+		edgePool.freeObject(edge);
+		edgePool.freeObject(r);
+		usedEdgePairs--;
+	}
+
+	void computeInternal(int32_t start, int32_t end, IntermediateHull &result);
+
+	bool mergeProjection(IntermediateHull &h0, IntermediateHull &h1, Vertex *&c0, Vertex *&c1);
+
+	void merge(IntermediateHull &h0, IntermediateHull &h1);
+
+	btVector3 toBtVector(const Point32 &v);
+
+	btVector3 getBtNormal(Face *face);
+
+	bool shiftFace(Face *face, btScalar amount, btAlignedObjectArray<Vertex *> stack);
+
+public:
+	Vertex *vertexList;
+
+	void compute(const void *coords, bool doubleCoords, int32_t stride, int32_t count);
+
+	btVector3 getCoordinates(const Vertex *v);
+
+	btScalar shrink(btScalar amount, btScalar clampAmount);
+};
+
+btConvexHullInternal::Int128 btConvexHullInternal::Int128::operator*(int64_t b) const {
+	bool negative = (int64_t)high < 0;
+	Int128 a = negative ? -*this : *this;
+	if (b < 0) {
+		negative = !negative;
+		b = -b;
+	}
+	Int128 result = mul(a.low, (uint64_t)b);
+	result.high += a.high * (uint64_t)b;
+	return negative ? -result : result;
+}
+
+btConvexHullInternal::Int128 btConvexHullInternal::Int128::mul(int64_t a, int64_t b) {
+	Int128 result;
+
+#ifdef USE_X86_64_ASM
+	__asm__("imulq %[b]"
+			: "=a"(result.low), "=d"(result.high)
+			: "0"(a), [b] "r"(b)
+			: "cc");
+	return result;
+
+#else
+	bool negative = a < 0;
+	if (negative) {
+		a = -a;
+	}
+	if (b < 0) {
+		negative = !negative;
+		b = -b;
+	}
+	DMul<uint64_t, uint32_t>::mul((uint64_t)a, (uint64_t)b, result.low, result.high);
+	return negative ? -result : result;
+#endif
+}
+
+btConvexHullInternal::Int128 btConvexHullInternal::Int128::mul(uint64_t a, uint64_t b) {
+	Int128 result;
+
+#ifdef USE_X86_64_ASM
+	__asm__("mulq %[b]"
+			: "=a"(result.low), "=d"(result.high)
+			: "0"(a), [b] "r"(b)
+			: "cc");
+
+#else
+	DMul<uint64_t, uint32_t>::mul(a, b, result.low, result.high);
+#endif
+
+	return result;
+}
+
+int32_t btConvexHullInternal::Rational64::compare(const Rational64 &b) const {
+	if (sign != b.sign) {
+		return sign - b.sign;
+	} else if (sign == 0) {
+		return 0;
+	}
+
+	//	return (numerator * b.denominator > b.numerator * denominator) ? sign : (numerator * b.denominator < b.numerator * denominator) ? -sign : 0;
+
+#ifdef USE_X86_64_ASM
+
+	int32_t result;
+	int64_t tmp;
+	int64_t dummy;
+	__asm__("mulq %[bn]\n\t"
+			"movq %%rax, %[tmp]\n\t"
+			"movq %%rdx, %%rbx\n\t"
+			"movq %[tn], %%rax\n\t"
+			"mulq %[bd]\n\t"
+			"subq %[tmp], %%rax\n\t"
+			"sbbq %%rbx, %%rdx\n\t" // rdx:rax contains 128-bit-difference "numerator*b.denominator - b.numerator*denominator"
+			"setnsb %%bh\n\t" // bh=1 if difference is non-negative, bh=0 otherwise
+			"orq %%rdx, %%rax\n\t"
+			"setnzb %%bl\n\t" // bl=1 if difference if non-zero, bl=0 if it is zero
+			"decb %%bh\n\t" // now bx=0x0000 if difference is zero, 0xff01 if it is negative, 0x0001 if it is positive (i.e., same sign as difference)
+			"shll $16, %%ebx\n\t" // ebx has same sign as difference
+			: "=&b"(result), [tmp] "=&r"(tmp), "=a"(dummy)
+			: "a"(denominator), [bn] "g"(b.numerator), [tn] "g"(numerator), [bd] "g"(b.denominator)
+			: "%rdx", "cc");
+	return result ? result ^ sign // if sign is +1, only bit 0 of result is inverted, which does not change the sign of result (and cannot result in zero)
+					// if sign is -1, all bits of result are inverted, which changes the sign of result (and again cannot result in zero)
+					:
+					0;
+
+#else
+
+	return sign * Int128::mul(m_numerator, b.m_denominator).ucmp(Int128::mul(m_denominator, b.m_numerator));
+
+#endif
+}
+
+int32_t btConvexHullInternal::Rational128::compare(const Rational128 &b) const {
+	if (sign != b.sign) {
+		return sign - b.sign;
+	} else if (sign == 0) {
+		return 0;
+	}
+	if (isInt64) {
+		return -b.compare(sign * (int64_t)numerator.low);
+	}
+
+	Int128 nbdLow, nbdHigh, dbnLow, dbnHigh;
+	DMul<Int128, uint64_t>::mul(numerator, b.denominator, nbdLow, nbdHigh);
+	DMul<Int128, uint64_t>::mul(denominator, b.numerator, dbnLow, dbnHigh);
+
+	int32_t cmp = nbdHigh.ucmp(dbnHigh);
+	if (cmp) {
+		return cmp * sign;
+	}
+	return nbdLow.ucmp(dbnLow) * sign;
+}
+
+int32_t btConvexHullInternal::Rational128::compare(int64_t b) const {
+	if (isInt64) {
+		int64_t a = sign * (int64_t)numerator.low;
+		return (a > b) ? 1 : (a < b) ? -1 : 0;
+	}
+	if (b > 0) {
+		if (sign <= 0) {
+			return -1;
+		}
+	} else if (b < 0) {
+		if (sign >= 0) {
+			return 1;
+		}
+		b = -b;
+	} else {
+		return sign;
+	}
+
+	return numerator.ucmp(denominator * b) * sign;
+}
+
+btConvexHullInternal::Edge *btConvexHullInternal::newEdgePair(Vertex *from, Vertex *to) {
+	btAssert(from && to);
+	Edge *e = edgePool.newObject();
+	Edge *r = edgePool.newObject();
+	e->reverse = r;
+	r->reverse = e;
+	e->copy = mergeStamp;
+	r->copy = mergeStamp;
+	e->target = to;
+	r->target = from;
+	e->face = NULL;
+	r->face = NULL;
+	usedEdgePairs++;
+	if (usedEdgePairs > maxUsedEdgePairs) {
+		maxUsedEdgePairs = usedEdgePairs;
+	}
+	return e;
+}
+
+bool btConvexHullInternal::mergeProjection(IntermediateHull &h0, IntermediateHull &h1, Vertex *&c0, Vertex *&c1) {
+	Vertex *v0 = h0.maxYx;
+	Vertex *v1 = h1.minYx;
+	if ((v0->point.x == v1->point.x) && (v0->point.y == v1->point.y)) {
+		btAssert(v0->point.z < v1->point.z);
+		Vertex *v1p = v1->prev;
+		if (v1p == v1) {
+			c0 = v0;
+			if (v1->edges) {
+				btAssert(v1->edges->next == v1->edges);
+				v1 = v1->edges->target;
+				btAssert(v1->edges->next == v1->edges);
+			}
+			c1 = v1;
+			return false;
+		}
+		Vertex *v1n = v1->next;
+		v1p->next = v1n;
+		v1n->prev = v1p;
+		if (v1 == h1.minXy) {
+			if ((v1n->point.x < v1p->point.x) || ((v1n->point.x == v1p->point.x) && (v1n->point.y < v1p->point.y))) {
+				h1.minXy = v1n;
+			} else {
+				h1.minXy = v1p;
+			}
+		}
+		if (v1 == h1.maxXy) {
+			if ((v1n->point.x > v1p->point.x) || ((v1n->point.x == v1p->point.x) && (v1n->point.y > v1p->point.y))) {
+				h1.maxXy = v1n;
+			} else {
+				h1.maxXy = v1p;
+			}
+		}
+	}
+
+	v0 = h0.maxXy;
+	v1 = h1.maxXy;
+	Vertex *v00 = NULL;
+	Vertex *v10 = NULL;
+	int32_t sign = 1;
+
+	for (int32_t side = 0; side <= 1; side++) {
+		int32_t dx = (v1->point.x - v0->point.x) * sign;
+		if (dx > 0) {
+			while (true) {
+				int32_t dy = v1->point.y - v0->point.y;
+
+				Vertex *w0 = side ? v0->next : v0->prev;
+				if (w0 != v0) {
+					int32_t dx0 = (w0->point.x - v0->point.x) * sign;
+					int32_t dy0 = w0->point.y - v0->point.y;
+					if ((dy0 <= 0) && ((dx0 == 0) || ((dx0 < 0) && (dy0 * dx <= dy * dx0)))) {
+						v0 = w0;
+						dx = (v1->point.x - v0->point.x) * sign;
+						continue;
+					}
+				}
+
+				Vertex *w1 = side ? v1->next : v1->prev;
+				if (w1 != v1) {
+					int32_t dx1 = (w1->point.x - v1->point.x) * sign;
+					int32_t dy1 = w1->point.y - v1->point.y;
+					int32_t dxn = (w1->point.x - v0->point.x) * sign;
+					if ((dxn > 0) && (dy1 < 0) && ((dx1 == 0) || ((dx1 < 0) && (dy1 * dx < dy * dx1)))) {
+						v1 = w1;
+						dx = dxn;
+						continue;
+					}
+				}
+
+				break;
+			}
+		} else if (dx < 0) {
+			while (true) {
+				int32_t dy = v1->point.y - v0->point.y;
+
+				Vertex *w1 = side ? v1->prev : v1->next;
+				if (w1 != v1) {
+					int32_t dx1 = (w1->point.x - v1->point.x) * sign;
+					int32_t dy1 = w1->point.y - v1->point.y;
+					if ((dy1 >= 0) && ((dx1 == 0) || ((dx1 < 0) && (dy1 * dx <= dy * dx1)))) {
+						v1 = w1;
+						dx = (v1->point.x - v0->point.x) * sign;
+						continue;
+					}
+				}
+
+				Vertex *w0 = side ? v0->prev : v0->next;
+				if (w0 != v0) {
+					int32_t dx0 = (w0->point.x - v0->point.x) * sign;
+					int32_t dy0 = w0->point.y - v0->point.y;
+					int32_t dxn = (v1->point.x - w0->point.x) * sign;
+					if ((dxn < 0) && (dy0 > 0) && ((dx0 == 0) || ((dx0 < 0) && (dy0 * dx < dy * dx0)))) {
+						v0 = w0;
+						dx = dxn;
+						continue;
+					}
+				}
+
+				break;
+			}
+		} else {
+			int32_t x = v0->point.x;
+			int32_t y0 = v0->point.y;
+			Vertex *w0 = v0;
+			Vertex *t;
+			while (((t = side ? w0->next : w0->prev) != v0) && (t->point.x == x) && (t->point.y <= y0)) {
+				w0 = t;
+				y0 = t->point.y;
+			}
+			v0 = w0;
+
+			int32_t y1 = v1->point.y;
+			Vertex *w1 = v1;
+			while (((t = side ? w1->prev : w1->next) != v1) && (t->point.x == x) && (t->point.y >= y1)) {
+				w1 = t;
+				y1 = t->point.y;
+			}
+			v1 = w1;
+		}
+
+		if (side == 0) {
+			v00 = v0;
+			v10 = v1;
+
+			v0 = h0.minXy;
+			v1 = h1.minXy;
+			sign = -1;
+		}
+	}
+
+	v0->prev = v1;
+	v1->next = v0;
+
+	v00->next = v10;
+	v10->prev = v00;
+
+	if (h1.minXy->point.x < h0.minXy->point.x) {
+		h0.minXy = h1.minXy;
+	}
+	if (h1.maxXy->point.x >= h0.maxXy->point.x) {
+		h0.maxXy = h1.maxXy;
+	}
+
+	h0.maxYx = h1.maxYx;
+
+	c0 = v00;
+	c1 = v10;
+
+	return true;
+}
+
+void btConvexHullInternal::computeInternal(int32_t start, int32_t end, IntermediateHull &result) {
+	int32_t n = end - start;
+	switch (n) {
+		case 0:
+			result.minXy = NULL;
+			result.maxXy = NULL;
+			result.minYx = NULL;
+			result.maxYx = NULL;
+			return;
+		case 2: {
+			Vertex *v = originalVertices[start];
+			Vertex *w = v + 1;
+			if (v->point != w->point) {
+				int32_t dx = v->point.x - w->point.x;
+				int32_t dy = v->point.y - w->point.y;
+
+				if ((dx == 0) && (dy == 0)) {
+					if (v->point.z > w->point.z) {
+						Vertex *t = w;
+						w = v;
+						v = t;
+					}
+					btAssert(v->point.z < w->point.z);
+					v->next = v;
+					v->prev = v;
+					result.minXy = v;
+					result.maxXy = v;
+					result.minYx = v;
+					result.maxYx = v;
+				} else {
+					v->next = w;
+					v->prev = w;
+					w->next = v;
+					w->prev = v;
+
+					if ((dx < 0) || ((dx == 0) && (dy < 0))) {
+						result.minXy = v;
+						result.maxXy = w;
+					} else {
+						result.minXy = w;
+						result.maxXy = v;
+					}
+
+					if ((dy < 0) || ((dy == 0) && (dx < 0))) {
+						result.minYx = v;
+						result.maxYx = w;
+					} else {
+						result.minYx = w;
+						result.maxYx = v;
+					}
+				}
+
+				Edge *e = newEdgePair(v, w);
+				e->link(e);
+				v->edges = e;
+
+				e = e->reverse;
+				e->link(e);
+				w->edges = e;
+
+				return;
+			}
+		}
+		// lint -fallthrough
+		case 1: {
+			Vertex *v = originalVertices[start];
+			v->edges = NULL;
+			v->next = v;
+			v->prev = v;
+
+			result.minXy = v;
+			result.maxXy = v;
+			result.minYx = v;
+			result.maxYx = v;
+
+			return;
+		}
+	}
+
+	int32_t split0 = start + n / 2;
+	Point32 p = originalVertices[split0 - 1]->point;
+	int32_t split1 = split0;
+	while ((split1 < end) && (originalVertices[split1]->point == p)) {
+		split1++;
+	}
+	computeInternal(start, split0, result);
+	IntermediateHull hull1;
+	computeInternal(split1, end, hull1);
+#ifdef DEBUG_CONVEX_HULL
+	printf("\n\nMerge\n");
+	result.print();
+	hull1.print();
+#endif
+	merge(result, hull1);
+#ifdef DEBUG_CONVEX_HULL
+	printf("\n  Result\n");
+	result.print();
+#endif
+}
+
+#ifdef DEBUG_CONVEX_HULL
+void btConvexHullInternal::IntermediateHull::print() {
+	printf("    Hull\n");
+	for (Vertex *v = minXy; v;) {
+		printf("      ");
+		v->print();
+		if (v == maxXy) {
+			printf(" maxXy");
+		}
+		if (v == minYx) {
+			printf(" minYx");
+		}
+		if (v == maxYx) {
+			printf(" maxYx");
+		}
+		if (v->next->prev != v) {
+			printf(" Inconsistency");
+		}
+		printf("\n");
+		v = v->next;
+		if (v == minXy) {
+			break;
+		}
+	}
+	if (minXy) {
+		minXy->copy = (minXy->copy == -1) ? -2 : -1;
+		minXy->printGraph();
+	}
+}
+
+void btConvexHullInternal::Vertex::printGraph() {
+	print();
+	printf("\nEdges\n");
+	Edge *e = edges;
+	if (e) {
+		do {
+			e->print();
+			printf("\n");
+			e = e->next;
+		} while (e != edges);
+		do {
+			Vertex *v = e->target;
+			if (v->copy != copy) {
+				v->copy = copy;
+				v->printGraph();
+			}
+			e = e->next;
+		} while (e != edges);
+	}
+}
+#endif
+
+btConvexHullInternal::Orientation btConvexHullInternal::getOrientation(const Edge *prev, const Edge *next, const Point32 &s, const Point32 &t) {
+	btAssert(prev->reverse->target == next->reverse->target);
+	if (prev->next == next) {
+		if (prev->prev == next) {
+			Point64 n = t.cross(s);
+			Point64 m = (*prev->target - *next->reverse->target).cross(*next->target - *next->reverse->target);
+			btAssert(!m.isZero());
+			int64_t dot = n.dot(m);
+			btAssert(dot != 0);
+			return (dot > 0) ? COUNTER_CLOCKWISE : CLOCKWISE;
+		}
+		return COUNTER_CLOCKWISE;
+	} else if (prev->prev == next) {
+		return CLOCKWISE;
+	} else {
+		return NONE;
+	}
+}
+
+btConvexHullInternal::Edge *btConvexHullInternal::findMaxAngle(bool ccw, const Vertex *start, const Point32 &s, const Point64 &rxs, const Point64 &sxrxs, Rational64 &minCot) {
+	Edge *minEdge = NULL;
+
+#ifdef DEBUG_CONVEX_HULL
+	printf("find max edge for %d\n", start->point.index);
+#endif
+	Edge *e = start->edges;
+	if (e) {
+		do {
+			if (e->copy > mergeStamp) {
+				Point32 t = *e->target - *start;
+				Rational64 cot(t.dot(sxrxs), t.dot(rxs));
+#ifdef DEBUG_CONVEX_HULL
+				printf("      Angle is %f (%d) for ", (float)btAtan(cot.toScalar()), (int32_t)cot.isNaN());
+				e->print();
+#endif
+				if (cot.isNaN()) {
+					btAssert(ccw ? (t.dot(s) < 0) : (t.dot(s) > 0));
+				} else {
+					int32_t cmp;
+					if (minEdge == NULL) {
+						minCot = cot;
+						minEdge = e;
+					} else if ((cmp = cot.compare(minCot)) < 0) {
+						minCot = cot;
+						minEdge = e;
+					} else if ((cmp == 0) && (ccw == (getOrientation(minEdge, e, s, t) == COUNTER_CLOCKWISE))) {
+						minEdge = e;
+					}
+				}
+#ifdef DEBUG_CONVEX_HULL
+				printf("\n");
+#endif
+			}
+			e = e->next;
+		} while (e != start->edges);
+	}
+	return minEdge;
+}
+
+void btConvexHullInternal::findEdgeForCoplanarFaces(Vertex *c0, Vertex *c1, Edge *&e0, Edge *&e1, Vertex *stop0, Vertex *stop1) {
+	Edge *start0 = e0;
+	Edge *start1 = e1;
+	Point32 et0 = start0 ? start0->target->point : c0->point;
+	Point32 et1 = start1 ? start1->target->point : c1->point;
+	Point32 s = c1->point - c0->point;
+	Point64 normal = ((start0 ? start0 : start1)->target->point - c0->point).cross(s);
+	int64_t dist = c0->point.dot(normal);
+	btAssert(!start1 || (start1->target->point.dot(normal) == dist));
+	Point64 perp = s.cross(normal);
+	btAssert(!perp.isZero());
+
+#ifdef DEBUG_CONVEX_HULL
+	printf("   Advancing %d %d  (%p %p, %d %d)\n", c0->point.index, c1->point.index, start0, start1, start0 ? start0->target->point.index : -1, start1 ? start1->target->point.index : -1);
+#endif
+
+	int64_t maxDot0 = et0.dot(perp);
+	if (e0) {
+		while (e0->target != stop0) {
+			Edge *e = e0->reverse->prev;
+			if (e->target->point.dot(normal) < dist) {
+				break;
+			}
+			btAssert(e->target->point.dot(normal) == dist);
+			if (e->copy == mergeStamp) {
+				break;
+			}
+			int64_t dot = e->target->point.dot(perp);
+			if (dot <= maxDot0) {
+				break;
+			}
+			maxDot0 = dot;
+			e0 = e;
+			et0 = e->target->point;
+		}
+	}
+
+	int64_t maxDot1 = et1.dot(perp);
+	if (e1) {
+		while (e1->target != stop1) {
+			Edge *e = e1->reverse->next;
+			if (e->target->point.dot(normal) < dist) {
+				break;
+			}
+			btAssert(e->target->point.dot(normal) == dist);
+			if (e->copy == mergeStamp) {
+				break;
+			}
+			int64_t dot = e->target->point.dot(perp);
+			if (dot <= maxDot1) {
+				break;
+			}
+			maxDot1 = dot;
+			e1 = e;
+			et1 = e->target->point;
+		}
+	}
+
+#ifdef DEBUG_CONVEX_HULL
+	printf("   Starting at %d %d\n", et0.index, et1.index);
+#endif
+
+	int64_t dx = maxDot1 - maxDot0;
+	if (dx > 0) {
+		while (true) {
+			int64_t dy = (et1 - et0).dot(s);
+
+			if (e0 && (e0->target != stop0)) {
+				Edge *f0 = e0->next->reverse;
+				if (f0->copy > mergeStamp) {
+					int64_t dx0 = (f0->target->point - et0).dot(perp);
+					int64_t dy0 = (f0->target->point - et0).dot(s);
+					if ((dx0 == 0) ? (dy0 < 0) : ((dx0 < 0) && (Rational64(dy0, dx0).compare(Rational64(dy, dx)) >= 0))) {
+						et0 = f0->target->point;
+						dx = (et1 - et0).dot(perp);
+						e0 = (e0 == start0) ? NULL : f0;
+						continue;
+					}
+				}
+			}
+
+			if (e1 && (e1->target != stop1)) {
+				Edge *f1 = e1->reverse->next;
+				if (f1->copy > mergeStamp) {
+					Point32 d1 = f1->target->point - et1;
+					if (d1.dot(normal) == 0) {
+						int64_t dx1 = d1.dot(perp);
+						int64_t dy1 = d1.dot(s);
+						int64_t dxn = (f1->target->point - et0).dot(perp);
+						if ((dxn > 0) && ((dx1 == 0) ? (dy1 < 0) : ((dx1 < 0) && (Rational64(dy1, dx1).compare(Rational64(dy, dx)) > 0)))) {
+							e1 = f1;
+							et1 = e1->target->point;
+							dx = dxn;
+							continue;
+						}
+					} else {
+						btAssert((e1 == start1) && (d1.dot(normal) < 0));
+					}
+				}
+			}
+
+			break;
+		}
+	} else if (dx < 0) {
+		while (true) {
+			int64_t dy = (et1 - et0).dot(s);
+
+			if (e1 && (e1->target != stop1)) {
+				Edge *f1 = e1->prev->reverse;
+				if (f1->copy > mergeStamp) {
+					int64_t dx1 = (f1->target->point - et1).dot(perp);
+					int64_t dy1 = (f1->target->point - et1).dot(s);
+					if ((dx1 == 0) ? (dy1 > 0) : ((dx1 < 0) && (Rational64(dy1, dx1).compare(Rational64(dy, dx)) <= 0))) {
+						et1 = f1->target->point;
+						dx = (et1 - et0).dot(perp);
+						e1 = (e1 == start1) ? NULL : f1;
+						continue;
+					}
+				}
+			}
+
+			if (e0 && (e0->target != stop0)) {
+				Edge *f0 = e0->reverse->prev;
+				if (f0->copy > mergeStamp) {
+					Point32 d0 = f0->target->point - et0;
+					if (d0.dot(normal) == 0) {
+						int64_t dx0 = d0.dot(perp);
+						int64_t dy0 = d0.dot(s);
+						int64_t dxn = (et1 - f0->target->point).dot(perp);
+						if ((dxn < 0) && ((dx0 == 0) ? (dy0 > 0) : ((dx0 < 0) && (Rational64(dy0, dx0).compare(Rational64(dy, dx)) < 0)))) {
+							e0 = f0;
+							et0 = e0->target->point;
+							dx = dxn;
+							continue;
+						}
+					} else {
+						btAssert((e0 == start0) && (d0.dot(normal) < 0));
+					}
+				}
+			}
+
+			break;
+		}
+	}
+#ifdef DEBUG_CONVEX_HULL
+	printf("   Advanced edges to %d %d\n", et0.index, et1.index);
+#endif
+}
+
+void btConvexHullInternal::merge(IntermediateHull &h0, IntermediateHull &h1) {
+	if (!h1.maxXy) {
+		return;
+	}
+	if (!h0.maxXy) {
+		h0 = h1;
+		return;
+	}
+
+	mergeStamp--;
+
+	Vertex *c0 = NULL;
+	Edge *toPrev0 = NULL;
+	Edge *firstNew0 = NULL;
+	Edge *pendingHead0 = NULL;
+	Edge *pendingTail0 = NULL;
+	Vertex *c1 = NULL;
+	Edge *toPrev1 = NULL;
+	Edge *firstNew1 = NULL;
+	Edge *pendingHead1 = NULL;
+	Edge *pendingTail1 = NULL;
+	Point32 prevPoint;
+
+	if (mergeProjection(h0, h1, c0, c1)) {
+		Point32 s = *c1 - *c0;
+		Point64 normal = Point32(0, 0, -1).cross(s);
+		Point64 t = s.cross(normal);
+		btAssert(!t.isZero());
+
+		Edge *e = c0->edges;
+		Edge *start0 = NULL;
+		if (e) {
+			do {
+				int64_t dot = (*e->target - *c0).dot(normal);
+				btAssert(dot <= 0);
+				if ((dot == 0) && ((*e->target - *c0).dot(t) > 0)) {
+					if (!start0 || (getOrientation(start0, e, s, Point32(0, 0, -1)) == CLOCKWISE)) {
+						start0 = e;
+					}
+				}
+				e = e->next;
+			} while (e != c0->edges);
+		}
+
+		e = c1->edges;
+		Edge *start1 = NULL;
+		if (e) {
+			do {
+				int64_t dot = (*e->target - *c1).dot(normal);
+				btAssert(dot <= 0);
+				if ((dot == 0) && ((*e->target - *c1).dot(t) > 0)) {
+					if (!start1 || (getOrientation(start1, e, s, Point32(0, 0, -1)) == COUNTER_CLOCKWISE)) {
+						start1 = e;
+					}
+				}
+				e = e->next;
+			} while (e != c1->edges);
+		}
+
+		if (start0 || start1) {
+			findEdgeForCoplanarFaces(c0, c1, start0, start1, NULL, NULL);
+			if (start0) {
+				c0 = start0->target;
+			}
+			if (start1) {
+				c1 = start1->target;
+			}
+		}
+
+		prevPoint = c1->point;
+		prevPoint.z++;
+	} else {
+		prevPoint = c1->point;
+		prevPoint.x++;
+	}
+
+	Vertex *first0 = c0;
+	Vertex *first1 = c1;
+	bool firstRun = true;
+
+	while (true) {
+		Point32 s = *c1 - *c0;
+		Point32 r = prevPoint - c0->point;
+		Point64 rxs = r.cross(s);
+		Point64 sxrxs = s.cross(rxs);
+
+#ifdef DEBUG_CONVEX_HULL
+		printf("\n  Checking %d %d\n", c0->point.index, c1->point.index);
+#endif
+		Rational64 minCot0(0, 0);
+		Edge *min0 = findMaxAngle(false, c0, s, rxs, sxrxs, minCot0);
+		Rational64 minCot1(0, 0);
+		Edge *min1 = findMaxAngle(true, c1, s, rxs, sxrxs, minCot1);
+		if (!min0 && !min1) {
+			Edge *e = newEdgePair(c0, c1);
+			e->link(e);
+			c0->edges = e;
+
+			e = e->reverse;
+			e->link(e);
+			c1->edges = e;
+			return;
+		} else {
+			int32_t cmp = !min0 ? 1 : !min1 ? -1 : minCot0.compare(minCot1);
+#ifdef DEBUG_CONVEX_HULL
+			printf("    -> Result %d\n", cmp);
+#endif
+			if (firstRun || ((cmp >= 0) ? !minCot1.isNegativeInfinity() : !minCot0.isNegativeInfinity())) {
+				Edge *e = newEdgePair(c0, c1);
+				if (pendingTail0) {
+					pendingTail0->prev = e;
+				} else {
+					pendingHead0 = e;
+				}
+				e->next = pendingTail0;
+				pendingTail0 = e;
+
+				e = e->reverse;
+				if (pendingTail1) {
+					pendingTail1->next = e;
+				} else {
+					pendingHead1 = e;
+				}
+				e->prev = pendingTail1;
+				pendingTail1 = e;
+			}
+
+			Edge *e0 = min0;
+			Edge *e1 = min1;
+
+#ifdef DEBUG_CONVEX_HULL
+			printf("   Found min edges to %d %d\n", e0 ? e0->target->point.index : -1, e1 ? e1->target->point.index : -1);
+#endif
+
+			if (cmp == 0) {
+				findEdgeForCoplanarFaces(c0, c1, e0, e1, NULL, NULL);
+			}
+
+			if ((cmp >= 0) && e1) {
+				if (toPrev1) {
+					for (Edge *e = toPrev1->next, *n = NULL; e != min1; e = n) {
+						n = e->next;
+						removeEdgePair(e);
+					}
+				}
+
+				if (pendingTail1) {
+					if (toPrev1) {
+						toPrev1->link(pendingHead1);
+					} else {
+						min1->prev->link(pendingHead1);
+						firstNew1 = pendingHead1;
+					}
+					pendingTail1->link(min1);
+					pendingHead1 = NULL;
+					pendingTail1 = NULL;
+				} else if (!toPrev1) {
+					firstNew1 = min1;
+				}
+
+				prevPoint = c1->point;
+				c1 = e1->target;
+				toPrev1 = e1->reverse;
+			}
+
+			if ((cmp <= 0) && e0) {
+				if (toPrev0) {
+					for (Edge *e = toPrev0->prev, *n = NULL; e != min0; e = n) {
+						n = e->prev;
+						removeEdgePair(e);
+					}
+				}
+
+				if (pendingTail0) {
+					if (toPrev0) {
+						pendingHead0->link(toPrev0);
+					} else {
+						pendingHead0->link(min0->next);
+						firstNew0 = pendingHead0;
+					}
+					min0->link(pendingTail0);
+					pendingHead0 = NULL;
+					pendingTail0 = NULL;
+				} else if (!toPrev0) {
+					firstNew0 = min0;
+				}
+
+				prevPoint = c0->point;
+				c0 = e0->target;
+				toPrev0 = e0->reverse;
+			}
+		}
+
+		if ((c0 == first0) && (c1 == first1)) {
+			if (toPrev0 == NULL) {
+				pendingHead0->link(pendingTail0);
+				c0->edges = pendingTail0;
+			} else {
+				for (Edge *e = toPrev0->prev, *n = NULL; e != firstNew0; e = n) {
+					n = e->prev;
+					removeEdgePair(e);
+				}
+				if (pendingTail0) {
+					pendingHead0->link(toPrev0);
+					firstNew0->link(pendingTail0);
+				}
+			}
+
+			if (toPrev1 == NULL) {
+				pendingTail1->link(pendingHead1);
+				c1->edges = pendingTail1;
+			} else {
+				for (Edge *e = toPrev1->next, *n = NULL; e != firstNew1; e = n) {
+					n = e->next;
+					removeEdgePair(e);
+				}
+				if (pendingTail1) {
+					toPrev1->link(pendingHead1);
+					pendingTail1->link(firstNew1);
+				}
+			}
+
+			return;
+		}
+
+		firstRun = false;
+	}
+}
+
+static bool pointCmp(const btConvexHullInternal::Point32 &p, const btConvexHullInternal::Point32 &q) {
+	return (p.y < q.y) || ((p.y == q.y) && ((p.x < q.x) || ((p.x == q.x) && (p.z < q.z))));
+}
+
+void btConvexHullInternal::compute(const void *coords, bool doubleCoords, int32_t stride, int32_t count) {
+	btVector3 min(btScalar(1e30), btScalar(1e30), btScalar(1e30)), max(btScalar(-1e30), btScalar(-1e30), btScalar(-1e30));
+	const char *ptr = (const char *)coords;
+	if (doubleCoords) {
+		for (int32_t i = 0; i < count; i++) {
+			const double *v = (const double *)ptr;
+			btVector3 p((btScalar)v[0], (btScalar)v[1], (btScalar)v[2]);
+			ptr += stride;
+			min.setMin(p);
+			max.setMax(p);
+		}
+	} else {
+		for (int32_t i = 0; i < count; i++) {
+			const float *v = (const float *)ptr;
+			btVector3 p(v[0], v[1], v[2]);
+			ptr += stride;
+			min.setMin(p);
+			max.setMax(p);
+		}
+	}
+
+	btVector3 s = max - min;
+	maxAxis = s.maxAxis();
+	minAxis = s.minAxis();
+	if (minAxis == maxAxis) {
+		minAxis = (maxAxis + 1) % 3;
+	}
+	medAxis = 3 - maxAxis - minAxis;
+
+	s /= btScalar(10216);
+	if (((medAxis + 1) % 3) != maxAxis) {
+		s *= -1;
+	}
+	scaling = s;
+
+	if (s[0] != 0) {
+		s[0] = btScalar(1) / s[0];
+	}
+	if (s[1] != 0) {
+		s[1] = btScalar(1) / s[1];
+	}
+	if (s[2] != 0) {
+		s[2] = btScalar(1) / s[2];
+	}
+
+	center = (min + max) * btScalar(0.5);
+
+	btAlignedObjectArray<Point32> points;
+	points.resize(count);
+	ptr = (const char *)coords;
+	if (doubleCoords) {
+		for (int32_t i = 0; i < count; i++) {
+			const double *v = (const double *)ptr;
+			btVector3 p((btScalar)v[0], (btScalar)v[1], (btScalar)v[2]);
+			ptr += stride;
+			p = (p - center) * s;
+			points[i].x = (int32_t)p[medAxis];
+			points[i].y = (int32_t)p[maxAxis];
+			points[i].z = (int32_t)p[minAxis];
+			points[i].index = i;
+		}
+	} else {
+		for (int32_t i = 0; i < count; i++) {
+			const float *v = (const float *)ptr;
+			btVector3 p(v[0], v[1], v[2]);
+			ptr += stride;
+			p = (p - center) * s;
+			points[i].x = (int32_t)p[medAxis];
+			points[i].y = (int32_t)p[maxAxis];
+			points[i].z = (int32_t)p[minAxis];
+			points[i].index = i;
+		}
+	}
+	points.quickSort(pointCmp);
+
+	vertexPool.reset();
+	vertexPool.setArraySize(count);
+	originalVertices.resize(count);
+	for (int32_t i = 0; i < count; i++) {
+		Vertex *v = vertexPool.newObject();
+		v->edges = NULL;
+		v->point = points[i];
+		v->copy = -1;
+		originalVertices[i] = v;
+	}
+
+	points.clear();
+
+	edgePool.reset();
+	edgePool.setArraySize(6 * count);
+
+	usedEdgePairs = 0;
+	maxUsedEdgePairs = 0;
+
+	mergeStamp = -3;
+
+	IntermediateHull hull;
+	computeInternal(0, count, hull);
+	vertexList = hull.minXy;
+#ifdef DEBUG_CONVEX_HULL
+	printf("max. edges %d (3v = %d)", maxUsedEdgePairs, 3 * count);
+#endif
+}
+
+btVector3 btConvexHullInternal::toBtVector(const Point32 &v) {
+	btVector3 p;
+	p[medAxis] = btScalar(v.x);
+	p[maxAxis] = btScalar(v.y);
+	p[minAxis] = btScalar(v.z);
+	return p * scaling;
+}
+
+btVector3 btConvexHullInternal::getBtNormal(Face *face) {
+	return toBtVector(face->dir0).cross(toBtVector(face->dir1)).normalized();
+}
+
+btVector3 btConvexHullInternal::getCoordinates(const Vertex *v) {
+	btVector3 p;
+	p[medAxis] = v->xvalue();
+	p[maxAxis] = v->yvalue();
+	p[minAxis] = v->zvalue();
+	return p * scaling + center;
+}
+
+btScalar btConvexHullInternal::shrink(btScalar amount, btScalar clampAmount) {
+	if (!vertexList) {
+		return 0;
+	}
+	int32_t stamp = --mergeStamp;
+	btAlignedObjectArray<Vertex *> stack;
+	vertexList->copy = stamp;
+	stack.push_back(vertexList);
+	btAlignedObjectArray<Face *> faces;
+
+	Point32 ref = vertexList->point;
+	Int128 hullCenterX(0, 0);
+	Int128 hullCenterY(0, 0);
+	Int128 hullCenterZ(0, 0);
+	Int128 volume(0, 0);
+
+	while (stack.size() > 0) {
+		Vertex *v = stack[stack.size() - 1];
+		stack.pop_back();
+		Edge *e = v->edges;
+		if (e) {
+			do {
+				if (e->target->copy != stamp) {
+					e->target->copy = stamp;
+					stack.push_back(e->target);
+				}
+				if (e->copy != stamp) {
+					Face *face = facePool.newObject();
+					face->init(e->target, e->reverse->prev->target, v);
+					faces.push_back(face);
+					Edge *f = e;
+
+					Vertex *a = NULL;
+					Vertex *b = NULL;
+					do {
+						if (a && b) {
+							int64_t vol = (v->point - ref).dot((a->point - ref).cross(b->point - ref));
+							btAssert(vol >= 0);
+							Point32 c = v->point + a->point + b->point + ref;
+							hullCenterX += vol * c.x;
+							hullCenterY += vol * c.y;
+							hullCenterZ += vol * c.z;
+							volume += vol;
+						}
+
+						btAssert(f->copy != stamp);
+						f->copy = stamp;
+						f->face = face;
+
+						a = b;
+						b = f->target;
+
+						f = f->reverse->prev;
+					} while (f != e);
+				}
+				e = e->next;
+			} while (e != v->edges);
+		}
+	}
+
+	if (volume.getSign() <= 0) {
+		return 0;
+	}
+
+	btVector3 hullCenter;
+	hullCenter[medAxis] = hullCenterX.toScalar();
+	hullCenter[maxAxis] = hullCenterY.toScalar();
+	hullCenter[minAxis] = hullCenterZ.toScalar();
+	hullCenter /= 4 * volume.toScalar();
+	hullCenter *= scaling;
+
+	int32_t faceCount = faces.size();
+
+	if (clampAmount > 0) {
+		btScalar minDist = SIMD_INFINITY;
+		for (int32_t i = 0; i < faceCount; i++) {
+			btVector3 normal = getBtNormal(faces[i]);
+			btScalar dist = normal.dot(toBtVector(faces[i]->origin) - hullCenter);
+			if (dist < minDist) {
+				minDist = dist;
+			}
+		}
+
+		if (minDist <= 0) {
+			return 0;
+		}
+
+		amount = btMin(amount, minDist * clampAmount);
+	}
+
+	uint32_t seed = 243703;
+	for (int32_t i = 0; i < faceCount; i++, seed = 1664525 * seed + 1013904223) {
+		btSwap(faces[i], faces[seed % faceCount]);
+	}
+
+	for (int32_t i = 0; i < faceCount; i++) {
+		if (!shiftFace(faces[i], amount, stack)) {
+			return -amount;
+		}
+	}
+
+	return amount;
+}
+
+bool btConvexHullInternal::shiftFace(Face *face, btScalar amount, btAlignedObjectArray<Vertex *> stack) {
+	btVector3 origShift = getBtNormal(face) * -amount;
+	if (scaling[0] != 0) {
+		origShift[0] /= scaling[0];
+	}
+	if (scaling[1] != 0) {
+		origShift[1] /= scaling[1];
+	}
+	if (scaling[2] != 0) {
+		origShift[2] /= scaling[2];
+	}
+	Point32 shift((int32_t)origShift[medAxis], (int32_t)origShift[maxAxis], (int32_t)origShift[minAxis]);
+	if (shift.isZero()) {
+		return true;
+	}
+	Point64 normal = face->getNormal();
+#ifdef DEBUG_CONVEX_HULL
+	printf("\nShrinking face (%d %d %d) (%d %d %d) (%d %d %d) by (%d %d %d)\n",
+			face->origin.x, face->origin.y, face->origin.z, face->dir0.x, face->dir0.y, face->dir0.z, face->dir1.x, face->dir1.y, face->dir1.z, shift.x, shift.y, shift.z);
+#endif
+	int64_t origDot = face->origin.dot(normal);
+	Point32 shiftedOrigin = face->origin + shift;
+	int64_t shiftedDot = shiftedOrigin.dot(normal);
+	btAssert(shiftedDot <= origDot);
+	if (shiftedDot >= origDot) {
+		return false;
+	}
+
+	Edge *intersection = NULL;
+
+	Edge *startEdge = face->nearbyVertex->edges;
+#ifdef DEBUG_CONVEX_HULL
+	printf("Start edge is ");
+	startEdge->print();
+	printf(", normal is (%lld %lld %lld), shifted dot is %lld\n", normal.x, normal.y, normal.z, shiftedDot);
+#endif
+	Rational128 optDot = face->nearbyVertex->dot(normal);
+	int32_t cmp = optDot.compare(shiftedDot);
+#ifdef SHOW_ITERATIONS
+	int32_t n = 0;
+#endif
+	if (cmp >= 0) {
+		Edge *e = startEdge;
+		do {
+#ifdef SHOW_ITERATIONS
+			n++;
+#endif
+			Rational128 dot = e->target->dot(normal);
+			btAssert(dot.compare(origDot) <= 0);
+#ifdef DEBUG_CONVEX_HULL
+			printf("Moving downwards, edge is ");
+			e->print();
+			printf(", dot is %f (%f %lld)\n", (float)dot.toScalar(), (float)optDot.toScalar(), shiftedDot);
+#endif
+			if (dot.compare(optDot) < 0) {
+				int32_t c = dot.compare(shiftedDot);
+				optDot = dot;
+				e = e->reverse;
+				startEdge = e;
+				if (c < 0) {
+					intersection = e;
+					break;
+				}
+				cmp = c;
+			}
+			e = e->prev;
+		} while (e != startEdge);
+
+		if (!intersection) {
+			return false;
+		}
+	} else {
+		Edge *e = startEdge;
+		do {
+#ifdef SHOW_ITERATIONS
+			n++;
+#endif
+			Rational128 dot = e->target->dot(normal);
+			btAssert(dot.compare(origDot) <= 0);
+#ifdef DEBUG_CONVEX_HULL
+			printf("Moving upwards, edge is ");
+			e->print();
+			printf(", dot is %f (%f %lld)\n", (float)dot.toScalar(), (float)optDot.toScalar(), shiftedDot);
+#endif
+			if (dot.compare(optDot) > 0) {
+				cmp = dot.compare(shiftedDot);
+				if (cmp >= 0) {
+					intersection = e;
+					break;
+				}
+				optDot = dot;
+				e = e->reverse;
+				startEdge = e;
+			}
+			e = e->prev;
+		} while (e != startEdge);
+
+		if (!intersection) {
+			return true;
+		}
+	}
+
+#ifdef SHOW_ITERATIONS
+	printf("Needed %d iterations to find initial intersection\n", n);
+#endif
+
+	if (cmp == 0) {
+		Edge *e = intersection->reverse->next;
+#ifdef SHOW_ITERATIONS
+		n = 0;
+#endif
+		while (e->target->dot(normal).compare(shiftedDot) <= 0) {
+#ifdef SHOW_ITERATIONS
+			n++;
+#endif
+			e = e->next;
+			if (e == intersection->reverse) {
+				return true;
+			}
+#ifdef DEBUG_CONVEX_HULL
+			printf("Checking for outwards edge, current edge is ");
+			e->print();
+			printf("\n");
+#endif
+		}
+#ifdef SHOW_ITERATIONS
+		printf("Needed %d iterations to check for complete containment\n", n);
+#endif
+	}
+
+	Edge *firstIntersection = NULL;
+	Edge *faceEdge = NULL;
+	Edge *firstFaceEdge = NULL;
+
+#ifdef SHOW_ITERATIONS
+	int32_t m = 0;
+#endif
+	while (true) {
+#ifdef SHOW_ITERATIONS
+		m++;
+#endif
+#ifdef DEBUG_CONVEX_HULL
+		printf("Intersecting edge is ");
+		intersection->print();
+		printf("\n");
+#endif
+		if (cmp == 0) {
+			Edge *e = intersection->reverse->next;
+			startEdge = e;
+#ifdef SHOW_ITERATIONS
+			n = 0;
+#endif
+			while (true) {
+#ifdef SHOW_ITERATIONS
+				n++;
+#endif
+				if (e->target->dot(normal).compare(shiftedDot) >= 0) {
+					break;
+				}
+				intersection = e->reverse;
+				e = e->next;
+				if (e == startEdge) {
+					return true;
+				}
+			}
+#ifdef SHOW_ITERATIONS
+			printf("Needed %d iterations to advance intersection\n", n);
+#endif
+		}
+
+#ifdef DEBUG_CONVEX_HULL
+		printf("Advanced intersecting edge to ");
+		intersection->print();
+		printf(", cmp = %d\n", cmp);
+#endif
+
+		if (!firstIntersection) {
+			firstIntersection = intersection;
+		} else if (intersection == firstIntersection) {
+			break;
+		}
+
+		int32_t prevCmp = cmp;
+		Edge *prevIntersection = intersection;
+		Edge *prevFaceEdge = faceEdge;
+
+		Edge *e = intersection->reverse;
+#ifdef SHOW_ITERATIONS
+		n = 0;
+#endif
+		while (true) {
+#ifdef SHOW_ITERATIONS
+			n++;
+#endif
+			e = e->reverse->prev;
+			btAssert(e != intersection->reverse);
+			cmp = e->target->dot(normal).compare(shiftedDot);
+#ifdef DEBUG_CONVEX_HULL
+			printf("Testing edge ");
+			e->print();
+			printf(" -> cmp = %d\n", cmp);
+#endif
+			if (cmp >= 0) {
+				intersection = e;
+				break;
+			}
+		}
+#ifdef SHOW_ITERATIONS
+		printf("Needed %d iterations to find other intersection of face\n", n);
+#endif
+
+		if (cmp > 0) {
+			Vertex *removed = intersection->target;
+			e = intersection->reverse;
+			if (e->prev == e) {
+				removed->edges = NULL;
+			} else {
+				removed->edges = e->prev;
+				e->prev->link(e->next);
+				e->link(e);
+			}
+#ifdef DEBUG_CONVEX_HULL
+			printf("1: Removed part contains (%d %d %d)\n", removed->point.x, removed->point.y, removed->point.z);
+#endif
+
+			Point64 n0 = intersection->face->getNormal();
+			Point64 n1 = intersection->reverse->face->getNormal();
+			int64_t m00 = face->dir0.dot(n0);
+			int64_t m01 = face->dir1.dot(n0);
+			int64_t m10 = face->dir0.dot(n1);
+			int64_t m11 = face->dir1.dot(n1);
+			int64_t r0 = (intersection->face->origin - shiftedOrigin).dot(n0);
+			int64_t r1 = (intersection->reverse->face->origin - shiftedOrigin).dot(n1);
+			Int128 det = Int128::mul(m00, m11) - Int128::mul(m01, m10);
+			btAssert(det.getSign() != 0);
+			Vertex *v = vertexPool.newObject();
+			v->point.index = -1;
+			v->copy = -1;
+			v->point128 = PointR128(Int128::mul(face->dir0.x * r0, m11) - Int128::mul(face->dir0.x * r1, m01) + Int128::mul(face->dir1.x * r1, m00) - Int128::mul(face->dir1.x * r0, m10) + det * shiftedOrigin.x,
+					Int128::mul(face->dir0.y * r0, m11) - Int128::mul(face->dir0.y * r1, m01) + Int128::mul(face->dir1.y * r1, m00) - Int128::mul(face->dir1.y * r0, m10) + det * shiftedOrigin.y,
+					Int128::mul(face->dir0.z * r0, m11) - Int128::mul(face->dir0.z * r1, m01) + Int128::mul(face->dir1.z * r1, m00) - Int128::mul(face->dir1.z * r0, m10) + det * shiftedOrigin.z,
+					det);
+			v->point.x = (int32_t)v->point128.xvalue();
+			v->point.y = (int32_t)v->point128.yvalue();
+			v->point.z = (int32_t)v->point128.zvalue();
+			intersection->target = v;
+			v->edges = e;
+
+			stack.push_back(v);
+			stack.push_back(removed);
+			stack.push_back(NULL);
+		}
+
+		if (cmp || prevCmp || (prevIntersection->reverse->next->target != intersection->target)) {
+			faceEdge = newEdgePair(prevIntersection->target, intersection->target);
+			if (prevCmp == 0) {
+				faceEdge->link(prevIntersection->reverse->next);
+			}
+			if ((prevCmp == 0) || prevFaceEdge) {
+				prevIntersection->reverse->link(faceEdge);
+			}
+			if (cmp == 0) {
+				intersection->reverse->prev->link(faceEdge->reverse);
+			}
+			faceEdge->reverse->link(intersection->reverse);
+		} else {
+			faceEdge = prevIntersection->reverse->next;
+		}
+
+		if (prevFaceEdge) {
+			if (prevCmp > 0) {
+				faceEdge->link(prevFaceEdge->reverse);
+			} else if (faceEdge != prevFaceEdge->reverse) {
+				stack.push_back(prevFaceEdge->target);
+				while (faceEdge->next != prevFaceEdge->reverse) {
+					Vertex *removed = faceEdge->next->target;
+					removeEdgePair(faceEdge->next);
+					stack.push_back(removed);
+#ifdef DEBUG_CONVEX_HULL
+					printf("2: Removed part contains (%d %d %d)\n", removed->point.x, removed->point.y, removed->point.z);
+#endif
+				}
+				stack.push_back(NULL);
+			}
+		}
+		faceEdge->face = face;
+		faceEdge->reverse->face = intersection->face;
+
+		if (!firstFaceEdge) {
+			firstFaceEdge = faceEdge;
+		}
+	}
+#ifdef SHOW_ITERATIONS
+	printf("Needed %d iterations to process all intersections\n", m);
+#endif
+
+	if (cmp > 0) {
+		firstFaceEdge->reverse->target = faceEdge->target;
+		firstIntersection->reverse->link(firstFaceEdge);
+		firstFaceEdge->link(faceEdge->reverse);
+	} else if (firstFaceEdge != faceEdge->reverse) {
+		stack.push_back(faceEdge->target);
+		while (firstFaceEdge->next != faceEdge->reverse) {
+			Vertex *removed = firstFaceEdge->next->target;
+			removeEdgePair(firstFaceEdge->next);
+			stack.push_back(removed);
+#ifdef DEBUG_CONVEX_HULL
+			printf("3: Removed part contains (%d %d %d)\n", removed->point.x, removed->point.y, removed->point.z);
+#endif
+		}
+		stack.push_back(NULL);
+	}
+
+	btAssert(stack.size() > 0);
+	vertexList = stack[0];
+
+#ifdef DEBUG_CONVEX_HULL
+	printf("Removing part\n");
+#endif
+#ifdef SHOW_ITERATIONS
+	n = 0;
+#endif
+	int32_t pos = 0;
+	while (pos < stack.size()) {
+		int32_t end = stack.size();
+		while (pos < end) {
+			Vertex *kept = stack[pos++];
+#ifdef DEBUG_CONVEX_HULL
+			kept->print();
+#endif
+			bool deeper = false;
+			Vertex *removed;
+			while ((removed = stack[pos++]) != NULL) {
+#ifdef SHOW_ITERATIONS
+				n++;
+#endif
+				kept->receiveNearbyFaces(removed);
+				while (removed->edges) {
+					if (!deeper) {
+						deeper = true;
+						stack.push_back(kept);
+					}
+					stack.push_back(removed->edges->target);
+					removeEdgePair(removed->edges);
+				}
+			}
+			if (deeper) {
+				stack.push_back(NULL);
+			}
+		}
+	}
+#ifdef SHOW_ITERATIONS
+	printf("Needed %d iterations to remove part\n", n);
+#endif
+
+	stack.resize(0);
+	face->origin = shiftedOrigin;
+
+	return true;
+}
+
+static int32_t getVertexCopy(btConvexHullInternal::Vertex *vertex, btAlignedObjectArray<btConvexHullInternal::Vertex *> &vertices) {
+	int32_t index = vertex->copy;
+	if (index < 0) {
+		index = vertices.size();
+		vertex->copy = index;
+		vertices.push_back(vertex);
+#ifdef DEBUG_CONVEX_HULL
+		printf("Vertex %d gets index *%d\n", vertex->point.index, index);
+#endif
+	}
+	return index;
+}
+
+btScalar btConvexHullComputer::compute(const void *coords, bool doubleCoords, int32_t stride, int32_t count, btScalar shrink, btScalar shrinkClamp) {
+	if (count <= 0) {
+		vertices.clear();
+		edges.clear();
+		faces.clear();
+		return 0;
+	}
+
+	btConvexHullInternal hull;
+	hull.compute(coords, doubleCoords, stride, count);
+
+	btScalar shift = 0;
+	if ((shrink > 0) && ((shift = hull.shrink(shrink, shrinkClamp)) < 0)) {
+		vertices.clear();
+		edges.clear();
+		faces.clear();
+		return shift;
+	}
+
+	vertices.resize(0);
+	edges.resize(0);
+	faces.resize(0);
+
+	btAlignedObjectArray<btConvexHullInternal::Vertex *> oldVertices;
+	getVertexCopy(hull.vertexList, oldVertices);
+	int32_t copied = 0;
+	while (copied < oldVertices.size()) {
+		btConvexHullInternal::Vertex *v = oldVertices[copied];
+		vertices.push_back(hull.getCoordinates(v));
+		btConvexHullInternal::Edge *firstEdge = v->edges;
+		if (firstEdge) {
+			int32_t firstCopy = -1;
+			int32_t prevCopy = -1;
+			btConvexHullInternal::Edge *e = firstEdge;
+			do {
+				if (e->copy < 0) {
+					int32_t s = edges.size();
+					edges.push_back(Edge());
+					edges.push_back(Edge());
+					Edge *c = &edges[s];
+					Edge *r = &edges[s + 1];
+					e->copy = s;
+					e->reverse->copy = s + 1;
+					c->reverse = 1;
+					r->reverse = -1;
+					c->targetVertex = getVertexCopy(e->target, oldVertices);
+					r->targetVertex = copied;
+#ifdef DEBUG_CONVEX_HULL
+					printf("      CREATE: Vertex *%d has edge to *%d\n", copied, c->getTargetVertex());
+#endif
+				}
+				if (prevCopy >= 0) {
+					edges[e->copy].next = prevCopy - e->copy;
+				} else {
+					firstCopy = e->copy;
+				}
+				prevCopy = e->copy;
+				e = e->next;
+			} while (e != firstEdge);
+			edges[firstCopy].next = prevCopy - firstCopy;
+		}
+		copied++;
+	}
+
+	for (int32_t i = 0; i < copied; i++) {
+		btConvexHullInternal::Vertex *v = oldVertices[i];
+		btConvexHullInternal::Edge *firstEdge = v->edges;
+		if (firstEdge) {
+			btConvexHullInternal::Edge *e = firstEdge;
+			do {
+				if (e->copy >= 0) {
+#ifdef DEBUG_CONVEX_HULL
+					printf("Vertex *%d has edge to *%d\n", i, edges[e->copy].getTargetVertex());
+#endif
+					faces.push_back(e->copy);
+					btConvexHullInternal::Edge *f = e;
+					do {
+#ifdef DEBUG_CONVEX_HULL
+						printf("   Face *%d\n", edges[f->copy].getTargetVertex());
+#endif
+						f->copy = -1;
+						f = f->reverse->prev;
+					} while (f != e);
+				}
+				e = e->next;
+			} while (e != firstEdge);
+		}
+	}
+
+	return shift;
+}
+
+//GODOT ADDITION
+};
+//
diff --git a/thirdparty/vhacd/src/vhacdICHull.cpp b/thirdparty/vhacd/src/vhacdICHull.cpp
new file mode 100644
index 0000000000..989587c124
--- /dev/null
+++ b/thirdparty/vhacd/src/vhacdICHull.cpp
@@ -0,0 +1,731 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include "vhacdICHull.h"
+#include <limits>
+
+#ifdef _MSC_VER
+#pragma warning(disable:4456 4706)
+#endif
+
+
+namespace VHACD {
+const double ICHull::sc_eps = 1.0e-15;
+const int32_t ICHull::sc_dummyIndex = std::numeric_limits<int32_t>::max();
+ICHull::ICHull()
+{
+    m_isFlat = false;
+}
+bool ICHull::AddPoints(const Vec3<double>* points, size_t nPoints)
+{
+    if (!points) {
+        return false;
+    }
+    CircularListElement<TMMVertex>* vertex = NULL;
+    for (size_t i = 0; i < nPoints; i++) {
+        vertex = m_mesh.AddVertex();
+        vertex->GetData().m_pos.X() = points[i].X();
+        vertex->GetData().m_pos.Y() = points[i].Y();
+        vertex->GetData().m_pos.Z() = points[i].Z();
+        vertex->GetData().m_name = static_cast<int32_t>(i);
+    }
+    return true;
+}
+bool ICHull::AddPoint(const Vec3<double>& point, int32_t id)
+{
+    if (AddPoints(&point, 1)) {
+        m_mesh.m_vertices.GetData().m_name = id;
+        return true;
+    }
+    return false;
+}
+
+ICHullError ICHull::Process()
+{
+    uint32_t addedPoints = 0;
+    if (m_mesh.GetNVertices() < 3) {
+        return ICHullErrorNotEnoughPoints;
+    }
+    if (m_mesh.GetNVertices() == 3) {
+        m_isFlat = true;
+        CircularListElement<TMMTriangle>* t1 = m_mesh.AddTriangle();
+        CircularListElement<TMMTriangle>* t2 = m_mesh.AddTriangle();
+        CircularListElement<TMMVertex>* v0 = m_mesh.m_vertices.GetHead();
+        CircularListElement<TMMVertex>* v1 = v0->GetNext();
+        CircularListElement<TMMVertex>* v2 = v1->GetNext();
+        // Compute the normal to the plane
+        Vec3<double> p0 = v0->GetData().m_pos;
+        Vec3<double> p1 = v1->GetData().m_pos;
+        Vec3<double> p2 = v2->GetData().m_pos;
+        m_normal = (p1 - p0) ^ (p2 - p0);
+        m_normal.Normalize();
+        t1->GetData().m_vertices[0] = v0;
+        t1->GetData().m_vertices[1] = v1;
+        t1->GetData().m_vertices[2] = v2;
+        t2->GetData().m_vertices[0] = v1;
+        t2->GetData().m_vertices[1] = v2;
+        t2->GetData().m_vertices[2] = v2;
+        return ICHullErrorOK;
+    }
+    if (m_isFlat) {
+        m_mesh.m_edges.Clear();
+        m_mesh.m_triangles.Clear();
+        m_isFlat = false;
+    }
+    if (m_mesh.GetNTriangles() == 0) // we have to create the first polyhedron
+    {
+        ICHullError res = DoubleTriangle();
+        if (res != ICHullErrorOK) {
+            return res;
+        }
+        else {
+            addedPoints += 3;
+        }
+    }
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    // go to the first added and not processed vertex
+    while (!(vertices.GetHead()->GetPrev()->GetData().m_tag)) {
+        vertices.Prev();
+    }
+    while (!vertices.GetData().m_tag) // not processed
+    {
+        vertices.GetData().m_tag = true;
+        if (ProcessPoint()) {
+            addedPoints++;
+            CleanUp(addedPoints);
+            vertices.Next();
+            if (!GetMesh().CheckConsistancy()) {
+                size_t nV = m_mesh.GetNVertices();
+                CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+                for (size_t v = 0; v < nV; ++v) {
+                    if (vertices.GetData().m_name == sc_dummyIndex) {
+                        vertices.Delete();
+                        break;
+                    }
+                    vertices.Next();
+                }
+                return ICHullErrorInconsistent;
+            }
+        }
+    }
+    if (m_isFlat) {
+        SArray<CircularListElement<TMMTriangle>*> trianglesToDuplicate;
+        size_t nT = m_mesh.GetNTriangles();
+        for (size_t f = 0; f < nT; f++) {
+            TMMTriangle& currentTriangle = m_mesh.m_triangles.GetHead()->GetData();
+            if (currentTriangle.m_vertices[0]->GetData().m_name == sc_dummyIndex || currentTriangle.m_vertices[1]->GetData().m_name == sc_dummyIndex || currentTriangle.m_vertices[2]->GetData().m_name == sc_dummyIndex) {
+                m_trianglesToDelete.PushBack(m_mesh.m_triangles.GetHead());
+                for (int32_t k = 0; k < 3; k++) {
+                    for (int32_t h = 0; h < 2; h++) {
+                        if (currentTriangle.m_edges[k]->GetData().m_triangles[h] == m_mesh.m_triangles.GetHead()) {
+                            currentTriangle.m_edges[k]->GetData().m_triangles[h] = 0;
+                            break;
+                        }
+                    }
+                }
+            }
+            else {
+                trianglesToDuplicate.PushBack(m_mesh.m_triangles.GetHead());
+            }
+            m_mesh.m_triangles.Next();
+        }
+        size_t nE = m_mesh.GetNEdges();
+        for (size_t e = 0; e < nE; e++) {
+            TMMEdge& currentEdge = m_mesh.m_edges.GetHead()->GetData();
+            if (currentEdge.m_triangles[0] == 0 && currentEdge.m_triangles[1] == 0) {
+                m_edgesToDelete.PushBack(m_mesh.m_edges.GetHead());
+            }
+            m_mesh.m_edges.Next();
+        }
+        size_t nV = m_mesh.GetNVertices();
+        CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+        for (size_t v = 0; v < nV; ++v) {
+            if (vertices.GetData().m_name == sc_dummyIndex) {
+                vertices.Delete();
+            }
+            else {
+                vertices.GetData().m_tag = false;
+                vertices.Next();
+            }
+        }
+        CleanEdges();
+        CleanTriangles();
+        CircularListElement<TMMTriangle>* newTriangle;
+        for (size_t t = 0; t < trianglesToDuplicate.Size(); t++) {
+            newTriangle = m_mesh.AddTriangle();
+            newTriangle->GetData().m_vertices[0] = trianglesToDuplicate[t]->GetData().m_vertices[1];
+            newTriangle->GetData().m_vertices[1] = trianglesToDuplicate[t]->GetData().m_vertices[0];
+            newTriangle->GetData().m_vertices[2] = trianglesToDuplicate[t]->GetData().m_vertices[2];
+        }
+    }
+    return ICHullErrorOK;
+}
+ICHullError ICHull::Process(const uint32_t nPointsCH,
+    const double minVolume)
+{
+    uint32_t addedPoints = 0;
+    if (nPointsCH < 3 || m_mesh.GetNVertices() < 3) {
+        return ICHullErrorNotEnoughPoints;
+    }
+    if (m_mesh.GetNVertices() == 3) {
+        m_isFlat = true;
+        CircularListElement<TMMTriangle>* t1 = m_mesh.AddTriangle();
+        CircularListElement<TMMTriangle>* t2 = m_mesh.AddTriangle();
+        CircularListElement<TMMVertex>* v0 = m_mesh.m_vertices.GetHead();
+        CircularListElement<TMMVertex>* v1 = v0->GetNext();
+        CircularListElement<TMMVertex>* v2 = v1->GetNext();
+        // Compute the normal to the plane
+        Vec3<double> p0 = v0->GetData().m_pos;
+        Vec3<double> p1 = v1->GetData().m_pos;
+        Vec3<double> p2 = v2->GetData().m_pos;
+        m_normal = (p1 - p0) ^ (p2 - p0);
+        m_normal.Normalize();
+        t1->GetData().m_vertices[0] = v0;
+        t1->GetData().m_vertices[1] = v1;
+        t1->GetData().m_vertices[2] = v2;
+        t2->GetData().m_vertices[0] = v1;
+        t2->GetData().m_vertices[1] = v0;
+        t2->GetData().m_vertices[2] = v2;
+        return ICHullErrorOK;
+    }
+
+    if (m_isFlat) {
+        m_mesh.m_triangles.Clear();
+        m_mesh.m_edges.Clear();
+        m_isFlat = false;
+    }
+
+    if (m_mesh.GetNTriangles() == 0) // we have to create the first polyhedron
+    {
+        ICHullError res = DoubleTriangle();
+        if (res != ICHullErrorOK) {
+            return res;
+        }
+        else {
+            addedPoints += 3;
+        }
+    }
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    while (!vertices.GetData().m_tag && addedPoints < nPointsCH) // not processed
+    {
+        if (!FindMaxVolumePoint((addedPoints > 4) ? minVolume : 0.0)) {
+            break;
+        }
+        vertices.GetData().m_tag = true;
+        if (ProcessPoint()) {
+            addedPoints++;
+            CleanUp(addedPoints);
+            if (!GetMesh().CheckConsistancy()) {
+                size_t nV = m_mesh.GetNVertices();
+                CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+                for (size_t v = 0; v < nV; ++v) {
+                    if (vertices.GetData().m_name == sc_dummyIndex) {
+                        vertices.Delete();
+                        break;
+                    }
+                    vertices.Next();
+                }
+                return ICHullErrorInconsistent;
+            }
+            vertices.Next();
+        }
+    }
+    // delete remaining points
+    while (!vertices.GetData().m_tag) {
+        vertices.Delete();
+    }
+    if (m_isFlat) {
+        SArray<CircularListElement<TMMTriangle>*> trianglesToDuplicate;
+        size_t nT = m_mesh.GetNTriangles();
+        for (size_t f = 0; f < nT; f++) {
+            TMMTriangle& currentTriangle = m_mesh.m_triangles.GetHead()->GetData();
+            if (currentTriangle.m_vertices[0]->GetData().m_name == sc_dummyIndex || currentTriangle.m_vertices[1]->GetData().m_name == sc_dummyIndex || currentTriangle.m_vertices[2]->GetData().m_name == sc_dummyIndex) {
+                m_trianglesToDelete.PushBack(m_mesh.m_triangles.GetHead());
+                for (int32_t k = 0; k < 3; k++) {
+                    for (int32_t h = 0; h < 2; h++) {
+                        if (currentTriangle.m_edges[k]->GetData().m_triangles[h] == m_mesh.m_triangles.GetHead()) {
+                            currentTriangle.m_edges[k]->GetData().m_triangles[h] = 0;
+                            break;
+                        }
+                    }
+                }
+            }
+            else {
+                trianglesToDuplicate.PushBack(m_mesh.m_triangles.GetHead());
+            }
+            m_mesh.m_triangles.Next();
+        }
+        size_t nE = m_mesh.GetNEdges();
+        for (size_t e = 0; e < nE; e++) {
+            TMMEdge& currentEdge = m_mesh.m_edges.GetHead()->GetData();
+            if (currentEdge.m_triangles[0] == 0 && currentEdge.m_triangles[1] == 0) {
+                m_edgesToDelete.PushBack(m_mesh.m_edges.GetHead());
+            }
+            m_mesh.m_edges.Next();
+        }
+        size_t nV = m_mesh.GetNVertices();
+        CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+        for (size_t v = 0; v < nV; ++v) {
+            if (vertices.GetData().m_name == sc_dummyIndex) {
+                vertices.Delete();
+            }
+            else {
+                vertices.GetData().m_tag = false;
+                vertices.Next();
+            }
+        }
+        CleanEdges();
+        CleanTriangles();
+        CircularListElement<TMMTriangle>* newTriangle;
+        for (size_t t = 0; t < trianglesToDuplicate.Size(); t++) {
+            newTriangle = m_mesh.AddTriangle();
+            newTriangle->GetData().m_vertices[0] = trianglesToDuplicate[t]->GetData().m_vertices[1];
+            newTriangle->GetData().m_vertices[1] = trianglesToDuplicate[t]->GetData().m_vertices[0];
+            newTriangle->GetData().m_vertices[2] = trianglesToDuplicate[t]->GetData().m_vertices[2];
+        }
+    }
+    return ICHullErrorOK;
+}
+bool ICHull::FindMaxVolumePoint(const double minVolume)
+{
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    CircularListElement<TMMVertex>* vMaxVolume = 0;
+    CircularListElement<TMMVertex>* vHeadPrev = vertices.GetHead()->GetPrev();
+
+    double maxVolume = minVolume;
+    double volume = 0.0;
+    while (!vertices.GetData().m_tag) // not processed
+    {
+        if (ComputePointVolume(volume, false)) {
+            if (maxVolume < volume) {
+                maxVolume = volume;
+                vMaxVolume = vertices.GetHead();
+            }
+            vertices.Next();
+        }
+    }
+    CircularListElement<TMMVertex>* vHead = vHeadPrev->GetNext();
+    vertices.GetHead() = vHead;
+    if (!vMaxVolume) {
+        return false;
+    }
+    if (vMaxVolume != vHead) {
+        Vec3<double> pos = vHead->GetData().m_pos;
+        int32_t id = vHead->GetData().m_name;
+        vHead->GetData().m_pos = vMaxVolume->GetData().m_pos;
+        vHead->GetData().m_name = vMaxVolume->GetData().m_name;
+        vMaxVolume->GetData().m_pos = pos;
+        vHead->GetData().m_name = id;
+    }
+    return true;
+}
+ICHullError ICHull::DoubleTriangle()
+{
+    // find three non colinear points
+    m_isFlat = false;
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    CircularListElement<TMMVertex>* v0 = vertices.GetHead();
+    while (Colinear(v0->GetData().m_pos,
+        v0->GetNext()->GetData().m_pos,
+        v0->GetNext()->GetNext()->GetData().m_pos)) {
+        if ((v0 = v0->GetNext()) == vertices.GetHead()) {
+            return ICHullErrorCoplanarPoints;
+        }
+    }
+    CircularListElement<TMMVertex>* v1 = v0->GetNext();
+    CircularListElement<TMMVertex>* v2 = v1->GetNext();
+    // mark points as processed
+    v0->GetData().m_tag = v1->GetData().m_tag = v2->GetData().m_tag = true;
+
+    // create two triangles
+    CircularListElement<TMMTriangle>* f0 = MakeFace(v0, v1, v2, 0);
+    MakeFace(v2, v1, v0, f0);
+
+    // find a fourth non-coplanar point to form tetrahedron
+    CircularListElement<TMMVertex>* v3 = v2->GetNext();
+    vertices.GetHead() = v3;
+
+    double vol = ComputeVolume4(v0->GetData().m_pos, v1->GetData().m_pos, v2->GetData().m_pos, v3->GetData().m_pos);
+    while (fabs(vol) < sc_eps && !v3->GetNext()->GetData().m_tag) {
+        v3 = v3->GetNext();
+        vol = ComputeVolume4(v0->GetData().m_pos, v1->GetData().m_pos, v2->GetData().m_pos, v3->GetData().m_pos);
+    }
+    if (fabs(vol) < sc_eps) {
+        // compute the barycenter
+        Vec3<double> bary(0.0, 0.0, 0.0);
+        CircularListElement<TMMVertex>* vBary = v0;
+        do {
+            bary += vBary->GetData().m_pos;
+        } while ((vBary = vBary->GetNext()) != v0);
+        bary /= static_cast<double>(vertices.GetSize());
+
+        // Compute the normal to the plane
+        Vec3<double> p0 = v0->GetData().m_pos;
+        Vec3<double> p1 = v1->GetData().m_pos;
+        Vec3<double> p2 = v2->GetData().m_pos;
+        m_normal = (p1 - p0) ^ (p2 - p0);
+        m_normal.Normalize();
+        // add dummy vertex placed at (bary + normal)
+        vertices.GetHead() = v2;
+        Vec3<double> newPt = bary + m_normal;
+        AddPoint(newPt, sc_dummyIndex);
+        m_isFlat = true;
+        return ICHullErrorOK;
+    }
+    else if (v3 != vertices.GetHead()) {
+        TMMVertex temp;
+        temp.m_name = v3->GetData().m_name;
+        temp.m_pos = v3->GetData().m_pos;
+        v3->GetData().m_name = vertices.GetHead()->GetData().m_name;
+        v3->GetData().m_pos = vertices.GetHead()->GetData().m_pos;
+        vertices.GetHead()->GetData().m_name = temp.m_name;
+        vertices.GetHead()->GetData().m_pos = temp.m_pos;
+    }
+    return ICHullErrorOK;
+}
+CircularListElement<TMMTriangle>* ICHull::MakeFace(CircularListElement<TMMVertex>* v0,
+    CircularListElement<TMMVertex>* v1,
+    CircularListElement<TMMVertex>* v2,
+    CircularListElement<TMMTriangle>* fold)
+{
+    CircularListElement<TMMEdge>* e0;
+    CircularListElement<TMMEdge>* e1;
+    CircularListElement<TMMEdge>* e2;
+    int32_t index = 0;
+    if (!fold) // if first face to be created
+    {
+        e0 = m_mesh.AddEdge(); // create the three edges
+        e1 = m_mesh.AddEdge();
+        e2 = m_mesh.AddEdge();
+    }
+    else // otherwise re-use existing edges (in reverse order)
+    {
+        e0 = fold->GetData().m_edges[2];
+        e1 = fold->GetData().m_edges[1];
+        e2 = fold->GetData().m_edges[0];
+        index = 1;
+    }
+    e0->GetData().m_vertices[0] = v0;
+    e0->GetData().m_vertices[1] = v1;
+    e1->GetData().m_vertices[0] = v1;
+    e1->GetData().m_vertices[1] = v2;
+    e2->GetData().m_vertices[0] = v2;
+    e2->GetData().m_vertices[1] = v0;
+    // create the new face
+    CircularListElement<TMMTriangle>* f = m_mesh.AddTriangle();
+    f->GetData().m_edges[0] = e0;
+    f->GetData().m_edges[1] = e1;
+    f->GetData().m_edges[2] = e2;
+    f->GetData().m_vertices[0] = v0;
+    f->GetData().m_vertices[1] = v1;
+    f->GetData().m_vertices[2] = v2;
+    // link edges to face f
+    e0->GetData().m_triangles[index] = e1->GetData().m_triangles[index] = e2->GetData().m_triangles[index] = f;
+    return f;
+}
+CircularListElement<TMMTriangle>* ICHull::MakeConeFace(CircularListElement<TMMEdge>* e, CircularListElement<TMMVertex>* p)
+{
+    // create two new edges if they don't already exist
+    CircularListElement<TMMEdge>* newEdges[2];
+    for (int32_t i = 0; i < 2; ++i) {
+        if (!(newEdges[i] = e->GetData().m_vertices[i]->GetData().m_duplicate)) { // if the edge doesn't exits add it and mark the vertex as duplicated
+            newEdges[i] = m_mesh.AddEdge();
+            newEdges[i]->GetData().m_vertices[0] = e->GetData().m_vertices[i];
+            newEdges[i]->GetData().m_vertices[1] = p;
+            e->GetData().m_vertices[i]->GetData().m_duplicate = newEdges[i];
+        }
+    }
+    // make the new face
+    CircularListElement<TMMTriangle>* newFace = m_mesh.AddTriangle();
+    newFace->GetData().m_edges[0] = e;
+    newFace->GetData().m_edges[1] = newEdges[0];
+    newFace->GetData().m_edges[2] = newEdges[1];
+    MakeCCW(newFace, e, p);
+    for (int32_t i = 0; i < 2; ++i) {
+        for (int32_t j = 0; j < 2; ++j) {
+            if (!newEdges[i]->GetData().m_triangles[j]) {
+                newEdges[i]->GetData().m_triangles[j] = newFace;
+                break;
+            }
+        }
+    }
+    return newFace;
+}
+bool ICHull::ComputePointVolume(double& totalVolume, bool markVisibleFaces)
+{
+    // mark visible faces
+    CircularListElement<TMMTriangle>* fHead = m_mesh.GetTriangles().GetHead();
+    CircularListElement<TMMTriangle>* f = fHead;
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    CircularListElement<TMMVertex>* vertex0 = vertices.GetHead();
+    bool visible = false;
+    Vec3<double> pos0 = Vec3<double>(vertex0->GetData().m_pos.X(),
+        vertex0->GetData().m_pos.Y(),
+        vertex0->GetData().m_pos.Z());
+    double vol = 0.0;
+    totalVolume = 0.0;
+    Vec3<double> ver0, ver1, ver2;
+    do {
+        ver0.X() = f->GetData().m_vertices[0]->GetData().m_pos.X();
+        ver0.Y() = f->GetData().m_vertices[0]->GetData().m_pos.Y();
+        ver0.Z() = f->GetData().m_vertices[0]->GetData().m_pos.Z();
+        ver1.X() = f->GetData().m_vertices[1]->GetData().m_pos.X();
+        ver1.Y() = f->GetData().m_vertices[1]->GetData().m_pos.Y();
+        ver1.Z() = f->GetData().m_vertices[1]->GetData().m_pos.Z();
+        ver2.X() = f->GetData().m_vertices[2]->GetData().m_pos.X();
+        ver2.Y() = f->GetData().m_vertices[2]->GetData().m_pos.Y();
+        ver2.Z() = f->GetData().m_vertices[2]->GetData().m_pos.Z();
+        vol = ComputeVolume4(ver0, ver1, ver2, pos0);
+        if (vol < -sc_eps) {
+            vol = fabs(vol);
+            totalVolume += vol;
+            if (markVisibleFaces) {
+                f->GetData().m_visible = true;
+                m_trianglesToDelete.PushBack(f);
+            }
+            visible = true;
+        }
+        f = f->GetNext();
+    } while (f != fHead);
+
+    if (m_trianglesToDelete.Size() == m_mesh.m_triangles.GetSize()) {
+        for (size_t i = 0; i < m_trianglesToDelete.Size(); i++) {
+            m_trianglesToDelete[i]->GetData().m_visible = false;
+        }
+        visible = false;
+    }
+    // if no faces visible from p then p is inside the hull
+    if (!visible && markVisibleFaces) {
+        vertices.Delete();
+        m_trianglesToDelete.Resize(0);
+        return false;
+    }
+    return true;
+}
+bool ICHull::ProcessPoint()
+{
+    double totalVolume = 0.0;
+    if (!ComputePointVolume(totalVolume, true)) {
+        return false;
+    }
+    // Mark edges in interior of visible region for deletion.
+    // Create a new face based on each border edge
+    CircularListElement<TMMVertex>* v0 = m_mesh.GetVertices().GetHead();
+    CircularListElement<TMMEdge>* eHead = m_mesh.GetEdges().GetHead();
+    CircularListElement<TMMEdge>* e = eHead;
+    CircularListElement<TMMEdge>* tmp = 0;
+    int32_t nvisible = 0;
+    m_edgesToDelete.Resize(0);
+    m_edgesToUpdate.Resize(0);
+    do {
+        tmp = e->GetNext();
+        nvisible = 0;
+        for (int32_t k = 0; k < 2; k++) {
+            if (e->GetData().m_triangles[k]->GetData().m_visible) {
+                nvisible++;
+            }
+        }
+        if (nvisible == 2) {
+            m_edgesToDelete.PushBack(e);
+        }
+        else if (nvisible == 1) {
+            e->GetData().m_newFace = MakeConeFace(e, v0);
+            m_edgesToUpdate.PushBack(e);
+        }
+        e = tmp;
+    } while (e != eHead);
+    return true;
+}
+bool ICHull::MakeCCW(CircularListElement<TMMTriangle>* f,
+    CircularListElement<TMMEdge>* e,
+    CircularListElement<TMMVertex>* v)
+{
+    // the visible face adjacent to e
+    CircularListElement<TMMTriangle>* fv;
+    if (e->GetData().m_triangles[0]->GetData().m_visible) {
+        fv = e->GetData().m_triangles[0];
+    }
+    else {
+        fv = e->GetData().m_triangles[1];
+    }
+
+    //  set vertex[0] and vertex[1] to have the same orientation as the corresponding vertices of fv.
+    int32_t i; // index of e->m_vertices[0] in fv
+    CircularListElement<TMMVertex>* v0 = e->GetData().m_vertices[0];
+    CircularListElement<TMMVertex>* v1 = e->GetData().m_vertices[1];
+    for (i = 0; fv->GetData().m_vertices[i] != v0; i++)
+        ;
+
+    if (fv->GetData().m_vertices[(i + 1) % 3] != e->GetData().m_vertices[1]) {
+        f->GetData().m_vertices[0] = v1;
+        f->GetData().m_vertices[1] = v0;
+    }
+    else {
+        f->GetData().m_vertices[0] = v0;
+        f->GetData().m_vertices[1] = v1;
+        // swap edges
+        CircularListElement<TMMEdge>* tmp = f->GetData().m_edges[0];
+        f->GetData().m_edges[0] = f->GetData().m_edges[1];
+        f->GetData().m_edges[1] = tmp;
+    }
+    f->GetData().m_vertices[2] = v;
+    return true;
+}
+bool ICHull::CleanUp(uint32_t& addedPoints)
+{
+    bool r0 = CleanEdges();
+    bool r1 = CleanTriangles();
+    bool r2 = CleanVertices(addedPoints);
+    return r0 && r1 && r2;
+}
+bool ICHull::CleanEdges()
+{
+    // integrate the new faces into the data structure
+    CircularListElement<TMMEdge>* e;
+    const size_t ne_update = m_edgesToUpdate.Size();
+    for (size_t i = 0; i < ne_update; ++i) {
+        e = m_edgesToUpdate[i];
+        if (e->GetData().m_newFace) {
+            if (e->GetData().m_triangles[0]->GetData().m_visible) {
+                e->GetData().m_triangles[0] = e->GetData().m_newFace;
+            }
+            else {
+                e->GetData().m_triangles[1] = e->GetData().m_newFace;
+            }
+            e->GetData().m_newFace = 0;
+        }
+    }
+    // delete edges maked for deletion
+    CircularList<TMMEdge>& edges = m_mesh.GetEdges();
+    const size_t ne_delete = m_edgesToDelete.Size();
+    for (size_t i = 0; i < ne_delete; ++i) {
+        edges.Delete(m_edgesToDelete[i]);
+    }
+    m_edgesToDelete.Resize(0);
+    m_edgesToUpdate.Resize(0);
+    return true;
+}
+bool ICHull::CleanTriangles()
+{
+    CircularList<TMMTriangle>& triangles = m_mesh.GetTriangles();
+    const size_t nt_delete = m_trianglesToDelete.Size();
+    for (size_t i = 0; i < nt_delete; ++i) {
+        triangles.Delete(m_trianglesToDelete[i]);
+    }
+    m_trianglesToDelete.Resize(0);
+    return true;
+}
+bool ICHull::CleanVertices(uint32_t& addedPoints)
+{
+    // mark all vertices incident to some undeleted edge as on the hull
+    CircularList<TMMEdge>& edges = m_mesh.GetEdges();
+    CircularListElement<TMMEdge>* e = edges.GetHead();
+    size_t nE = edges.GetSize();
+    for (size_t i = 0; i < nE; i++) {
+        e->GetData().m_vertices[0]->GetData().m_onHull = true;
+        e->GetData().m_vertices[1]->GetData().m_onHull = true;
+        e = e->GetNext();
+    }
+    // delete all the vertices that have been processed but are not on the hull
+    CircularList<TMMVertex>& vertices = m_mesh.GetVertices();
+    CircularListElement<TMMVertex>* vHead = vertices.GetHead();
+    CircularListElement<TMMVertex>* v = vHead;
+    v = v->GetPrev();
+    do {
+        if (v->GetData().m_tag && !v->GetData().m_onHull) {
+            CircularListElement<TMMVertex>* tmp = v->GetPrev();
+            vertices.Delete(v);
+            v = tmp;
+            addedPoints--;
+        }
+        else {
+            v->GetData().m_duplicate = 0;
+            v->GetData().m_onHull = false;
+            v = v->GetPrev();
+        }
+    } while (v->GetData().m_tag && v != vHead);
+    return true;
+}
+void ICHull::Clear()
+{
+    m_mesh.Clear();
+    m_edgesToDelete.Resize(0);
+    m_edgesToUpdate.Resize(0);
+    m_trianglesToDelete.Resize(0);
+    m_isFlat = false;
+}
+const ICHull& ICHull::operator=(ICHull& rhs)
+{
+    if (&rhs != this) {
+        m_mesh.Copy(rhs.m_mesh);
+        m_edgesToDelete = rhs.m_edgesToDelete;
+        m_edgesToUpdate = rhs.m_edgesToUpdate;
+        m_trianglesToDelete = rhs.m_trianglesToDelete;
+        m_isFlat = rhs.m_isFlat;
+    }
+    return (*this);
+}
+bool ICHull::IsInside(const Vec3<double>& pt0, const double eps)
+{
+    const Vec3<double> pt(pt0.X(), pt0.Y(), pt0.Z());
+    if (m_isFlat) {
+        size_t nT = m_mesh.m_triangles.GetSize();
+        Vec3<double> ver0, ver1, ver2, a, b, c;
+        double u, v;
+        for (size_t t = 0; t < nT; t++) {
+            ver0.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.X();
+            ver0.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.Y();
+            ver0.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.Z();
+            ver1.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.X();
+            ver1.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.Y();
+            ver1.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.Z();
+            ver2.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.X();
+            ver2.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.Y();
+            ver2.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.Z();
+            a = ver1 - ver0;
+            b = ver2 - ver0;
+            c = pt - ver0;
+            u = c * a;
+            v = c * b;
+            if (u >= 0.0 && u <= 1.0 && v >= 0.0 && u + v <= 1.0) {
+                return true;
+            }
+            m_mesh.m_triangles.Next();
+        }
+        return false;
+    }
+    else {
+        size_t nT = m_mesh.m_triangles.GetSize();
+        Vec3<double> ver0, ver1, ver2;
+        double vol;
+        for (size_t t = 0; t < nT; t++) {
+            ver0.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.X();
+            ver0.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.Y();
+            ver0.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[0]->GetData().m_pos.Z();
+            ver1.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.X();
+            ver1.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.Y();
+            ver1.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[1]->GetData().m_pos.Z();
+            ver2.X() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.X();
+            ver2.Y() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.Y();
+            ver2.Z() = m_mesh.m_triangles.GetHead()->GetData().m_vertices[2]->GetData().m_pos.Z();
+            vol = ComputeVolume4(ver0, ver1, ver2, pt);
+            if (vol < eps) {
+                return false;
+            }
+            m_mesh.m_triangles.Next();
+        }
+        return true;
+    }
+}
+}
diff --git a/thirdparty/vhacd/src/vhacdManifoldMesh.cpp b/thirdparty/vhacd/src/vhacdManifoldMesh.cpp
new file mode 100644
index 0000000000..7aac9c0d17
--- /dev/null
+++ b/thirdparty/vhacd/src/vhacdManifoldMesh.cpp
@@ -0,0 +1,202 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include "vhacdManifoldMesh.h"
+namespace VHACD {
+TMMVertex::TMMVertex(void)
+{
+    Initialize();
+}
+void TMMVertex::Initialize()
+{
+    m_name = 0;
+    m_id = 0;
+    m_duplicate = 0;
+    m_onHull = false;
+    m_tag = false;
+}
+
+TMMVertex::~TMMVertex(void)
+{
+}
+TMMEdge::TMMEdge(void)
+{
+    Initialize();
+}
+void TMMEdge::Initialize()
+{
+    m_id = 0;
+    m_triangles[0] = m_triangles[1] = m_newFace = 0;
+    m_vertices[0] = m_vertices[1] = 0;
+}
+TMMEdge::~TMMEdge(void)
+{
+}
+void TMMTriangle::Initialize()
+{
+    m_id = 0;
+    for (int32_t i = 0; i < 3; i++) {
+        m_edges[i] = 0;
+        m_vertices[0] = 0;
+    }
+    m_visible = false;
+}
+TMMTriangle::TMMTriangle(void)
+{
+    Initialize();
+}
+TMMTriangle::~TMMTriangle(void)
+{
+}
+TMMesh::TMMesh()
+{
+}
+TMMesh::~TMMesh(void)
+{
+}
+void TMMesh::GetIFS(Vec3<double>* const points, Vec3<int32_t>* const triangles)
+{
+    size_t nV = m_vertices.GetSize();
+    size_t nT = m_triangles.GetSize();
+
+    for (size_t v = 0; v < nV; v++) {
+        points[v] = m_vertices.GetData().m_pos;
+        m_vertices.GetData().m_id = v;
+        m_vertices.Next();
+    }
+    for (size_t f = 0; f < nT; f++) {
+        TMMTriangle& currentTriangle = m_triangles.GetData();
+        triangles[f].X() = static_cast<int32_t>(currentTriangle.m_vertices[0]->GetData().m_id);
+        triangles[f].Y() = static_cast<int32_t>(currentTriangle.m_vertices[1]->GetData().m_id);
+        triangles[f].Z() = static_cast<int32_t>(currentTriangle.m_vertices[2]->GetData().m_id);
+        m_triangles.Next();
+    }
+}
+void TMMesh::Clear()
+{
+    m_vertices.Clear();
+    m_edges.Clear();
+    m_triangles.Clear();
+}
+void TMMesh::Copy(TMMesh& mesh)
+{
+    Clear();
+    // updating the id's
+    size_t nV = mesh.m_vertices.GetSize();
+    size_t nE = mesh.m_edges.GetSize();
+    size_t nT = mesh.m_triangles.GetSize();
+    for (size_t v = 0; v < nV; v++) {
+        mesh.m_vertices.GetData().m_id = v;
+        mesh.m_vertices.Next();
+    }
+    for (size_t e = 0; e < nE; e++) {
+        mesh.m_edges.GetData().m_id = e;
+        mesh.m_edges.Next();
+    }
+    for (size_t f = 0; f < nT; f++) {
+        mesh.m_triangles.GetData().m_id = f;
+        mesh.m_triangles.Next();
+    }
+    // copying data
+    m_vertices = mesh.m_vertices;
+    m_edges = mesh.m_edges;
+    m_triangles = mesh.m_triangles;
+
+    // generate mapping
+    CircularListElement<TMMVertex>** vertexMap = new CircularListElement<TMMVertex>*[nV];
+    CircularListElement<TMMEdge>** edgeMap = new CircularListElement<TMMEdge>*[nE];
+    CircularListElement<TMMTriangle>** triangleMap = new CircularListElement<TMMTriangle>*[nT];
+    for (size_t v = 0; v < nV; v++) {
+        vertexMap[v] = m_vertices.GetHead();
+        m_vertices.Next();
+    }
+    for (size_t e = 0; e < nE; e++) {
+        edgeMap[e] = m_edges.GetHead();
+        m_edges.Next();
+    }
+    for (size_t f = 0; f < nT; f++) {
+        triangleMap[f] = m_triangles.GetHead();
+        m_triangles.Next();
+    }
+
+    // updating pointers
+    for (size_t v = 0; v < nV; v++) {
+        if (vertexMap[v]->GetData().m_duplicate) {
+            vertexMap[v]->GetData().m_duplicate = edgeMap[vertexMap[v]->GetData().m_duplicate->GetData().m_id];
+        }
+    }
+    for (size_t e = 0; e < nE; e++) {
+        if (edgeMap[e]->GetData().m_newFace) {
+            edgeMap[e]->GetData().m_newFace = triangleMap[edgeMap[e]->GetData().m_newFace->GetData().m_id];
+        }
+        if (nT > 0) {
+            for (int32_t f = 0; f < 2; f++) {
+                if (edgeMap[e]->GetData().m_triangles[f]) {
+                    edgeMap[e]->GetData().m_triangles[f] = triangleMap[edgeMap[e]->GetData().m_triangles[f]->GetData().m_id];
+                }
+            }
+        }
+        for (int32_t v = 0; v < 2; v++) {
+            if (edgeMap[e]->GetData().m_vertices[v]) {
+                edgeMap[e]->GetData().m_vertices[v] = vertexMap[edgeMap[e]->GetData().m_vertices[v]->GetData().m_id];
+            }
+        }
+    }
+    for (size_t f = 0; f < nT; f++) {
+        if (nE > 0) {
+            for (int32_t e = 0; e < 3; e++) {
+                if (triangleMap[f]->GetData().m_edges[e]) {
+                    triangleMap[f]->GetData().m_edges[e] = edgeMap[triangleMap[f]->GetData().m_edges[e]->GetData().m_id];
+                }
+            }
+        }
+        for (int32_t v = 0; v < 3; v++) {
+            if (triangleMap[f]->GetData().m_vertices[v]) {
+                triangleMap[f]->GetData().m_vertices[v] = vertexMap[triangleMap[f]->GetData().m_vertices[v]->GetData().m_id];
+            }
+        }
+    }
+    delete[] vertexMap;
+    delete[] edgeMap;
+    delete[] triangleMap;
+}
+bool TMMesh::CheckConsistancy()
+{
+    size_t nE = m_edges.GetSize();
+    size_t nT = m_triangles.GetSize();
+    for (size_t e = 0; e < nE; e++) {
+        for (int32_t f = 0; f < 2; f++) {
+            if (!m_edges.GetHead()->GetData().m_triangles[f]) {
+                return false;
+            }
+        }
+        m_edges.Next();
+    }
+    for (size_t f = 0; f < nT; f++) {
+        for (int32_t e = 0; e < 3; e++) {
+            int32_t found = 0;
+            for (int32_t k = 0; k < 2; k++) {
+                if (m_triangles.GetHead()->GetData().m_edges[e]->GetData().m_triangles[k] == m_triangles.GetHead()) {
+                    found++;
+                }
+            }
+            if (found != 1) {
+                return false;
+            }
+        }
+        m_triangles.Next();
+    }
+    return true;
+}
+}
\ No newline at end of file
diff --git a/thirdparty/vhacd/src/vhacdMesh.cpp b/thirdparty/vhacd/src/vhacdMesh.cpp
new file mode 100644
index 0000000000..5d03989fda
--- /dev/null
+++ b/thirdparty/vhacd/src/vhacdMesh.cpp
@@ -0,0 +1,376 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _CRT_SECURE_NO_WARNINGS
+#define _CRT_SECURE_NO_WARNINGS
+#endif
+
+#include "btConvexHullComputer.h"
+#include "vhacdMesh.h"
+#include "FloatMath.h"
+#include <fstream>
+#include <iosfwd>
+#include <iostream>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string>
+
+namespace VHACD {
+Mesh::Mesh()
+{
+    m_diag = 1.0;
+}
+Mesh::~Mesh()
+{
+}
+
+Vec3<double>& Mesh::ComputeCenter(void)
+{
+	const size_t nV = GetNPoints();
+	if (nV)
+	{
+		double center[3];
+		uint32_t pcount = uint32_t(GetNPoints());
+		const double *points = GetPoints();
+		uint32_t tcount = uint32_t(GetNTriangles());
+		const uint32_t *indices = (const uint32_t *)GetTriangles();
+		FLOAT_MATH::fm_computeCentroid(pcount, points, tcount, indices, center);
+		m_center.X() = center[0];
+		m_center.Y() = center[1];
+		m_center.Z() = center[2];
+		m_minBB = GetPoint(0);
+		m_maxBB = GetPoint(0);
+		for (size_t v = 1; v < nV; v++)
+		{
+			Vec3<double> p = GetPoint(v);
+			if (p.X() < m_minBB.X())
+			{
+				m_minBB.X() = p.X();
+			}
+			if (p.Y() < m_minBB.Y())
+			{
+				m_minBB.Y() = p.Y();
+			}
+			if (p.Z() < m_minBB.Z())
+			{
+				m_minBB.Z() = p.Z();
+			}
+			if (p.X() > m_maxBB.X())
+			{
+				m_maxBB.X() = p.X();
+			}
+			if (p.Y() > m_maxBB.Y())
+			{
+				m_maxBB.Y() = p.Y();
+			}
+			if (p.Z() > m_maxBB.Z())
+			{
+				m_maxBB.Z() = p.Z();
+			}
+		}
+	}
+	return m_center;
+}
+
+double Mesh::ComputeVolume() const
+{
+    const size_t nV = GetNPoints();
+    const size_t nT = GetNTriangles();
+    if (nV == 0 || nT == 0) {
+        return 0.0;
+    }
+
+    Vec3<double> bary(0.0, 0.0, 0.0);
+    for (size_t v = 0; v < nV; v++) {
+        bary += GetPoint(v);
+    }
+    bary /= static_cast<double>(nV);
+
+    Vec3<double> ver0, ver1, ver2;
+    double totalVolume = 0.0;
+    for (int32_t t = 0; t < int32_t(nT); t++) {
+        const Vec3<int32_t>& tri = GetTriangle(t);
+        ver0 = GetPoint(tri[0]);
+        ver1 = GetPoint(tri[1]);
+        ver2 = GetPoint(tri[2]);
+        totalVolume += ComputeVolume4(ver0, ver1, ver2, bary);
+    }
+    return totalVolume / 6.0;
+}
+
+void Mesh::ComputeConvexHull(const double* const pts,
+    const size_t nPts)
+{
+    ResizePoints(0);
+    ResizeTriangles(0);
+    btConvexHullComputer ch;
+    ch.compute(pts, 3 * sizeof(double), (int32_t)nPts, -1.0, -1.0);
+    for (int32_t v = 0; v < ch.vertices.size(); v++) {
+        AddPoint(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+    }
+    const int32_t nt = ch.faces.size();
+    for (int32_t t = 0; t < nt; ++t) {
+        const btConvexHullComputer::Edge* sourceEdge = &(ch.edges[ch.faces[t]]);
+        int32_t a = sourceEdge->getSourceVertex();
+        int32_t b = sourceEdge->getTargetVertex();
+        const btConvexHullComputer::Edge* edge = sourceEdge->getNextEdgeOfFace();
+        int32_t c = edge->getTargetVertex();
+        while (c != a) {
+            AddTriangle(Vec3<int32_t>(a, b, c));
+            edge = edge->getNextEdgeOfFace();
+            b = c;
+            c = edge->getTargetVertex();
+        }
+    }
+}
+void Mesh::Clip(const Plane& plane,
+    SArray<Vec3<double> >& positivePart,
+    SArray<Vec3<double> >& negativePart) const
+{
+    const size_t nV = GetNPoints();
+    if (nV == 0) {
+        return;
+    }
+    double d;
+    for (size_t v = 0; v < nV; v++) {
+        const Vec3<double>& pt = GetPoint(v);
+        d = plane.m_a * pt[0] + plane.m_b * pt[1] + plane.m_c * pt[2] + plane.m_d;
+        if (d > 0.0) {
+            positivePart.PushBack(pt);
+        }
+        else if (d < 0.0) {
+            negativePart.PushBack(pt);
+        }
+        else {
+            positivePart.PushBack(pt);
+            negativePart.PushBack(pt);
+        }
+    }
+}
+bool Mesh::IsInside(const Vec3<double>& pt) const
+{
+    const size_t nV = GetNPoints();
+    const size_t nT = GetNTriangles();
+    if (nV == 0 || nT == 0) {
+        return false;
+    }
+    Vec3<double> ver0, ver1, ver2;
+    double volume;
+    for (int32_t t = 0; t < int32_t(nT); t++) {
+        const Vec3<int32_t>& tri = GetTriangle(t);
+        ver0 = GetPoint(tri[0]);
+        ver1 = GetPoint(tri[1]);
+        ver2 = GetPoint(tri[2]);
+        volume = ComputeVolume4(ver0, ver1, ver2, pt);
+        if (volume < 0.0) {
+            return false;
+        }
+    }
+    return true;
+}
+double Mesh::ComputeDiagBB()
+{
+    const size_t nPoints = GetNPoints();
+    if (nPoints == 0)
+        return 0.0;
+    Vec3<double> minBB = m_points[0];
+    Vec3<double> maxBB = m_points[0];
+    double x, y, z;
+    for (size_t v = 1; v < nPoints; v++) {
+        x = m_points[v][0];
+        y = m_points[v][1];
+        z = m_points[v][2];
+        if (x < minBB[0])
+            minBB[0] = x;
+        else if (x > maxBB[0])
+            maxBB[0] = x;
+        if (y < minBB[1])
+            minBB[1] = y;
+        else if (y > maxBB[1])
+            maxBB[1] = y;
+        if (z < minBB[2])
+            minBB[2] = z;
+        else if (z > maxBB[2])
+            maxBB[2] = z;
+    }
+    return (m_diag = (maxBB - minBB).GetNorm());
+}
+
+#ifdef VHACD_DEBUG_MESH
+bool Mesh::SaveVRML2(const std::string& fileName) const
+{
+    std::ofstream fout(fileName.c_str());
+    if (fout.is_open()) {
+        const Material material;
+
+        if (SaveVRML2(fout, material)) {
+            fout.close();
+            return true;
+        }
+        return false;
+    }
+    return false;
+}
+bool Mesh::SaveVRML2(std::ofstream& fout, const Material& material) const
+{
+    if (fout.is_open()) {
+        fout.setf(std::ios::fixed, std::ios::floatfield);
+        fout.setf(std::ios::showpoint);
+        fout.precision(6);
+        size_t nV = m_points.Size();
+        size_t nT = m_triangles.Size();
+        fout << "#VRML V2.0 utf8" << std::endl;
+        fout << "" << std::endl;
+        fout << "# Vertices: " << nV << std::endl;
+        fout << "# Triangles: " << nT << std::endl;
+        fout << "" << std::endl;
+        fout << "Group {" << std::endl;
+        fout << "    children [" << std::endl;
+        fout << "        Shape {" << std::endl;
+        fout << "            appearance Appearance {" << std::endl;
+        fout << "                material Material {" << std::endl;
+        fout << "                    diffuseColor " << material.m_diffuseColor[0] << " "
+             << material.m_diffuseColor[1] << " "
+             << material.m_diffuseColor[2] << std::endl;
+        fout << "                    ambientIntensity " << material.m_ambientIntensity << std::endl;
+        fout << "                    specularColor " << material.m_specularColor[0] << " "
+             << material.m_specularColor[1] << " "
+             << material.m_specularColor[2] << std::endl;
+        fout << "                    emissiveColor " << material.m_emissiveColor[0] << " "
+             << material.m_emissiveColor[1] << " "
+             << material.m_emissiveColor[2] << std::endl;
+        fout << "                    shininess " << material.m_shininess << std::endl;
+        fout << "                    transparency " << material.m_transparency << std::endl;
+        fout << "                }" << std::endl;
+        fout << "            }" << std::endl;
+        fout << "            geometry IndexedFaceSet {" << std::endl;
+        fout << "                ccw TRUE" << std::endl;
+        fout << "                solid TRUE" << std::endl;
+        fout << "                convex TRUE" << std::endl;
+        if (nV > 0) {
+            fout << "                coord DEF co Coordinate {" << std::endl;
+            fout << "                    point [" << std::endl;
+            for (size_t v = 0; v < nV; v++) {
+                fout << "                        " << m_points[v][0] << " "
+                     << m_points[v][1] << " "
+                     << m_points[v][2] << "," << std::endl;
+            }
+            fout << "                    ]" << std::endl;
+            fout << "                }" << std::endl;
+        }
+        if (nT > 0) {
+            fout << "                coordIndex [ " << std::endl;
+            for (size_t f = 0; f < nT; f++) {
+                fout << "                        " << m_triangles[f][0] << ", "
+                     << m_triangles[f][1] << ", "
+                     << m_triangles[f][2] << ", -1," << std::endl;
+            }
+            fout << "                ]" << std::endl;
+        }
+        fout << "            }" << std::endl;
+        fout << "        }" << std::endl;
+        fout << "    ]" << std::endl;
+        fout << "}" << std::endl;
+        return true;
+    }
+    return false;
+}
+bool Mesh::SaveOFF(const std::string& fileName) const
+{
+    std::ofstream fout(fileName.c_str());
+    if (fout.is_open()) {
+        size_t nV = m_points.Size();
+        size_t nT = m_triangles.Size();
+        fout << "OFF" << std::endl;
+        fout << nV << " " << nT << " " << 0 << std::endl;
+        for (size_t v = 0; v < nV; v++) {
+            fout << m_points[v][0] << " "
+                 << m_points[v][1] << " "
+                 << m_points[v][2] << std::endl;
+        }
+        for (size_t f = 0; f < nT; f++) {
+            fout << "3 " << m_triangles[f][0] << " "
+                 << m_triangles[f][1] << " "
+                 << m_triangles[f][2] << std::endl;
+        }
+        fout.close();
+        return true;
+    }
+    return false;
+}
+
+bool Mesh::LoadOFF(const std::string& fileName, bool invert)
+{
+    FILE* fid = fopen(fileName.c_str(), "r");
+    if (fid) {
+        const std::string strOFF("OFF");
+        char temp[1024];
+        fscanf(fid, "%s", temp);
+        if (std::string(temp) != strOFF) {
+            fclose(fid);
+            return false;
+        }
+        else {
+            int32_t nv = 0;
+            int32_t nf = 0;
+            int32_t ne = 0;
+            fscanf(fid, "%i", &nv);
+            fscanf(fid, "%i", &nf);
+            fscanf(fid, "%i", &ne);
+            m_points.Resize(nv);
+            m_triangles.Resize(nf);
+            Vec3<double> coord;
+            float x, y, z;
+            for (int32_t p = 0; p < nv; p++) {
+                fscanf(fid, "%f", &x);
+                fscanf(fid, "%f", &y);
+                fscanf(fid, "%f", &z);
+                m_points[p][0] = x;
+                m_points[p][1] = y;
+                m_points[p][2] = z;
+            }
+            int32_t i, j, k, s;
+            for (int32_t t = 0; t < nf; ++t) {
+                fscanf(fid, "%i", &s);
+                if (s == 3) {
+                    fscanf(fid, "%i", &i);
+                    fscanf(fid, "%i", &j);
+                    fscanf(fid, "%i", &k);
+                    m_triangles[t][0] = i;
+                    if (invert) {
+                        m_triangles[t][1] = k;
+                        m_triangles[t][2] = j;
+                    }
+                    else {
+                        m_triangles[t][1] = j;
+                        m_triangles[t][2] = k;
+                    }
+                }
+                else // Fix me: support only triangular meshes
+                {
+                    for (int32_t h = 0; h < s; ++h)
+                        fscanf(fid, "%i", &s);
+                }
+            }
+            fclose(fid);
+        }
+    }
+    else {
+        return false;
+    }
+    return true;
+}
+#endif // VHACD_DEBUG_MESH
+}
diff --git a/thirdparty/vhacd/src/vhacdRaycastMesh.cpp b/thirdparty/vhacd/src/vhacdRaycastMesh.cpp
new file mode 100644
index 0000000000..e8b9435d5f
--- /dev/null
+++ b/thirdparty/vhacd/src/vhacdRaycastMesh.cpp
@@ -0,0 +1,208 @@
+#include "vhacdRaycastMesh.h"
+#include <math.h>
+#include <assert.h>
+
+namespace RAYCAST_MESH
+{
+
+/* a = b - c */
+#define vector(a,b,c) \
+	(a)[0] = (b)[0] - (c)[0];	\
+	(a)[1] = (b)[1] - (c)[1];	\
+	(a)[2] = (b)[2] - (c)[2];
+
+#define innerProduct(v,q) \
+	((v)[0] * (q)[0] + \
+	(v)[1] * (q)[1] + \
+	(v)[2] * (q)[2])
+
+#define crossProduct(a,b,c) \
+	(a)[0] = (b)[1] * (c)[2] - (c)[1] * (b)[2]; \
+	(a)[1] = (b)[2] * (c)[0] - (c)[2] * (b)[0]; \
+	(a)[2] = (b)[0] * (c)[1] - (c)[0] * (b)[1];
+
+
+static inline bool rayIntersectsTriangle(const double *p,const double *d,const double *v0,const double *v1,const double *v2,double &t)
+{
+	double e1[3],e2[3],h[3],s[3],q[3];
+	double a,f,u,v;
+
+	vector(e1,v1,v0);
+	vector(e2,v2,v0);
+	crossProduct(h,d,e2);
+	a = innerProduct(e1,h);
+
+	if (a > -0.00001 && a < 0.00001)
+		return(false);
+
+	f = 1/a;
+	vector(s,p,v0);
+	u = f * (innerProduct(s,h));
+
+	if (u < 0.0 || u > 1.0)
+		return(false);
+
+	crossProduct(q,s,e1);
+	v = f * innerProduct(d,q);
+	if (v < 0.0 || u + v > 1.0)
+		return(false);
+	// at this stage we can compute t to find out where
+	// the intersection point is on the line
+	t = f * innerProduct(e2,q);
+	if (t > 0) // ray intersection
+		return(true);
+	else // this means that there is a line intersection
+		// but not a ray intersection
+		return (false);
+}
+
+static double getPointDistance(const double *p1, const double *p2)
+{
+	double dx = p1[0] - p2[0];
+	double dy = p1[1] - p2[1];
+	double dz = p1[2] - p2[2];
+	return sqrt(dx*dx + dy*dy + dz*dz);
+}
+
+class MyRaycastMesh : public VHACD::RaycastMesh
+{
+public:
+
+    template <class T>
+	MyRaycastMesh(uint32_t vcount,
+                  const T *vertices,
+                  uint32_t tcount,
+                  const uint32_t *indices)
+	{
+        mVcount = vcount;
+        mVertices = new double[mVcount * 3];
+        for (uint32_t i = 0; i < mVcount; i++)
+        {
+            mVertices[i * 3 + 0] = vertices[0];
+            mVertices[i * 3 + 1] = vertices[1];
+            mVertices[i * 3 + 2] = vertices[2];
+            vertices += 3;
+        }
+        mTcount = tcount;
+        mIndices = new uint32_t[mTcount * 3];
+        for (uint32_t i = 0; i < mTcount; i++)
+        {
+            mIndices[i * 3 + 0] = indices[0];
+            mIndices[i * 3 + 1] = indices[1];
+            mIndices[i * 3 + 2] = indices[2];
+            indices += 3;
+        }
+	}
+
+
+	~MyRaycastMesh(void)
+	{
+        delete[]mVertices;
+        delete[]mIndices;
+	}
+
+	virtual void release(void)
+	{
+		delete this;
+	}
+
+	virtual bool raycast(const double *from,			// The starting point of the raycast
+		const double *to,				// The ending point of the raycast
+		const double *closestToPoint,	// The point to match the nearest hit location (can just be the 'from' location of no specific point)
+		double *hitLocation,			// The point where the ray hit nearest to the 'closestToPoint' location
+		double *hitDistance) final		// The distance the ray traveled to the hit location
+	{
+		bool ret = false;
+
+		double dir[3];
+
+		dir[0] = to[0] - from[0];
+		dir[1] = to[1] - from[1];
+		dir[2] = to[2] - from[2];
+
+		double distance = sqrt( dir[0]*dir[0] + dir[1]*dir[1]+dir[2]*dir[2] );
+		if ( distance < 0.0000000001f ) return false;
+		double recipDistance = 1.0f / distance;
+		dir[0]*=recipDistance;
+		dir[1]*=recipDistance;
+		dir[2]*=recipDistance;
+		const uint32_t *indices = mIndices;
+		const double *vertices = mVertices;
+		double nearestDistance = distance;
+
+		for (uint32_t tri=0; tri<mTcount; tri++)
+		{
+			uint32_t i1 = indices[tri*3+0];
+			uint32_t i2 = indices[tri*3+1];
+			uint32_t i3 = indices[tri*3+2];
+
+			const double *p1 = &vertices[i1*3];
+			const double *p2 = &vertices[i2*3];
+			const double *p3 = &vertices[i3*3];
+
+			double t;
+			if ( rayIntersectsTriangle(from,dir,p1,p2,p3,t))
+			{
+				double hitPos[3];
+
+				hitPos[0] = from[0] + dir[0] * t;
+				hitPos[1] = from[1] + dir[1] * t;
+				hitPos[2] = from[2] + dir[2] * t;
+
+				double pointDistance = getPointDistance(hitPos, closestToPoint);
+
+				if (pointDistance < nearestDistance )
+				{
+					nearestDistance = pointDistance;
+					if ( hitLocation )
+					{
+						hitLocation[0] = hitPos[0];
+						hitLocation[1] = hitPos[1];
+						hitLocation[2] = hitPos[2];
+					}
+					if ( hitDistance )
+					{
+						*hitDistance = pointDistance;
+					}
+					ret = true;
+				}
+			}
+		}
+		return ret;
+	}
+
+	uint32_t		mVcount;
+	double	        *mVertices;
+	uint32_t		mTcount;
+	uint32_t	    *mIndices;
+};
+
+};
+
+
+
+using namespace RAYCAST_MESH;
+
+namespace VHACD
+{
+
+    RaycastMesh * RaycastMesh::createRaycastMesh(uint32_t vcount,		// The number of vertices in the source triangle mesh
+        const double *vertices,		// The array of vertex positions in the format x1,y1,z1..x2,y2,z2.. etc.
+        uint32_t tcount,		// The number of triangles in the source triangle mesh
+        const uint32_t *indices) // The triangle indices in the format of i1,i2,i3 ... i4,i5,i6, ...
+    {
+        MyRaycastMesh *m = new MyRaycastMesh(vcount, vertices, tcount, indices);
+        return static_cast<RaycastMesh *>(m);
+    }
+
+    RaycastMesh * RaycastMesh::createRaycastMesh(uint32_t vcount,		// The number of vertices in the source triangle mesh
+        const float *vertices,		// The array of vertex positions in the format x1,y1,z1..x2,y2,z2.. etc.
+        uint32_t tcount,		// The number of triangles in the source triangle mesh
+        const uint32_t *indices) // The triangle indices in the format of i1,i2,i3 ... i4,i5,i6, ...
+    {
+        MyRaycastMesh *m = new MyRaycastMesh(vcount, vertices, tcount, indices);
+        return static_cast<RaycastMesh *>(m);
+    }
+
+
+} // end of VHACD namespace
\ No newline at end of file
diff --git a/thirdparty/vhacd/src/vhacdVolume.cpp b/thirdparty/vhacd/src/vhacdVolume.cpp
new file mode 100644
index 0000000000..a250e5acaa
--- /dev/null
+++ b/thirdparty/vhacd/src/vhacdVolume.cpp
@@ -0,0 +1,1626 @@
+/* Copyright (c) 2011 Khaled Mamou (kmamou at gmail dot com)
+ All rights reserved.
+ 
+ 
+ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+ 
+ 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+ 
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+ 
+ 3. The names of the contributors may not be used to endorse or promote products derived from this software without specific prior written permission.
+ 
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _CRT_SECURE_NO_WARNINGS
+#define _CRT_SECURE_NO_WARNINGS
+#endif
+
+#include "btConvexHullComputer.h"
+#include "vhacdVolume.h"
+#include <algorithm>
+#include <float.h>
+#include <math.h>
+#include <queue>
+#include <string.h>
+
+#ifdef _MSC_VER
+#pragma warning(disable:4458 4100)
+#endif
+
+
+namespace VHACD {
+/********************************************************/
+/* AABB-triangle overlap test code                      */
+/* by Tomas Akenine-M�ller                              */
+/* Function: int32_t triBoxOverlap(float boxcenter[3],      */
+/*          float boxhalfsize[3],float triverts[3][3]); */
+/* History:                                             */
+/*   2001-03-05: released the code in its first version */
+/*   2001-06-18: changed the order of the tests, faster */
+/*                                                      */
+/* Acknowledgement: Many thanks to Pierre Terdiman for  */
+/* suggestions and discussions on how to optimize code. */
+/* Thanks to David Hunt for finding a ">="-bug!         */
+/********************************************************/
+
+#define X 0
+#define Y 1
+#define Z 2
+#define FINDMINMAX(x0, x1, x2, min, max) \
+    min = max = x0;                      \
+    if (x1 < min)                        \
+        min = x1;                        \
+    if (x1 > max)                        \
+        max = x1;                        \
+    if (x2 < min)                        \
+        min = x2;                        \
+    if (x2 > max)                        \
+        max = x2;
+
+#define AXISTEST_X01(a, b, fa, fb)                   \
+    p0 = a * v0[Y] - b * v0[Z];                      \
+    p2 = a * v2[Y] - b * v2[Z];                      \
+    if (p0 < p2) {                                   \
+        min = p0;                                    \
+        max = p2;                                    \
+    }                                                \
+    else {                                           \
+        min = p2;                                    \
+        max = p0;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[Y] + fb * boxhalfsize[Z]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+#define AXISTEST_X2(a, b, fa, fb)                    \
+    p0 = a * v0[Y] - b * v0[Z];                      \
+    p1 = a * v1[Y] - b * v1[Z];                      \
+    if (p0 < p1) {                                   \
+        min = p0;                                    \
+        max = p1;                                    \
+    }                                                \
+    else {                                           \
+        min = p1;                                    \
+        max = p0;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[Y] + fb * boxhalfsize[Z]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+#define AXISTEST_Y02(a, b, fa, fb)                   \
+    p0 = -a * v0[X] + b * v0[Z];                     \
+    p2 = -a * v2[X] + b * v2[Z];                     \
+    if (p0 < p2) {                                   \
+        min = p0;                                    \
+        max = p2;                                    \
+    }                                                \
+    else {                                           \
+        min = p2;                                    \
+        max = p0;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[X] + fb * boxhalfsize[Z]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+#define AXISTEST_Y1(a, b, fa, fb)                    \
+    p0 = -a * v0[X] + b * v0[Z];                     \
+    p1 = -a * v1[X] + b * v1[Z];                     \
+    if (p0 < p1) {                                   \
+        min = p0;                                    \
+        max = p1;                                    \
+    }                                                \
+    else {                                           \
+        min = p1;                                    \
+        max = p0;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[X] + fb * boxhalfsize[Z]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+#define AXISTEST_Z12(a, b, fa, fb)                   \
+    p1 = a * v1[X] - b * v1[Y];                      \
+    p2 = a * v2[X] - b * v2[Y];                      \
+    if (p2 < p1) {                                   \
+        min = p2;                                    \
+        max = p1;                                    \
+    }                                                \
+    else {                                           \
+        min = p1;                                    \
+        max = p2;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[X] + fb * boxhalfsize[Y]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+#define AXISTEST_Z0(a, b, fa, fb)                    \
+    p0 = a * v0[X] - b * v0[Y];                      \
+    p1 = a * v1[X] - b * v1[Y];                      \
+    if (p0 < p1) {                                   \
+        min = p0;                                    \
+        max = p1;                                    \
+    }                                                \
+    else {                                           \
+        min = p1;                                    \
+        max = p0;                                    \
+    }                                                \
+    rad = fa * boxhalfsize[X] + fb * boxhalfsize[Y]; \
+    if (min > rad || max < -rad)                     \
+        return 0;
+
+int32_t PlaneBoxOverlap(const Vec3<double>& normal,
+    const Vec3<double>& vert,
+    const Vec3<double>& maxbox)
+{
+    int32_t q;
+    Vec3<double> vmin, vmax;
+    double v;
+    for (q = X; q <= Z; q++) {
+        v = vert[q];
+        if (normal[q] > 0.0) {
+            vmin[q] = -maxbox[q] - v;
+            vmax[q] = maxbox[q] - v;
+        }
+        else {
+            vmin[q] = maxbox[q] - v;
+            vmax[q] = -maxbox[q] - v;
+        }
+    }
+    if (normal * vmin > 0.0)
+        return 0;
+    if (normal * vmax >= 0.0)
+        return 1;
+    return 0;
+}
+
+int32_t TriBoxOverlap(const Vec3<double>& boxcenter,
+    const Vec3<double>& boxhalfsize,
+    const Vec3<double>& triver0,
+    const Vec3<double>& triver1,
+    const Vec3<double>& triver2)
+{
+    /*    use separating axis theorem to test overlap between triangle and box */
+    /*    need to test for overlap in these directions: */
+    /*    1) the {x,y,z}-directions (actually, since we use the AABB of the triangle */
+    /*       we do not even need to test these) */
+    /*    2) normal of the triangle */
+    /*    3) crossproduct(edge from tri, {x,y,z}-directin) */
+    /*       this gives 3x3=9 more tests */
+
+    Vec3<double> v0, v1, v2;
+    double min, max, p0, p1, p2, rad, fex, fey, fez; // -NJMP- "d" local variable removed
+    Vec3<double> normal, e0, e1, e2;
+
+    /* This is the fastest branch on Sun */
+    /* move everything so that the boxcenter is in (0,0,0) */
+
+    v0 = triver0 - boxcenter;
+    v1 = triver1 - boxcenter;
+    v2 = triver2 - boxcenter;
+
+    /* compute triangle edges */
+    e0 = v1 - v0; /* tri edge 0 */
+    e1 = v2 - v1; /* tri edge 1 */
+    e2 = v0 - v2; /* tri edge 2 */
+
+    /* Bullet 3:  */
+    /*  test the 9 tests first (this was faster) */
+    fex = fabs(e0[X]);
+    fey = fabs(e0[Y]);
+    fez = fabs(e0[Z]);
+
+    AXISTEST_X01(e0[Z], e0[Y], fez, fey);
+    AXISTEST_Y02(e0[Z], e0[X], fez, fex);
+    AXISTEST_Z12(e0[Y], e0[X], fey, fex);
+
+    fex = fabs(e1[X]);
+    fey = fabs(e1[Y]);
+    fez = fabs(e1[Z]);
+
+    AXISTEST_X01(e1[Z], e1[Y], fez, fey);
+    AXISTEST_Y02(e1[Z], e1[X], fez, fex);
+    AXISTEST_Z0(e1[Y], e1[X], fey, fex);
+
+    fex = fabs(e2[X]);
+    fey = fabs(e2[Y]);
+    fez = fabs(e2[Z]);
+
+    AXISTEST_X2(e2[Z], e2[Y], fez, fey);
+    AXISTEST_Y1(e2[Z], e2[X], fez, fex);
+    AXISTEST_Z12(e2[Y], e2[X], fey, fex);
+
+    /* Bullet 1: */
+    /*  first test overlap in the {x,y,z}-directions */
+    /*  find min, max of the triangle each direction, and test for overlap in */
+    /*  that direction -- this is equivalent to testing a minimal AABB around */
+    /*  the triangle against the AABB */
+
+    /* test in X-direction */
+    FINDMINMAX(v0[X], v1[X], v2[X], min, max);
+    if (min > boxhalfsize[X] || max < -boxhalfsize[X])
+        return 0;
+
+    /* test in Y-direction */
+    FINDMINMAX(v0[Y], v1[Y], v2[Y], min, max);
+    if (min > boxhalfsize[Y] || max < -boxhalfsize[Y])
+        return 0;
+
+    /* test in Z-direction */
+    FINDMINMAX(v0[Z], v1[Z], v2[Z], min, max);
+    if (min > boxhalfsize[Z] || max < -boxhalfsize[Z])
+        return 0;
+
+    /* Bullet 2: */
+    /*  test if the box intersects the plane of the triangle */
+    /*  compute plane equation of triangle: normal*x+d=0 */
+    normal = e0 ^ e1;
+
+    if (!PlaneBoxOverlap(normal, v0, boxhalfsize))
+        return 0;
+    return 1; /* box and triangle overlaps */
+}
+
+// Slightly modified version of  Stan Melax's code for 3x3 matrix diagonalization (Thanks Stan!)
+// source: http://www.melax.com/diag.html?attredirects=0
+void Diagonalize(const double (&A)[3][3], double (&Q)[3][3], double (&D)[3][3])
+{
+    // A must be a symmetric matrix.
+    // returns Q and D such that
+    // Diagonal matrix D = QT * A * Q;  and  A = Q*D*QT
+    const int32_t maxsteps = 24; // certainly wont need that many.
+    int32_t k0, k1, k2;
+    double o[3], m[3];
+    double q[4] = { 0.0, 0.0, 0.0, 1.0 };
+    double jr[4];
+    double sqw, sqx, sqy, sqz;
+    double tmp1, tmp2, mq;
+    double AQ[3][3];
+    double thet, sgn, t, c;
+    for (int32_t i = 0; i < maxsteps; ++i) {
+        // quat to matrix
+        sqx = q[0] * q[0];
+        sqy = q[1] * q[1];
+        sqz = q[2] * q[2];
+        sqw = q[3] * q[3];
+        Q[0][0] = (sqx - sqy - sqz + sqw);
+        Q[1][1] = (-sqx + sqy - sqz + sqw);
+        Q[2][2] = (-sqx - sqy + sqz + sqw);
+        tmp1 = q[0] * q[1];
+        tmp2 = q[2] * q[3];
+        Q[1][0] = 2.0 * (tmp1 + tmp2);
+        Q[0][1] = 2.0 * (tmp1 - tmp2);
+        tmp1 = q[0] * q[2];
+        tmp2 = q[1] * q[3];
+        Q[2][0] = 2.0 * (tmp1 - tmp2);
+        Q[0][2] = 2.0 * (tmp1 + tmp2);
+        tmp1 = q[1] * q[2];
+        tmp2 = q[0] * q[3];
+        Q[2][1] = 2.0 * (tmp1 + tmp2);
+        Q[1][2] = 2.0 * (tmp1 - tmp2);
+
+        // AQ = A * Q
+        AQ[0][0] = Q[0][0] * A[0][0] + Q[1][0] * A[0][1] + Q[2][0] * A[0][2];
+        AQ[0][1] = Q[0][1] * A[0][0] + Q[1][1] * A[0][1] + Q[2][1] * A[0][2];
+        AQ[0][2] = Q[0][2] * A[0][0] + Q[1][2] * A[0][1] + Q[2][2] * A[0][2];
+        AQ[1][0] = Q[0][0] * A[0][1] + Q[1][0] * A[1][1] + Q[2][0] * A[1][2];
+        AQ[1][1] = Q[0][1] * A[0][1] + Q[1][1] * A[1][1] + Q[2][1] * A[1][2];
+        AQ[1][2] = Q[0][2] * A[0][1] + Q[1][2] * A[1][1] + Q[2][2] * A[1][2];
+        AQ[2][0] = Q[0][0] * A[0][2] + Q[1][0] * A[1][2] + Q[2][0] * A[2][2];
+        AQ[2][1] = Q[0][1] * A[0][2] + Q[1][1] * A[1][2] + Q[2][1] * A[2][2];
+        AQ[2][2] = Q[0][2] * A[0][2] + Q[1][2] * A[1][2] + Q[2][2] * A[2][2];
+        // D = Qt * AQ
+        D[0][0] = AQ[0][0] * Q[0][0] + AQ[1][0] * Q[1][0] + AQ[2][0] * Q[2][0];
+        D[0][1] = AQ[0][0] * Q[0][1] + AQ[1][0] * Q[1][1] + AQ[2][0] * Q[2][1];
+        D[0][2] = AQ[0][0] * Q[0][2] + AQ[1][0] * Q[1][2] + AQ[2][0] * Q[2][2];
+        D[1][0] = AQ[0][1] * Q[0][0] + AQ[1][1] * Q[1][0] + AQ[2][1] * Q[2][0];
+        D[1][1] = AQ[0][1] * Q[0][1] + AQ[1][1] * Q[1][1] + AQ[2][1] * Q[2][1];
+        D[1][2] = AQ[0][1] * Q[0][2] + AQ[1][1] * Q[1][2] + AQ[2][1] * Q[2][2];
+        D[2][0] = AQ[0][2] * Q[0][0] + AQ[1][2] * Q[1][0] + AQ[2][2] * Q[2][0];
+        D[2][1] = AQ[0][2] * Q[0][1] + AQ[1][2] * Q[1][1] + AQ[2][2] * Q[2][1];
+        D[2][2] = AQ[0][2] * Q[0][2] + AQ[1][2] * Q[1][2] + AQ[2][2] * Q[2][2];
+        o[0] = D[1][2];
+        o[1] = D[0][2];
+        o[2] = D[0][1];
+        m[0] = fabs(o[0]);
+        m[1] = fabs(o[1]);
+        m[2] = fabs(o[2]);
+
+        k0 = (m[0] > m[1] && m[0] > m[2]) ? 0 : (m[1] > m[2]) ? 1 : 2; // index of largest element of offdiag
+        k1 = (k0 + 1) % 3;
+        k2 = (k0 + 2) % 3;
+        if (o[k0] == 0.0) {
+            break; // diagonal already
+        }
+        thet = (D[k2][k2] - D[k1][k1]) / (2.0 * o[k0]);
+        sgn = (thet > 0.0) ? 1.0 : -1.0;
+        thet *= sgn; // make it positive
+        t = sgn / (thet + ((thet < 1.E6) ? sqrt(thet * thet + 1.0) : thet)); // sign(T)/(|T|+sqrt(T^2+1))
+        c = 1.0 / sqrt(t * t + 1.0); //  c= 1/(t^2+1) , t=s/c
+        if (c == 1.0) {
+            break; // no room for improvement - reached machine precision.
+        }
+        jr[0] = jr[1] = jr[2] = jr[3] = 0.0;
+        jr[k0] = sgn * sqrt((1.0 - c) / 2.0); // using 1/2 angle identity sin(a/2) = sqrt((1-cos(a))/2)
+        jr[k0] *= -1.0; // since our quat-to-matrix convention was for v*M instead of M*v
+        jr[3] = sqrt(1.0 - jr[k0] * jr[k0]);
+        if (jr[3] == 1.0) {
+            break; // reached limits of floating point precision
+        }
+        q[0] = (q[3] * jr[0] + q[0] * jr[3] + q[1] * jr[2] - q[2] * jr[1]);
+        q[1] = (q[3] * jr[1] - q[0] * jr[2] + q[1] * jr[3] + q[2] * jr[0]);
+        q[2] = (q[3] * jr[2] + q[0] * jr[1] - q[1] * jr[0] + q[2] * jr[3]);
+        q[3] = (q[3] * jr[3] - q[0] * jr[0] - q[1] * jr[1] - q[2] * jr[2]);
+        mq = sqrt(q[0] * q[0] + q[1] * q[1] + q[2] * q[2] + q[3] * q[3]);
+        q[0] /= mq;
+        q[1] /= mq;
+        q[2] /= mq;
+        q[3] /= mq;
+    }
+}
+const double TetrahedronSet::EPS = 0.0000000000001;
+VoxelSet::VoxelSet()
+{
+    m_minBB[0] = m_minBB[1] = m_minBB[2] = 0.0;
+    m_minBBVoxels[0] = m_minBBVoxels[1] = m_minBBVoxels[2] = 0;
+    m_maxBBVoxels[0] = m_maxBBVoxels[1] = m_maxBBVoxels[2] = 1;
+    m_minBBPts[0] = m_minBBPts[1] = m_minBBPts[2] = 0;
+    m_maxBBPts[0] = m_maxBBPts[1] = m_maxBBPts[2] = 1;
+    m_barycenter[0] = m_barycenter[1] = m_barycenter[2] = 0;
+    m_barycenterPCA[0] = m_barycenterPCA[1] = m_barycenterPCA[2] = 0.0;
+    m_scale = 1.0;
+    m_unitVolume = 1.0;
+    m_numVoxelsOnSurface = 0;
+    m_numVoxelsInsideSurface = 0;
+    memset(m_Q, 0, sizeof(double) * 9);
+    memset(m_D, 0, sizeof(double) * 9);
+}
+VoxelSet::~VoxelSet(void)
+{
+}
+void VoxelSet::ComputeBB()
+{
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    for (int32_t h = 0; h < 3; ++h) {
+        m_minBBVoxels[h] = m_voxels[0].m_coord[h];
+        m_maxBBVoxels[h] = m_voxels[0].m_coord[h];
+    }
+    Vec3<double> bary(0.0);
+    for (size_t p = 0; p < nVoxels; ++p) {
+        for (int32_t h = 0; h < 3; ++h) {
+            bary[h] += m_voxels[p].m_coord[h];
+            if (m_minBBVoxels[h] > m_voxels[p].m_coord[h])
+                m_minBBVoxels[h] = m_voxels[p].m_coord[h];
+            if (m_maxBBVoxels[h] < m_voxels[p].m_coord[h])
+                m_maxBBVoxels[h] = m_voxels[p].m_coord[h];
+        }
+    }
+    bary /= (double)nVoxels;
+    for (int32_t h = 0; h < 3; ++h) {
+        m_minBBPts[h] = m_minBBVoxels[h] * m_scale + m_minBB[h];
+        m_maxBBPts[h] = m_maxBBVoxels[h] * m_scale + m_minBB[h];
+        m_barycenter[h] = (short)(bary[h] + 0.5);
+    }
+}
+void VoxelSet::ComputeConvexHull(Mesh& meshCH, const size_t sampling) const
+{
+    const size_t CLUSTER_SIZE = 65536;
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+
+    SArray<Vec3<double> > cpoints;
+
+    Vec3<double>* points = new Vec3<double>[CLUSTER_SIZE];
+    size_t p = 0;
+    size_t s = 0;
+    short i, j, k;
+    while (p < nVoxels) {
+        size_t q = 0;
+        while (q < CLUSTER_SIZE && p < nVoxels) {
+            if (m_voxels[p].m_data == PRIMITIVE_ON_SURFACE) {
+                ++s;
+                if (s == sampling) {
+                    s = 0;
+                    i = m_voxels[p].m_coord[0];
+                    j = m_voxels[p].m_coord[1];
+                    k = m_voxels[p].m_coord[2];
+                    Vec3<double> p0((i - 0.5) * m_scale, (j - 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p1((i + 0.5) * m_scale, (j - 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p2((i + 0.5) * m_scale, (j + 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p3((i - 0.5) * m_scale, (j + 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p4((i - 0.5) * m_scale, (j - 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p5((i + 0.5) * m_scale, (j - 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p6((i + 0.5) * m_scale, (j + 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p7((i - 0.5) * m_scale, (j + 0.5) * m_scale, (k + 0.5) * m_scale);
+                    points[q++] = p0 + m_minBB;
+                    points[q++] = p1 + m_minBB;
+                    points[q++] = p2 + m_minBB;
+                    points[q++] = p3 + m_minBB;
+                    points[q++] = p4 + m_minBB;
+                    points[q++] = p5 + m_minBB;
+                    points[q++] = p6 + m_minBB;
+                    points[q++] = p7 + m_minBB;
+                }
+            }
+            ++p;
+        }
+        btConvexHullComputer ch;
+        ch.compute((double*)points, 3 * sizeof(double), (int32_t)q, -1.0, -1.0);
+        for (int32_t v = 0; v < ch.vertices.size(); v++) {
+            cpoints.PushBack(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+        }
+    }
+    delete[] points;
+
+    points = cpoints.Data();
+    btConvexHullComputer ch;
+    ch.compute((double*)points, 3 * sizeof(double), (int32_t)cpoints.Size(), -1.0, -1.0);
+    meshCH.ResizePoints(0);
+    meshCH.ResizeTriangles(0);
+    for (int32_t v = 0; v < ch.vertices.size(); v++) {
+        meshCH.AddPoint(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+    }
+    const int32_t nt = ch.faces.size();
+    for (int32_t t = 0; t < nt; ++t) {
+        const btConvexHullComputer::Edge* sourceEdge = &(ch.edges[ch.faces[t]]);
+        int32_t a = sourceEdge->getSourceVertex();
+        int32_t b = sourceEdge->getTargetVertex();
+        const btConvexHullComputer::Edge* edge = sourceEdge->getNextEdgeOfFace();
+        int32_t c = edge->getTargetVertex();
+        while (c != a) {
+            meshCH.AddTriangle(Vec3<int32_t>(a, b, c));
+            edge = edge->getNextEdgeOfFace();
+            b = c;
+            c = edge->getTargetVertex();
+        }
+    }
+}
+void VoxelSet::GetPoints(const Voxel& voxel,
+    Vec3<double>* const pts) const
+{
+    short i = voxel.m_coord[0];
+    short j = voxel.m_coord[1];
+    short k = voxel.m_coord[2];
+    pts[0][0] = (i - 0.5) * m_scale + m_minBB[0];
+    pts[1][0] = (i + 0.5) * m_scale + m_minBB[0];
+    pts[2][0] = (i + 0.5) * m_scale + m_minBB[0];
+    pts[3][0] = (i - 0.5) * m_scale + m_minBB[0];
+    pts[4][0] = (i - 0.5) * m_scale + m_minBB[0];
+    pts[5][0] = (i + 0.5) * m_scale + m_minBB[0];
+    pts[6][0] = (i + 0.5) * m_scale + m_minBB[0];
+    pts[7][0] = (i - 0.5) * m_scale + m_minBB[0];
+    pts[0][1] = (j - 0.5) * m_scale + m_minBB[1];
+    pts[1][1] = (j - 0.5) * m_scale + m_minBB[1];
+    pts[2][1] = (j + 0.5) * m_scale + m_minBB[1];
+    pts[3][1] = (j + 0.5) * m_scale + m_minBB[1];
+    pts[4][1] = (j - 0.5) * m_scale + m_minBB[1];
+    pts[5][1] = (j - 0.5) * m_scale + m_minBB[1];
+    pts[6][1] = (j + 0.5) * m_scale + m_minBB[1];
+    pts[7][1] = (j + 0.5) * m_scale + m_minBB[1];
+    pts[0][2] = (k - 0.5) * m_scale + m_minBB[2];
+    pts[1][2] = (k - 0.5) * m_scale + m_minBB[2];
+    pts[2][2] = (k - 0.5) * m_scale + m_minBB[2];
+    pts[3][2] = (k - 0.5) * m_scale + m_minBB[2];
+    pts[4][2] = (k + 0.5) * m_scale + m_minBB[2];
+    pts[5][2] = (k + 0.5) * m_scale + m_minBB[2];
+    pts[6][2] = (k + 0.5) * m_scale + m_minBB[2];
+    pts[7][2] = (k + 0.5) * m_scale + m_minBB[2];
+}
+void VoxelSet::Intersect(const Plane& plane,
+    SArray<Vec3<double> >* const positivePts,
+    SArray<Vec3<double> >* const negativePts,
+    const size_t sampling) const
+{
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    const double d0 = m_scale;
+    double d;
+    Vec3<double> pts[8];
+    Vec3<double> pt;
+    Voxel voxel;
+    size_t sp = 0;
+    size_t sn = 0;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        voxel = m_voxels[v];
+        pt = GetPoint(voxel);
+        d = plane.m_a * pt[0] + plane.m_b * pt[1] + plane.m_c * pt[2] + plane.m_d;
+        //            if      (d >= 0.0 && d <= d0) positivePts->PushBack(pt);
+        //            else if (d < 0.0 && -d <= d0) negativePts->PushBack(pt);
+        if (d >= 0.0) {
+            if (d <= d0) {
+                GetPoints(voxel, pts);
+                for (int32_t k = 0; k < 8; ++k) {
+                    positivePts->PushBack(pts[k]);
+                }
+            }
+            else {
+                if (++sp == sampling) {
+                    //                        positivePts->PushBack(pt);
+                    GetPoints(voxel, pts);
+                    for (int32_t k = 0; k < 8; ++k) {
+                        positivePts->PushBack(pts[k]);
+                    }
+                    sp = 0;
+                }
+            }
+        }
+        else {
+            if (-d <= d0) {
+                GetPoints(voxel, pts);
+                for (int32_t k = 0; k < 8; ++k) {
+                    negativePts->PushBack(pts[k]);
+                }
+            }
+            else {
+                if (++sn == sampling) {
+                    //                        negativePts->PushBack(pt);
+                    GetPoints(voxel, pts);
+                    for (int32_t k = 0; k < 8; ++k) {
+                        negativePts->PushBack(pts[k]);
+                    }
+                    sn = 0;
+                }
+            }
+        }
+    }
+}
+void VoxelSet::ComputeExteriorPoints(const Plane& plane,
+    const Mesh& mesh,
+    SArray<Vec3<double> >* const exteriorPts) const
+{
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    double d;
+    Vec3<double> pt;
+    Vec3<double> pts[8];
+    Voxel voxel;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        voxel = m_voxels[v];
+        pt = GetPoint(voxel);
+        d = plane.m_a * pt[0] + plane.m_b * pt[1] + plane.m_c * pt[2] + plane.m_d;
+        if (d >= 0.0) {
+            if (!mesh.IsInside(pt)) {
+                GetPoints(voxel, pts);
+                for (int32_t k = 0; k < 8; ++k) {
+                    exteriorPts->PushBack(pts[k]);
+                }
+            }
+        }
+    }
+}
+void VoxelSet::ComputeClippedVolumes(const Plane& plane,
+    double& positiveVolume,
+    double& negativeVolume) const
+{
+    negativeVolume = 0.0;
+    positiveVolume = 0.0;
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    double d;
+    Vec3<double> pt;
+    size_t nPositiveVoxels = 0;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        pt = GetPoint(m_voxels[v]);
+        d = plane.m_a * pt[0] + plane.m_b * pt[1] + plane.m_c * pt[2] + plane.m_d;
+        nPositiveVoxels += (d >= 0.0);
+    }
+    size_t nNegativeVoxels = nVoxels - nPositiveVoxels;
+    positiveVolume = m_unitVolume * nPositiveVoxels;
+    negativeVolume = m_unitVolume * nNegativeVoxels;
+}
+void VoxelSet::SelectOnSurface(PrimitiveSet* const onSurfP) const
+{
+    VoxelSet* const onSurf = (VoxelSet*)onSurfP;
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+
+    for (int32_t h = 0; h < 3; ++h) {
+        onSurf->m_minBB[h] = m_minBB[h];
+    }
+    onSurf->m_voxels.Resize(0);
+    onSurf->m_scale = m_scale;
+    onSurf->m_unitVolume = m_unitVolume;
+    onSurf->m_numVoxelsOnSurface = 0;
+    onSurf->m_numVoxelsInsideSurface = 0;
+    Voxel voxel;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        voxel = m_voxels[v];
+        if (voxel.m_data == PRIMITIVE_ON_SURFACE) {
+            onSurf->m_voxels.PushBack(voxel);
+            ++onSurf->m_numVoxelsOnSurface;
+        }
+    }
+}
+void VoxelSet::Clip(const Plane& plane,
+    PrimitiveSet* const positivePartP,
+    PrimitiveSet* const negativePartP) const
+{
+    VoxelSet* const positivePart = (VoxelSet*)positivePartP;
+    VoxelSet* const negativePart = (VoxelSet*)negativePartP;
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+
+    for (int32_t h = 0; h < 3; ++h) {
+        negativePart->m_minBB[h] = positivePart->m_minBB[h] = m_minBB[h];
+    }
+    positivePart->m_voxels.Resize(0);
+    negativePart->m_voxels.Resize(0);
+    positivePart->m_voxels.Allocate(nVoxels);
+    negativePart->m_voxels.Allocate(nVoxels);
+    negativePart->m_scale = positivePart->m_scale = m_scale;
+    negativePart->m_unitVolume = positivePart->m_unitVolume = m_unitVolume;
+    negativePart->m_numVoxelsOnSurface = positivePart->m_numVoxelsOnSurface = 0;
+    negativePart->m_numVoxelsInsideSurface = positivePart->m_numVoxelsInsideSurface = 0;
+
+    double d;
+    Vec3<double> pt;
+    Voxel voxel;
+    const double d0 = m_scale;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        voxel = m_voxels[v];
+        pt = GetPoint(voxel);
+        d = plane.m_a * pt[0] + plane.m_b * pt[1] + plane.m_c * pt[2] + plane.m_d;
+        if (d >= 0.0) {
+            if (voxel.m_data == PRIMITIVE_ON_SURFACE || d <= d0) {
+                voxel.m_data = PRIMITIVE_ON_SURFACE;
+                positivePart->m_voxels.PushBack(voxel);
+                ++positivePart->m_numVoxelsOnSurface;
+            }
+            else {
+                positivePart->m_voxels.PushBack(voxel);
+                ++positivePart->m_numVoxelsInsideSurface;
+            }
+        }
+        else {
+            if (voxel.m_data == PRIMITIVE_ON_SURFACE || -d <= d0) {
+                voxel.m_data = PRIMITIVE_ON_SURFACE;
+                negativePart->m_voxels.PushBack(voxel);
+                ++negativePart->m_numVoxelsOnSurface;
+            }
+            else {
+                negativePart->m_voxels.PushBack(voxel);
+                ++negativePart->m_numVoxelsInsideSurface;
+            }
+        }
+    }
+}
+void VoxelSet::Convert(Mesh& mesh, const VOXEL_VALUE value) const
+{
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    Voxel voxel;
+    Vec3<double> pts[8];
+    for (size_t v = 0; v < nVoxels; ++v) {
+        voxel = m_voxels[v];
+        if (voxel.m_data == value) {
+            GetPoints(voxel, pts);
+            int32_t s = (int32_t)mesh.GetNPoints();
+            for (int32_t k = 0; k < 8; ++k) {
+                mesh.AddPoint(pts[k]);
+            }
+            mesh.AddTriangle(Vec3<int32_t>(s + 0, s + 2, s + 1));
+            mesh.AddTriangle(Vec3<int32_t>(s + 0, s + 3, s + 2));
+            mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 5, s + 6));
+            mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 6, s + 7));
+            mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 6, s + 2));
+            mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 2, s + 3));
+            mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 1, s + 5));
+            mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 0, s + 1));
+            mesh.AddTriangle(Vec3<int32_t>(s + 6, s + 5, s + 1));
+            mesh.AddTriangle(Vec3<int32_t>(s + 6, s + 1, s + 2));
+            mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 0, s + 4));
+            mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 3, s + 0));
+        }
+    }
+}
+void VoxelSet::ComputePrincipalAxes()
+{
+    const size_t nVoxels = m_voxels.Size();
+    if (nVoxels == 0)
+        return;
+    m_barycenterPCA[0] = m_barycenterPCA[1] = m_barycenterPCA[2] = 0.0;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        Voxel& voxel = m_voxels[v];
+        m_barycenterPCA[0] += voxel.m_coord[0];
+        m_barycenterPCA[1] += voxel.m_coord[1];
+        m_barycenterPCA[2] += voxel.m_coord[2];
+    }
+    m_barycenterPCA /= (double)nVoxels;
+
+    double covMat[3][3] = { { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 } };
+    double x, y, z;
+    for (size_t v = 0; v < nVoxels; ++v) {
+        Voxel& voxel = m_voxels[v];
+        x = voxel.m_coord[0] - m_barycenter[0];
+        y = voxel.m_coord[1] - m_barycenter[1];
+        z = voxel.m_coord[2] - m_barycenter[2];
+        covMat[0][0] += x * x;
+        covMat[1][1] += y * y;
+        covMat[2][2] += z * z;
+        covMat[0][1] += x * y;
+        covMat[0][2] += x * z;
+        covMat[1][2] += y * z;
+    }
+    covMat[0][0] /= nVoxels;
+    covMat[1][1] /= nVoxels;
+    covMat[2][2] /= nVoxels;
+    covMat[0][1] /= nVoxels;
+    covMat[0][2] /= nVoxels;
+    covMat[1][2] /= nVoxels;
+    covMat[1][0] = covMat[0][1];
+    covMat[2][0] = covMat[0][2];
+    covMat[2][1] = covMat[1][2];
+    Diagonalize(covMat, m_Q, m_D);
+}
+Volume::Volume()
+{
+    m_dim[0] = m_dim[1] = m_dim[2] = 0;
+    m_minBB[0] = m_minBB[1] = m_minBB[2] = 0.0;
+    m_maxBB[0] = m_maxBB[1] = m_maxBB[2] = 1.0;
+    m_numVoxelsOnSurface = 0;
+    m_numVoxelsInsideSurface = 0;
+    m_numVoxelsOutsideSurface = 0;
+    m_scale = 1.0;
+    m_data = 0;
+}
+Volume::~Volume(void)
+{
+    delete[] m_data;
+}
+void Volume::Allocate()
+{
+    delete[] m_data;
+    size_t size = m_dim[0] * m_dim[1] * m_dim[2];
+    m_data = new unsigned char[size];
+    memset(m_data, PRIMITIVE_UNDEFINED, sizeof(unsigned char) * size);
+}
+void Volume::Free()
+{
+    delete[] m_data;
+    m_data = 0;
+}
+void Volume::FillOutsideSurface(const size_t i0,
+    const size_t j0,
+    const size_t k0,
+    const size_t i1,
+    const size_t j1,
+    const size_t k1)
+{
+    const short neighbours[6][3] = { { 1, 0, 0 },
+        { 0, 1, 0 },
+        { 0, 0, 1 },
+        { -1, 0, 0 },
+        { 0, -1, 0 },
+        { 0, 0, -1 } };
+    std::queue<Vec3<short> > fifo;
+    Vec3<short> current;
+    short a, b, c;
+    for (size_t i = i0; i < i1; ++i) {
+        for (size_t j = j0; j < j1; ++j) {
+            for (size_t k = k0; k < k1; ++k) {
+
+                if (GetVoxel(i, j, k) == PRIMITIVE_UNDEFINED) {
+                    current[0] = (short)i;
+                    current[1] = (short)j;
+                    current[2] = (short)k;
+                    fifo.push(current);
+                    GetVoxel(current[0], current[1], current[2]) = PRIMITIVE_OUTSIDE_SURFACE;
+                    ++m_numVoxelsOutsideSurface;
+                    while (fifo.size() > 0) {
+                        current = fifo.front();
+                        fifo.pop();
+                        for (int32_t h = 0; h < 6; ++h) {
+                            a = current[0] + neighbours[h][0];
+                            b = current[1] + neighbours[h][1];
+                            c = current[2] + neighbours[h][2];
+                            if (a < 0 || a >= (int32_t)m_dim[0] || b < 0 || b >= (int32_t)m_dim[1] || c < 0 || c >= (int32_t)m_dim[2]) {
+                                continue;
+                            }
+                            unsigned char& v = GetVoxel(a, b, c);
+                            if (v == PRIMITIVE_UNDEFINED) {
+                                v = PRIMITIVE_OUTSIDE_SURFACE;
+                                ++m_numVoxelsOutsideSurface;
+                                fifo.push(Vec3<short>(a, b, c));
+                            }
+                        }
+                    }
+                }
+            }
+        }
+    }
+}
+void Volume::FillInsideSurface()
+{
+    const size_t i0 = m_dim[0];
+    const size_t j0 = m_dim[1];
+    const size_t k0 = m_dim[2];
+    for (size_t i = 0; i < i0; ++i) {
+        for (size_t j = 0; j < j0; ++j) {
+            for (size_t k = 0; k < k0; ++k) {
+                unsigned char& v = GetVoxel(i, j, k);
+                if (v == PRIMITIVE_UNDEFINED) {
+                    v = PRIMITIVE_INSIDE_SURFACE;
+                    ++m_numVoxelsInsideSurface;
+                }
+            }
+        }
+    }
+}
+void Volume::Convert(Mesh& mesh, const VOXEL_VALUE value) const
+{
+    const size_t i0 = m_dim[0];
+    const size_t j0 = m_dim[1];
+    const size_t k0 = m_dim[2];
+    for (size_t i = 0; i < i0; ++i) {
+        for (size_t j = 0; j < j0; ++j) {
+            for (size_t k = 0; k < k0; ++k) {
+                const unsigned char& voxel = GetVoxel(i, j, k);
+                if (voxel == value) {
+                    Vec3<double> p0((i - 0.5) * m_scale, (j - 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p1((i + 0.5) * m_scale, (j - 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p2((i + 0.5) * m_scale, (j + 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p3((i - 0.5) * m_scale, (j + 0.5) * m_scale, (k - 0.5) * m_scale);
+                    Vec3<double> p4((i - 0.5) * m_scale, (j - 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p5((i + 0.5) * m_scale, (j - 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p6((i + 0.5) * m_scale, (j + 0.5) * m_scale, (k + 0.5) * m_scale);
+                    Vec3<double> p7((i - 0.5) * m_scale, (j + 0.5) * m_scale, (k + 0.5) * m_scale);
+                    int32_t s = (int32_t)mesh.GetNPoints();
+                    mesh.AddPoint(p0 + m_minBB);
+                    mesh.AddPoint(p1 + m_minBB);
+                    mesh.AddPoint(p2 + m_minBB);
+                    mesh.AddPoint(p3 + m_minBB);
+                    mesh.AddPoint(p4 + m_minBB);
+                    mesh.AddPoint(p5 + m_minBB);
+                    mesh.AddPoint(p6 + m_minBB);
+                    mesh.AddPoint(p7 + m_minBB);
+                    mesh.AddTriangle(Vec3<int32_t>(s + 0, s + 2, s + 1));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 0, s + 3, s + 2));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 5, s + 6));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 6, s + 7));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 6, s + 2));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 2, s + 3));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 1, s + 5));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 4, s + 0, s + 1));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 6, s + 5, s + 1));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 6, s + 1, s + 2));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 0, s + 4));
+                    mesh.AddTriangle(Vec3<int32_t>(s + 7, s + 3, s + 0));
+                }
+            }
+        }
+    }
+}
+void Volume::Convert(VoxelSet& vset) const
+{
+    for (int32_t h = 0; h < 3; ++h) {
+        vset.m_minBB[h] = m_minBB[h];
+    }
+    vset.m_voxels.Allocate(m_numVoxelsInsideSurface + m_numVoxelsOnSurface);
+    vset.m_scale = m_scale;
+    vset.m_unitVolume = m_scale * m_scale * m_scale;
+    const short i0 = (short)m_dim[0];
+    const short j0 = (short)m_dim[1];
+    const short k0 = (short)m_dim[2];
+    Voxel voxel;
+    vset.m_numVoxelsOnSurface = 0;
+    vset.m_numVoxelsInsideSurface = 0;
+    for (short i = 0; i < i0; ++i) {
+        for (short j = 0; j < j0; ++j) {
+            for (short k = 0; k < k0; ++k) {
+                const unsigned char& value = GetVoxel(i, j, k);
+                if (value == PRIMITIVE_INSIDE_SURFACE) {
+                    voxel.m_coord[0] = i;
+                    voxel.m_coord[1] = j;
+                    voxel.m_coord[2] = k;
+                    voxel.m_data = PRIMITIVE_INSIDE_SURFACE;
+                    vset.m_voxels.PushBack(voxel);
+                    ++vset.m_numVoxelsInsideSurface;
+                }
+                else if (value == PRIMITIVE_ON_SURFACE) {
+                    voxel.m_coord[0] = i;
+                    voxel.m_coord[1] = j;
+                    voxel.m_coord[2] = k;
+                    voxel.m_data = PRIMITIVE_ON_SURFACE;
+                    vset.m_voxels.PushBack(voxel);
+                    ++vset.m_numVoxelsOnSurface;
+                }
+            }
+        }
+    }
+}
+
+void Volume::Convert(TetrahedronSet& tset) const
+{
+    tset.m_tetrahedra.Allocate(5 * (m_numVoxelsInsideSurface + m_numVoxelsOnSurface));
+    tset.m_scale = m_scale;
+    const short i0 = (short)m_dim[0];
+    const short j0 = (short)m_dim[1];
+    const short k0 = (short)m_dim[2];
+    tset.m_numTetrahedraOnSurface = 0;
+    tset.m_numTetrahedraInsideSurface = 0;
+    Tetrahedron tetrahedron;
+    for (short i = 0; i < i0; ++i) {
+        for (short j = 0; j < j0; ++j) {
+            for (short k = 0; k < k0; ++k) {
+                const unsigned char& value = GetVoxel(i, j, k);
+                if (value == PRIMITIVE_INSIDE_SURFACE || value == PRIMITIVE_ON_SURFACE) {
+                    tetrahedron.m_data = value;
+                    Vec3<double> p1((i - 0.5) * m_scale + m_minBB[0], (j - 0.5) * m_scale + m_minBB[1], (k - 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p2((i + 0.5) * m_scale + m_minBB[0], (j - 0.5) * m_scale + m_minBB[1], (k - 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p3((i + 0.5) * m_scale + m_minBB[0], (j + 0.5) * m_scale + m_minBB[1], (k - 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p4((i - 0.5) * m_scale + m_minBB[0], (j + 0.5) * m_scale + m_minBB[1], (k - 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p5((i - 0.5) * m_scale + m_minBB[0], (j - 0.5) * m_scale + m_minBB[1], (k + 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p6((i + 0.5) * m_scale + m_minBB[0], (j - 0.5) * m_scale + m_minBB[1], (k + 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p7((i + 0.5) * m_scale + m_minBB[0], (j + 0.5) * m_scale + m_minBB[1], (k + 0.5) * m_scale + m_minBB[2]);
+                    Vec3<double> p8((i - 0.5) * m_scale + m_minBB[0], (j + 0.5) * m_scale + m_minBB[1], (k + 0.5) * m_scale + m_minBB[2]);
+
+                    tetrahedron.m_pts[0] = p2;
+                    tetrahedron.m_pts[1] = p4;
+                    tetrahedron.m_pts[2] = p7;
+                    tetrahedron.m_pts[3] = p5;
+                    tset.m_tetrahedra.PushBack(tetrahedron);
+
+                    tetrahedron.m_pts[0] = p6;
+                    tetrahedron.m_pts[1] = p2;
+                    tetrahedron.m_pts[2] = p7;
+                    tetrahedron.m_pts[3] = p5;
+                    tset.m_tetrahedra.PushBack(tetrahedron);
+
+                    tetrahedron.m_pts[0] = p3;
+                    tetrahedron.m_pts[1] = p4;
+                    tetrahedron.m_pts[2] = p7;
+                    tetrahedron.m_pts[3] = p2;
+                    tset.m_tetrahedra.PushBack(tetrahedron);
+
+                    tetrahedron.m_pts[0] = p1;
+                    tetrahedron.m_pts[1] = p4;
+                    tetrahedron.m_pts[2] = p2;
+                    tetrahedron.m_pts[3] = p5;
+                    tset.m_tetrahedra.PushBack(tetrahedron);
+
+                    tetrahedron.m_pts[0] = p8;
+                    tetrahedron.m_pts[1] = p5;
+                    tetrahedron.m_pts[2] = p7;
+                    tetrahedron.m_pts[3] = p4;
+                    tset.m_tetrahedra.PushBack(tetrahedron);
+                    if (value == PRIMITIVE_INSIDE_SURFACE) {
+                        tset.m_numTetrahedraInsideSurface += 5;
+                    }
+                    else {
+                        tset.m_numTetrahedraOnSurface += 5;
+                    }
+                }
+            }
+        }
+    }
+}
+
+void Volume::AlignToPrincipalAxes(double (&rot)[3][3]) const
+{
+    const short i0 = (short)m_dim[0];
+    const short j0 = (short)m_dim[1];
+    const short k0 = (short)m_dim[2];
+    Vec3<double> barycenter(0.0);
+    size_t nVoxels = 0;
+    for (short i = 0; i < i0; ++i) {
+        for (short j = 0; j < j0; ++j) {
+            for (short k = 0; k < k0; ++k) {
+                const unsigned char& value = GetVoxel(i, j, k);
+                if (value == PRIMITIVE_INSIDE_SURFACE || value == PRIMITIVE_ON_SURFACE) {
+                    barycenter[0] += i;
+                    barycenter[1] += j;
+                    barycenter[2] += k;
+                    ++nVoxels;
+                }
+            }
+        }
+    }
+    barycenter /= (double)nVoxels;
+
+    double covMat[3][3] = { { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 } };
+    double x, y, z;
+    for (short i = 0; i < i0; ++i) {
+        for (short j = 0; j < j0; ++j) {
+            for (short k = 0; k < k0; ++k) {
+                const unsigned char& value = GetVoxel(i, j, k);
+                if (value == PRIMITIVE_INSIDE_SURFACE || value == PRIMITIVE_ON_SURFACE) {
+                    x = i - barycenter[0];
+                    y = j - barycenter[1];
+                    z = k - barycenter[2];
+                    covMat[0][0] += x * x;
+                    covMat[1][1] += y * y;
+                    covMat[2][2] += z * z;
+                    covMat[0][1] += x * y;
+                    covMat[0][2] += x * z;
+                    covMat[1][2] += y * z;
+                }
+            }
+        }
+    }
+    covMat[1][0] = covMat[0][1];
+    covMat[2][0] = covMat[0][2];
+    covMat[2][1] = covMat[1][2];
+    double D[3][3];
+    Diagonalize(covMat, rot, D);
+}
+TetrahedronSet::TetrahedronSet()
+{
+    m_minBB[0] = m_minBB[1] = m_minBB[2] = 0.0;
+    m_maxBB[0] = m_maxBB[1] = m_maxBB[2] = 1.0;
+    m_barycenter[0] = m_barycenter[1] = m_barycenter[2] = 0.0;
+    m_scale = 1.0;
+    m_numTetrahedraOnSurface = 0;
+    m_numTetrahedraInsideSurface = 0;
+    memset(m_Q, 0, sizeof(double) * 9);
+    memset(m_D, 0, sizeof(double) * 9);
+}
+TetrahedronSet::~TetrahedronSet(void)
+{
+}
+void TetrahedronSet::ComputeBB()
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+
+    for (int32_t h = 0; h < 3; ++h) {
+        m_minBB[h] = m_maxBB[h] = m_tetrahedra[0].m_pts[0][h];
+        m_barycenter[h] = 0.0;
+    }
+    for (size_t p = 0; p < nTetrahedra; ++p) {
+        for (int32_t i = 0; i < 4; ++i) {
+            for (int32_t h = 0; h < 3; ++h) {
+                if (m_minBB[h] > m_tetrahedra[p].m_pts[i][h])
+                    m_minBB[h] = m_tetrahedra[p].m_pts[i][h];
+                if (m_maxBB[h] < m_tetrahedra[p].m_pts[i][h])
+                    m_maxBB[h] = m_tetrahedra[p].m_pts[i][h];
+                m_barycenter[h] += m_tetrahedra[p].m_pts[i][h];
+            }
+        }
+    }
+    m_barycenter /= (double)(4 * nTetrahedra);
+}
+void TetrahedronSet::ComputeConvexHull(Mesh& meshCH, const size_t sampling) const
+{
+    const size_t CLUSTER_SIZE = 65536;
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+
+    SArray<Vec3<double> > cpoints;
+
+    Vec3<double>* points = new Vec3<double>[CLUSTER_SIZE];
+    size_t p = 0;
+    while (p < nTetrahedra) {
+        size_t q = 0;
+        size_t s = 0;
+        while (q < CLUSTER_SIZE && p < nTetrahedra) {
+            if (m_tetrahedra[p].m_data == PRIMITIVE_ON_SURFACE) {
+                ++s;
+                if (s == sampling) {
+                    s = 0;
+                    for (int32_t a = 0; a < 4; ++a) {
+                        points[q++] = m_tetrahedra[p].m_pts[a];
+                        for (int32_t xx = 0; xx < 3; ++xx) {
+                            assert(m_tetrahedra[p].m_pts[a][xx] + EPS >= m_minBB[xx]);
+                            assert(m_tetrahedra[p].m_pts[a][xx] <= m_maxBB[xx] + EPS);
+                        }
+                    }
+                }
+            }
+            ++p;
+        }
+        btConvexHullComputer ch;
+        ch.compute((double*)points, 3 * sizeof(double), (int32_t)q, -1.0, -1.0);
+        for (int32_t v = 0; v < ch.vertices.size(); v++) {
+            cpoints.PushBack(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+        }
+    }
+    delete[] points;
+
+    points = cpoints.Data();
+    btConvexHullComputer ch;
+    ch.compute((double*)points, 3 * sizeof(double), (int32_t)cpoints.Size(), -1.0, -1.0);
+    meshCH.ResizePoints(0);
+    meshCH.ResizeTriangles(0);
+    for (int32_t v = 0; v < ch.vertices.size(); v++) {
+        meshCH.AddPoint(Vec3<double>(ch.vertices[v].getX(), ch.vertices[v].getY(), ch.vertices[v].getZ()));
+    }
+    const int32_t nt = ch.faces.size();
+    for (int32_t t = 0; t < nt; ++t) {
+        const btConvexHullComputer::Edge* sourceEdge = &(ch.edges[ch.faces[t]]);
+        int32_t a = sourceEdge->getSourceVertex();
+        int32_t b = sourceEdge->getTargetVertex();
+        const btConvexHullComputer::Edge* edge = sourceEdge->getNextEdgeOfFace();
+        int32_t c = edge->getTargetVertex();
+        while (c != a) {
+            meshCH.AddTriangle(Vec3<int32_t>(a, b, c));
+            edge = edge->getNextEdgeOfFace();
+            b = c;
+            c = edge->getTargetVertex();
+        }
+    }
+}
+inline bool TetrahedronSet::Add(Tetrahedron& tetrahedron)
+{
+    double v = ComputeVolume4(tetrahedron.m_pts[0], tetrahedron.m_pts[1], tetrahedron.m_pts[2], tetrahedron.m_pts[3]);
+
+    const double EPS = 0.0000000001;
+    if (fabs(v) < EPS) {
+        return false;
+    }
+    else if (v < 0.0) {
+        Vec3<double> tmp = tetrahedron.m_pts[0];
+        tetrahedron.m_pts[0] = tetrahedron.m_pts[1];
+        tetrahedron.m_pts[1] = tmp;
+    }
+
+    for (int32_t a = 0; a < 4; ++a) {
+        for (int32_t xx = 0; xx < 3; ++xx) {
+            assert(tetrahedron.m_pts[a][xx] + EPS >= m_minBB[xx]);
+            assert(tetrahedron.m_pts[a][xx] <= m_maxBB[xx] + EPS);
+        }
+    }
+    m_tetrahedra.PushBack(tetrahedron);
+    return true;
+}
+
+void TetrahedronSet::AddClippedTetrahedra(const Vec3<double> (&pts)[10], const int32_t nPts)
+{
+    const int32_t tetF[4][3] = { { 0, 1, 2 }, { 2, 1, 3 }, { 3, 1, 0 }, { 3, 0, 2 } };
+    if (nPts < 4) {
+        return;
+    }
+    else if (nPts == 4) {
+        Tetrahedron tetrahedron;
+        tetrahedron.m_data = PRIMITIVE_ON_SURFACE;
+        tetrahedron.m_pts[0] = pts[0];
+        tetrahedron.m_pts[1] = pts[1];
+        tetrahedron.m_pts[2] = pts[2];
+        tetrahedron.m_pts[3] = pts[3];
+        if (Add(tetrahedron)) {
+            ++m_numTetrahedraOnSurface;
+        }
+    }
+    else if (nPts == 5) {
+        const int32_t tet[15][4] = {
+            { 0, 1, 2, 3 }, { 1, 2, 3, 4 }, { 0, 2, 3, 4 }, { 0, 1, 3, 4 }, { 0, 1, 2, 4 },
+        };
+        const int32_t rem[5] = { 4, 0, 1, 2, 3 };
+        double maxVol = 0.0;
+        int32_t h0 = -1;
+        Tetrahedron tetrahedron0;
+        tetrahedron0.m_data = PRIMITIVE_ON_SURFACE;
+        for (int32_t h = 0; h < 5; ++h) {
+            double v = ComputeVolume4(pts[tet[h][0]], pts[tet[h][1]], pts[tet[h][2]], pts[tet[h][3]]);
+            if (v > maxVol) {
+                h0 = h;
+                tetrahedron0.m_pts[0] = pts[tet[h][0]];
+                tetrahedron0.m_pts[1] = pts[tet[h][1]];
+                tetrahedron0.m_pts[2] = pts[tet[h][2]];
+                tetrahedron0.m_pts[3] = pts[tet[h][3]];
+                maxVol = v;
+            }
+            else if (-v > maxVol) {
+                h0 = h;
+                tetrahedron0.m_pts[0] = pts[tet[h][1]];
+                tetrahedron0.m_pts[1] = pts[tet[h][0]];
+                tetrahedron0.m_pts[2] = pts[tet[h][2]];
+                tetrahedron0.m_pts[3] = pts[tet[h][3]];
+                maxVol = -v;
+            }
+        }
+        if (h0 == -1)
+            return;
+        if (Add(tetrahedron0)) {
+            ++m_numTetrahedraOnSurface;
+        }
+        else {
+            return;
+        }
+        int32_t a = rem[h0];
+        maxVol = 0.0;
+        int32_t h1 = -1;
+        Tetrahedron tetrahedron1;
+        tetrahedron1.m_data = PRIMITIVE_ON_SURFACE;
+        for (int32_t h = 0; h < 4; ++h) {
+            double v = ComputeVolume4(pts[a], tetrahedron0.m_pts[tetF[h][0]], tetrahedron0.m_pts[tetF[h][1]], tetrahedron0.m_pts[tetF[h][2]]);
+            if (v > maxVol) {
+                h1 = h;
+                tetrahedron1.m_pts[0] = pts[a];
+                tetrahedron1.m_pts[1] = tetrahedron0.m_pts[tetF[h][0]];
+                tetrahedron1.m_pts[2] = tetrahedron0.m_pts[tetF[h][1]];
+                tetrahedron1.m_pts[3] = tetrahedron0.m_pts[tetF[h][2]];
+                maxVol = v;
+            }
+        }
+        if (h1 == -1 && Add(tetrahedron1)) {
+            ++m_numTetrahedraOnSurface;
+        }
+    }
+    else if (nPts == 6) {
+
+        const int32_t tet[15][4] = { { 2, 3, 4, 5 }, { 1, 3, 4, 5 }, { 1, 2, 4, 5 }, { 1, 2, 3, 5 }, { 1, 2, 3, 4 },
+            { 0, 3, 4, 5 }, { 0, 2, 4, 5 }, { 0, 2, 3, 5 }, { 0, 2, 3, 4 }, { 0, 1, 4, 5 },
+            { 0, 1, 3, 5 }, { 0, 1, 3, 4 }, { 0, 1, 2, 5 }, { 0, 1, 2, 4 }, { 0, 1, 2, 3 } };
+        const int32_t rem[15][2] = { { 0, 1 }, { 0, 2 }, { 0, 3 }, { 0, 4 }, { 0, 5 },
+            { 1, 2 }, { 1, 3 }, { 1, 4 }, { 1, 5 }, { 2, 3 },
+            { 2, 4 }, { 2, 5 }, { 3, 4 }, { 3, 5 }, { 4, 5 } };
+        double maxVol = 0.0;
+        int32_t h0 = -1;
+        Tetrahedron tetrahedron0;
+        tetrahedron0.m_data = PRIMITIVE_ON_SURFACE;
+        for (int32_t h = 0; h < 15; ++h) {
+            double v = ComputeVolume4(pts[tet[h][0]], pts[tet[h][1]], pts[tet[h][2]], pts[tet[h][3]]);
+            if (v > maxVol) {
+                h0 = h;
+                tetrahedron0.m_pts[0] = pts[tet[h][0]];
+                tetrahedron0.m_pts[1] = pts[tet[h][1]];
+                tetrahedron0.m_pts[2] = pts[tet[h][2]];
+                tetrahedron0.m_pts[3] = pts[tet[h][3]];
+                maxVol = v;
+            }
+            else if (-v > maxVol) {
+                h0 = h;
+                tetrahedron0.m_pts[0] = pts[tet[h][1]];
+                tetrahedron0.m_pts[1] = pts[tet[h][0]];
+                tetrahedron0.m_pts[2] = pts[tet[h][2]];
+                tetrahedron0.m_pts[3] = pts[tet[h][3]];
+                maxVol = -v;
+            }
+        }
+        if (h0 == -1)
+            return;
+        if (Add(tetrahedron0)) {
+            ++m_numTetrahedraOnSurface;
+        }
+        else {
+            return;
+        }
+
+        int32_t a0 = rem[h0][0];
+        int32_t a1 = rem[h0][1];
+        int32_t h1 = -1;
+        Tetrahedron tetrahedron1;
+        tetrahedron1.m_data = PRIMITIVE_ON_SURFACE;
+        maxVol = 0.0;
+        for (int32_t h = 0; h < 4; ++h) {
+            double v = ComputeVolume4(pts[a0], tetrahedron0.m_pts[tetF[h][0]], tetrahedron0.m_pts[tetF[h][1]], tetrahedron0.m_pts[tetF[h][2]]);
+            if (v > maxVol) {
+                h1 = h;
+                tetrahedron1.m_pts[0] = pts[a0];
+                tetrahedron1.m_pts[1] = tetrahedron0.m_pts[tetF[h][0]];
+                tetrahedron1.m_pts[2] = tetrahedron0.m_pts[tetF[h][1]];
+                tetrahedron1.m_pts[3] = tetrahedron0.m_pts[tetF[h][2]];
+                maxVol = v;
+            }
+        }
+        if (h1 != -1 && Add(tetrahedron1)) {
+            ++m_numTetrahedraOnSurface;
+        }
+        else {
+            h1 = -1;
+        }
+        maxVol = 0.0;
+        int32_t h2 = -1;
+        Tetrahedron tetrahedron2;
+        tetrahedron2.m_data = PRIMITIVE_ON_SURFACE;
+        for (int32_t h = 0; h < 4; ++h) {
+            double v = ComputeVolume4(pts[a0], tetrahedron0.m_pts[tetF[h][0]], tetrahedron0.m_pts[tetF[h][1]], tetrahedron0.m_pts[tetF[h][2]]);
+            if (h == h1)
+                continue;
+            if (v > maxVol) {
+                h2 = h;
+                tetrahedron2.m_pts[0] = pts[a1];
+                tetrahedron2.m_pts[1] = tetrahedron0.m_pts[tetF[h][0]];
+                tetrahedron2.m_pts[2] = tetrahedron0.m_pts[tetF[h][1]];
+                tetrahedron2.m_pts[3] = tetrahedron0.m_pts[tetF[h][2]];
+                maxVol = v;
+            }
+        }
+        if (h1 != -1) {
+            for (int32_t h = 0; h < 4; ++h) {
+                double v = ComputeVolume4(pts[a1], tetrahedron1.m_pts[tetF[h][0]], tetrahedron1.m_pts[tetF[h][1]], tetrahedron1.m_pts[tetF[h][2]]);
+                if (h == 1)
+                    continue;
+                if (v > maxVol) {
+                    h2 = h;
+                    tetrahedron2.m_pts[0] = pts[a1];
+                    tetrahedron2.m_pts[1] = tetrahedron1.m_pts[tetF[h][0]];
+                    tetrahedron2.m_pts[2] = tetrahedron1.m_pts[tetF[h][1]];
+                    tetrahedron2.m_pts[3] = tetrahedron1.m_pts[tetF[h][2]];
+                    maxVol = v;
+                }
+            }
+        }
+        if (h2 != -1 && Add(tetrahedron2)) {
+            ++m_numTetrahedraOnSurface;
+        }
+    }
+    else {
+        assert(0);
+    }
+}
+
+void TetrahedronSet::Intersect(const Plane& plane,
+    SArray<Vec3<double> >* const positivePts,
+    SArray<Vec3<double> >* const negativePts,
+    const size_t sampling) const
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+}
+void TetrahedronSet::ComputeExteriorPoints(const Plane& plane,
+    const Mesh& mesh,
+    SArray<Vec3<double> >* const exteriorPts) const
+{
+}
+void TetrahedronSet::ComputeClippedVolumes(const Plane& plane,
+    double& positiveVolume,
+    double& negativeVolume) const
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+}
+
+void TetrahedronSet::SelectOnSurface(PrimitiveSet* const onSurfP) const
+{
+    TetrahedronSet* const onSurf = (TetrahedronSet*)onSurfP;
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    onSurf->m_tetrahedra.Resize(0);
+    onSurf->m_scale = m_scale;
+    onSurf->m_numTetrahedraOnSurface = 0;
+    onSurf->m_numTetrahedraInsideSurface = 0;
+    onSurf->m_barycenter = m_barycenter;
+    onSurf->m_minBB = m_minBB;
+    onSurf->m_maxBB = m_maxBB;
+    for (int32_t i = 0; i < 3; ++i) {
+        for (int32_t j = 0; j < 3; ++j) {
+            onSurf->m_Q[i][j] = m_Q[i][j];
+            onSurf->m_D[i][j] = m_D[i][j];
+        }
+    }
+    Tetrahedron tetrahedron;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        tetrahedron = m_tetrahedra[v];
+        if (tetrahedron.m_data == PRIMITIVE_ON_SURFACE) {
+            onSurf->m_tetrahedra.PushBack(tetrahedron);
+            ++onSurf->m_numTetrahedraOnSurface;
+        }
+    }
+}
+void TetrahedronSet::Clip(const Plane& plane,
+    PrimitiveSet* const positivePartP,
+    PrimitiveSet* const negativePartP) const
+{
+    TetrahedronSet* const positivePart = (TetrahedronSet*)positivePartP;
+    TetrahedronSet* const negativePart = (TetrahedronSet*)negativePartP;
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    positivePart->m_tetrahedra.Resize(0);
+    negativePart->m_tetrahedra.Resize(0);
+    positivePart->m_tetrahedra.Allocate(nTetrahedra);
+    negativePart->m_tetrahedra.Allocate(nTetrahedra);
+    negativePart->m_scale = positivePart->m_scale = m_scale;
+    negativePart->m_numTetrahedraOnSurface = positivePart->m_numTetrahedraOnSurface = 0;
+    negativePart->m_numTetrahedraInsideSurface = positivePart->m_numTetrahedraInsideSurface = 0;
+    negativePart->m_barycenter = m_barycenter;
+    positivePart->m_barycenter = m_barycenter;
+    negativePart->m_minBB = m_minBB;
+    positivePart->m_minBB = m_minBB;
+    negativePart->m_maxBB = m_maxBB;
+    positivePart->m_maxBB = m_maxBB;
+    for (int32_t i = 0; i < 3; ++i) {
+        for (int32_t j = 0; j < 3; ++j) {
+            negativePart->m_Q[i][j] = positivePart->m_Q[i][j] = m_Q[i][j];
+            negativePart->m_D[i][j] = positivePart->m_D[i][j] = m_D[i][j];
+        }
+    }
+
+    Tetrahedron tetrahedron;
+    double delta, alpha;
+    int32_t sign[4];
+    int32_t npos, nneg;
+    Vec3<double> posPts[10];
+    Vec3<double> negPts[10];
+    Vec3<double> P0, P1, M;
+    const Vec3<double> n(plane.m_a, plane.m_b, plane.m_c);
+    const int32_t edges[6][2] = { { 0, 1 }, { 0, 2 }, { 0, 3 }, { 1, 2 }, { 1, 3 }, { 2, 3 } };
+    double dist;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        tetrahedron = m_tetrahedra[v];
+        npos = nneg = 0;
+        for (int32_t i = 0; i < 4; ++i) {
+            dist = plane.m_a * tetrahedron.m_pts[i][0] + plane.m_b * tetrahedron.m_pts[i][1] + plane.m_c * tetrahedron.m_pts[i][2] + plane.m_d;
+            if (dist > 0.0) {
+                sign[i] = 1;
+                posPts[npos] = tetrahedron.m_pts[i];
+                ++npos;
+            }
+            else {
+                sign[i] = -1;
+                negPts[nneg] = tetrahedron.m_pts[i];
+                ++nneg;
+            }
+        }
+
+        if (npos == 4) {
+            positivePart->Add(tetrahedron);
+            if (tetrahedron.m_data == PRIMITIVE_ON_SURFACE) {
+                ++positivePart->m_numTetrahedraOnSurface;
+            }
+            else {
+                ++positivePart->m_numTetrahedraInsideSurface;
+            }
+        }
+        else if (nneg == 4) {
+            negativePart->Add(tetrahedron);
+            if (tetrahedron.m_data == PRIMITIVE_ON_SURFACE) {
+                ++negativePart->m_numTetrahedraOnSurface;
+            }
+            else {
+                ++negativePart->m_numTetrahedraInsideSurface;
+            }
+        }
+        else {
+            int32_t nnew = 0;
+            for (int32_t j = 0; j < 6; ++j) {
+                if (sign[edges[j][0]] * sign[edges[j][1]] == -1) {
+                    P0 = tetrahedron.m_pts[edges[j][0]];
+                    P1 = tetrahedron.m_pts[edges[j][1]];
+                    delta = (P0 - P1) * n;
+                    alpha = -(plane.m_d + (n * P1)) / delta;
+                    assert(alpha >= 0.0 && alpha <= 1.0);
+                    M = alpha * P0 + (1 - alpha) * P1;
+                    for (int32_t xx = 0; xx < 3; ++xx) {
+                        assert(M[xx] + EPS >= m_minBB[xx]);
+                        assert(M[xx] <= m_maxBB[xx] + EPS);
+                    }
+                    posPts[npos++] = M;
+                    negPts[nneg++] = M;
+                    ++nnew;
+                }
+            }
+            negativePart->AddClippedTetrahedra(negPts, nneg);
+            positivePart->AddClippedTetrahedra(posPts, npos);
+        }
+    }
+}
+void TetrahedronSet::Convert(Mesh& mesh, const VOXEL_VALUE value) const
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        const Tetrahedron& tetrahedron = m_tetrahedra[v];
+        if (tetrahedron.m_data == value) {
+            int32_t s = (int32_t)mesh.GetNPoints();
+            mesh.AddPoint(tetrahedron.m_pts[0]);
+            mesh.AddPoint(tetrahedron.m_pts[1]);
+            mesh.AddPoint(tetrahedron.m_pts[2]);
+            mesh.AddPoint(tetrahedron.m_pts[3]);
+            mesh.AddTriangle(Vec3<int32_t>(s + 0, s + 1, s + 2));
+            mesh.AddTriangle(Vec3<int32_t>(s + 2, s + 1, s + 3));
+            mesh.AddTriangle(Vec3<int32_t>(s + 3, s + 1, s + 0));
+            mesh.AddTriangle(Vec3<int32_t>(s + 3, s + 0, s + 2));
+        }
+    }
+}
+const double TetrahedronSet::ComputeVolume() const
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return 0.0;
+    double volume = 0.0;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        const Tetrahedron& tetrahedron = m_tetrahedra[v];
+        volume += fabs(ComputeVolume4(tetrahedron.m_pts[0], tetrahedron.m_pts[1], tetrahedron.m_pts[2], tetrahedron.m_pts[3]));
+    }
+    return volume / 6.0;
+}
+const double TetrahedronSet::ComputeMaxVolumeError() const
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return 0.0;
+    double volume = 0.0;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        const Tetrahedron& tetrahedron = m_tetrahedra[v];
+        if (tetrahedron.m_data == PRIMITIVE_ON_SURFACE) {
+            volume += fabs(ComputeVolume4(tetrahedron.m_pts[0], tetrahedron.m_pts[1], tetrahedron.m_pts[2], tetrahedron.m_pts[3]));
+        }
+    }
+    return volume / 6.0;
+}
+void TetrahedronSet::RevertAlignToPrincipalAxes()
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    double x, y, z;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        Tetrahedron& tetrahedron = m_tetrahedra[v];
+        for (int32_t i = 0; i < 4; ++i) {
+            x = tetrahedron.m_pts[i][0] - m_barycenter[0];
+            y = tetrahedron.m_pts[i][1] - m_barycenter[1];
+            z = tetrahedron.m_pts[i][2] - m_barycenter[2];
+            tetrahedron.m_pts[i][0] = m_Q[0][0] * x + m_Q[0][1] * y + m_Q[0][2] * z + m_barycenter[0];
+            tetrahedron.m_pts[i][1] = m_Q[1][0] * x + m_Q[1][1] * y + m_Q[1][2] * z + m_barycenter[1];
+            tetrahedron.m_pts[i][2] = m_Q[2][0] * x + m_Q[2][1] * y + m_Q[2][2] * z + m_barycenter[2];
+        }
+    }
+    ComputeBB();
+}
+void TetrahedronSet::ComputePrincipalAxes()
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    double covMat[3][3] = { { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 },
+        { 0.0, 0.0, 0.0 } };
+    double x, y, z;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        Tetrahedron& tetrahedron = m_tetrahedra[v];
+        for (int32_t i = 0; i < 4; ++i) {
+            x = tetrahedron.m_pts[i][0] - m_barycenter[0];
+            y = tetrahedron.m_pts[i][1] - m_barycenter[1];
+            z = tetrahedron.m_pts[i][2] - m_barycenter[2];
+            covMat[0][0] += x * x;
+            covMat[1][1] += y * y;
+            covMat[2][2] += z * z;
+            covMat[0][1] += x * y;
+            covMat[0][2] += x * z;
+            covMat[1][2] += y * z;
+        }
+    }
+    double n = nTetrahedra * 4.0;
+    covMat[0][0] /= n;
+    covMat[1][1] /= n;
+    covMat[2][2] /= n;
+    covMat[0][1] /= n;
+    covMat[0][2] /= n;
+    covMat[1][2] /= n;
+    covMat[1][0] = covMat[0][1];
+    covMat[2][0] = covMat[0][2];
+    covMat[2][1] = covMat[1][2];
+    Diagonalize(covMat, m_Q, m_D);
+}
+void TetrahedronSet::AlignToPrincipalAxes()
+{
+    const size_t nTetrahedra = m_tetrahedra.Size();
+    if (nTetrahedra == 0)
+        return;
+    double x, y, z;
+    for (size_t v = 0; v < nTetrahedra; ++v) {
+        Tetrahedron& tetrahedron = m_tetrahedra[v];
+        for (int32_t i = 0; i < 4; ++i) {
+            x = tetrahedron.m_pts[i][0] - m_barycenter[0];
+            y = tetrahedron.m_pts[i][1] - m_barycenter[1];
+            z = tetrahedron.m_pts[i][2] - m_barycenter[2];
+            tetrahedron.m_pts[i][0] = m_Q[0][0] * x + m_Q[1][0] * y + m_Q[2][0] * z + m_barycenter[0];
+            tetrahedron.m_pts[i][1] = m_Q[0][1] * x + m_Q[1][1] * y + m_Q[2][1] * z + m_barycenter[1];
+            tetrahedron.m_pts[i][2] = m_Q[0][2] * x + m_Q[1][2] * y + m_Q[2][2] * z + m_barycenter[2];
+        }
+    }
+    ComputeBB();
+}
+}