//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
static const char* satClipKernelsCL =
	"#define TRIANGLE_NUM_CONVEX_FACES 5\n"
	"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
	"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
	"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
	"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
	"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
	"#ifdef cl_ext_atomic_counters_32\n"
	"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
	"#else\n"
	"#define counter32_t volatile __global int*\n"
	"#endif\n"
	"#define GET_GROUP_IDX get_group_id(0)\n"
	"#define GET_LOCAL_IDX get_local_id(0)\n"
	"#define GET_GLOBAL_IDX get_global_id(0)\n"
	"#define GET_GROUP_SIZE get_local_size(0)\n"
	"#define GET_NUM_GROUPS get_num_groups(0)\n"
	"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
	"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
	"#define AtomInc(x) atom_inc(&(x))\n"
	"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
	"#define AppendInc(x, out) out = atomic_inc(x)\n"
	"#define AtomAdd(x, value) atom_add(&(x), value)\n"
	"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
	"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
	"#define max2 max\n"
	"#define min2 min\n"
	"typedef unsigned int u32;\n"
	"#ifndef B3_CONTACT4DATA_H\n"
	"#define B3_CONTACT4DATA_H\n"
	"#ifndef B3_FLOAT4_H\n"
	"#define B3_FLOAT4_H\n"
	"#ifndef B3_PLATFORM_DEFINITIONS_H\n"
	"#define B3_PLATFORM_DEFINITIONS_H\n"
	"struct MyTest\n"
	"{\n"
	"	int bla;\n"
	"};\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
	"#define B3_LARGE_FLOAT 1e18f\n"
	"#define B3_INFINITY 1e18f\n"
	"#define b3Assert(a)\n"
	"#define b3ConstArray(a) __global const a*\n"
	"#define b3AtomicInc atomic_inc\n"
	"#define b3AtomicAdd atomic_add\n"
	"#define b3Fabs fabs\n"
	"#define b3Sqrt native_sqrt\n"
	"#define b3Sin native_sin\n"
	"#define b3Cos native_cos\n"
	"#define B3_STATIC\n"
	"#endif\n"
	"#endif\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"	typedef float4	b3Float4;\n"
	"	#define b3Float4ConstArg const b3Float4\n"
	"	#define b3MakeFloat4 (float4)\n"
	"	float b3Dot3F4(b3Float4ConstArg v0,b3Float4ConstArg v1)\n"
	"	{\n"
	"		float4 a1 = b3MakeFloat4(v0.xyz,0.f);\n"
	"		float4 b1 = b3MakeFloat4(v1.xyz,0.f);\n"
	"		return dot(a1, b1);\n"
	"	}\n"
	"	b3Float4 b3Cross3(b3Float4ConstArg v0,b3Float4ConstArg v1)\n"
	"	{\n"
	"		float4 a1 = b3MakeFloat4(v0.xyz,0.f);\n"
	"		float4 b1 = b3MakeFloat4(v1.xyz,0.f);\n"
	"		return cross(a1, b1);\n"
	"	}\n"
	"	#define b3MinFloat4 min\n"
	"	#define b3MaxFloat4 max\n"
	"	#define b3Normalized(a) normalize(a)\n"
	"#endif \n"
	"		\n"
	"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
	"{\n"
	"	if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6)	\n"
	"		return false;\n"
	"	return true;\n"
	"}\n"
	"inline int    b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
	"{\n"
	"    float maxDot = -B3_INFINITY;\n"
	"    int i = 0;\n"
	"    int ptIndex = -1;\n"
	"    for( i = 0; i < vecLen; i++ )\n"
	"    {\n"
	"        float dot = b3Dot3F4(vecArray[i],vec);\n"
	"            \n"
	"        if( dot > maxDot )\n"
	"        {\n"
	"            maxDot = dot;\n"
	"            ptIndex = i;\n"
	"        }\n"
	"    }\n"
	"	b3Assert(ptIndex>=0);\n"
	"    if (ptIndex<0)\n"
	"	{\n"
	"		ptIndex = 0;\n"
	"	}\n"
	"    *dotOut = maxDot;\n"
	"    return ptIndex;\n"
	"}\n"
	"#endif //B3_FLOAT4_H\n"
	"typedef  struct b3Contact4Data b3Contact4Data_t;\n"
	"struct b3Contact4Data\n"
	"{\n"
	"	b3Float4	m_worldPosB[4];\n"
	"//	b3Float4	m_localPosA[4];\n"
	"//	b3Float4	m_localPosB[4];\n"
	"	b3Float4	m_worldNormalOnB;	//	w: m_nPoints\n"
	"	unsigned short  m_restituitionCoeffCmp;\n"
	"	unsigned short  m_frictionCoeffCmp;\n"
	"	int m_batchIdx;\n"
	"	int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
	"	int m_bodyBPtrAndSignBit;\n"
	"	int	m_childIndexA;\n"
	"	int	m_childIndexB;\n"
	"	int m_unused1;\n"
	"	int m_unused2;\n"
	"};\n"
	"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
	"{\n"
	"	return (int)contact->m_worldNormalOnB.w;\n"
	"};\n"
	"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
	"{\n"
	"	contact->m_worldNormalOnB.w = (float)numPoints;\n"
	"};\n"
	"#endif //B3_CONTACT4DATA_H\n"
	"#ifndef B3_CONVEX_POLYHEDRON_DATA_H\n"
	"#define B3_CONVEX_POLYHEDRON_DATA_H\n"
	"#ifndef B3_FLOAT4_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_FLOAT4_H\n"
	"#ifndef B3_QUAT_H\n"
	"#define B3_QUAT_H\n"
	"#ifndef B3_PLATFORM_DEFINITIONS_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif\n"
	"#endif\n"
	"#ifndef B3_FLOAT4_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_FLOAT4_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"	typedef float4	b3Quat;\n"
	"	#define b3QuatConstArg const b3Quat\n"
	"	\n"
	"	\n"
	"inline float4 b3FastNormalize4(float4 v)\n"
	"{\n"
	"	v = (float4)(v.xyz,0.f);\n"
	"	return fast_normalize(v);\n"
	"}\n"
	"	\n"
	"inline b3Quat b3QuatMul(b3Quat a, b3Quat b);\n"
	"inline b3Quat b3QuatNormalized(b3QuatConstArg in);\n"
	"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
	"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
	"inline b3Quat b3QuatInverse(b3QuatConstArg q);\n"
	"inline b3Quat b3QuatMul(b3QuatConstArg a, b3QuatConstArg b)\n"
	"{\n"
	"	b3Quat ans;\n"
	"	ans = b3Cross3( a, b );\n"
	"	ans += a.w*b+b.w*a;\n"
	"//	ans.w = a.w*b.w - (a.x*b.x+a.y*b.y+a.z*b.z);\n"
	"	ans.w = a.w*b.w - b3Dot3F4(a, b);\n"
	"	return ans;\n"
	"}\n"
	"inline b3Quat b3QuatNormalized(b3QuatConstArg in)\n"
	"{\n"
	"	b3Quat q;\n"
	"	q=in;\n"
	"	//return b3FastNormalize4(in);\n"
	"	float len = native_sqrt(dot(q, q));\n"
	"	if(len > 0.f)\n"
	"	{\n"
	"		q *= 1.f / len;\n"
	"	}\n"
	"	else\n"
	"	{\n"
	"		q.x = q.y = q.z = 0.f;\n"
	"		q.w = 1.f;\n"
	"	}\n"
	"	return q;\n"
	"}\n"
	"inline float4 b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec)\n"
	"{\n"
	"	b3Quat qInv = b3QuatInvert( q );\n"
	"	float4 vcpy = vec;\n"
	"	vcpy.w = 0.f;\n"
	"	float4 out = b3QuatMul(b3QuatMul(q,vcpy),qInv);\n"
	"	return out;\n"
	"}\n"
	"inline b3Quat b3QuatInverse(b3QuatConstArg q)\n"
	"{\n"
	"	return (b3Quat)(-q.xyz, q.w);\n"
	"}\n"
	"inline b3Quat b3QuatInvert(b3QuatConstArg q)\n"
	"{\n"
	"	return (b3Quat)(-q.xyz, q.w);\n"
	"}\n"
	"inline float4 b3QuatInvRotate(b3QuatConstArg q, b3QuatConstArg vec)\n"
	"{\n"
	"	return b3QuatRotate( b3QuatInvert( q ), vec );\n"
	"}\n"
	"inline b3Float4 b3TransformPoint(b3Float4ConstArg point, b3Float4ConstArg translation, b3QuatConstArg  orientation)\n"
	"{\n"
	"	return b3QuatRotate( orientation, point ) + (translation);\n"
	"}\n"
	"	\n"
	"#endif \n"
	"#endif //B3_QUAT_H\n"
	"typedef struct b3GpuFace b3GpuFace_t;\n"
	"struct b3GpuFace\n"
	"{\n"
	"	b3Float4 m_plane;\n"
	"	int m_indexOffset;\n"
	"	int m_numIndices;\n"
	"	int m_unusedPadding1;\n"
	"	int m_unusedPadding2;\n"
	"};\n"
	"typedef struct b3ConvexPolyhedronData b3ConvexPolyhedronData_t;\n"
	"struct b3ConvexPolyhedronData\n"
	"{\n"
	"	b3Float4		m_localCenter;\n"
	"	b3Float4		m_extents;\n"
	"	b3Float4		mC;\n"
	"	b3Float4		mE;\n"
	"	float			m_radius;\n"
	"	int	m_faceOffset;\n"
	"	int m_numFaces;\n"
	"	int	m_numVertices;\n"
	"	int m_vertexOffset;\n"
	"	int	m_uniqueEdgesOffset;\n"
	"	int	m_numUniqueEdges;\n"
	"	int m_unused;\n"
	"};\n"
	"#endif //B3_CONVEX_POLYHEDRON_DATA_H\n"
	"#ifndef B3_COLLIDABLE_H\n"
	"#define B3_COLLIDABLE_H\n"
	"#ifndef B3_FLOAT4_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_FLOAT4_H\n"
	"#ifndef B3_QUAT_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_QUAT_H\n"
	"enum b3ShapeTypes\n"
	"{\n"
	"	SHAPE_HEIGHT_FIELD=1,\n"
	"	SHAPE_CONVEX_HULL=3,\n"
	"	SHAPE_PLANE=4,\n"
	"	SHAPE_CONCAVE_TRIMESH=5,\n"
	"	SHAPE_COMPOUND_OF_CONVEX_HULLS=6,\n"
	"	SHAPE_SPHERE=7,\n"
	"	MAX_NUM_SHAPE_TYPES,\n"
	"};\n"
	"typedef struct b3Collidable b3Collidable_t;\n"
	"struct b3Collidable\n"
	"{\n"
	"	union {\n"
	"		int m_numChildShapes;\n"
	"		int m_bvhIndex;\n"
	"	};\n"
	"	union\n"
	"	{\n"
	"		float m_radius;\n"
	"		int	m_compoundBvhIndex;\n"
	"	};\n"
	"	int m_shapeType;\n"
	"	int m_shapeIndex;\n"
	"};\n"
	"typedef struct b3GpuChildShape b3GpuChildShape_t;\n"
	"struct b3GpuChildShape\n"
	"{\n"
	"	b3Float4	m_childPosition;\n"
	"	b3Quat		m_childOrientation;\n"
	"	int m_shapeIndex;\n"
	"	int m_unused0;\n"
	"	int m_unused1;\n"
	"	int m_unused2;\n"
	"};\n"
	"struct b3CompoundOverlappingPair\n"
	"{\n"
	"	int m_bodyIndexA;\n"
	"	int m_bodyIndexB;\n"
	"//	int	m_pairType;\n"
	"	int m_childShapeIndexA;\n"
	"	int m_childShapeIndexB;\n"
	"};\n"
	"#endif //B3_COLLIDABLE_H\n"
	"#ifndef B3_RIGIDBODY_DATA_H\n"
	"#define B3_RIGIDBODY_DATA_H\n"
	"#ifndef B3_FLOAT4_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_FLOAT4_H\n"
	"#ifndef B3_QUAT_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_QUAT_H\n"
	"#ifndef B3_MAT3x3_H\n"
	"#define B3_MAT3x3_H\n"
	"#ifndef B3_QUAT_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"#endif \n"
	"#endif //B3_QUAT_H\n"
	"#ifdef __cplusplus\n"
	"#else\n"
	"typedef struct\n"
	"{\n"
	"	b3Float4 m_row[3];\n"
	"}b3Mat3x3;\n"
	"#define b3Mat3x3ConstArg const b3Mat3x3\n"
	"#define b3GetRow(m,row) (m.m_row[row])\n"
	"inline b3Mat3x3 b3QuatGetRotationMatrix(b3Quat quat)\n"
	"{\n"
	"	b3Float4 quat2 = (b3Float4)(quat.x*quat.x, quat.y*quat.y, quat.z*quat.z, 0.f);\n"
	"	b3Mat3x3 out;\n"
	"	out.m_row[0].x=1-2*quat2.y-2*quat2.z;\n"
	"	out.m_row[0].y=2*quat.x*quat.y-2*quat.w*quat.z;\n"
	"	out.m_row[0].z=2*quat.x*quat.z+2*quat.w*quat.y;\n"
	"	out.m_row[0].w = 0.f;\n"
	"	out.m_row[1].x=2*quat.x*quat.y+2*quat.w*quat.z;\n"
	"	out.m_row[1].y=1-2*quat2.x-2*quat2.z;\n"
	"	out.m_row[1].z=2*quat.y*quat.z-2*quat.w*quat.x;\n"
	"	out.m_row[1].w = 0.f;\n"
	"	out.m_row[2].x=2*quat.x*quat.z-2*quat.w*quat.y;\n"
	"	out.m_row[2].y=2*quat.y*quat.z+2*quat.w*quat.x;\n"
	"	out.m_row[2].z=1-2*quat2.x-2*quat2.y;\n"
	"	out.m_row[2].w = 0.f;\n"
	"	return out;\n"
	"}\n"
	"inline b3Mat3x3 b3AbsoluteMat3x3(b3Mat3x3ConstArg matIn)\n"
	"{\n"
	"	b3Mat3x3 out;\n"
	"	out.m_row[0] = fabs(matIn.m_row[0]);\n"
	"	out.m_row[1] = fabs(matIn.m_row[1]);\n"
	"	out.m_row[2] = fabs(matIn.m_row[2]);\n"
	"	return out;\n"
	"}\n"
	"__inline\n"
	"b3Mat3x3 mtZero();\n"
	"__inline\n"
	"b3Mat3x3 mtIdentity();\n"
	"__inline\n"
	"b3Mat3x3 mtTranspose(b3Mat3x3 m);\n"
	"__inline\n"
	"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b);\n"
	"__inline\n"
	"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b);\n"
	"__inline\n"
	"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b);\n"
	"__inline\n"
	"b3Mat3x3 mtZero()\n"
	"{\n"
	"	b3Mat3x3 m;\n"
	"	m.m_row[0] = (b3Float4)(0.f);\n"
	"	m.m_row[1] = (b3Float4)(0.f);\n"
	"	m.m_row[2] = (b3Float4)(0.f);\n"
	"	return m;\n"
	"}\n"
	"__inline\n"
	"b3Mat3x3 mtIdentity()\n"
	"{\n"
	"	b3Mat3x3 m;\n"
	"	m.m_row[0] = (b3Float4)(1,0,0,0);\n"
	"	m.m_row[1] = (b3Float4)(0,1,0,0);\n"
	"	m.m_row[2] = (b3Float4)(0,0,1,0);\n"
	"	return m;\n"
	"}\n"
	"__inline\n"
	"b3Mat3x3 mtTranspose(b3Mat3x3 m)\n"
	"{\n"
	"	b3Mat3x3 out;\n"
	"	out.m_row[0] = (b3Float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
	"	out.m_row[1] = (b3Float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
	"	out.m_row[2] = (b3Float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
	"	return out;\n"
	"}\n"
	"__inline\n"
	"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b)\n"
	"{\n"
	"	b3Mat3x3 transB;\n"
	"	transB = mtTranspose( b );\n"
	"	b3Mat3x3 ans;\n"
	"	//	why this doesn't run when 0ing in the for{}\n"
	"	a.m_row[0].w = 0.f;\n"
	"	a.m_row[1].w = 0.f;\n"
	"	a.m_row[2].w = 0.f;\n"
	"	for(int i=0; i<3; i++)\n"
	"	{\n"
	"//	a.m_row[i].w = 0.f;\n"
	"		ans.m_row[i].x = b3Dot3F4(a.m_row[i],transB.m_row[0]);\n"
	"		ans.m_row[i].y = b3Dot3F4(a.m_row[i],transB.m_row[1]);\n"
	"		ans.m_row[i].z = b3Dot3F4(a.m_row[i],transB.m_row[2]);\n"
	"		ans.m_row[i].w = 0.f;\n"
	"	}\n"
	"	return ans;\n"
	"}\n"
	"__inline\n"
	"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b)\n"
	"{\n"
	"	b3Float4 ans;\n"
	"	ans.x = b3Dot3F4( a.m_row[0], b );\n"
	"	ans.y = b3Dot3F4( a.m_row[1], b );\n"
	"	ans.z = b3Dot3F4( a.m_row[2], b );\n"
	"	ans.w = 0.f;\n"
	"	return ans;\n"
	"}\n"
	"__inline\n"
	"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b)\n"
	"{\n"
	"	b3Float4 colx = b3MakeFloat4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
	"	b3Float4 coly = b3MakeFloat4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
	"	b3Float4 colz = b3MakeFloat4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
	"	b3Float4 ans;\n"
	"	ans.x = b3Dot3F4( a, colx );\n"
	"	ans.y = b3Dot3F4( a, coly );\n"
	"	ans.z = b3Dot3F4( a, colz );\n"
	"	return ans;\n"
	"}\n"
	"#endif\n"
	"#endif //B3_MAT3x3_H\n"
	"typedef struct b3RigidBodyData b3RigidBodyData_t;\n"
	"struct b3RigidBodyData\n"
	"{\n"
	"	b3Float4				m_pos;\n"
	"	b3Quat					m_quat;\n"
	"	b3Float4				m_linVel;\n"
	"	b3Float4				m_angVel;\n"
	"	int 					m_collidableIdx;\n"
	"	float 				m_invMass;\n"
	"	float 				m_restituitionCoeff;\n"
	"	float 				m_frictionCoeff;\n"
	"};\n"
	"typedef struct b3InertiaData b3InertiaData_t;\n"
	"struct b3InertiaData\n"
	"{\n"
	"	b3Mat3x3 m_invInertiaWorld;\n"
	"	b3Mat3x3 m_initInvInertia;\n"
	"};\n"
	"#endif //B3_RIGIDBODY_DATA_H\n"
	"	\n"
	"#define GET_NPOINTS(x) (x).m_worldNormalOnB.w\n"
	"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
	"#define make_float4 (float4)\n"
	"#define make_float2 (float2)\n"
	"#define make_uint4 (uint4)\n"
	"#define make_int4 (int4)\n"
	"#define make_uint2 (uint2)\n"
	"#define make_int2 (int2)\n"
	"__inline\n"
	"float fastDiv(float numerator, float denominator)\n"
	"{\n"
	"	return native_divide(numerator, denominator);	\n"
	"//	return numerator/denominator;	\n"
	"}\n"
	"__inline\n"
	"float4 fastDiv4(float4 numerator, float4 denominator)\n"
	"{\n"
	"	return native_divide(numerator, denominator);	\n"
	"}\n"
	"__inline\n"
	"float4 cross3(float4 a, float4 b)\n"
	"{\n"
	"	return cross(a,b);\n"
	"}\n"
	"//#define dot3F4 dot\n"
	"__inline\n"
	"float dot3F4(float4 a, float4 b)\n"
	"{\n"
	"	float4 a1 = make_float4(a.xyz,0.f);\n"
	"	float4 b1 = make_float4(b.xyz,0.f);\n"
	"	return dot(a1, b1);\n"
	"}\n"
	"__inline\n"
	"float4 fastNormalize4(float4 v)\n"
	"{\n"
	"	return fast_normalize(v);\n"
	"}\n"
	"///////////////////////////////////////\n"
	"//	Quaternion\n"
	"///////////////////////////////////////\n"
	"typedef float4 Quaternion;\n"
	"__inline\n"
	"Quaternion qtMul(Quaternion a, Quaternion b);\n"
	"__inline\n"
	"Quaternion qtNormalize(Quaternion in);\n"
	"__inline\n"
	"float4 qtRotate(Quaternion q, float4 vec);\n"
	"__inline\n"
	"Quaternion qtInvert(Quaternion q);\n"
	"__inline\n"
	"Quaternion qtMul(Quaternion a, Quaternion b)\n"
	"{\n"
	"	Quaternion ans;\n"
	"	ans = cross3( a, b );\n"
	"	ans += a.w*b+b.w*a;\n"
	"//	ans.w = a.w*b.w - (a.x*b.x+a.y*b.y+a.z*b.z);\n"
	"	ans.w = a.w*b.w - dot3F4(a, b);\n"
	"	return ans;\n"
	"}\n"
	"__inline\n"
	"Quaternion qtNormalize(Quaternion in)\n"
	"{\n"
	"	return fastNormalize4(in);\n"
	"//	in /= length( in );\n"
	"//	return in;\n"
	"}\n"
	"__inline\n"
	"float4 qtRotate(Quaternion q, float4 vec)\n"
	"{\n"
	"	Quaternion qInv = qtInvert( q );\n"
	"	float4 vcpy = vec;\n"
	"	vcpy.w = 0.f;\n"
	"	float4 out = qtMul(qtMul(q,vcpy),qInv);\n"
	"	return out;\n"
	"}\n"
	"__inline\n"
	"Quaternion qtInvert(Quaternion q)\n"
	"{\n"
	"	return (Quaternion)(-q.xyz, q.w);\n"
	"}\n"
	"__inline\n"
	"float4 qtInvRotate(const Quaternion q, float4 vec)\n"
	"{\n"
	"	return qtRotate( qtInvert( q ), vec );\n"
	"}\n"
	"__inline\n"
	"float4 transform(const float4* p, const float4* translation, const Quaternion* orientation)\n"
	"{\n"
	"	return qtRotate( *orientation, *p ) + (*translation);\n"
	"}\n"
	"__inline\n"
	"float4 normalize3(const float4 a)\n"
	"{\n"
	"	float4 n = make_float4(a.x, a.y, a.z, 0.f);\n"
	"	return fastNormalize4( n );\n"
	"}\n"
	"__inline float4 lerp3(const float4 a,const float4 b, float  t)\n"
	"{\n"
	"	return make_float4(	a.x + (b.x - a.x) * t,\n"
	"						a.y + (b.y - a.y) * t,\n"
	"						a.z + (b.z - a.z) * t,\n"
	"						0.f);\n"
	"}\n"
	"// Clips a face to the back of a plane, return the number of vertices out, stored in ppVtxOut\n"
	"int clipFaceGlobal(__global const float4* pVtxIn, int numVertsIn, float4 planeNormalWS,float planeEqWS, __global float4* ppVtxOut)\n"
	"{\n"
	"	\n"
	"	int ve;\n"
	"	float ds, de;\n"
	"	int numVertsOut = 0;\n"
	"    //double-check next test\n"
	"    	if (numVertsIn < 2)\n"
	"    		return 0;\n"
	"    \n"
	"	float4 firstVertex=pVtxIn[numVertsIn-1];\n"
	"	float4 endVertex = pVtxIn[0];\n"
	"	\n"
	"	ds = dot3F4(planeNormalWS,firstVertex)+planeEqWS;\n"
	"    \n"
	"	for (ve = 0; ve < numVertsIn; ve++)\n"
	"	{\n"
	"		endVertex=pVtxIn[ve];\n"
	"		de = dot3F4(planeNormalWS,endVertex)+planeEqWS;\n"
	"		if (ds<0)\n"
	"		{\n"
	"			if (de<0)\n"
	"			{\n"
	"				// Start < 0, end < 0, so output endVertex\n"
	"				ppVtxOut[numVertsOut++] = endVertex;\n"
	"			}\n"
	"			else\n"
	"			{\n"
	"				// Start < 0, end >= 0, so output intersection\n"
	"				ppVtxOut[numVertsOut++] = lerp3(firstVertex, endVertex,(ds * 1.f/(ds - de)) );\n"
	"			}\n"
	"		}\n"
	"		else\n"
	"		{\n"
	"			if (de<0)\n"
	"			{\n"
	"				// Start >= 0, end < 0 so output intersection and end\n"
	"				ppVtxOut[numVertsOut++] = lerp3(firstVertex, endVertex,(ds * 1.f/(ds - de)) );\n"
	"				ppVtxOut[numVertsOut++] = endVertex;\n"
	"			}\n"
	"		}\n"
	"		firstVertex = endVertex;\n"
	"		ds = de;\n"
	"	}\n"
	"	return numVertsOut;\n"
	"}\n"
	"// Clips a face to the back of a plane, return the number of vertices out, stored in ppVtxOut\n"
	"int clipFace(const float4* pVtxIn, int numVertsIn, float4 planeNormalWS,float planeEqWS, float4* ppVtxOut)\n"
	"{\n"
	"	\n"
	"	int ve;\n"
	"	float ds, de;\n"
	"	int numVertsOut = 0;\n"
	"//double-check next test\n"
	"	if (numVertsIn < 2)\n"
	"		return 0;\n"
	"	float4 firstVertex=pVtxIn[numVertsIn-1];\n"
	"	float4 endVertex = pVtxIn[0];\n"
	"	\n"
	"	ds = dot3F4(planeNormalWS,firstVertex)+planeEqWS;\n"
	"	for (ve = 0; ve < numVertsIn; ve++)\n"
	"	{\n"
	"		endVertex=pVtxIn[ve];\n"
	"		de = dot3F4(planeNormalWS,endVertex)+planeEqWS;\n"
	"		if (ds<0)\n"
	"		{\n"
	"			if (de<0)\n"
	"			{\n"
	"				// Start < 0, end < 0, so output endVertex\n"
	"				ppVtxOut[numVertsOut++] = endVertex;\n"
	"			}\n"
	"			else\n"
	"			{\n"
	"				// Start < 0, end >= 0, so output intersection\n"
	"				ppVtxOut[numVertsOut++] = lerp3(firstVertex, endVertex,(ds * 1.f/(ds - de)) );\n"
	"			}\n"
	"		}\n"
	"		else\n"
	"		{\n"
	"			if (de<0)\n"
	"			{\n"
	"				// Start >= 0, end < 0 so output intersection and end\n"
	"				ppVtxOut[numVertsOut++] = lerp3(firstVertex, endVertex,(ds * 1.f/(ds - de)) );\n"
	"				ppVtxOut[numVertsOut++] = endVertex;\n"
	"			}\n"
	"		}\n"
	"		firstVertex = endVertex;\n"
	"		ds = de;\n"
	"	}\n"
	"	return numVertsOut;\n"
	"}\n"
	"int clipFaceAgainstHull(const float4 separatingNormal, __global const b3ConvexPolyhedronData_t* hullA,  \n"
	"	const float4 posA, const Quaternion ornA, float4* worldVertsB1, int numWorldVertsB1,\n"
	"	float4* worldVertsB2, int capacityWorldVertsB2,\n"
	"	const float minDist, float maxDist,\n"
	"	__global const float4* vertices,\n"
	"	__global const b3GpuFace_t* faces,\n"
	"	__global const int* indices,\n"
	"	float4* contactsOut,\n"
	"	int contactCapacity)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"	float4* pVtxIn = worldVertsB1;\n"
	"	float4* pVtxOut = worldVertsB2;\n"
	"	\n"
	"	int numVertsIn = numWorldVertsB1;\n"
	"	int numVertsOut = 0;\n"
	"	int closestFaceA=-1;\n"
	"	{\n"
	"		float dmin = FLT_MAX;\n"
	"		for(int face=0;face<hullA->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(\n"
	"				faces[hullA->m_faceOffset+face].m_plane.x, \n"
	"				faces[hullA->m_faceOffset+face].m_plane.y, \n"
	"				faces[hullA->m_faceOffset+face].m_plane.z,0.f);\n"
	"			const float4 faceANormalWS = qtRotate(ornA,Normal);\n"
	"		\n"
	"			float d = dot3F4(faceANormalWS,separatingNormal);\n"
	"			if (d < dmin)\n"
	"			{\n"
	"				dmin = d;\n"
	"				closestFaceA = face;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	if (closestFaceA<0)\n"
	"		return numContactsOut;\n"
	"	b3GpuFace_t polyA = faces[hullA->m_faceOffset+closestFaceA];\n"
	"	// clip polygon to back of planes of all faces of hull A that are adjacent to witness face\n"
	"	int numVerticesA = polyA.m_numIndices;\n"
	"	for(int e0=0;e0<numVerticesA;e0++)\n"
	"	{\n"
	"		const float4 a = vertices[hullA->m_vertexOffset+indices[polyA.m_indexOffset+e0]];\n"
	"		const float4 b = vertices[hullA->m_vertexOffset+indices[polyA.m_indexOffset+((e0+1)%numVerticesA)]];\n"
	"		const float4 edge0 = a - b;\n"
	"		const float4 WorldEdge0 = qtRotate(ornA,edge0);\n"
	"		float4 planeNormalA = make_float4(polyA.m_plane.x,polyA.m_plane.y,polyA.m_plane.z,0.f);\n"
	"		float4 worldPlaneAnormal1 = qtRotate(ornA,planeNormalA);\n"
	"		float4 planeNormalWS1 = -cross3(WorldEdge0,worldPlaneAnormal1);\n"
	"		float4 worldA1 = transform(&a,&posA,&ornA);\n"
	"		float planeEqWS1 = -dot3F4(worldA1,planeNormalWS1);\n"
	"		\n"
	"		float4 planeNormalWS = planeNormalWS1;\n"
	"		float planeEqWS=planeEqWS1;\n"
	"		\n"
	"		//clip face\n"
	"		//clipFace(*pVtxIn, *pVtxOut,planeNormalWS,planeEqWS);\n"
	"		numVertsOut = clipFace(pVtxIn, numVertsIn, planeNormalWS,planeEqWS, pVtxOut);\n"
	"		//btSwap(pVtxIn,pVtxOut);\n"
	"		float4* tmp = pVtxOut;\n"
	"		pVtxOut = pVtxIn;\n"
	"		pVtxIn = tmp;\n"
	"		numVertsIn = numVertsOut;\n"
	"		numVertsOut = 0;\n"
	"	}\n"
	"	\n"
	"	// only keep points that are behind the witness face\n"
	"	{\n"
	"		float4 localPlaneNormal  = make_float4(polyA.m_plane.x,polyA.m_plane.y,polyA.m_plane.z,0.f);\n"
	"		float localPlaneEq = polyA.m_plane.w;\n"
	"		float4 planeNormalWS = qtRotate(ornA,localPlaneNormal);\n"
	"		float planeEqWS=localPlaneEq-dot3F4(planeNormalWS,posA);\n"
	"		for (int i=0;i<numVertsIn;i++)\n"
	"		{\n"
	"			float depth = dot3F4(planeNormalWS,pVtxIn[i])+planeEqWS;\n"
	"			if (depth <=minDist)\n"
	"			{\n"
	"				depth = minDist;\n"
	"			}\n"
	"			if (depth <=maxDist)\n"
	"			{\n"
	"				float4 pointInWorld = pVtxIn[i];\n"
	"				//resultOut.addContactPoint(separatingNormal,point,depth);\n"
	"				contactsOut[numContactsOut++] = make_float4(pointInWorld.x,pointInWorld.y,pointInWorld.z,depth);\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	return numContactsOut;\n"
	"}\n"
	"int clipFaceAgainstHullLocalA(const float4 separatingNormal, const b3ConvexPolyhedronData_t* hullA,  \n"
	"	const float4 posA, const Quaternion ornA, float4* worldVertsB1, int numWorldVertsB1,\n"
	"	float4* worldVertsB2, int capacityWorldVertsB2,\n"
	"	const float minDist, float maxDist,\n"
	"	const float4* verticesA,\n"
	"	const b3GpuFace_t* facesA,\n"
	"	const int* indicesA,\n"
	"	__global const float4* verticesB,\n"
	"	__global const b3GpuFace_t* facesB,\n"
	"	__global const int* indicesB,\n"
	"	float4* contactsOut,\n"
	"	int contactCapacity)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"	float4* pVtxIn = worldVertsB1;\n"
	"	float4* pVtxOut = worldVertsB2;\n"
	"	\n"
	"	int numVertsIn = numWorldVertsB1;\n"
	"	int numVertsOut = 0;\n"
	"	int closestFaceA=-1;\n"
	"	{\n"
	"		float dmin = FLT_MAX;\n"
	"		for(int face=0;face<hullA->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(\n"
	"				facesA[hullA->m_faceOffset+face].m_plane.x, \n"
	"				facesA[hullA->m_faceOffset+face].m_plane.y, \n"
	"				facesA[hullA->m_faceOffset+face].m_plane.z,0.f);\n"
	"			const float4 faceANormalWS = qtRotate(ornA,Normal);\n"
	"		\n"
	"			float d = dot3F4(faceANormalWS,separatingNormal);\n"
	"			if (d < dmin)\n"
	"			{\n"
	"				dmin = d;\n"
	"				closestFaceA = face;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	if (closestFaceA<0)\n"
	"		return numContactsOut;\n"
	"	b3GpuFace_t polyA = facesA[hullA->m_faceOffset+closestFaceA];\n"
	"	// clip polygon to back of planes of all faces of hull A that are adjacent to witness face\n"
	"	int numVerticesA = polyA.m_numIndices;\n"
	"	for(int e0=0;e0<numVerticesA;e0++)\n"
	"	{\n"
	"		const float4 a = verticesA[hullA->m_vertexOffset+indicesA[polyA.m_indexOffset+e0]];\n"
	"		const float4 b = verticesA[hullA->m_vertexOffset+indicesA[polyA.m_indexOffset+((e0+1)%numVerticesA)]];\n"
	"		const float4 edge0 = a - b;\n"
	"		const float4 WorldEdge0 = qtRotate(ornA,edge0);\n"
	"		float4 planeNormalA = make_float4(polyA.m_plane.x,polyA.m_plane.y,polyA.m_plane.z,0.f);\n"
	"		float4 worldPlaneAnormal1 = qtRotate(ornA,planeNormalA);\n"
	"		float4 planeNormalWS1 = -cross3(WorldEdge0,worldPlaneAnormal1);\n"
	"		float4 worldA1 = transform(&a,&posA,&ornA);\n"
	"		float planeEqWS1 = -dot3F4(worldA1,planeNormalWS1);\n"
	"		\n"
	"		float4 planeNormalWS = planeNormalWS1;\n"
	"		float planeEqWS=planeEqWS1;\n"
	"		\n"
	"		//clip face\n"
	"		//clipFace(*pVtxIn, *pVtxOut,planeNormalWS,planeEqWS);\n"
	"		numVertsOut = clipFace(pVtxIn, numVertsIn, planeNormalWS,planeEqWS, pVtxOut);\n"
	"		//btSwap(pVtxIn,pVtxOut);\n"
	"		float4* tmp = pVtxOut;\n"
	"		pVtxOut = pVtxIn;\n"
	"		pVtxIn = tmp;\n"
	"		numVertsIn = numVertsOut;\n"
	"		numVertsOut = 0;\n"
	"	}\n"
	"	\n"
	"	// only keep points that are behind the witness face\n"
	"	{\n"
	"		float4 localPlaneNormal  = make_float4(polyA.m_plane.x,polyA.m_plane.y,polyA.m_plane.z,0.f);\n"
	"		float localPlaneEq = polyA.m_plane.w;\n"
	"		float4 planeNormalWS = qtRotate(ornA,localPlaneNormal);\n"
	"		float planeEqWS=localPlaneEq-dot3F4(planeNormalWS,posA);\n"
	"		for (int i=0;i<numVertsIn;i++)\n"
	"		{\n"
	"			float depth = dot3F4(planeNormalWS,pVtxIn[i])+planeEqWS;\n"
	"			if (depth <=minDist)\n"
	"			{\n"
	"				depth = minDist;\n"
	"			}\n"
	"			if (depth <=maxDist)\n"
	"			{\n"
	"				float4 pointInWorld = pVtxIn[i];\n"
	"				//resultOut.addContactPoint(separatingNormal,point,depth);\n"
	"				contactsOut[numContactsOut++] = make_float4(pointInWorld.x,pointInWorld.y,pointInWorld.z,depth);\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	return numContactsOut;\n"
	"}\n"
	"int	clipHullAgainstHull(const float4 separatingNormal,\n"
	"	__global const b3ConvexPolyhedronData_t* hullA, __global const b3ConvexPolyhedronData_t* hullB, \n"
	"	const float4 posA, const Quaternion ornA,const float4 posB, const Quaternion ornB, \n"
	"	float4* worldVertsB1, float4* worldVertsB2, int capacityWorldVerts,\n"
	"	const float minDist, float maxDist,\n"
	"	__global const float4* vertices,\n"
	"	__global const b3GpuFace_t* faces,\n"
	"	__global const int* indices,\n"
	"	float4*	localContactsOut,\n"
	"	int localContactCapacity)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"	int numWorldVertsB1= 0;\n"
	"	int closestFaceB=-1;\n"
	"	float dmax = -FLT_MAX;\n"
	"	{\n"
	"		for(int face=0;face<hullB->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(faces[hullB->m_faceOffset+face].m_plane.x, \n"
	"				faces[hullB->m_faceOffset+face].m_plane.y, faces[hullB->m_faceOffset+face].m_plane.z,0.f);\n"
	"			const float4 WorldNormal = qtRotate(ornB, Normal);\n"
	"			float d = dot3F4(WorldNormal,separatingNormal);\n"
	"			if (d > dmax)\n"
	"			{\n"
	"				dmax = d;\n"
	"				closestFaceB = face;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	{\n"
	"		const b3GpuFace_t polyB = faces[hullB->m_faceOffset+closestFaceB];\n"
	"		const int numVertices = polyB.m_numIndices;\n"
	"		for(int e0=0;e0<numVertices;e0++)\n"
	"		{\n"
	"			const float4 b = vertices[hullB->m_vertexOffset+indices[polyB.m_indexOffset+e0]];\n"
	"			worldVertsB1[numWorldVertsB1++] = transform(&b,&posB,&ornB);\n"
	"		}\n"
	"	}\n"
	"	if (closestFaceB>=0)\n"
	"	{\n"
	"		numContactsOut = clipFaceAgainstHull(separatingNormal, hullA, \n"
	"				posA,ornA,\n"
	"				worldVertsB1,numWorldVertsB1,worldVertsB2,capacityWorldVerts, minDist, maxDist,vertices,\n"
	"				faces,\n"
	"				indices,localContactsOut,localContactCapacity);\n"
	"	}\n"
	"	return numContactsOut;\n"
	"}\n"
	"int	clipHullAgainstHullLocalA(const float4 separatingNormal,\n"
	"	const b3ConvexPolyhedronData_t* hullA, __global const b3ConvexPolyhedronData_t* hullB, \n"
	"	const float4 posA, const Quaternion ornA,const float4 posB, const Quaternion ornB, \n"
	"	float4* worldVertsB1, float4* worldVertsB2, int capacityWorldVerts,\n"
	"	const float minDist, float maxDist,\n"
	"	const float4* verticesA,\n"
	"	const b3GpuFace_t* facesA,\n"
	"	const int* indicesA,\n"
	"	__global const float4* verticesB,\n"
	"	__global const b3GpuFace_t* facesB,\n"
	"	__global const int* indicesB,\n"
	"	float4*	localContactsOut,\n"
	"	int localContactCapacity)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"	int numWorldVertsB1= 0;\n"
	"	int closestFaceB=-1;\n"
	"	float dmax = -FLT_MAX;\n"
	"	{\n"
	"		for(int face=0;face<hullB->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(facesB[hullB->m_faceOffset+face].m_plane.x, \n"
	"				facesB[hullB->m_faceOffset+face].m_plane.y, facesB[hullB->m_faceOffset+face].m_plane.z,0.f);\n"
	"			const float4 WorldNormal = qtRotate(ornB, Normal);\n"
	"			float d = dot3F4(WorldNormal,separatingNormal);\n"
	"			if (d > dmax)\n"
	"			{\n"
	"				dmax = d;\n"
	"				closestFaceB = face;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"	{\n"
	"		const b3GpuFace_t polyB = facesB[hullB->m_faceOffset+closestFaceB];\n"
	"		const int numVertices = polyB.m_numIndices;\n"
	"		for(int e0=0;e0<numVertices;e0++)\n"
	"		{\n"
	"			const float4 b = verticesB[hullB->m_vertexOffset+indicesB[polyB.m_indexOffset+e0]];\n"
	"			worldVertsB1[numWorldVertsB1++] = transform(&b,&posB,&ornB);\n"
	"		}\n"
	"	}\n"
	"	if (closestFaceB>=0)\n"
	"	{\n"
	"		numContactsOut = clipFaceAgainstHullLocalA(separatingNormal, hullA, \n"
	"				posA,ornA,\n"
	"				worldVertsB1,numWorldVertsB1,worldVertsB2,capacityWorldVerts, minDist, maxDist,\n"
	"				verticesA,facesA,indicesA,\n"
	"				verticesB,facesB,indicesB,\n"
	"				localContactsOut,localContactCapacity);\n"
	"	}\n"
	"	return numContactsOut;\n"
	"}\n"
	"#define PARALLEL_SUM(v, n) for(int j=1; j<n; j++) v[0] += v[j];\n"
	"#define PARALLEL_DO(execution, n) for(int ie=0; ie<n; ie++){execution;}\n"
	"#define REDUCE_MAX(v, n) {int i=0; for(int offset=0; offset<n; offset++) v[i] = (v[i].y > v[i+offset].y)? v[i]: v[i+offset]; }\n"
	"#define REDUCE_MIN(v, n) {int i=0; for(int offset=0; offset<n; offset++) v[i] = (v[i].y < v[i+offset].y)? v[i]: v[i+offset]; }\n"
	"int extractManifoldSequentialGlobal(__global const float4* p, int nPoints, float4 nearNormal, int4* contactIdx)\n"
	"{\n"
	"	if( nPoints == 0 )\n"
	"        return 0;\n"
	"    \n"
	"    if (nPoints <=4)\n"
	"        return nPoints;\n"
	"    \n"
	"    \n"
	"    if (nPoints >64)\n"
	"        nPoints = 64;\n"
	"    \n"
	"	float4 center = make_float4(0.f);\n"
	"	{\n"
	"		\n"
	"		for (int i=0;i<nPoints;i++)\n"
	"			center += p[i];\n"
	"		center /= (float)nPoints;\n"
	"	}\n"
	"    \n"
	"	\n"
	"    \n"
	"	//	sample 4 directions\n"
	"    \n"
	"    float4 aVector = p[0] - center;\n"
	"    float4 u = cross3( nearNormal, aVector );\n"
	"    float4 v = cross3( nearNormal, u );\n"
	"    u = normalize3( u );\n"
	"    v = normalize3( v );\n"
	"    \n"
	"    \n"
	"    //keep point with deepest penetration\n"
	"    float minW= FLT_MAX;\n"
	"    \n"
	"    int minIndex=-1;\n"
	"    \n"
	"    float4 maxDots;\n"
	"    maxDots.x = FLT_MIN;\n"
	"    maxDots.y = FLT_MIN;\n"
	"    maxDots.z = FLT_MIN;\n"
	"    maxDots.w = FLT_MIN;\n"
	"    \n"
	"    //	idx, distance\n"
	"    for(int ie = 0; ie<nPoints; ie++ )\n"
	"    {\n"
	"        if (p[ie].w<minW)\n"
	"        {\n"
	"            minW = p[ie].w;\n"
	"            minIndex=ie;\n"
	"        }\n"
	"        float f;\n"
	"        float4 r = p[ie]-center;\n"
	"        f = dot3F4( u, r );\n"
	"        if (f<maxDots.x)\n"
	"        {\n"
	"            maxDots.x = f;\n"
	"            contactIdx[0].x = ie;\n"
	"        }\n"
	"        \n"
	"        f = dot3F4( -u, r );\n"
	"        if (f<maxDots.y)\n"
	"        {\n"
	"            maxDots.y = f;\n"
	"            contactIdx[0].y = ie;\n"
	"        }\n"
	"        \n"
	"        \n"
	"        f = dot3F4( v, r );\n"
	"        if (f<maxDots.z)\n"
	"        {\n"
	"            maxDots.z = f;\n"
	"            contactIdx[0].z = ie;\n"
	"        }\n"
	"        \n"
	"        f = dot3F4( -v, r );\n"
	"        if (f<maxDots.w)\n"
	"        {\n"
	"            maxDots.w = f;\n"
	"            contactIdx[0].w = ie;\n"
	"        }\n"
	"        \n"
	"    }\n"
	"    \n"
	"    if (contactIdx[0].x != minIndex && contactIdx[0].y != minIndex && contactIdx[0].z != minIndex && contactIdx[0].w != minIndex)\n"
	"    {\n"
	"        //replace the first contact with minimum (todo: replace contact with least penetration)\n"
	"        contactIdx[0].x = minIndex;\n"
	"    }\n"
	"    \n"
	"    return 4;\n"
	"    \n"
	"}\n"
	"int extractManifoldSequentialGlobalFake(__global const float4* p, int nPoints, float4 nearNormal, int* contactIdx)\n"
	"{\n"
	"    contactIdx[0] = 0;\n"
	"    contactIdx[1] = 1;\n"
	"    contactIdx[2] = 2;\n"
	"    contactIdx[3] = 3;\n"
	"    \n"
	"	if( nPoints == 0 ) return 0;\n"
	"    \n"
	"	nPoints = min2( nPoints, 4 );\n"
	"    return nPoints;\n"
	"    \n"
	"}\n"
	"int extractManifoldSequential(const float4* p, int nPoints, float4 nearNormal, int* contactIdx)\n"
	"{\n"
	"	if( nPoints == 0 ) return 0;\n"
	"	nPoints = min2( nPoints, 64 );\n"
	"	float4 center = make_float4(0.f);\n"
	"	{\n"
	"		float4 v[64];\n"
	"		for (int i=0;i<nPoints;i++)\n"
	"			v[i] = p[i];\n"
	"		//memcpy( v, p, nPoints*sizeof(float4) );\n"
	"		PARALLEL_SUM( v, nPoints );\n"
	"		center = v[0]/(float)nPoints;\n"
	"	}\n"
	"	\n"
	"	{	//	sample 4 directions\n"
	"		if( nPoints < 4 )\n"
	"		{\n"
	"			for(int i=0; i<nPoints; i++) \n"
	"				contactIdx[i] = i;\n"
	"			return nPoints;\n"
	"		}\n"
	"		float4 aVector = p[0] - center;\n"
	"		float4 u = cross3( nearNormal, aVector );\n"
	"		float4 v = cross3( nearNormal, u );\n"
	"		u = normalize3( u );\n"
	"		v = normalize3( v );\n"
	"		int idx[4];\n"
	"		float2 max00 = make_float2(0,FLT_MAX);\n"
	"		{\n"
	"			//	idx, distance\n"
	"			{\n"
	"				{\n"
	"					int4 a[64];\n"
	"					for(int ie = 0; ie<nPoints; ie++ )\n"
	"					{\n"
	"						\n"
	"						\n"
	"						float f;\n"
	"						float4 r = p[ie]-center;\n"
	"						f = dot3F4( u, r );\n"
	"						a[ie].x = ((*(u32*)&f) & 0xffffff00) | (0xff & ie);\n"
	"						f = dot3F4( -u, r );\n"
	"						a[ie].y = ((*(u32*)&f) & 0xffffff00) | (0xff & ie);\n"
	"						f = dot3F4( v, r );\n"
	"						a[ie].z = ((*(u32*)&f) & 0xffffff00) | (0xff & ie);\n"
	"						f = dot3F4( -v, r );\n"
	"						a[ie].w = ((*(u32*)&f) & 0xffffff00) | (0xff & ie);\n"
	"					}\n"
	"					for(int ie=0; ie<nPoints; ie++)\n"
	"					{\n"
	"						a[0].x = (a[0].x > a[ie].x )? a[0].x: a[ie].x;\n"
	"						a[0].y = (a[0].y > a[ie].y )? a[0].y: a[ie].y;\n"
	"						a[0].z = (a[0].z > a[ie].z )? a[0].z: a[ie].z;\n"
	"						a[0].w = (a[0].w > a[ie].w )? a[0].w: a[ie].w;\n"
	"					}\n"
	"					idx[0] = (int)a[0].x & 0xff;\n"
	"					idx[1] = (int)a[0].y & 0xff;\n"
	"					idx[2] = (int)a[0].z & 0xff;\n"
	"					idx[3] = (int)a[0].w & 0xff;\n"
	"				}\n"
	"			}\n"
	"			{\n"
	"				float2 h[64];\n"
	"				PARALLEL_DO( h[ie] = make_float2((float)ie, p[ie].w), nPoints );\n"
	"				REDUCE_MIN( h, nPoints );\n"
	"				max00 = h[0];\n"
	"			}\n"
	"		}\n"
	"		contactIdx[0] = idx[0];\n"
	"		contactIdx[1] = idx[1];\n"
	"		contactIdx[2] = idx[2];\n"
	"		contactIdx[3] = idx[3];\n"
	"		return 4;\n"
	"	}\n"
	"}\n"
	"__kernel void   extractManifoldAndAddContactKernel(__global const int4* pairs, \n"
	"																	__global const b3RigidBodyData_t* rigidBodies, \n"
	"																	__global const float4* closestPointsWorld,\n"
	"																	__global const float4* separatingNormalsWorld,\n"
	"																	__global const int* contactCounts,\n"
	"																	__global const int* contactOffsets,\n"
	"																	__global struct b3Contact4Data* restrict contactsOut,\n"
	"																	counter32_t nContactsOut,\n"
	"																	int contactCapacity,\n"
	"																	int numPairs,\n"
	"																	int pairIndex\n"
	"																	)\n"
	"{\n"
	"	int idx = get_global_id(0);\n"
	"	\n"
	"	if (idx<numPairs)\n"
	"	{\n"
	"		float4 normal = separatingNormalsWorld[idx];\n"
	"		int nPoints = contactCounts[idx];\n"
	"		__global const float4* pointsIn = &closestPointsWorld[contactOffsets[idx]];\n"
	"		float4 localPoints[64];\n"
	"		for (int i=0;i<nPoints;i++)\n"
	"		{\n"
	"			localPoints[i] = pointsIn[i];\n"
	"		}\n"
	"		int contactIdx[4];// = {-1,-1,-1,-1};\n"
	"		contactIdx[0] = -1;\n"
	"		contactIdx[1] = -1;\n"
	"		contactIdx[2] = -1;\n"
	"		contactIdx[3] = -1;\n"
	"		int nContacts = extractManifoldSequential(localPoints, nPoints, normal, contactIdx);\n"
	"		int dstIdx;\n"
	"		AppendInc( nContactsOut, dstIdx );\n"
	"		if (dstIdx<contactCapacity)\n"
	"		{\n"
	"			__global struct b3Contact4Data* c = contactsOut + dstIdx;\n"
	"			c->m_worldNormalOnB = -normal;\n"
	"			c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"			c->m_batchIdx = idx;\n"
	"			int bodyA = pairs[pairIndex].x;\n"
	"			int bodyB = pairs[pairIndex].y;\n"
	"			c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0 ? -bodyA:bodyA;\n"
	"			c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0 ? -bodyB:bodyB;\n"
	"			c->m_childIndexA = -1;\n"
	"			c->m_childIndexB = -1;\n"
	"			for (int i=0;i<nContacts;i++)\n"
	"			{\n"
	"				c->m_worldPosB[i] = localPoints[contactIdx[i]];\n"
	"			}\n"
	"			GET_NPOINTS(*c) = nContacts;\n"
	"		}\n"
	"	}\n"
	"}\n"
	"void	trInverse(float4 translationIn, Quaternion orientationIn,\n"
	"		float4* translationOut, Quaternion* orientationOut)\n"
	"{\n"
	"	*orientationOut = qtInvert(orientationIn);\n"
	"	*translationOut = qtRotate(*orientationOut, -translationIn);\n"
	"}\n"
	"void	trMul(float4 translationA, Quaternion orientationA,\n"
	"						float4 translationB, Quaternion orientationB,\n"
	"		float4* translationOut, Quaternion* orientationOut)\n"
	"{\n"
	"	*orientationOut = qtMul(orientationA,orientationB);\n"
	"	*translationOut = transform(&translationB,&translationA,&orientationA);\n"
	"}\n"
	"__kernel void   clipHullHullKernel( __global int4* pairs, \n"
	"																					__global const b3RigidBodyData_t* rigidBodies, \n"
	"																					__global const b3Collidable_t* collidables,\n"
	"																					__global const b3ConvexPolyhedronData_t* convexShapes, \n"
	"																					__global const float4* vertices,\n"
	"																					__global const float4* uniqueEdges,\n"
	"																					__global const b3GpuFace_t* faces,\n"
	"																					__global const int* indices,\n"
	"																					__global const float4* separatingNormals,\n"
	"																					__global const int* hasSeparatingAxis,\n"
	"																					__global struct b3Contact4Data* restrict globalContactsOut,\n"
	"																					counter32_t nGlobalContactsOut,\n"
	"																					int numPairs,\n"
	"																					int contactCapacity)\n"
	"{\n"
	"	int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"	float4 worldVertsB1[64];\n"
	"	float4 worldVertsB2[64];\n"
	"	int capacityWorldVerts = 64;	\n"
	"	float4 localContactsOut[64];\n"
	"	int localContactCapacity=64;\n"
	"	\n"
	"	float minDist = -1e30f;\n"
	"	float maxDist = 0.02f;\n"
	"	if (i<numPairs)\n"
	"	{\n"
	"		int bodyIndexA = pairs[i].x;\n"
	"		int bodyIndexB = pairs[i].y;\n"
	"			\n"
	"		int collidableIndexA = rigidBodies[bodyIndexA].m_collidableIdx;\n"
	"		int collidableIndexB = rigidBodies[bodyIndexB].m_collidableIdx;\n"
	"		if (hasSeparatingAxis[i])\n"
	"		{\n"
	"			\n"
	"			int shapeIndexA = collidables[collidableIndexA].m_shapeIndex;\n"
	"			int shapeIndexB = collidables[collidableIndexB].m_shapeIndex;\n"
	"			\n"
	"		\n"
	"			int numLocalContactsOut = clipHullAgainstHull(separatingNormals[i],\n"
	"														&convexShapes[shapeIndexA], &convexShapes[shapeIndexB],\n"
	"														rigidBodies[bodyIndexA].m_pos,rigidBodies[bodyIndexA].m_quat,\n"
	"													  rigidBodies[bodyIndexB].m_pos,rigidBodies[bodyIndexB].m_quat,\n"
	"													  worldVertsB1,worldVertsB2,capacityWorldVerts,\n"
	"														minDist, maxDist,\n"
	"														vertices,faces,indices,\n"
	"														localContactsOut,localContactCapacity);\n"
	"												\n"
	"		if (numLocalContactsOut>0)\n"
	"		{\n"
	"				float4 normal = -separatingNormals[i];\n"
	"				int nPoints = numLocalContactsOut;\n"
	"				float4* pointsIn = localContactsOut;\n"
	"				int contactIdx[4];// = {-1,-1,-1,-1};\n"
	"				contactIdx[0] = -1;\n"
	"				contactIdx[1] = -1;\n"
	"				contactIdx[2] = -1;\n"
	"				contactIdx[3] = -1;\n"
	"		\n"
	"				int nReducedContacts = extractManifoldSequential(pointsIn, nPoints, normal, contactIdx);\n"
	"		\n"
	"				\n"
	"				int mprContactIndex = pairs[pairIndex].z;\n"
	"				int dstIdx = mprContactIndex;\n"
	"				if (dstIdx<0)\n"
	"				{\n"
	"					AppendInc( nGlobalContactsOut, dstIdx );\n"
	"				}\n"
	"				if (dstIdx<contactCapacity)\n"
	"				{\n"
	"					pairs[pairIndex].z = dstIdx;\n"
	"					__global struct b3Contact4Data* c = globalContactsOut+ dstIdx;\n"
	"					c->m_worldNormalOnB = -normal;\n"
	"					c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"					c->m_batchIdx = pairIndex;\n"
	"					int bodyA = pairs[pairIndex].x;\n"
	"					int bodyB = pairs[pairIndex].y;\n"
	"					c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0?-bodyA:bodyA;\n"
	"					c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0?-bodyB:bodyB;\n"
	"					c->m_childIndexA = -1;\n"
	"					c->m_childIndexB = -1;\n"
	"					for (int i=0;i<nReducedContacts;i++)\n"
	"					{\n"
	"					//this condition means: overwrite contact point, unless at index i==0 we have a valid 'mpr' contact\n"
	"						if (i>0||(mprContactIndex<0))\n"
	"						{\n"
	"							c->m_worldPosB[i] = pointsIn[contactIdx[i]];\n"
	"						}\n"
	"					}\n"
	"					GET_NPOINTS(*c) = nReducedContacts;\n"
	"				}\n"
	"				\n"
	"			}//		if (numContactsOut>0)\n"
	"		}//		if (hasSeparatingAxis[i])\n"
	"	}//	if (i<numPairs)\n"
	"}\n"
	"__kernel void   clipCompoundsHullHullKernel( __global const int4* gpuCompoundPairs, \n"
	"																					__global const b3RigidBodyData_t* rigidBodies, \n"
	"																					__global const b3Collidable_t* collidables,\n"
	"																					__global const b3ConvexPolyhedronData_t* convexShapes, \n"
	"																					__global const float4* vertices,\n"
	"																					__global const float4* uniqueEdges,\n"
	"																					__global const b3GpuFace_t* faces,\n"
	"																					__global const int* indices,\n"
	"																					__global const b3GpuChildShape_t* gpuChildShapes,\n"
	"																					__global const float4* gpuCompoundSepNormalsOut,\n"
	"																					__global const int* gpuHasCompoundSepNormalsOut,\n"
	"																					__global struct b3Contact4Data* restrict globalContactsOut,\n"
	"																					counter32_t nGlobalContactsOut,\n"
	"																					int numCompoundPairs, int maxContactCapacity)\n"
	"{\n"
	"	int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"	float4 worldVertsB1[64];\n"
	"	float4 worldVertsB2[64];\n"
	"	int capacityWorldVerts = 64;	\n"
	"	float4 localContactsOut[64];\n"
	"	int localContactCapacity=64;\n"
	"	\n"
	"	float minDist = -1e30f;\n"
	"	float maxDist = 0.02f;\n"
	"	if (i<numCompoundPairs)\n"
	"	{\n"
	"		if (gpuHasCompoundSepNormalsOut[i])\n"
	"		{\n"
	"			int bodyIndexA = gpuCompoundPairs[i].x;\n"
	"			int bodyIndexB = gpuCompoundPairs[i].y;\n"
	"			\n"
	"			int childShapeIndexA = gpuCompoundPairs[i].z;\n"
	"			int childShapeIndexB = gpuCompoundPairs[i].w;\n"
	"			\n"
	"			int collidableIndexA = -1;\n"
	"			int collidableIndexB = -1;\n"
	"			\n"
	"			float4 ornA = rigidBodies[bodyIndexA].m_quat;\n"
	"			float4 posA = rigidBodies[bodyIndexA].m_pos;\n"
	"			\n"
	"			float4 ornB = rigidBodies[bodyIndexB].m_quat;\n"
	"			float4 posB = rigidBodies[bodyIndexB].m_pos;\n"
	"								\n"
	"			if (childShapeIndexA >= 0)\n"
	"			{\n"
	"				collidableIndexA = gpuChildShapes[childShapeIndexA].m_shapeIndex;\n"
	"				float4 childPosA = gpuChildShapes[childShapeIndexA].m_childPosition;\n"
	"				float4 childOrnA = gpuChildShapes[childShapeIndexA].m_childOrientation;\n"
	"				float4 newPosA = qtRotate(ornA,childPosA)+posA;\n"
	"				float4 newOrnA = qtMul(ornA,childOrnA);\n"
	"				posA = newPosA;\n"
	"				ornA = newOrnA;\n"
	"			} else\n"
	"			{\n"
	"				collidableIndexA = rigidBodies[bodyIndexA].m_collidableIdx;\n"
	"			}\n"
	"			\n"
	"			if (childShapeIndexB>=0)\n"
	"			{\n"
	"				collidableIndexB = gpuChildShapes[childShapeIndexB].m_shapeIndex;\n"
	"				float4 childPosB = gpuChildShapes[childShapeIndexB].m_childPosition;\n"
	"				float4 childOrnB = gpuChildShapes[childShapeIndexB].m_childOrientation;\n"
	"				float4 newPosB = transform(&childPosB,&posB,&ornB);\n"
	"				float4 newOrnB = qtMul(ornB,childOrnB);\n"
	"				posB = newPosB;\n"
	"				ornB = newOrnB;\n"
	"			} else\n"
	"			{\n"
	"				collidableIndexB = rigidBodies[bodyIndexB].m_collidableIdx;	\n"
	"			}\n"
	"			\n"
	"			int shapeIndexA = collidables[collidableIndexA].m_shapeIndex;\n"
	"			int shapeIndexB = collidables[collidableIndexB].m_shapeIndex;\n"
	"		\n"
	"			int numLocalContactsOut = clipHullAgainstHull(gpuCompoundSepNormalsOut[i],\n"
	"														&convexShapes[shapeIndexA], &convexShapes[shapeIndexB],\n"
	"														posA,ornA,\n"
	"													  posB,ornB,\n"
	"													  worldVertsB1,worldVertsB2,capacityWorldVerts,\n"
	"														minDist, maxDist,\n"
	"														vertices,faces,indices,\n"
	"														localContactsOut,localContactCapacity);\n"
	"												\n"
	"		if (numLocalContactsOut>0)\n"
	"		{\n"
	"				float4 normal = -gpuCompoundSepNormalsOut[i];\n"
	"				int nPoints = numLocalContactsOut;\n"
	"				float4* pointsIn = localContactsOut;\n"
	"				int contactIdx[4];// = {-1,-1,-1,-1};\n"
	"				contactIdx[0] = -1;\n"
	"				contactIdx[1] = -1;\n"
	"				contactIdx[2] = -1;\n"
	"				contactIdx[3] = -1;\n"
	"		\n"
	"				int nReducedContacts = extractManifoldSequential(pointsIn, nPoints, normal, contactIdx);\n"
	"		\n"
	"				int dstIdx;\n"
	"				AppendInc( nGlobalContactsOut, dstIdx );\n"
	"				if ((dstIdx+nReducedContacts) < maxContactCapacity)\n"
	"				{\n"
	"					__global struct b3Contact4Data* c = globalContactsOut+ dstIdx;\n"
	"					c->m_worldNormalOnB = -normal;\n"
	"					c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"					c->m_batchIdx = pairIndex;\n"
	"					int bodyA = gpuCompoundPairs[pairIndex].x;\n"
	"					int bodyB = gpuCompoundPairs[pairIndex].y;\n"
	"					c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0?-bodyA:bodyA;\n"
	"					c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0?-bodyB:bodyB;\n"
	"					c->m_childIndexA = childShapeIndexA;\n"
	"					c->m_childIndexB = childShapeIndexB;\n"
	"					for (int i=0;i<nReducedContacts;i++)\n"
	"					{\n"
	"						c->m_worldPosB[i] = pointsIn[contactIdx[i]];\n"
	"					}\n"
	"					GET_NPOINTS(*c) = nReducedContacts;\n"
	"				}\n"
	"				\n"
	"			}//		if (numContactsOut>0)\n"
	"		}//		if (gpuHasCompoundSepNormalsOut[i])\n"
	"	}//	if (i<numCompoundPairs)\n"
	"}\n"
	"__kernel void   sphereSphereCollisionKernel( __global const int4* pairs, \n"
	"																					__global const b3RigidBodyData_t* rigidBodies, \n"
	"																					__global const b3Collidable_t* collidables,\n"
	"																					__global const float4* separatingNormals,\n"
	"																					__global const int* hasSeparatingAxis,\n"
	"																					__global struct b3Contact4Data* restrict globalContactsOut,\n"
	"																					counter32_t nGlobalContactsOut,\n"
	"																					int contactCapacity,\n"
	"																					int numPairs)\n"
	"{\n"
	"	int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"	if (i<numPairs)\n"
	"	{\n"
	"		int bodyIndexA = pairs[i].x;\n"
	"		int bodyIndexB = pairs[i].y;\n"
	"			\n"
	"		int collidableIndexA = rigidBodies[bodyIndexA].m_collidableIdx;\n"
	"		int collidableIndexB = rigidBodies[bodyIndexB].m_collidableIdx;\n"
	"		if (collidables[collidableIndexA].m_shapeType == SHAPE_SPHERE &&\n"
	"			collidables[collidableIndexB].m_shapeType == SHAPE_SPHERE)\n"
	"		{\n"
	"			//sphere-sphere\n"
	"			float radiusA = collidables[collidableIndexA].m_radius;\n"
	"			float radiusB = collidables[collidableIndexB].m_radius;\n"
	"			float4 posA = rigidBodies[bodyIndexA].m_pos;\n"
	"			float4 posB = rigidBodies[bodyIndexB].m_pos;\n"
	"			float4 diff = posA-posB;\n"
	"			float len = length(diff);\n"
	"			\n"
	"			///iff distance positive, don't generate a new contact\n"
	"			if ( len <= (radiusA+radiusB))\n"
	"			{\n"
	"				///distance (negative means penetration)\n"
	"				float dist = len - (radiusA+radiusB);\n"
	"				float4 normalOnSurfaceB = make_float4(1.f,0.f,0.f,0.f);\n"
	"				if (len > 0.00001)\n"
	"				{\n"
	"					normalOnSurfaceB = diff / len;\n"
	"				}\n"
	"				float4 contactPosB = posB + normalOnSurfaceB*radiusB;\n"
	"				contactPosB.w = dist;\n"
	"								\n"
	"				int dstIdx;\n"
	"				AppendInc( nGlobalContactsOut, dstIdx );\n"
	"				if (dstIdx < contactCapacity)\n"
	"				{\n"
	"					__global struct b3Contact4Data* c = &globalContactsOut[dstIdx];\n"
	"					c->m_worldNormalOnB = -normalOnSurfaceB;\n"
	"					c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"					c->m_batchIdx = pairIndex;\n"
	"					int bodyA = pairs[pairIndex].x;\n"
	"					int bodyB = pairs[pairIndex].y;\n"
	"					c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0?-bodyA:bodyA;\n"
	"					c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0?-bodyB:bodyB;\n"
	"					c->m_worldPosB[0] = contactPosB;\n"
	"					c->m_childIndexA = -1;\n"
	"					c->m_childIndexB = -1;\n"
	"					GET_NPOINTS(*c) = 1;\n"
	"				}//if (dstIdx < numPairs)\n"
	"			}//if ( len <= (radiusA+radiusB))\n"
	"		}//SHAPE_SPHERE SHAPE_SPHERE\n"
	"	}//if (i<numPairs)\n"
	"}				\n"
	"__kernel void   clipHullHullConcaveConvexKernel( __global int4* concavePairsIn,\n"
	"																					__global const b3RigidBodyData_t* rigidBodies, \n"
	"																					__global const b3Collidable_t* collidables,\n"
	"																					__global const b3ConvexPolyhedronData_t* convexShapes, \n"
	"																					__global const float4* vertices,\n"
	"																					__global const float4* uniqueEdges,\n"
	"																					__global const b3GpuFace_t* faces,\n"
	"																					__global const int* indices,\n"
	"																					__global const b3GpuChildShape_t* gpuChildShapes,\n"
	"																					__global const float4* separatingNormals,\n"
	"																					__global struct b3Contact4Data* restrict globalContactsOut,\n"
	"																					counter32_t nGlobalContactsOut,\n"
	"																					int contactCapacity,\n"
	"																					int numConcavePairs)\n"
	"{\n"
	"	int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"	float4 worldVertsB1[64];\n"
	"	float4 worldVertsB2[64];\n"
	"	int capacityWorldVerts = 64;	\n"
	"	float4 localContactsOut[64];\n"
	"	int localContactCapacity=64;\n"
	"	\n"
	"	float minDist = -1e30f;\n"
	"	float maxDist = 0.02f;\n"
	"	if (i<numConcavePairs)\n"
	"	{\n"
	"		//negative value means that the pair is invalid\n"
	"		if (concavePairsIn[i].w<0)\n"
	"			return;\n"
	"		int bodyIndexA = concavePairsIn[i].x;\n"
	"		int bodyIndexB = concavePairsIn[i].y;\n"
	"		int f = concavePairsIn[i].z;\n"
	"		int childShapeIndexA = f;\n"
	"		\n"
	"		int collidableIndexA = rigidBodies[bodyIndexA].m_collidableIdx;\n"
	"		int collidableIndexB = rigidBodies[bodyIndexB].m_collidableIdx;\n"
	"		\n"
	"		int shapeIndexA = collidables[collidableIndexA].m_shapeIndex;\n"
	"		int shapeIndexB = collidables[collidableIndexB].m_shapeIndex;\n"
	"		\n"
	"		///////////////////////////////////////////////////////////////\n"
	"		\n"
	"	\n"
	"		bool overlap = false;\n"
	"		\n"
	"		b3ConvexPolyhedronData_t convexPolyhedronA;\n"
	"	//add 3 vertices of the triangle\n"
	"		convexPolyhedronA.m_numVertices = 3;\n"
	"		convexPolyhedronA.m_vertexOffset = 0;\n"
	"		float4	localCenter = make_float4(0.f,0.f,0.f,0.f);\n"
	"		b3GpuFace_t face = faces[convexShapes[shapeIndexA].m_faceOffset+f];\n"
	"		\n"
	"		float4 verticesA[3];\n"
	"		for (int i=0;i<3;i++)\n"
	"		{\n"
	"			int index = indices[face.m_indexOffset+i];\n"
	"			float4 vert = vertices[convexShapes[shapeIndexA].m_vertexOffset+index];\n"
	"			verticesA[i] = vert;\n"
	"			localCenter += vert;\n"
	"		}\n"
	"		float dmin = FLT_MAX;\n"
	"		int localCC=0;\n"
	"		//a triangle has 3 unique edges\n"
	"		convexPolyhedronA.m_numUniqueEdges = 3;\n"
	"		convexPolyhedronA.m_uniqueEdgesOffset = 0;\n"
	"		float4 uniqueEdgesA[3];\n"
	"		\n"
	"		uniqueEdgesA[0] = (verticesA[1]-verticesA[0]);\n"
	"		uniqueEdgesA[1] = (verticesA[2]-verticesA[1]);\n"
	"		uniqueEdgesA[2] = (verticesA[0]-verticesA[2]);\n"
	"		convexPolyhedronA.m_faceOffset = 0;\n"
	"                                  \n"
	"		float4 normal = make_float4(face.m_plane.x,face.m_plane.y,face.m_plane.z,0.f);\n"
	"                             \n"
	"		b3GpuFace_t facesA[TRIANGLE_NUM_CONVEX_FACES];\n"
	"		int indicesA[3+3+2+2+2];\n"
	"		int curUsedIndices=0;\n"
	"		int fidx=0;\n"
	"		//front size of triangle\n"
	"		{\n"
	"			facesA[fidx].m_indexOffset=curUsedIndices;\n"
	"			indicesA[0] = 0;\n"
	"			indicesA[1] = 1;\n"
	"			indicesA[2] = 2;\n"
	"			curUsedIndices+=3;\n"
	"			float c = face.m_plane.w;\n"
	"			facesA[fidx].m_plane.x = normal.x;\n"
	"			facesA[fidx].m_plane.y = normal.y;\n"
	"			facesA[fidx].m_plane.z = normal.z;\n"
	"			facesA[fidx].m_plane.w = c;\n"
	"			facesA[fidx].m_numIndices=3;\n"
	"		}\n"
	"		fidx++;\n"
	"		//back size of triangle\n"
	"		{\n"
	"			facesA[fidx].m_indexOffset=curUsedIndices;\n"
	"			indicesA[3]=2;\n"
	"			indicesA[4]=1;\n"
	"			indicesA[5]=0;\n"
	"			curUsedIndices+=3;\n"
	"			float c = dot3F4(normal,verticesA[0]);\n"
	"			float c1 = -face.m_plane.w;\n"
	"			facesA[fidx].m_plane.x = -normal.x;\n"
	"			facesA[fidx].m_plane.y = -normal.y;\n"
	"			facesA[fidx].m_plane.z = -normal.z;\n"
	"			facesA[fidx].m_plane.w = c;\n"
	"			facesA[fidx].m_numIndices=3;\n"
	"		}\n"
	"		fidx++;\n"
	"		bool addEdgePlanes = true;\n"
	"		if (addEdgePlanes)\n"
	"		{\n"
	"			int numVertices=3;\n"
	"			int prevVertex = numVertices-1;\n"
	"			for (int i=0;i<numVertices;i++)\n"
	"			{\n"
	"				float4 v0 = verticesA[i];\n"
	"				float4 v1 = verticesA[prevVertex];\n"
	"                                            \n"
	"				float4 edgeNormal = normalize(cross(normal,v1-v0));\n"
	"				float c = -dot3F4(edgeNormal,v0);\n"
	"				facesA[fidx].m_numIndices = 2;\n"
	"				facesA[fidx].m_indexOffset=curUsedIndices;\n"
	"				indicesA[curUsedIndices++]=i;\n"
	"				indicesA[curUsedIndices++]=prevVertex;\n"
	"                                            \n"
	"				facesA[fidx].m_plane.x = edgeNormal.x;\n"
	"				facesA[fidx].m_plane.y = edgeNormal.y;\n"
	"				facesA[fidx].m_plane.z = edgeNormal.z;\n"
	"				facesA[fidx].m_plane.w = c;\n"
	"				fidx++;\n"
	"				prevVertex = i;\n"
	"			}\n"
	"		}\n"
	"		convexPolyhedronA.m_numFaces = TRIANGLE_NUM_CONVEX_FACES;\n"
	"		convexPolyhedronA.m_localCenter = localCenter*(1.f/3.f);\n"
	"		float4 posA = rigidBodies[bodyIndexA].m_pos;\n"
	"		posA.w = 0.f;\n"
	"		float4 posB = rigidBodies[bodyIndexB].m_pos;\n"
	"		posB.w = 0.f;\n"
	"		float4 ornA = rigidBodies[bodyIndexA].m_quat;\n"
	"		float4 ornB =rigidBodies[bodyIndexB].m_quat;\n"
	"		float4 sepAxis = separatingNormals[i];\n"
	"		\n"
	"		int shapeTypeB = collidables[collidableIndexB].m_shapeType;\n"
	"		int childShapeIndexB =-1;\n"
	"		if (shapeTypeB==SHAPE_COMPOUND_OF_CONVEX_HULLS)\n"
	"		{\n"
	"			///////////////////\n"
	"			///compound shape support\n"
	"			\n"
	"			childShapeIndexB = concavePairsIn[pairIndex].w;\n"
	"			int childColIndexB = gpuChildShapes[childShapeIndexB].m_shapeIndex;\n"
	"			shapeIndexB = collidables[childColIndexB].m_shapeIndex;\n"
	"			float4 childPosB = gpuChildShapes[childShapeIndexB].m_childPosition;\n"
	"			float4 childOrnB = gpuChildShapes[childShapeIndexB].m_childOrientation;\n"
	"			float4 newPosB = transform(&childPosB,&posB,&ornB);\n"
	"			float4 newOrnB = qtMul(ornB,childOrnB);\n"
	"			posB = newPosB;\n"
	"			ornB = newOrnB;\n"
	"			\n"
	"		}\n"
	"		\n"
	"		////////////////////////////////////////\n"
	"		\n"
	"		\n"
	"		\n"
	"		int numLocalContactsOut = clipHullAgainstHullLocalA(sepAxis,\n"
	"														&convexPolyhedronA, &convexShapes[shapeIndexB],\n"
	"														posA,ornA,\n"
	"													  posB,ornB,\n"
	"													  worldVertsB1,worldVertsB2,capacityWorldVerts,\n"
	"														minDist, maxDist,\n"
	"														&verticesA,&facesA,&indicesA,\n"
	"														vertices,faces,indices,\n"
	"														localContactsOut,localContactCapacity);\n"
	"												\n"
	"		if (numLocalContactsOut>0)\n"
	"		{\n"
	"			float4 normal = -separatingNormals[i];\n"
	"			int nPoints = numLocalContactsOut;\n"
	"			float4* pointsIn = localContactsOut;\n"
	"			int contactIdx[4];// = {-1,-1,-1,-1};\n"
	"			contactIdx[0] = -1;\n"
	"			contactIdx[1] = -1;\n"
	"			contactIdx[2] = -1;\n"
	"			contactIdx[3] = -1;\n"
	"	\n"
	"			int nReducedContacts = extractManifoldSequential(pointsIn, nPoints, normal, contactIdx);\n"
	"	\n"
	"			int dstIdx;\n"
	"			AppendInc( nGlobalContactsOut, dstIdx );\n"
	"			if (dstIdx<contactCapacity)\n"
	"			{\n"
	"				__global struct b3Contact4Data* c = globalContactsOut+ dstIdx;\n"
	"				c->m_worldNormalOnB = -normal;\n"
	"				c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"				c->m_batchIdx = pairIndex;\n"
	"				int bodyA = concavePairsIn[pairIndex].x;\n"
	"				int bodyB = concavePairsIn[pairIndex].y;\n"
	"				c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0?-bodyA:bodyA;\n"
	"				c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0?-bodyB:bodyB;\n"
	"				c->m_childIndexA = childShapeIndexA;\n"
	"				c->m_childIndexB = childShapeIndexB;\n"
	"				for (int i=0;i<nReducedContacts;i++)\n"
	"				{\n"
	"					c->m_worldPosB[i] = pointsIn[contactIdx[i]];\n"
	"				}\n"
	"				GET_NPOINTS(*c) = nReducedContacts;\n"
	"			}\n"
	"				\n"
	"		}//		if (numContactsOut>0)\n"
	"	}//	if (i<numPairs)\n"
	"}\n"
	"int	findClippingFaces(const float4 separatingNormal,\n"
	"                      __global const b3ConvexPolyhedronData_t* hullA, __global const b3ConvexPolyhedronData_t* hullB,\n"
	"                      const float4 posA, const Quaternion ornA,const float4 posB, const Quaternion ornB,\n"
	"                       __global float4* worldVertsA1,\n"
	"                      __global float4* worldNormalsA1,\n"
	"                      __global float4* worldVertsB1,\n"
	"                      int capacityWorldVerts,\n"
	"                      const float minDist, float maxDist,\n"
	"                      __global const float4* vertices,\n"
	"                      __global const b3GpuFace_t* faces,\n"
	"                      __global const int* indices,\n"
	"                      __global int4* clippingFaces, int pairIndex)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"	int numWorldVertsB1= 0;\n"
	"    \n"
	"    \n"
	"	int closestFaceB=-1;\n"
	"	float dmax = -FLT_MAX;\n"
	"    \n"
	"	{\n"
	"		for(int face=0;face<hullB->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(faces[hullB->m_faceOffset+face].m_plane.x,\n"
	"                                              faces[hullB->m_faceOffset+face].m_plane.y, faces[hullB->m_faceOffset+face].m_plane.z,0.f);\n"
	"			const float4 WorldNormal = qtRotate(ornB, Normal);\n"
	"			float d = dot3F4(WorldNormal,separatingNormal);\n"
	"			if (d > dmax)\n"
	"			{\n"
	"				dmax = d;\n"
	"				closestFaceB = face;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"    \n"
	"	{\n"
	"		const b3GpuFace_t polyB = faces[hullB->m_faceOffset+closestFaceB];\n"
	"		const int numVertices = polyB.m_numIndices;\n"
	"		for(int e0=0;e0<numVertices;e0++)\n"
	"		{\n"
	"			const float4 b = vertices[hullB->m_vertexOffset+indices[polyB.m_indexOffset+e0]];\n"
	"			worldVertsB1[pairIndex*capacityWorldVerts+numWorldVertsB1++] = transform(&b,&posB,&ornB);\n"
	"		}\n"
	"	}\n"
	"    \n"
	"    int closestFaceA=-1;\n"
	"	{\n"
	"		float dmin = FLT_MAX;\n"
	"		for(int face=0;face<hullA->m_numFaces;face++)\n"
	"		{\n"
	"			const float4 Normal = make_float4(\n"
	"                                              faces[hullA->m_faceOffset+face].m_plane.x,\n"
	"                                              faces[hullA->m_faceOffset+face].m_plane.y,\n"
	"                                              faces[hullA->m_faceOffset+face].m_plane.z,\n"
	"                                              0.f);\n"
	"			const float4 faceANormalWS = qtRotate(ornA,Normal);\n"
	"            \n"
	"			float d = dot3F4(faceANormalWS,separatingNormal);\n"
	"			if (d < dmin)\n"
	"			{\n"
	"				dmin = d;\n"
	"				closestFaceA = face;\n"
	"                worldNormalsA1[pairIndex] = faceANormalWS;\n"
	"			}\n"
	"		}\n"
	"	}\n"
	"    \n"
	"    int numVerticesA = faces[hullA->m_faceOffset+closestFaceA].m_numIndices;\n"
	"	for(int e0=0;e0<numVerticesA;e0++)\n"
	"	{\n"
	"        const float4 a = vertices[hullA->m_vertexOffset+indices[faces[hullA->m_faceOffset+closestFaceA].m_indexOffset+e0]];\n"
	"        worldVertsA1[pairIndex*capacityWorldVerts+e0] = transform(&a, &posA,&ornA);\n"
	"    }\n"
	"    \n"
	"    clippingFaces[pairIndex].x = closestFaceA;\n"
	"    clippingFaces[pairIndex].y = closestFaceB;\n"
	"    clippingFaces[pairIndex].z = numVerticesA;\n"
	"    clippingFaces[pairIndex].w = numWorldVertsB1;\n"
	"    \n"
	"    \n"
	"	return numContactsOut;\n"
	"}\n"
	"int clipFaces(__global float4* worldVertsA1,\n"
	"              __global float4* worldNormalsA1,\n"
	"              __global float4* worldVertsB1,\n"
	"              __global float4* worldVertsB2, \n"
	"              int capacityWorldVertsB2,\n"
	"              const float minDist, float maxDist,\n"
	"              __global int4* clippingFaces,\n"
	"              int pairIndex)\n"
	"{\n"
	"	int numContactsOut = 0;\n"
	"    \n"
	"    int closestFaceA = clippingFaces[pairIndex].x;\n"
	"    int closestFaceB = clippingFaces[pairIndex].y;\n"
	"	int numVertsInA = clippingFaces[pairIndex].z;\n"
	"	int numVertsInB = clippingFaces[pairIndex].w;\n"
	"    \n"
	"	int numVertsOut = 0;\n"
	"    \n"
	"	if (closestFaceA<0)\n"
	"		return numContactsOut;\n"
	"    \n"
	"    __global float4* pVtxIn = &worldVertsB1[pairIndex*capacityWorldVertsB2];\n"
	"    __global float4* pVtxOut = &worldVertsB2[pairIndex*capacityWorldVertsB2];\n"
	"    \n"
	"    \n"
	"	\n"
	"	// clip polygon to back of planes of all faces of hull A that are adjacent to witness face\n"
	"    \n"
	"	for(int e0=0;e0<numVertsInA;e0++)\n"
	"	{\n"
	"		const float4 aw = worldVertsA1[pairIndex*capacityWorldVertsB2+e0];\n"
	"		const float4 bw = worldVertsA1[pairIndex*capacityWorldVertsB2+((e0+1)%numVertsInA)];\n"
	"		const float4 WorldEdge0 = aw - bw;\n"
	"		float4 worldPlaneAnormal1 = worldNormalsA1[pairIndex];\n"
	"		float4 planeNormalWS1 = -cross3(WorldEdge0,worldPlaneAnormal1);\n"
	"		float4 worldA1 = aw;\n"
	"		float planeEqWS1 = -dot3F4(worldA1,planeNormalWS1);\n"
	"		float4 planeNormalWS = planeNormalWS1;\n"
	"		float planeEqWS=planeEqWS1;\n"
	"		numVertsOut = clipFaceGlobal(pVtxIn, numVertsInB, planeNormalWS,planeEqWS, pVtxOut);\n"
	"		__global float4* tmp = pVtxOut;\n"
	"		pVtxOut = pVtxIn;\n"
	"		pVtxIn = tmp;\n"
	"		numVertsInB = numVertsOut;\n"
	"		numVertsOut = 0;\n"
	"	}\n"
	"    \n"
	"    //float4 planeNormalWS = worldNormalsA1[pairIndex];\n"
	"    //float planeEqWS=-dot3F4(planeNormalWS,worldVertsA1[pairIndex*capacityWorldVertsB2]);\n"
	"    \n"
	"    /*for (int i=0;i<numVertsInB;i++)\n"
	"    {\n"
	"        pVtxOut[i] = pVtxIn[i];\n"
	"    }*/\n"
	"    \n"
	"    \n"
	"    \n"
	"    \n"
	"    //numVertsInB=0;\n"
	"	\n"
	"    float4 planeNormalWS = worldNormalsA1[pairIndex];\n"
	"    float planeEqWS=-dot3F4(planeNormalWS,worldVertsA1[pairIndex*capacityWorldVertsB2]);\n"
	"    for (int i=0;i<numVertsInB;i++)\n"
	"    {\n"
	"        float depth = dot3F4(planeNormalWS,pVtxIn[i])+planeEqWS;\n"
	"        if (depth <=minDist)\n"
	"        {\n"
	"            depth = minDist;\n"
	"        }\n"
	"        \n"
	"        if (depth <=maxDist)\n"
	"        {\n"
	"            float4 pointInWorld = pVtxIn[i];\n"
	"            pVtxOut[numContactsOut++] = make_float4(pointInWorld.x,pointInWorld.y,pointInWorld.z,depth);\n"
	"        }\n"
	"    }\n"
	"   \n"
	"    clippingFaces[pairIndex].w =numContactsOut;\n"
	"   \n"
	"    \n"
	"	return numContactsOut;\n"
	"}\n"
	"__kernel void   findClippingFacesKernel(  __global const int4* pairs,\n"
	"                                        __global const b3RigidBodyData_t* rigidBodies,\n"
	"                                        __global const b3Collidable_t* collidables,\n"
	"                                        __global const b3ConvexPolyhedronData_t* convexShapes,\n"
	"                                        __global const float4* vertices,\n"
	"                                        __global const float4* uniqueEdges,\n"
	"                                        __global const b3GpuFace_t* faces,\n"
	"                                        __global const int* indices,\n"
	"                                        __global const float4* separatingNormals,\n"
	"                                        __global const int* hasSeparatingAxis,\n"
	"                                        __global int4* clippingFacesOut,\n"
	"                                        __global float4* worldVertsA1,\n"
	"                                        __global float4* worldNormalsA1,\n"
	"                                        __global float4* worldVertsB1,\n"
	"                                        int capacityWorldVerts,\n"
	"                                        int numPairs\n"
	"                                        )\n"
	"{\n"
	"    \n"
	"	int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"    \n"
	"	\n"
	"	float minDist = -1e30f;\n"
	"	float maxDist = 0.02f;\n"
	"    \n"
	"	if (i<numPairs)\n"
	"	{\n"
	"        \n"
	"		if (hasSeparatingAxis[i])\n"
	"		{\n"
	"            \n"
	"			int bodyIndexA = pairs[i].x;\n"
	"			int bodyIndexB = pairs[i].y;\n"
	"			\n"
	"			int collidableIndexA = rigidBodies[bodyIndexA].m_collidableIdx;\n"
	"			int collidableIndexB = rigidBodies[bodyIndexB].m_collidableIdx;\n"
	"			\n"
	"			int shapeIndexA = collidables[collidableIndexA].m_shapeIndex;\n"
	"			int shapeIndexB = collidables[collidableIndexB].m_shapeIndex;\n"
	"			\n"
	"            \n"
	"            \n"
	"			int numLocalContactsOut = findClippingFaces(separatingNormals[i],\n"
	"                                                        &convexShapes[shapeIndexA], &convexShapes[shapeIndexB],\n"
	"                                                        rigidBodies[bodyIndexA].m_pos,rigidBodies[bodyIndexA].m_quat,\n"
	"                                                        rigidBodies[bodyIndexB].m_pos,rigidBodies[bodyIndexB].m_quat,\n"
	"                                                        worldVertsA1,\n"
	"                                                        worldNormalsA1,\n"
	"                                                        worldVertsB1,capacityWorldVerts,\n"
	"                                                        minDist, maxDist,\n"
	"                                                        vertices,faces,indices,\n"
	"                                                        clippingFacesOut,i);\n"
	"            \n"
	"            \n"
	"		}//		if (hasSeparatingAxis[i])\n"
	"	}//	if (i<numPairs)\n"
	"    \n"
	"}\n"
	"__kernel void   clipFacesAndFindContactsKernel(    __global const float4* separatingNormals,\n"
	"                                                   __global const int* hasSeparatingAxis,\n"
	"                                                   __global int4* clippingFacesOut,\n"
	"                                                   __global float4* worldVertsA1,\n"
	"                                                   __global float4* worldNormalsA1,\n"
	"                                                   __global float4* worldVertsB1,\n"
	"                                                   __global float4* worldVertsB2,\n"
	"                                                    int vertexFaceCapacity,\n"
	"                                                   int numPairs,\n"
	"					                                        int debugMode\n"
	"                                                   )\n"
	"{\n"
	"    int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"    \n"
	"	float minDist = -1e30f;\n"
	"	float maxDist = 0.02f;\n"
	"    \n"
	"	if (i<numPairs)\n"
	"	{\n"
	"        \n"
	"		if (hasSeparatingAxis[i])\n"
	"		{\n"
	"            \n"
	"//			int bodyIndexA = pairs[i].x;\n"
	"	//		int bodyIndexB = pairs[i].y;\n"
	"		    \n"
	"            int numLocalContactsOut = 0;\n"
	"            int capacityWorldVertsB2 = vertexFaceCapacity;\n"
	"            \n"
	"            __global float4* pVtxIn = &worldVertsB1[pairIndex*capacityWorldVertsB2];\n"
	"            __global float4* pVtxOut = &worldVertsB2[pairIndex*capacityWorldVertsB2];\n"
	"            \n"
	"            {\n"
	"                __global int4* clippingFaces = clippingFacesOut;\n"
	"            \n"
	"                \n"
	"                int closestFaceA = clippingFaces[pairIndex].x;\n"
	"                int closestFaceB = clippingFaces[pairIndex].y;\n"
	"                int numVertsInA = clippingFaces[pairIndex].z;\n"
	"                int numVertsInB = clippingFaces[pairIndex].w;\n"
	"                \n"
	"                int numVertsOut = 0;\n"
	"                \n"
	"                if (closestFaceA>=0)\n"
	"                {\n"
	"                    \n"
	"                    \n"
	"                    \n"
	"                    // clip polygon to back of planes of all faces of hull A that are adjacent to witness face\n"
	"                    \n"
	"                    for(int e0=0;e0<numVertsInA;e0++)\n"
	"                    {\n"
	"                        const float4 aw = worldVertsA1[pairIndex*capacityWorldVertsB2+e0];\n"
	"                        const float4 bw = worldVertsA1[pairIndex*capacityWorldVertsB2+((e0+1)%numVertsInA)];\n"
	"                        const float4 WorldEdge0 = aw - bw;\n"
	"                        float4 worldPlaneAnormal1 = worldNormalsA1[pairIndex];\n"
	"                        float4 planeNormalWS1 = -cross3(WorldEdge0,worldPlaneAnormal1);\n"
	"                        float4 worldA1 = aw;\n"
	"                        float planeEqWS1 = -dot3F4(worldA1,planeNormalWS1);\n"
	"                        float4 planeNormalWS = planeNormalWS1;\n"
	"                        float planeEqWS=planeEqWS1;\n"
	"                        numVertsOut = clipFaceGlobal(pVtxIn, numVertsInB, planeNormalWS,planeEqWS, pVtxOut);\n"
	"                        __global float4* tmp = pVtxOut;\n"
	"                        pVtxOut = pVtxIn;\n"
	"                        pVtxIn = tmp;\n"
	"                        numVertsInB = numVertsOut;\n"
	"                        numVertsOut = 0;\n"
	"                    }\n"
	"                    \n"
	"                    float4 planeNormalWS = worldNormalsA1[pairIndex];\n"
	"                    float planeEqWS=-dot3F4(planeNormalWS,worldVertsA1[pairIndex*capacityWorldVertsB2]);\n"
	"                    \n"
	"                    for (int i=0;i<numVertsInB;i++)\n"
	"                    {\n"
	"                        float depth = dot3F4(planeNormalWS,pVtxIn[i])+planeEqWS;\n"
	"                        if (depth <=minDist)\n"
	"                        {\n"
	"                            depth = minDist;\n"
	"                        }\n"
	"                        \n"
	"                        if (depth <=maxDist)\n"
	"                        {\n"
	"                            float4 pointInWorld = pVtxIn[i];\n"
	"                            pVtxOut[numLocalContactsOut++] = make_float4(pointInWorld.x,pointInWorld.y,pointInWorld.z,depth);\n"
	"                        }\n"
	"                    }\n"
	"                    \n"
	"                }\n"
	"                clippingFaces[pairIndex].w =numLocalContactsOut;\n"
	"                \n"
	"            }\n"
	"            \n"
	"            for (int i=0;i<numLocalContactsOut;i++)\n"
	"                pVtxIn[i] = pVtxOut[i];\n"
	"                \n"
	"		}//		if (hasSeparatingAxis[i])\n"
	"	}//	if (i<numPairs)\n"
	"    \n"
	"}\n"
	"__kernel void   newContactReductionKernel( __global int4* pairs,\n"
	"                                                   __global const b3RigidBodyData_t* rigidBodies,\n"
	"                                                   __global const float4* separatingNormals,\n"
	"                                                   __global const int* hasSeparatingAxis,\n"
	"                                                   __global struct b3Contact4Data* globalContactsOut,\n"
	"                                                   __global int4* clippingFaces,\n"
	"                                                   __global float4* worldVertsB2,\n"
	"                                                   volatile __global int* nGlobalContactsOut,\n"
	"                                                   int vertexFaceCapacity,\n"
	"												   int contactCapacity,\n"
	"                                                   int numPairs\n"
	"                                                   )\n"
	"{\n"
	"    int i = get_global_id(0);\n"
	"	int pairIndex = i;\n"
	"	\n"
	"    int4 contactIdx;\n"
	"    contactIdx=make_int4(0,1,2,3);\n"
	"    \n"
	"	if (i<numPairs)\n"
	"	{\n"
	"        \n"
	"		if (hasSeparatingAxis[i])\n"
	"		{\n"
	"            \n"
	"			\n"
	"            \n"
	"            \n"
	"			int nPoints = clippingFaces[pairIndex].w;\n"
	"           \n"
	"            if (nPoints>0)\n"
	"            {\n"
	"                 __global float4* pointsIn = &worldVertsB2[pairIndex*vertexFaceCapacity];\n"
	"                float4 normal = -separatingNormals[i];\n"
	"                \n"
	"                int nReducedContacts = extractManifoldSequentialGlobal(pointsIn, nPoints, normal, &contactIdx);\n"
	"            \n"
	"				int mprContactIndex = pairs[pairIndex].z;\n"
	"                int dstIdx = mprContactIndex;\n"
	"				if (dstIdx<0)\n"
	"				{\n"
	"	                AppendInc( nGlobalContactsOut, dstIdx );\n"
	"				}\n"
	"//#if 0\n"
	"                \n"
	"				if (dstIdx < contactCapacity)\n"
	"				{\n"
	"					__global struct b3Contact4Data* c = &globalContactsOut[dstIdx];\n"
	"					c->m_worldNormalOnB = -normal;\n"
	"					c->m_restituitionCoeffCmp = (0.f*0xffff);c->m_frictionCoeffCmp = (0.7f*0xffff);\n"
	"					c->m_batchIdx = pairIndex;\n"
	"					int bodyA = pairs[pairIndex].x;\n"
	"					int bodyB = pairs[pairIndex].y;\n"
	"					pairs[pairIndex].w = dstIdx;\n"
	"					c->m_bodyAPtrAndSignBit = rigidBodies[bodyA].m_invMass==0?-bodyA:bodyA;\n"
	"					c->m_bodyBPtrAndSignBit = rigidBodies[bodyB].m_invMass==0?-bodyB:bodyB;\n"
	"                    c->m_childIndexA =-1;\n"
	"					c->m_childIndexB =-1;\n"
	"                    switch (nReducedContacts)\n"
	"                    {\n"
	"                        case 4:\n"
	"                            c->m_worldPosB[3] = pointsIn[contactIdx.w];\n"
	"                        case 3:\n"
	"                            c->m_worldPosB[2] = pointsIn[contactIdx.z];\n"
	"                        case 2:\n"
	"                            c->m_worldPosB[1] = pointsIn[contactIdx.y];\n"
	"                        case 1:\n"
	"							if (mprContactIndex<0)//test\n"
	"	                            c->m_worldPosB[0] = pointsIn[contactIdx.x];\n"
	"                        default:\n"
	"                        {\n"
	"                        }\n"
	"                    };\n"
	"                    \n"
	"					GET_NPOINTS(*c) = nReducedContacts;\n"
	"                    \n"
	"                 }\n"
	"                 \n"
	"                \n"
	"//#endif\n"
	"				\n"
	"			}//		if (numContactsOut>0)\n"
	"		}//		if (hasSeparatingAxis[i])\n"
	"	}//	if (i<numPairs)\n"
	"    \n"
	"    \n"
	"}\n";