ruby-bullet 0.0.2-x86-mingw32

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/MiniCL/MiniCLTaskWrap.cpp

@@ -0,0 +1,249 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2007 Erwin Coumans  http://bulletphysics.com
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include <MiniCL/cl_MiniCL_Defs.h>
+
+#define MSTRINGIFY(A) A
+#include "../OpenCLC10/ApplyForces.cl"
+#include "../OpenCLC10/Integrate.cl"
+#include "../OpenCLC10/PrepareLinks.cl"
+#include "../OpenCLC10/SolvePositions.cl"
+#include "../OpenCLC10/UpdateNodes.cl"
+#include "../OpenCLC10/UpdateNormals.cl"
+#include "../OpenCLC10/UpdatePositions.cl"
+#include "../OpenCLC10/UpdatePositionsFromVelocities.cl"
+#include "../OpenCLC10/VSolveLinks.cl"
+#include "../OpenCLC10/UpdateFixedVertexPositions.cl"
+//#include "../OpenCLC10/SolveCollisionsAndUpdateVelocities.cl"
+
+
+MINICL_REGISTER(PrepareLinksKernel)
+MINICL_REGISTER(VSolveLinksKernel)
+MINICL_REGISTER(UpdatePositionsFromVelocitiesKernel)
+MINICL_REGISTER(SolvePositionsFromLinksKernel)
+MINICL_REGISTER(updateVelocitiesFromPositionsWithVelocitiesKernel)
+MINICL_REGISTER(updateVelocitiesFromPositionsWithoutVelocitiesKernel)
+MINICL_REGISTER(IntegrateKernel)
+MINICL_REGISTER(ApplyForcesKernel)
+MINICL_REGISTER(ResetNormalsAndAreasKernel)
+MINICL_REGISTER(NormalizeNormalsAndAreasKernel)
+MINICL_REGISTER(UpdateSoftBodiesKernel)
+MINICL_REGISTER(UpdateFixedVertexPositions)
+
+float mydot3a(float4 a, float4 b)
+{
+   return a.x*b.x + a.y*b.y + a.z*b.z;
+}
+
+
+typedef struct 
+{
+	int firstObject;
+	int endObject;
+} CollisionObjectIndices;
+
+typedef struct 
+{
+	float4 shapeTransform[4]; // column major 4x4 matrix
+	float4 linearVelocity;
+	float4 angularVelocity;
+
+	int softBodyIdentifier;
+	int collisionShapeType;
+	
+
+	// Shape information
+	// Compressed from the union
+	float radius;
+	float halfHeight;
+	int upAxis;
+		
+	float margin;
+	float friction;
+
+	int padding0;
+	
+} CollisionShapeDescription;
+
+// From btBroadphaseProxy.h
+__constant int CAPSULE_SHAPE_PROXYTYPE = 10;
+
+// Multiply column-major matrix against vector
+float4 matrixVectorMul( float4 matrix[4], float4 vector )
+{
+	float4 returnVector;
+	float4 row0 = float4(matrix[0].x, matrix[1].x, matrix[2].x, matrix[3].x);
+	float4 row1 = float4(matrix[0].y, matrix[1].y, matrix[2].y, matrix[3].y);
+	float4 row2 = float4(matrix[0].z, matrix[1].z, matrix[2].z, matrix[3].z);
+	float4 row3 = float4(matrix[0].w, matrix[1].w, matrix[2].w, matrix[3].w);
+	returnVector.x = dot(row0, vector);
+	returnVector.y = dot(row1, vector);
+	returnVector.z = dot(row2, vector);
+	returnVector.w = dot(row3, vector);
+	return returnVector;
+}
+
+__kernel void 
+SolveCollisionsAndUpdateVelocitiesKernel( 
+	const int numNodes,
+	const float isolverdt,
+	__global int *g_vertexClothIdentifier,
+	__global float4 *g_vertexPreviousPositions,
+	__global float * g_perClothFriction,
+	__global float * g_clothDampingFactor,
+	__global CollisionObjectIndices * g_perClothCollisionObjectIndices,
+	__global CollisionShapeDescription * g_collisionObjectDetails,
+	__global float4 * g_vertexForces,
+	__global float4 *g_vertexVelocities,
+	__global float4 *g_vertexPositions GUID_ARG)
+{
+	int nodeID = get_global_id(0);
+	float4 forceOnVertex = (float4)(0.f, 0.f, 0.f, 0.f);
+	
+	if( get_global_id(0) < numNodes )
+	{	
+		int clothIdentifier = g_vertexClothIdentifier[nodeID];
+		
+		// Abort if this is not a valid cloth
+		if( clothIdentifier < 0 )
+			return;
+
+
+		float4 position (g_vertexPositions[nodeID].xyz, 1.f);
+		float4 previousPosition (g_vertexPreviousPositions[nodeID].xyz, 1.f);
+			
+		float clothFriction = g_perClothFriction[clothIdentifier];
+		float dampingFactor = g_clothDampingFactor[clothIdentifier];
+		float velocityCoefficient = (1.f - dampingFactor);		
+		float4 difference = position - previousPosition;
+		float4 velocity = difference*velocityCoefficient*isolverdt;
+		
+		CollisionObjectIndices collisionObjectIndices = g_perClothCollisionObjectIndices[clothIdentifier];
+	
+		int numObjects = collisionObjectIndices.endObject - collisionObjectIndices.firstObject;
+		
+		if( numObjects > 0 )
+		{
+			// We have some possible collisions to deal with
+			for( int collision = collisionObjectIndices.firstObject; collision < collisionObjectIndices.endObject; ++collision )
+			{
+				CollisionShapeDescription shapeDescription = g_collisionObjectDetails[collision];
+				float colliderFriction = shapeDescription.friction;
+
+				if( shapeDescription.collisionShapeType == CAPSULE_SHAPE_PROXYTYPE )
+				{
+					// Colliding with a capsule
+
+					float capsuleHalfHeight = shapeDescription.halfHeight;
+					float capsuleRadius = shapeDescription.radius;
+					float capsuleMargin = shapeDescription.margin;
+					int capsuleupAxis = shapeDescription.upAxis;
+
+					// Four columns of worldTransform matrix
+					float4 worldTransform[4];
+					worldTransform[0] = shapeDescription.shapeTransform[0];
+					worldTransform[1] = shapeDescription.shapeTransform[1];
+					worldTransform[2] = shapeDescription.shapeTransform[2];
+					worldTransform[3] = shapeDescription.shapeTransform[3];
+
+					// Correctly define capsule centerline vector 
+					float4 c1 (0.f, 0.f, 0.f, 1.f); 
+					float4 c2 (0.f, 0.f, 0.f, 1.f);
+					c1.x = select( 0.f, -capsuleHalfHeight, capsuleupAxis == 0 );
+					c1.y = select( 0.f, -capsuleHalfHeight, capsuleupAxis == 1 );
+					c1.z = select( 0.f, -capsuleHalfHeight, capsuleupAxis == 2 );
+					c2.x = -c1.x;
+					c2.y = -c1.y;
+					c2.z = -c1.z;
+
+
+					float4 worldC1 = matrixVectorMul(worldTransform, c1);
+					float4 worldC2 = matrixVectorMul(worldTransform, c2);
+					float4 segment = (worldC2 - worldC1);
+
+					// compute distance of tangent to vertex along line segment in capsule
+					float distanceAlongSegment = -( mydot3a( (worldC1 - position), segment ) / mydot3a(segment, segment) );
+
+					float4 closestPoint = (worldC1 + (segment * distanceAlongSegment));
+					float distanceFromLine = length(position - closestPoint);
+					float distanceFromC1 = length(worldC1 - position);
+					float distanceFromC2 = length(worldC2 - position);
+					
+					// Final distance from collision, point to push from, direction to push in
+					// for impulse force
+					float dist;
+					float4 normalVector;
+					if( distanceAlongSegment < 0 )
+					{
+						dist = distanceFromC1;
+						normalVector = float4(normalize(position - worldC1).xyz, 0.f);
+					} else if( distanceAlongSegment > 1.f ) {
+						dist = distanceFromC2;
+						normalVector = float4(normalize(position - worldC2).xyz, 0.f);	
+					} else {
+						dist = distanceFromLine;
+						normalVector = float4(normalize(position - closestPoint).xyz, 0.f);
+					}
+						
+					float4 colliderLinearVelocity = shapeDescription.linearVelocity;
+					float4 colliderAngularVelocity = shapeDescription.angularVelocity;
+					float4 velocityOfSurfacePoint = colliderLinearVelocity + cross(colliderAngularVelocity, position - float4(worldTransform[0].w, worldTransform[1].w, worldTransform[2].w, 0.f));
+
+					float minDistance = capsuleRadius + capsuleMargin;
+					
+					// In case of no collision, this is the value of velocity
+					velocity = (position - previousPosition) * velocityCoefficient * isolverdt;
+					
+					
+					// Check for a collision
+					if( dist < minDistance )
+					{
+						// Project back to surface along normal
+						position = position + float4(normalVector*(minDistance - dist)*0.9f);
+						velocity = (position - previousPosition) * velocityCoefficient * isolverdt;
+						float4 relativeVelocity = velocity - velocityOfSurfacePoint;
+
+						float4 p1 = normalize(cross(normalVector, segment));
+						float4 p2 = normalize(cross(p1, normalVector));
+						// Full friction is sum of velocities in each direction of plane
+						float4 frictionVector = p1*mydot3a(relativeVelocity, p1) + p2*mydot3a(relativeVelocity, p2);
+
+						// Real friction is peak friction corrected by friction coefficients
+						frictionVector = frictionVector * (colliderFriction*clothFriction);
+
+						float approachSpeed = dot(relativeVelocity, normalVector);
+
+						if( approachSpeed <= 0.0f )
+							forceOnVertex -= frictionVector;
+					}
+				}
+			}
+		}
+
+		g_vertexVelocities[nodeID] = float4(velocity.xyz, 0.f);	
+
+		// Update external force
+		g_vertexForces[nodeID] = float4(forceOnVertex.xyz, 0.f);
+
+		g_vertexPositions[nodeID] = float4(position.xyz, 0.f);
+	}
+}
+
+
+MINICL_REGISTER(SolveCollisionsAndUpdateVelocitiesKernel);
+
+
+
+

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/MiniCL/cmake_install.cmake

@@ -0,0 +1,29 @@
+# Install script for directory: C:/Users/abex/Documents/GitHub/ruby-bullet/deps/src/bullet-2.80-rev2531/src/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/MiniCL
+
+# Set the install prefix
+IF(NOT DEFINED CMAKE_INSTALL_PREFIX)
+  SET(CMAKE_INSTALL_PREFIX "..")
+ENDIF(NOT DEFINED CMAKE_INSTALL_PREFIX)
+STRING(REGEX REPLACE "/$" "" CMAKE_INSTALL_PREFIX "${CMAKE_INSTALL_PREFIX}")
+
+# Set the install configuration name.
+IF(NOT DEFINED CMAKE_INSTALL_CONFIG_NAME)
+  IF(BUILD_TYPE)
+    STRING(REGEX REPLACE "^[^A-Za-z0-9_]+" ""
+           CMAKE_INSTALL_CONFIG_NAME "${BUILD_TYPE}")
+  ELSE(BUILD_TYPE)
+    SET(CMAKE_INSTALL_CONFIG_NAME "Release")
+  ENDIF(BUILD_TYPE)
+  MESSAGE(STATUS "Install configuration: \"${CMAKE_INSTALL_CONFIG_NAME}\"")
+ENDIF(NOT DEFINED CMAKE_INSTALL_CONFIG_NAME)
+
+# Set the component getting installed.
+IF(NOT CMAKE_INSTALL_COMPONENT)
+  IF(COMPONENT)
+    MESSAGE(STATUS "Install component: \"${COMPONENT}\"")
+    SET(CMAKE_INSTALL_COMPONENT "${COMPONENT}")
+  ELSE(COMPONENT)
+    SET(CMAKE_INSTALL_COMPONENT)
+  ENDIF(COMPONENT)
+ENDIF(NOT CMAKE_INSTALL_COMPONENT)
+

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/NVidia/CMakeLists.txt

@@ -0,0 +1,81 @@
+
+ADD_DEFINITIONS(-DUSE_NVIDIA_OPENCL)
+ADD_DEFINITIONS(-DCL_PLATFORM_NVIDIA)
+
+INCLUDE_DIRECTORIES(
+	${BULLET_PHYSICS_SOURCE_DIR}/src
+	${NVIDIA_OPENCL_INCLUDES}
+)
+
+
+
+SET(BulletSoftBodyOpenCLSolvers_SRCS
+	../btSoftBodySolver_OpenCL.cpp
+	../btSoftBodySolver_OpenCLSIMDAware.cpp
+	../btSoftBodySolverOutputCLtoGL.cpp
+)
+
+SET(BulletSoftBodyOpenCLSolvers_HDRS
+	../btSoftBodySolver_OpenCL.h
+	../../Shared/btSoftBodySolverData.h
+	../btSoftBodySolverVertexData_OpenCL.h
+	../btSoftBodySolverTriangleData_OpenCL.h
+	../btSoftBodySolverLinkData_OpenCL.h
+	../btSoftBodySolverLinkData_OpenCLSIMDAware.h
+	../btSoftBodySolverBuffer_OpenCL.h
+	../btSoftBodySolverVertexBuffer_OpenGL.h
+	../btSoftBodySolverOutputCLtoGL.h
+)
+
+# OpenCL and HLSL Shaders.
+# Build rules generated to stringify these into headers
+# which are needed by some of the sources
+SET(BulletSoftBodyOpenCLSolvers_Shaders
+#	OutputToVertexArray
+	UpdateNormals
+	Integrate
+	UpdatePositions
+	UpdateNodes
+	SolvePositions
+	UpdatePositionsFromVelocities
+	ApplyForces
+	PrepareLinks
+	VSolveLinks
+)
+
+foreach(f ${BulletSoftBodyOpenCLSolvers_Shaders})
+    LIST(APPEND BulletSoftBodyOpenCLSolvers_OpenCLC "../OpenCLC10/${f}.cl")
+endforeach(f) 
+
+
+
+ADD_LIBRARY(BulletSoftBodySolvers_OpenCL_NVidia
+	${BulletSoftBodyOpenCLSolvers_SRCS} 
+	${BulletSoftBodyOpenCLSolvers_HDRS} 
+	${BulletSoftBodyOpenCLSolvers_OpenCLC}
+)
+
+SET_TARGET_PROPERTIES(BulletSoftBodySolvers_OpenCL_NVidia PROPERTIES VERSION ${BULLET_VERSION})
+SET_TARGET_PROPERTIES(BulletSoftBodySolvers_OpenCL_NVidia PROPERTIES SOVERSION ${BULLET_VERSION})
+IF (BUILD_SHARED_LIBS)
+	TARGET_LINK_LIBRARIES(BulletSoftBodySolvers_OpenCL_NVidia BulletSoftBody BulletDynamics)
+ENDIF (BUILD_SHARED_LIBS)
+
+
+IF (INSTALL_LIBS)
+	IF (NOT INTERNAL_CREATE_DISTRIBUTABLE_MSVC_PROJECTFILES)
+		IF (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} GREATER 2.5)
+			IF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
+				INSTALL(TARGETS BulletSoftBodySolvers_OpenCL_NVidia DESTINATION .)
+			ELSE (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
+				INSTALL(TARGETS BulletSoftBodySolvers_OpenCL_NVidia DESTINATION lib${LIB_SUFFIX})
+#headers are already installed by BulletMultiThreaded library
+			ENDIF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
+		ENDIF (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} GREATER 2.5)
+
+		IF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
+			SET_TARGET_PROPERTIES(BulletSoftBodySolvers_OpenCL_NVidia PROPERTIES FRAMEWORK true)
+			SET_TARGET_PROPERTIES(BulletSoftBodySolvers_OpenCL_NVidia PROPERTIES PUBLIC_HEADER "${BulletSoftBodyOpenCLSolvers_HDRS}")
+		ENDIF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
+	ENDIF (NOT INTERNAL_CREATE_DISTRIBUTABLE_MSVC_PROJECTFILES)
+ENDIF (INSTALL_LIBS)

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/NVidia/premake4.lua

@@ -0,0 +1,27 @@
+	
+hasCL = findOpenCL_NVIDIA()
+	
+if (hasCL) then
+	
+	project "BulletSoftBodySolvers_OpenCL_NVIDIA"
+		
+ 	defines { "USE_NVIDIA_OPENCL","CL_PLATFORM_NVIDIA"}
+
+	initOpenCL_NVIDIA()
+	
+	kind "StaticLib"
+	
+	targetdir "../../../../../lib"
+	
+	includedirs {
+		".",
+		"../../../..",
+		"../../../../../Glut"
+	}
+	files {
+		"../btSoftBodySolver_OpenCL.cpp",
+		"../btSoftBodySolver_OpenCLSIMDAware.cpp",
+		"../btSoftBodySolverOutputCLtoGL.cpp"
+	}
+
+end

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/OpenCLC10/ApplyForces.cl

@@ -0,0 +1,102 @@
+MSTRINGIFY(
+
+
+float adot3(float4 a, float4 b)
+{
+   return a.x*b.x + a.y*b.y + a.z*b.z;
+}
+
+float alength3(float4 a)
+{
+	a.w = 0;
+	return length(a);
+}
+
+float4 anormalize3(float4 a)
+{
+	a.w = 0;
+	return normalize(a);
+}
+
+float4 projectOnAxis( float4 v, float4 a )
+{
+	return (a*adot3(v, a));
+}
+
+__kernel void 
+ApplyForcesKernel(
+	const uint numNodes,
+	const float solverdt,
+	const float epsilon,
+	__global int * g_vertexClothIdentifier,
+	__global float4 * g_vertexNormal,
+	__global float * g_vertexArea,
+	__global float * g_vertexInverseMass,
+	__global float * g_clothLiftFactor,
+	__global float * g_clothDragFactor,
+	__global float4 * g_clothWindVelocity,
+	__global float4 * g_clothAcceleration,
+	__global float * g_clothMediumDensity,
+	__global float4 * g_vertexForceAccumulator,
+	__global float4 * g_vertexVelocity GUID_ARG)
+{
+	unsigned int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{		
+		int clothId  = g_vertexClothIdentifier[nodeID];
+		float nodeIM = g_vertexInverseMass[nodeID];
+		
+		if( nodeIM > 0.0f )
+		{
+			float4 nodeV  = g_vertexVelocity[nodeID];
+			float4 normal = g_vertexNormal[nodeID];
+			float area    = g_vertexArea[nodeID];
+			float4 nodeF  = g_vertexForceAccumulator[nodeID];
+			
+			// Read per-cloth values
+			float4 clothAcceleration = g_clothAcceleration[clothId];
+			float4 clothWindVelocity = g_clothWindVelocity[clothId];
+			float liftFactor = g_clothLiftFactor[clothId];
+			float dragFactor = g_clothDragFactor[clothId];
+			float mediumDensity = g_clothMediumDensity[clothId];
+		
+			// Apply the acceleration to the cloth rather than do this via a force
+			nodeV += (clothAcceleration*solverdt);
+
+			g_vertexVelocity[nodeID] = nodeV;
+
+			// Aerodynamics
+			float4 rel_v = nodeV - clothWindVelocity;
+			float rel_v_len = alength3(rel_v);
+			float rel_v2 = dot(rel_v, rel_v);
+			
+			if( rel_v2 > epsilon )
+			{
+				float4 rel_v_nrm = anormalize3(rel_v);
+				float4 nrm = normal;
+									
+				nrm = nrm * (dot(nrm, rel_v) < 0 ? -1.f : 1.f);
+
+				float4 fDrag = (float4)(0.f, 0.f, 0.f, 0.f);
+				float4 fLift = (float4)(0.f, 0.f, 0.f, 0.f);
+
+				float n_dot_v = dot(nrm, rel_v_nrm);
+
+				// drag force
+				if ( dragFactor > 0.f )
+					fDrag = 0.5f * dragFactor * mediumDensity * rel_v2 * area * n_dot_v * (-1.0f) * rel_v_nrm;
+
+				// lift force
+				// Check angle of attack
+				// cos(10�) = 0.98480
+				if ( 0 < n_dot_v && n_dot_v < 0.98480f)
+					fLift = 0.5f * liftFactor * mediumDensity * rel_v_len * area * sqrt(1.0f-n_dot_v*n_dot_v) * (cross(cross(nrm, rel_v_nrm), rel_v_nrm));
+				
+				nodeF += fDrag + fLift;
+					g_vertexForceAccumulator[nodeID] = nodeF;	
+			}
+		}
+	}
+}
+
+);

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/OpenCLC10/ComputeBounds.cl

@@ -0,0 +1,82 @@
+MSTRINGIFY(
+#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n
+#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n
+
+__kernel void
+ComputeBoundsKernel( 
+	const int numNodes,
+	const int numSoftBodies,
+	__global int * g_vertexClothIdentifier,
+	__global float4 * g_vertexPositions,
+	/* Unfortunately, to get the atomics below to work these arrays cannot be */
+	/* uint4, though that is the layout of the data */
+	/* Therefore this is little-endian-only code */
+	volatile __global uint * g_clothMinBounds,
+	volatile __global uint * g_clothMaxBounds,
+	volatile __local uint * clothMinBounds,
+	volatile __local uint * clothMaxBounds)
+{
+	// Init min and max bounds arrays
+	if( get_local_id(0) < numSoftBodies )
+	{
+		
+		clothMinBounds[get_local_id(0)*4] = UINT_MAX;
+		clothMinBounds[get_local_id(0)*4+1] = UINT_MAX;
+		clothMinBounds[get_local_id(0)*4+2] = UINT_MAX;
+		clothMinBounds[get_local_id(0)*4+3] = UINT_MAX;
+		clothMaxBounds[get_local_id(0)*4] = 0;
+		clothMaxBounds[get_local_id(0)*4+1] = 0;
+		clothMaxBounds[get_local_id(0)*4+2] = 0;
+		clothMaxBounds[get_local_id(0)*4+3] = 0;
+
+	}
+
+	barrier(CLK_LOCAL_MEM_FENCE);
+
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{	
+		int clothIdentifier = g_vertexClothIdentifier[nodeID];
+		if( clothIdentifier >= 0 )
+		{
+
+			float4 position = (float4)(g_vertexPositions[nodeID].xyz, 0.f);
+
+			/* Reinterpret position as uint */
+			uint4 positionUInt = (uint4)(as_uint(position.x), as_uint(position.y), as_uint(position.z), 0);
+		
+			/* Invert sign bit of positives and whole of negatives to allow comparison as unsigned ints */
+			positionUInt.x ^= (1+~(positionUInt.x >> 31) | 0x80000000);
+			positionUInt.y ^= (1+~(positionUInt.y >> 31) | 0x80000000);		
+			positionUInt.z ^= (1+~(positionUInt.z >> 31) | 0x80000000);
+		
+			// Min/max with the LDS values
+			atom_min(&(clothMinBounds[clothIdentifier*4]), positionUInt.x);
+			atom_min(&(clothMinBounds[clothIdentifier*4+1]), positionUInt.y);
+			atom_min(&(clothMinBounds[clothIdentifier*4+2]), positionUInt.z);
+
+			atom_max(&(clothMaxBounds[clothIdentifier*4]), positionUInt.x);
+			atom_max(&(clothMaxBounds[clothIdentifier*4+1]), positionUInt.y);
+			atom_max(&(clothMaxBounds[clothIdentifier*4+2]), positionUInt.z);
+		}
+	}
+	
+	barrier(CLK_LOCAL_MEM_FENCE);
+
+
+	/* Use global atomics to update the global versions of the data */
+	if( get_local_id(0) < numSoftBodies )
+	{
+		/*atom_min(&(g_clothMinBounds[get_local_id(0)].x), clothMinBounds[get_local_id(0)].x);*/
+		atom_min(&(g_clothMinBounds[get_local_id(0)*4]), clothMinBounds[get_local_id(0)*4]);
+		atom_min(&(g_clothMinBounds[get_local_id(0)*4+1]), clothMinBounds[get_local_id(0)*4+1]);
+		atom_min(&(g_clothMinBounds[get_local_id(0)*4+2]), clothMinBounds[get_local_id(0)*4+2]);
+
+		atom_max(&(g_clothMaxBounds[get_local_id(0)*4]), clothMaxBounds[get_local_id(0)*4]);		
+		atom_max(&(g_clothMaxBounds[get_local_id(0)*4+1]), clothMaxBounds[get_local_id(0)*4+1]);
+		atom_max(&(g_clothMaxBounds[get_local_id(0)*4+2]), clothMaxBounds[get_local_id(0)*4+2]);
+	}
+}
+
+
+);

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/OpenCLC10/Integrate.cl

@@ -0,0 +1,35 @@
+MSTRINGIFY(
+
+// Node indices for each link
+
+
+
+__kernel void
+IntegrateKernel( 
+	const int numNodes,
+	const float solverdt,
+	__global float * g_vertexInverseMasses,
+	__global float4 * g_vertexPositions,
+	__global float4 * g_vertexVelocity,
+	__global float4 * g_vertexPreviousPositions,
+	__global float4 * g_vertexForceAccumulator GUID_ARG)
+{
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{	
+		float4 position   = g_vertexPositions[nodeID];
+		float4 velocity   = g_vertexVelocity[nodeID];
+		float4 force      = g_vertexForceAccumulator[nodeID];
+		float inverseMass = g_vertexInverseMasses[nodeID];
+		
+		g_vertexPreviousPositions[nodeID] = position;
+		velocity += force * inverseMass * solverdt;
+		position += velocity * solverdt;
+		
+		g_vertexForceAccumulator[nodeID] = (float4)(0.f, 0.f, 0.f, 0.0f);
+		g_vertexPositions[nodeID]        = position;
+		g_vertexVelocity[nodeID]         = velocity;	
+	}
+}
+
+);

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/OpenCLC10/OutputToVertexArray.cl

@@ -0,0 +1,46 @@
+MSTRINGIFY(
+
+__kernel void 
+OutputToVertexArrayWithNormalsKernel( 
+	const int startNode, const int numNodes, __global float *g_vertexBuffer,
+	const int positionOffset, const int positionStride, const __global float4* g_vertexPositions, 
+	const int normalOffset, const int normalStride, const __global float4* g_vertexNormals  )
+{
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{			
+		float4 position = g_vertexPositions[nodeID + startNode];
+		float4 normal = g_vertexNormals[nodeID + startNode];
+		
+		// Stride should account for the float->float4 conversion
+		int positionDestination = nodeID * positionStride + positionOffset;		
+		g_vertexBuffer[positionDestination] = position.x;
+		g_vertexBuffer[positionDestination+1] = position.y;
+		g_vertexBuffer[positionDestination+2] = position.z;
+		
+		int normalDestination = nodeID * normalStride + normalOffset;
+		g_vertexBuffer[normalDestination] = normal.x;
+		g_vertexBuffer[normalDestination+1] = normal.y;
+		g_vertexBuffer[normalDestination+2] = normal.z;		
+	}
+}
+
+__kernel void 
+OutputToVertexArrayWithoutNormalsKernel(
+	const int startNode, const int numNodes, __global float *g_vertexBuffer,
+	const int positionOffset, const int positionStride, const __global float4* g_vertexPositions )
+{
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{			
+		float4 position = g_vertexPositions[nodeID + startNode];
+		
+		// Stride should account for the float->float4 conversion
+		int positionDestination = nodeID * positionStride + positionOffset;		
+		g_vertexBuffer[positionDestination] = position.x;
+		g_vertexBuffer[positionDestination+1] = position.y;
+		g_vertexBuffer[positionDestination+2] = position.z;		
+	}
+}
+
+);

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/OpenCLC10/PrepareLinks.cl

@@ -0,0 +1,38 @@
+MSTRINGIFY(
+
+
+
+__kernel void 
+PrepareLinksKernel( 
+	const int numLinks,
+	__global int2 * g_linksVertexIndices,
+	__global float * g_linksMassLSC,
+	__global float4 * g_nodesPreviousPosition,
+	__global float * g_linksLengthRatio,
+	__global float4 * g_linksCurrentLength GUID_ARG)
+{
+	int linkID = get_global_id(0);
+	if( linkID < numLinks )
+	{	
+		
+		int2 nodeIndices = g_linksVertexIndices[linkID];
+		int node0 = nodeIndices.x;
+		int node1 = nodeIndices.y;
+		
+		float4 nodePreviousPosition0 = g_nodesPreviousPosition[node0];
+		float4 nodePreviousPosition1 = g_nodesPreviousPosition[node1];
+
+		float massLSC = g_linksMassLSC[linkID];
+		
+		float4 linkCurrentLength = nodePreviousPosition1 - nodePreviousPosition0;
+		linkCurrentLength.w = 0.f;
+		
+		float linkLengthRatio = dot(linkCurrentLength, linkCurrentLength)*massLSC;
+		linkLengthRatio = 1.0f/linkLengthRatio;
+		
+		g_linksCurrentLength[linkID] = linkCurrentLength;
+		g_linksLengthRatio[linkID]   = linkLengthRatio;		
+	}
+}
+
+);