ruby-bullet 0.0.2-x86-mingw32

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

@@ -0,0 +1,39 @@
+MSTRINGIFY(
+
+
+__kernel void 
+updateVelocitiesFromPositionsWithVelocitiesKernel( 
+	int numNodes,
+	float isolverdt,
+	__global float4 * g_vertexPositions,
+	__global float4 * g_vertexPreviousPositions,
+	__global int * g_vertexClothIndices,
+	__global float *g_clothVelocityCorrectionCoefficients,
+	__global float * g_clothDampingFactor,
+	__global float4 * g_vertexVelocities,
+	__global float4 * g_vertexForces GUID_ARG)
+{
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{	
+		float4 position = g_vertexPositions[nodeID];
+		float4 previousPosition = g_vertexPreviousPositions[nodeID];
+		float4 velocity = g_vertexVelocities[nodeID];
+		int clothIndex = g_vertexClothIndices[nodeID];
+		float velocityCorrectionCoefficient = g_clothVelocityCorrectionCoefficients[clothIndex];
+		float dampingFactor = g_clothDampingFactor[clothIndex];
+		float velocityCoefficient = (1.f - dampingFactor);
+		
+		float4 difference = position - previousPosition;
+				
+		velocity += difference*velocityCorrectionCoefficient*isolverdt;
+		
+		// Damp the velocity
+		velocity *= velocityCoefficient;
+		
+		g_vertexVelocities[nodeID] = velocity;
+		g_vertexForces[nodeID] = (float4)(0.f, 0.f, 0.f, 0.f);								
+	}
+}
+
+);

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

@@ -0,0 +1,102 @@
+MSTRINGIFY(
+
+float length3(float4 a)
+{
+	a.w = 0;
+	return length(a);
+}
+
+float4 normalize3(float4 a)
+{
+	a.w = 0;
+	return normalize(a);
+}
+
+__kernel void 
+ResetNormalsAndAreasKernel(
+	const unsigned int numNodes,
+	__global float4 * g_vertexNormals,
+	__global float * g_vertexArea GUID_ARG)
+{
+	if( get_global_id(0) < numNodes )
+	{
+		g_vertexNormals[get_global_id(0)] = (float4)(0.0f, 0.0f, 0.0f, 0.0f);
+		g_vertexArea[get_global_id(0)]    = 0.0f;
+	}
+}
+
+
+__kernel void 
+UpdateSoftBodiesKernel(
+	const unsigned int startFace,
+	const unsigned int numFaces,
+	__global int4 * g_triangleVertexIndexSet,
+	__global float4 * g_vertexPositions,
+	__global float4 * g_vertexNormals,
+	__global float * g_vertexArea,
+	__global float4 * g_triangleNormals,
+	__global float * g_triangleArea GUID_ARG)
+{
+	int faceID = get_global_id(0) + startFace;
+	if( get_global_id(0) < numFaces )
+	{		
+		int4 triangleIndexSet = g_triangleVertexIndexSet[ faceID ];
+		int nodeIndex0 = triangleIndexSet.x;
+		int nodeIndex1 = triangleIndexSet.y;
+		int nodeIndex2 = triangleIndexSet.z;
+
+		float4 node0 = g_vertexPositions[nodeIndex0];
+		float4 node1 = g_vertexPositions[nodeIndex1];
+		float4 node2 = g_vertexPositions[nodeIndex2];
+		float4 nodeNormal0 = g_vertexNormals[nodeIndex0];
+		float4 nodeNormal1 = g_vertexNormals[nodeIndex1];
+		float4 nodeNormal2 = g_vertexNormals[nodeIndex2];
+		float vertexArea0 = g_vertexArea[nodeIndex0];
+		float vertexArea1 = g_vertexArea[nodeIndex1];
+		float vertexArea2 = g_vertexArea[nodeIndex2];
+		
+		float4 vector0 = node1 - node0;
+		float4 vector1 = node2 - node0;
+		
+		float4 faceNormal = cross(vector0, vector1);
+		float triangleArea = length(faceNormal);
+
+		nodeNormal0 = nodeNormal0 + faceNormal;
+		nodeNormal1 = nodeNormal1 + faceNormal;
+		nodeNormal2 = nodeNormal2 + faceNormal;
+		vertexArea0 = vertexArea0 + triangleArea;
+		vertexArea1 = vertexArea1 + triangleArea;
+		vertexArea2 = vertexArea2 + triangleArea;
+		
+		g_triangleNormals[faceID] = normalize3(faceNormal);
+		g_vertexNormals[nodeIndex0] = nodeNormal0;
+		g_vertexNormals[nodeIndex1] = nodeNormal1;
+		g_vertexNormals[nodeIndex2] = nodeNormal2;
+		g_triangleArea[faceID] = triangleArea;
+		g_vertexArea[nodeIndex0] = vertexArea0;
+		g_vertexArea[nodeIndex1] = vertexArea1;
+		g_vertexArea[nodeIndex2] = vertexArea2;
+	}
+}
+
+__kernel void 
+NormalizeNormalsAndAreasKernel( 
+	const unsigned int numNodes,
+	__global int * g_vertexTriangleCount,
+	__global float4 * g_vertexNormals,
+	__global float * g_vertexArea GUID_ARG)
+{
+	if( get_global_id(0) < numNodes )
+	{
+		float4 normal = g_vertexNormals[get_global_id(0)];
+		float area = g_vertexArea[get_global_id(0)];
+		int numTriangles = g_vertexTriangleCount[get_global_id(0)];
+		
+		float vectorLength = length3(normal);
+		
+		g_vertexNormals[get_global_id(0)] = normalize3(normal);
+		g_vertexArea[get_global_id(0)] = area/(float)(numTriangles);
+	}
+}
+
+);

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

@@ -0,0 +1,34 @@
+MSTRINGIFY(
+
+__kernel void 
+updateVelocitiesFromPositionsWithoutVelocitiesKernel( 
+	const int numNodes,
+	const float isolverdt,
+	__global float4 * g_vertexPositions,
+	__global float4 * g_vertexPreviousPositions,
+	__global int * g_vertexClothIndices,
+	__global float * g_clothDampingFactor,
+	__global float4 * g_vertexVelocities,
+	__global float4 * g_vertexForces GUID_ARG)
+
+{
+	int nodeID = get_global_id(0);
+	if( nodeID < numNodes )
+	{	
+		float4 position = g_vertexPositions[nodeID];
+		float4 previousPosition = g_vertexPreviousPositions[nodeID];
+		float4 velocity = g_vertexVelocities[nodeID];
+		int clothIndex = g_vertexClothIndices[nodeID];
+		float dampingFactor = g_clothDampingFactor[clothIndex];
+		float velocityCoefficient = (1.f - dampingFactor);
+		
+		float4 difference = position - previousPosition;
+				
+		velocity = difference*velocityCoefficient*isolverdt;		
+		
+		g_vertexVelocities[nodeID] = velocity;
+		g_vertexForces[nodeID] = (float4)(0.f, 0.f, 0.f, 0.f);								
+	}
+}
+
+);

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

@@ -0,0 +1,28 @@
+
+MSTRINGIFY(
+
+
+
+
+__kernel void 
+UpdatePositionsFromVelocitiesKernel( 
+	const int numNodes,
+	const float solverSDT,
+	__global float4 * g_vertexVelocities,
+	__global float4 * g_vertexPreviousPositions,
+	__global float4 * g_vertexCurrentPosition GUID_ARG)
+{
+	int vertexID = get_global_id(0);
+	if( vertexID < numNodes )
+	{	
+		float4 previousPosition = g_vertexPreviousPositions[vertexID];
+		float4 velocity         = g_vertexVelocities[vertexID];
+		
+		float4 newPosition      = previousPosition + velocity*solverSDT;
+		
+		g_vertexCurrentPosition[vertexID]   = newPosition;
+		g_vertexPreviousPositions[vertexID] = newPosition;
+	}
+}
+
+);

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

@@ -0,0 +1,45 @@
+MSTRINGIFY(
+
+__kernel void 
+VSolveLinksKernel( 
+	int startLink,
+	int numLinks,
+	float kst,
+	__global int2 * g_linksVertexIndices,
+	__global float * g_linksLengthRatio,
+	__global float4 * g_linksCurrentLength,
+	__global float * g_vertexInverseMass,
+	__global float4 * g_vertexVelocity GUID_ARG)
+{
+	int linkID = get_global_id(0) + startLink;
+	if( get_global_id(0) < numLinks )
+	{		
+		int2 nodeIndices = g_linksVertexIndices[linkID];
+		int node0 = nodeIndices.x;
+		int node1 = nodeIndices.y;
+		
+		float linkLengthRatio = g_linksLengthRatio[linkID];
+		float3 linkCurrentLength = g_linksCurrentLength[linkID].xyz;
+		
+		float3 vertexVelocity0 = g_vertexVelocity[node0].xyz;
+		float3 vertexVelocity1 = g_vertexVelocity[node1].xyz;
+
+		float vertexInverseMass0 = g_vertexInverseMass[node0];
+		float vertexInverseMass1 = g_vertexInverseMass[node1]; 
+
+		float3 nodeDifference = vertexVelocity0 - vertexVelocity1;
+		float dotResult = dot(linkCurrentLength, nodeDifference);
+		float j = -dotResult*linkLengthRatio*kst;
+		
+		float3 velocityChange0 = linkCurrentLength*(j*vertexInverseMass0);
+		float3 velocityChange1 = linkCurrentLength*(j*vertexInverseMass1);
+		
+		vertexVelocity0 += velocityChange0;
+		vertexVelocity1 -= velocityChange1;
+
+		g_vertexVelocity[node0] = (float4)(vertexVelocity0, 0.f);
+		g_vertexVelocity[node1] = (float4)(vertexVelocity1, 0.f);
+	}
+}
+
+);

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/btSoftBodySolverBuffer_OpenCL.h

@@ -0,0 +1,209 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#ifndef BT_SOFT_BODY_SOLVER_BUFFER_OPENCL_H
+#define BT_SOFT_BODY_SOLVER_BUFFER_OPENCL_H
+
+// OpenCL support
+
+#ifdef USE_MINICL
+	#include "MiniCL/cl.h"
+#else //USE_MINICL
+	#ifdef __APPLE__
+		#include <OpenCL/OpenCL.h>
+	#else
+		#include <CL/cl.h>
+	#endif //__APPLE__
+#endif//USE_MINICL
+
+#ifndef SAFE_RELEASE
+#define SAFE_RELEASE(p)      { if(p) { (p)->Release(); (p)=NULL; } }
+#endif
+
+template <typename ElementType> class btOpenCLBuffer
+{
+public:
+
+	cl_command_queue	m_cqCommandQue;
+	cl_context			m_clContext;
+	cl_mem				m_buffer;
+
+
+
+	btAlignedObjectArray< ElementType > * m_CPUBuffer;
+	
+	int  m_gpuSize;
+	bool m_onGPU;
+	bool m_readOnlyOnGPU;
+	bool m_allocated;
+
+
+	bool createBuffer( cl_mem* preexistingBuffer = 0)
+	{
+
+		cl_int err;
+		 
+
+		if( preexistingBuffer )
+		{
+			m_buffer = *preexistingBuffer;
+		} 
+		else {
+
+			cl_mem_flags flags= m_readOnlyOnGPU ? CL_MEM_READ_ONLY : CL_MEM_READ_WRITE;
+
+			size_t size = m_CPUBuffer->size() * sizeof(ElementType);
+			// At a minimum the buffer must exist
+			if( size == 0 )
+				size = sizeof(ElementType);
+			m_buffer = clCreateBuffer(m_clContext, flags, size, 0, &err);
+			if( err != CL_SUCCESS )
+			{
+				btAssert( "Buffer::Buffer(m_buffer)");
+			}
+		}
+
+		m_gpuSize = m_CPUBuffer->size();
+
+		return true;
+	}
+
+public:
+	btOpenCLBuffer( cl_command_queue	commandQue,cl_context ctx, btAlignedObjectArray< ElementType >* CPUBuffer, bool readOnly)
+		:m_cqCommandQue(commandQue),
+		m_clContext(ctx),
+		m_buffer(0),
+		m_CPUBuffer(CPUBuffer),
+		m_gpuSize(0),
+		m_onGPU(false),
+		m_readOnlyOnGPU(readOnly),
+		m_allocated(false)
+	{
+	}
+
+	~btOpenCLBuffer()
+	{
+		clReleaseMemObject(m_buffer);
+	}
+
+
+	bool moveToGPU()
+	{
+
+
+		cl_int err;
+
+		if( (m_CPUBuffer->size() != m_gpuSize) )
+		{
+			m_onGPU = false;
+		}
+
+		if( !m_allocated && m_CPUBuffer->size() == 0  )
+		{
+			// If it isn't on the GPU and yet there is no data on the CPU side this may cause a problem with some kernels.
+			// We should create *something* on the device side
+			if (!createBuffer()) {
+				return false;
+			}
+			m_allocated = true;
+		}
+
+		if( !m_onGPU && m_CPUBuffer->size() > 0 )
+		{
+			if (!m_allocated || (m_CPUBuffer->size() != m_gpuSize)) {
+				if (!createBuffer()) {
+					return false;
+				}
+				m_allocated = true;
+			}
+			
+			size_t size = m_CPUBuffer->size() * sizeof(ElementType);
+			err = clEnqueueWriteBuffer(m_cqCommandQue,m_buffer,
+				CL_FALSE,
+				0,
+				size, 
+				&((*m_CPUBuffer)[0]),0,0,0);
+			if( err != CL_SUCCESS )
+			{
+				btAssert( "CommandQueue::enqueueWriteBuffer(m_buffer)" );
+			}
+
+			m_onGPU = true;
+		}
+
+		return true;
+
+	}
+
+	bool moveFromGPU()
+	{
+
+		cl_int err;
+
+		if (m_CPUBuffer->size() > 0) {
+			if (m_onGPU && !m_readOnlyOnGPU) {
+				size_t size = m_CPUBuffer->size() * sizeof(ElementType);
+				err = clEnqueueReadBuffer(m_cqCommandQue,
+					m_buffer,
+					CL_TRUE,
+					0,
+					size,
+					&((*m_CPUBuffer)[0]),0,0,0);
+
+				if( err != CL_SUCCESS )
+				{
+					btAssert( "CommandQueue::enqueueReadBuffer(m_buffer)" );
+				}
+
+				m_onGPU = false;
+			}
+		}
+
+		return true;
+	}
+
+	bool copyFromGPU()
+	{
+
+		cl_int err;
+		size_t size = m_CPUBuffer->size() * sizeof(ElementType);
+
+		if (m_CPUBuffer->size() > 0) {
+			if (m_onGPU && !m_readOnlyOnGPU) {
+				err = clEnqueueReadBuffer(m_cqCommandQue,
+					m_buffer,
+					CL_TRUE,
+					0,size, 
+					&((*m_CPUBuffer)[0]),0,0,0);
+
+				if( err != CL_SUCCESS )
+				{
+					btAssert( "CommandQueue::enqueueReadBuffer(m_buffer)");
+				}
+
+			}
+		}
+
+		return true;
+	}
+
+	virtual void changedOnCPU()
+	{
+		m_onGPU = false;
+	}
+}; // class btOpenCLBuffer
+
+
+#endif // #ifndef BT_SOFT_BODY_SOLVER_BUFFER_OPENCL_H

data/deps/include/bullet/BulletMultiThreaded/GpuSoftBodySolvers/OpenCL/btSoftBodySolverLinkData_OpenCL.h

@@ -0,0 +1,99 @@
+/*
+Bullet Continuous Collision Detection and Physics Library
+Copyright (c) 2003-2006 Erwin Coumans  http://continuousphysics.com/Bullet/
+
+This software is provided 'as-is', without any express or implied warranty.
+In no event will the authors be held liable for any damages arising from the use of this software.
+Permission is granted to anyone to use this software for any purpose, 
+including commercial applications, and to alter it and redistribute it freely, 
+subject to the following restrictions:
+
+1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
+2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
+3. This notice may not be removed or altered from any source distribution.
+*/
+
+#include "BulletMultiThreaded/GpuSoftBodySolvers/Shared/btSoftBodySolverData.h"
+#include "btSoftBodySolverBuffer_OpenCL.h"
+
+
+#ifndef BT_SOFT_BODY_SOLVER_LINK_DATA_OPENCL_H
+#define BT_SOFT_BODY_SOLVER_LINK_DATA_OPENCL_H
+
+
+class btSoftBodyLinkDataOpenCL : public btSoftBodyLinkData
+{
+public:
+	bool				m_onGPU;
+
+	cl_command_queue	m_cqCommandQue;
+
+
+	btOpenCLBuffer<LinkNodePair> m_clLinks;
+	btOpenCLBuffer<float>							      m_clLinkStrength;
+	btOpenCLBuffer<float>								  m_clLinksMassLSC;
+	btOpenCLBuffer<float>								  m_clLinksRestLengthSquared;
+	btOpenCLBuffer<Vectormath::Aos::Vector3>			  m_clLinksCLength;
+	btOpenCLBuffer<float>								  m_clLinksLengthRatio;
+	btOpenCLBuffer<float>								  m_clLinksRestLength;
+	btOpenCLBuffer<float>								  m_clLinksMaterialLinearStiffnessCoefficient;
+
+	struct BatchPair
+	{
+		int start;
+		int length;
+
+		BatchPair() :
+			start(0),
+			length(0)
+		{
+		}
+
+		BatchPair( int s, int l ) : 
+			start( s ),
+			length( l )
+		{
+		}
+	};
+
+	/**
+	 * Link addressing information for each cloth.
+	 * Allows link locations to be computed independently of data batching.
+	 */
+	btAlignedObjectArray< int >							m_linkAddresses;
+
+	/**
+	 * Start and length values for computation batches over link data.
+	 */
+	btAlignedObjectArray< BatchPair >		m_batchStartLengths;
+
+	btSoftBodyLinkDataOpenCL(cl_command_queue queue, cl_context ctx);
+
+	virtual ~btSoftBodyLinkDataOpenCL();
+
+	/** Allocate enough space in all link-related arrays to fit numLinks links */
+	virtual void createLinks( int numLinks );
+	
+	/** Insert the link described into the correct data structures assuming space has already been allocated by a call to createLinks */
+	virtual void setLinkAt( 
+		const LinkDescription &link, 
+		int linkIndex );
+
+	virtual bool onAccelerator();
+
+	virtual bool moveToAccelerator();
+
+	virtual bool moveFromAccelerator();
+
+	/**
+	 * Generate (and later update) the batching for the entire link set.
+	 * This redoes a lot of work because it batches the entire set when each cloth is inserted.
+	 * In theory we could delay it until just before we need the cloth.
+	 * It's a one-off overhead, though, so that is a later optimisation.
+	 */
+	void generateBatches();
+};
+
+
+
+#endif // #ifndef BT_SOFT_BODY_SOLVER_LINK_DATA_OPENCL_H