ruby-bullet 0.0.2-x86-mingw32

data/deps/include/bullet/BulletMultiThreaded/SpuGatheringCollisionDispatcher.cpp

@@ -0,0 +1,276 @@
+/*
+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 "SpuGatheringCollisionDispatcher.h"
+#include "SpuCollisionTaskProcess.h"
+
+
+#include "BulletCollision/BroadphaseCollision/btOverlappingPairCache.h"
+#include "BulletCollision/CollisionDispatch/btEmptyCollisionAlgorithm.h"
+#include "SpuContactManifoldCollisionAlgorithm.h"
+#include "BulletCollision/CollisionDispatch/btCollisionObject.h"
+#include "BulletCollision/CollisionShapes/btCollisionShape.h"
+#include "LinearMath/btQuickprof.h"
+#include "BulletMultiThreaded/SpuNarrowPhaseCollisionTask/SpuCollisionShapes.h"
+
+
+
+
+
+SpuGatheringCollisionDispatcher::SpuGatheringCollisionDispatcher(class	btThreadSupportInterface*	threadInterface, unsigned int	maxNumOutstandingTasks,btCollisionConfiguration* collisionConfiguration)
+:btCollisionDispatcher(collisionConfiguration),
+m_spuCollisionTaskProcess(0),
+m_threadInterface(threadInterface),
+m_maxNumOutstandingTasks(maxNumOutstandingTasks)
+{
+	
+}
+
+
+bool	SpuGatheringCollisionDispatcher::supportsDispatchPairOnSpu(int proxyType0,int proxyType1)
+{
+	bool supported0 = (
+		(proxyType0 == BOX_SHAPE_PROXYTYPE) ||
+		(proxyType0 == TRIANGLE_SHAPE_PROXYTYPE) ||
+		(proxyType0 == SPHERE_SHAPE_PROXYTYPE) ||
+		(proxyType0 == CAPSULE_SHAPE_PROXYTYPE) ||
+		(proxyType0 == CYLINDER_SHAPE_PROXYTYPE) ||
+//		(proxyType0 == CONE_SHAPE_PROXYTYPE) ||
+		(proxyType0 == TRIANGLE_MESH_SHAPE_PROXYTYPE) ||
+		(proxyType0 == CONVEX_HULL_SHAPE_PROXYTYPE)||
+		(proxyType0 == STATIC_PLANE_PROXYTYPE)||
+		(proxyType0 == COMPOUND_SHAPE_PROXYTYPE)
+		);
+
+	bool supported1 = (
+		(proxyType1 == BOX_SHAPE_PROXYTYPE) ||
+		(proxyType1 == TRIANGLE_SHAPE_PROXYTYPE) ||
+		(proxyType1 == SPHERE_SHAPE_PROXYTYPE) ||
+		(proxyType1 == CAPSULE_SHAPE_PROXYTYPE) ||
+		(proxyType1 == CYLINDER_SHAPE_PROXYTYPE) ||
+//		(proxyType1 == CONE_SHAPE_PROXYTYPE) ||
+		(proxyType1 == TRIANGLE_MESH_SHAPE_PROXYTYPE) ||
+		(proxyType1 == CONVEX_HULL_SHAPE_PROXYTYPE) ||
+		(proxyType1 == STATIC_PLANE_PROXYTYPE) ||
+		(proxyType1 == COMPOUND_SHAPE_PROXYTYPE)
+		);
+
+	
+	return supported0 && supported1;
+}
+
+
+
+SpuGatheringCollisionDispatcher::~SpuGatheringCollisionDispatcher()
+{
+	if (m_spuCollisionTaskProcess)
+		delete m_spuCollisionTaskProcess;
+	
+}
+
+#include "stdio.h"
+
+
+
+///interface for iterating all overlapping collision pairs, no matter how those pairs are stored (array, set, map etc)
+///this is useful for the collision dispatcher.
+class btSpuCollisionPairCallback : public btOverlapCallback
+{
+	const btDispatcherInfo& m_dispatchInfo;
+	SpuGatheringCollisionDispatcher*	m_dispatcher;
+
+public:
+
+	btSpuCollisionPairCallback(const btDispatcherInfo& dispatchInfo, SpuGatheringCollisionDispatcher*	dispatcher)
+	:m_dispatchInfo(dispatchInfo),
+	m_dispatcher(dispatcher)
+	{
+	}
+
+	virtual bool	processOverlap(btBroadphasePair& collisionPair)
+	{
+
+
+		//PPU version
+		//(*m_dispatcher->getNearCallback())(collisionPair,*m_dispatcher,m_dispatchInfo);
+
+		//only support discrete collision detection for now, we could fallback on PPU/unoptimized version for TOI/CCD
+		btAssert(m_dispatchInfo.m_dispatchFunc == btDispatcherInfo::DISPATCH_DISCRETE);
+
+		//by default, Bullet will use this near callback
+		{
+			///userInfo is used to determine if the SPU has to handle this case or not (skip PPU tasks)
+			if (!collisionPair.m_internalTmpValue)
+			{
+				collisionPair.m_internalTmpValue = 1;
+			}
+			if (!collisionPair.m_algorithm)
+			{
+				btCollisionObject* colObj0 = (btCollisionObject*)collisionPair.m_pProxy0->m_clientObject;
+				btCollisionObject* colObj1 = (btCollisionObject*)collisionPair.m_pProxy1->m_clientObject;
+
+				btCollisionAlgorithmConstructionInfo ci;
+				ci.m_dispatcher1 = m_dispatcher;
+				ci.m_manifold = 0;
+
+				if (m_dispatcher->needsCollision(colObj0,colObj1))
+				{
+					int	proxyType0 = colObj0->getCollisionShape()->getShapeType();
+					int	proxyType1 = colObj1->getCollisionShape()->getShapeType();
+					bool supportsSpuDispatch = m_dispatcher->supportsDispatchPairOnSpu(proxyType0,proxyType1) 
+						&& ((colObj0->getCollisionFlags() & btCollisionObject::CF_DISABLE_SPU_COLLISION_PROCESSING) == 0)
+						&& ((colObj1->getCollisionFlags() & btCollisionObject::CF_DISABLE_SPU_COLLISION_PROCESSING) == 0);
+
+					if (proxyType0 == COMPOUND_SHAPE_PROXYTYPE)
+					{
+						btCompoundShape* compound = (btCompoundShape*)colObj0->getCollisionShape();
+						if (compound->getNumChildShapes()>MAX_SPU_COMPOUND_SUBSHAPES)
+						{
+							//printf("PPU fallback, compound->getNumChildShapes(%d)>%d\n",compound->getNumChildShapes(),MAX_SPU_COMPOUND_SUBSHAPES);
+							supportsSpuDispatch = false;
+						}
+					}
+
+					if (proxyType1 == COMPOUND_SHAPE_PROXYTYPE)
+					{
+						btCompoundShape* compound = (btCompoundShape*)colObj1->getCollisionShape();
+						if (compound->getNumChildShapes()>MAX_SPU_COMPOUND_SUBSHAPES)
+						{
+							//printf("PPU fallback, compound->getNumChildShapes(%d)>%d\n",compound->getNumChildShapes(),MAX_SPU_COMPOUND_SUBSHAPES);
+							supportsSpuDispatch = false;
+						}
+					}
+
+					if (supportsSpuDispatch)
+					{
+
+						int so = sizeof(SpuContactManifoldCollisionAlgorithm);
+#ifdef ALLOCATE_SEPARATELY
+						void* mem = btAlignedAlloc(so,16);//m_dispatcher->allocateCollisionAlgorithm(so);
+#else
+						void* mem = m_dispatcher->allocateCollisionAlgorithm(so);
+#endif
+						collisionPair.m_algorithm = new(mem) SpuContactManifoldCollisionAlgorithm(ci,colObj0,colObj1);
+						collisionPair.m_internalTmpValue =  2;
+					} else
+					{
+						collisionPair.m_algorithm = m_dispatcher->findAlgorithm(colObj0,colObj1);
+						collisionPair.m_internalTmpValue = 3;
+					}
+				} 
+			}
+		}
+		return false;
+	}
+};
+
+void	SpuGatheringCollisionDispatcher::dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,const btDispatcherInfo& dispatchInfo, btDispatcher* dispatcher) 
+{
+
+	if (dispatchInfo.m_enableSPU)
+	{
+		m_maxNumOutstandingTasks = m_threadInterface->getNumTasks();
+
+		{
+			BT_PROFILE("processAllOverlappingPairs");
+
+			if (!m_spuCollisionTaskProcess)
+				m_spuCollisionTaskProcess = new SpuCollisionTaskProcess(m_threadInterface,m_maxNumOutstandingTasks);
+		
+			m_spuCollisionTaskProcess->setNumTasks(m_maxNumOutstandingTasks);
+	//		printf("m_maxNumOutstandingTasks =%d\n",m_maxNumOutstandingTasks);
+
+			m_spuCollisionTaskProcess->initialize2(dispatchInfo.m_useEpa);
+			
+		
+			///modified version of btCollisionDispatcher::dispatchAllCollisionPairs:
+			{
+				btSpuCollisionPairCallback	collisionCallback(dispatchInfo,this);
+
+				pairCache->processAllOverlappingPairs(&collisionCallback,dispatcher);
+			}
+		}
+
+		//send one big batch
+		int numTotalPairs = pairCache->getNumOverlappingPairs();
+		if (numTotalPairs)
+		{
+			btBroadphasePair* pairPtr = pairCache->getOverlappingPairArrayPtr();
+			int i;
+			{
+				int pairRange =	SPU_BATCHSIZE_BROADPHASE_PAIRS;
+				if (numTotalPairs < (m_spuCollisionTaskProcess->getNumTasks()*SPU_BATCHSIZE_BROADPHASE_PAIRS))
+				{
+					pairRange = (numTotalPairs/m_spuCollisionTaskProcess->getNumTasks())+1;
+				}
+	
+				BT_PROFILE("addWorkToTask");
+				for (i=0;i<numTotalPairs;)
+				{
+					//Performance Hint: tweak this number during benchmarking
+					
+					int endIndex = (i+pairRange) < numTotalPairs ? i+pairRange : numTotalPairs;
+					m_spuCollisionTaskProcess->addWorkToTask(pairPtr,i,endIndex);
+					i = endIndex;
+				}
+			}
+			{
+				BT_PROFILE("PPU fallback");
+				//handle PPU fallback pairs
+				for (i=0;i<numTotalPairs;i++)
+				{
+					btBroadphasePair& collisionPair = pairPtr[i];
+					if (collisionPair.m_internalTmpValue == 3)
+					{
+						if (collisionPair.m_algorithm)
+						{
+							btCollisionObject* colObj0 = (btCollisionObject*)collisionPair.m_pProxy0->m_clientObject;
+							btCollisionObject* colObj1 = (btCollisionObject*)collisionPair.m_pProxy1->m_clientObject;
+	
+							if (dispatcher->needsCollision(colObj0,colObj1))
+							{
+								btManifoldResult contactPointResult(colObj0,colObj1);
+								
+								if (dispatchInfo.m_dispatchFunc == 		btDispatcherInfo::DISPATCH_DISCRETE)
+								{
+									//discrete collision detection query
+									collisionPair.m_algorithm->processCollision(colObj0,colObj1,dispatchInfo,&contactPointResult);
+								} else
+								{
+									//continuous collision detection query, time of impact (toi)
+									btScalar toi = collisionPair.m_algorithm->calculateTimeOfImpact(colObj0,colObj1,dispatchInfo,&contactPointResult);
+									if (dispatchInfo.m_timeOfImpact > toi)
+										dispatchInfo.m_timeOfImpact = toi;
+	
+								}
+							}
+						}
+					}
+				}
+			}
+		}
+		{
+			BT_PROFILE("flush2");
+			//make sure all SPU work is done
+			m_spuCollisionTaskProcess->flush2();
+		}
+
+	} else
+	{
+		///PPU fallback
+		///!Need to make sure to clear all 'algorithms' when switching between SPU and PPU
+		btCollisionDispatcher::dispatchAllCollisionPairs(pairCache,dispatchInfo,dispatcher);
+	}
+}

data/deps/include/bullet/BulletMultiThreaded/SpuGatheringCollisionDispatcher.h

@@ -0,0 +1,72 @@
+/*
+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.
+*/
+#ifndef BT_SPU_GATHERING_COLLISION__DISPATCHER_H
+#define BT_SPU_GATHERING_COLLISION__DISPATCHER_H
+
+#include "BulletCollision/CollisionDispatch/btCollisionDispatcher.h"
+
+
+///Tuning value to optimized SPU utilization 
+///Too small value means Task overhead is large compared to computation (too fine granularity)
+///Too big value might render some SPUs are idle, while a few other SPUs are doing all work.
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 8
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 16
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 64
+#define SPU_BATCHSIZE_BROADPHASE_PAIRS 128
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 256
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 512
+//#define SPU_BATCHSIZE_BROADPHASE_PAIRS 1024
+
+
+
+class SpuCollisionTaskProcess;
+
+///SpuGatheringCollisionDispatcher can use SPU to gather and calculate collision detection
+///Time of Impact, Closest Points and Penetration Depth.
+class SpuGatheringCollisionDispatcher : public btCollisionDispatcher
+{
+	
+	SpuCollisionTaskProcess*	m_spuCollisionTaskProcess;
+	
+protected:
+
+	class	btThreadSupportInterface*	m_threadInterface;
+
+	unsigned int	m_maxNumOutstandingTasks;
+	
+
+public:
+
+	//can be used by SPU collision algorithms	
+	SpuCollisionTaskProcess*	getSpuCollisionTaskProcess()
+	{
+			return m_spuCollisionTaskProcess;
+	}
+	
+	SpuGatheringCollisionDispatcher (class	btThreadSupportInterface*	threadInterface, unsigned int	maxNumOutstandingTasks,btCollisionConfiguration* collisionConfiguration);
+	
+	virtual ~SpuGatheringCollisionDispatcher();
+
+	bool	supportsDispatchPairOnSpu(int proxyType0,int proxyType1);
+
+	virtual void	dispatchAllCollisionPairs(btOverlappingPairCache* pairCache,const btDispatcherInfo& dispatchInfo,btDispatcher* dispatcher) ;
+
+};
+
+
+
+#endif //BT_SPU_GATHERING_COLLISION__DISPATCHER_H
+
+

data/deps/include/bullet/BulletMultiThreaded/SpuLibspe2Support.cpp

@@ -0,0 +1,257 @@
+/*
+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.
+*/
+
+#ifdef USE_LIBSPE2
+
+#include "SpuLibspe2Support.h"
+
+
+
+
+//SpuLibspe2Support helps to initialize/shutdown libspe2, start/stop SPU tasks and communication
+///Setup and initialize SPU/CELL/Libspe2
+SpuLibspe2Support::SpuLibspe2Support(spe_program_handle_t *speprog, int numThreads)
+{
+	this->program = speprog;
+	this->numThreads =  ((numThreads <= spe_cpu_info_get(SPE_COUNT_PHYSICAL_SPES, -1)) ? numThreads : spe_cpu_info_get(SPE_COUNT_PHYSICAL_SPES, -1));
+}
+
+///cleanup/shutdown Libspe2
+SpuLibspe2Support::~SpuLibspe2Support()
+{
+	
+	stopSPU();
+}
+
+
+
+///send messages to SPUs
+void SpuLibspe2Support::sendRequest(uint32_t uiCommand, uint32_t uiArgument0, uint32_t uiArgument1)
+{
+	spe_context_ptr_t context;
+	
+	switch (uiCommand)
+	{
+	case CMD_SAMPLE_TASK_COMMAND:
+	{
+		//get taskdescription
+		SpuSampleTaskDesc* taskDesc = (SpuSampleTaskDesc*) uiArgument0;
+
+		btAssert(taskDesc->m_taskId<m_activeSpuStatus.size());
+
+		//get status of SPU on which task should run
+		btSpuStatus&	spuStatus = m_activeSpuStatus[taskDesc->m_taskId];
+
+		//set data for spuStatus
+		spuStatus.m_commandId = uiCommand;
+		spuStatus.m_status = Spu_Status_Occupied; //set SPU as "occupied"
+		spuStatus.m_taskDesc.p = taskDesc; 
+		
+		//get context
+		context = data[taskDesc->m_taskId].context;
+		
+		
+		taskDesc->m_mainMemoryPtr = reinterpret_cast<uint64_t> (spuStatus.m_lsMemory.p);
+		
+
+		break;
+	}
+	case CMD_GATHER_AND_PROCESS_PAIRLIST:
+		{
+			//get taskdescription
+			SpuGatherAndProcessPairsTaskDesc* taskDesc = (SpuGatherAndProcessPairsTaskDesc*) uiArgument0;
+
+			btAssert(taskDesc->taskId<m_activeSpuStatus.size());
+
+			//get status of SPU on which task should run
+			btSpuStatus&	spuStatus = m_activeSpuStatus[taskDesc->taskId];
+
+			//set data for spuStatus
+			spuStatus.m_commandId = uiCommand;
+			spuStatus.m_status = Spu_Status_Occupied; //set SPU as "occupied"
+			spuStatus.m_taskDesc.p = taskDesc; 
+			
+			//get context
+			context = data[taskDesc->taskId].context;
+			
+			
+			taskDesc->m_lsMemory = (CollisionTask_LocalStoreMemory*)spuStatus.m_lsMemory.p;
+			
+			break;
+		}
+	default:
+		{
+			///not implemented
+			btAssert(0);
+		}
+
+	};
+
+	
+	//write taskdescription in mailbox
+	unsigned int event = Spu_Mailbox_Event_Task;
+	spe_in_mbox_write(context, &event, 1, SPE_MBOX_ANY_NONBLOCKING);
+
+}
+
+///check for messages from SPUs
+void SpuLibspe2Support::waitForResponse(unsigned int *puiArgument0, unsigned int *puiArgument1)
+{
+	///We should wait for (one of) the first tasks to finish (or other SPU messages), and report its response
+	
+	///A possible response can be 'yes, SPU handled it', or 'no, please do a PPU fallback'
+	
+	btAssert(m_activeSpuStatus.size());
+
+	
+	int last = -1;
+	
+	//find an active spu/thread
+	while(last < 0)
+	{
+		for (int i=0;i<m_activeSpuStatus.size();i++)
+		{
+			if ( m_activeSpuStatus[i].m_status == Spu_Status_Free)
+			{
+				last = i;
+				break;
+			}
+		}
+		if(last < 0)
+			sched_yield();
+	}
+
+
+
+	btSpuStatus& spuStatus = m_activeSpuStatus[last];
+
+	///need to find an active spu
+	btAssert(last>=0);
+
+	
+
+	*puiArgument0 = spuStatus.m_taskId;
+	*puiArgument1 = spuStatus.m_status;
+
+
+}
+
+
+void SpuLibspe2Support::startSPU()
+{
+	this->internal_startSPU();
+}
+
+
+
+///start the spus group (can be called at the beginning of each frame, to make sure that the right SPU program is loaded)
+void SpuLibspe2Support::internal_startSPU()
+{
+	m_activeSpuStatus.resize(numThreads);
+	
+	
+	for (int i=0; i < numThreads; i++)
+	{
+		
+		if(data[i].context == NULL) 
+		{
+					
+			 /* Create context */
+			if ((data[i].context = spe_context_create(0, NULL)) == NULL)
+			{
+			      perror ("Failed creating context");
+		          exit(1);
+			}
+	
+			/* Load program into context */
+			if(spe_program_load(data[i].context, this->program))
+			{
+			      perror ("Failed loading program");
+		          exit(1);
+			}
+			
+			m_activeSpuStatus[i].m_status = Spu_Status_Startup; 
+			m_activeSpuStatus[i].m_taskId = i; 
+			m_activeSpuStatus[i].m_commandId = 0; 
+			m_activeSpuStatus[i].m_lsMemory.p = NULL; 
+			
+			
+			data[i].entry = SPE_DEFAULT_ENTRY;
+			data[i].flags = 0;
+			data[i].argp.p = &m_activeSpuStatus[i];
+			data[i].envp.p = NULL;
+			
+		    /* Create thread for each SPE context */
+			if (pthread_create(&data[i].pthread, NULL, &ppu_pthread_function, &(data[i]) ))
+			{
+			      perror ("Failed creating thread");
+		          exit(1);
+			}
+			/*
+			else
+			{
+				printf("started thread %d\n",i);
+			}*/
+		}		
+	}
+	
+	
+	for (int i=0; i < numThreads; i++)
+	{
+		if(data[i].context != NULL) 
+		{
+			while( m_activeSpuStatus[i].m_status == Spu_Status_Startup)
+			{
+				// wait for spu to set up
+				sched_yield();
+			}
+			printf("Spu %d is ready\n", i);
+		}
+	}
+}
+
+///tell the task scheduler we are done with the SPU tasks
+void SpuLibspe2Support::stopSPU()
+{
+	// wait for all threads to finish 
+	int i;
+	for ( i = 0; i < this->numThreads; i++ ) 
+	{ 
+		
+		unsigned int event = Spu_Mailbox_Event_Shutdown;
+		spe_context_ptr_t context = data[i].context;
+		spe_in_mbox_write(context, &event, 1, SPE_MBOX_ALL_BLOCKING);
+		pthread_join (data[i].pthread, NULL); 
+		
+	} 
+	// close SPE program 
+	spe_image_close(program); 
+	// destroy SPE contexts 
+	for ( i = 0; i < this->numThreads; i++ ) 
+	{ 
+		if(data[i].context != NULL)
+		{
+			spe_context_destroy (data[i].context);
+		}
+	} 
+	
+	m_activeSpuStatus.clear();
+	
+}
+
+
+
+#endif //USE_LIBSPE2
+