added sample task, for multi threading tests

added union vec_float4 m_Vec128; for btVector3 for SPU and PPU.

Extras/BulletMultiThreaded/SpuSampleTask/SpuSampleTask.cpp

@@ -0,0 +1,75 @@
+/*
+Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
+
+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 "SpuSampleTask.h"
+#include "BulletDynamics/Dynamics/btRigidBody.h"
+#include "../PlatformDefinitions.h"
+#include "../SpuFakeDma.h"
+#include "LinearMath/btMinMax.h"
+#include <stdio.h>
+
+
+struct SampleTask_LocalStoreMemory
+{
+
+};
+
+
+
+//-- MAIN METHOD
+void processSampleTask(void* userPtr, void* lsMemory)
+{
+//	BT_PROFILE("processSampleTask");
+
+	SampleTask_LocalStoreMemory* localMemory = (SampleTask_LocalStoreMemory*)lsMemory;
+
+	SpuSampleTaskDesc* taskDescPtr = (SpuSampleTaskDesc*)userPtr;
+	SpuSampleTaskDesc& taskDesc = *taskDescPtr;
+
+	switch (taskDesc.m_sampleCommand)
+	{
+	case CMD_SAMPLE_INTEGRATE_BODIES:
+		{
+#ifdef __SPU__
+			spu_printf("hello SPU world\n");
+#else
+			printf("hello world\n");
+#endif
+			break;
+		}
+	default:
+		{
+
+		}
+	};
+}
+
+
+#if defined(__CELLOS_LV2__) || defined (LIBSPE2)
+
+ATTRIBUTE_ALIGNED16(SampleTask_LocalStoreMemory	gLocalStoreMemory);
+
+void* createSampleLocalStoreMemory()
+{
+	return &gLocalStoreMemory;
+}
+#else
+void* createSampleLocalStoreMemory()
+{
+        return new SampleTask_LocalStoreMemory;
+};
+
+#endif

Extras/BulletMultiThreaded/SpuSampleTask/SpuSampleTask.h

@@ -0,0 +1,53 @@
+/*
+Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
+
+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 SPU_SAMPLE_TASK_H
+#define SPU_SAMPLE_TASK_H
+
+#include "../PlatformDefinitions.h"
+#include "LinearMath/btScalar.h"
+#include "LinearMath/btVector3.h"
+#include "LinearMath/btMatrix3x3.h"
+
+#include "LinearMath/btAlignedAllocator.h"
+
+
+enum
+{
+	CMD_SAMPLE_INTEGRATE_BODIES = 1
+};
+
+
+
+ATTRIBUTE_ALIGNED16(struct) SpuSampleTaskDesc
+{
+	BT_DECLARE_ALIGNED_ALLOCATOR();
+
+	uint32_t						m_sampleCommand;
+	uint32_t						m_taskId;
+
+	uint64_t 	m_mainMemoryPtr;
+	int			m_sampleValue;
+	
+
+};
+
+
+void	processSampleTask(void* userPtr, void* lsMemory);
+void*	createSampleLocalStoreMemory();
+
+
+#endif //SPU_SAMPLE_TASK_H
+

Extras/BulletMultiThreaded/SpuSampleTaskProcess.cpp

@@ -105,7 +105,7 @@ void	SpuSampleTaskProcess::initialize()
 }
 
 
-void SpuSampleTaskProcess::issueTask(void* sampleMainMemPtr,int sampleValue)
+void SpuSampleTaskProcess::issueTask(void* sampleMainMemPtr,int sampleValue,int sampleCommand)
 {
 
 #ifdef DEBUG_SPU_TASK_SCHEDULING
@@ -123,13 +123,14 @@ void SpuSampleTaskProcess::issueTask(void* sampleMainMemPtr,int sampleValue)
 	
 		taskDesc.m_mainMemoryPtr = reinterpret_cast<uint64_t>(sampleMainMemPtr);
 		taskDesc.m_sampleValue = sampleValue;
+		taskDesc.m_sampleCommand = sampleCommand;
 
 		//some bookkeeping to recognize finished tasks
 		taskDesc.m_taskId = m_currentTask;
 	}
 
 
-	m_threadInterface->sendRequest(CMD_SAMPLE_TASK_COMMAND, (uint32_t) &taskDesc, m_currentTask);
+	m_threadInterface->sendRequest(1, (uint32_t) &taskDesc, m_currentTask);
 
 	// if all tasks busy, wait for spu event to clear the task.
 	

Extras/BulletMultiThreaded/SpuSampleTaskProcess.h

@@ -26,18 +26,8 @@ subject to the following restrictions:
 #include "LinearMath/btAlignedObjectArray.h"
 
 
-///SpuSampleTaskDesc
+#include "SpuSampleTask/SpuSampleTask.h"
-struct SpuSampleTaskDesc
-{
-	uint64_t 	m_mainMemoryPtr;
-	int			m_sampleValue;
 
-	uint16_t 	m_taskId;
-}
-#if defined(__CELLOS_LV2__) || defined(USE_LIBSPE2)
-__attribute__ ((aligned (16)))
-#endif
-;
 
 //just add your commands here, try to keep them globally unique for debugging purposes
 #define CMD_SAMPLE_TASK_COMMAND 10
@@ -76,7 +66,7 @@ public:
 	///call initialize in the beginning of the frame, before addCollisionPairToTask
 	void initialize();
 
-	void issueTask(void* sampleMainMemPtr,int sampleValue);
+	void issueTask(void* sampleMainMemPtr,int sampleValue,int sampleCommand);
 
 	///call flush to submit potential outstanding work to SPUs and wait for all involved SPUs to be finished
 	void flush();

src/LinearMath/btQuadWord.h

@@ -18,6 +18,14 @@ subject to the following restrictions:
 
 #include "btScalar.h"
 #include "btMinMax.h"
+#include <math.h>
+#ifdef __PPU__
+#include <altivec.h>
+#endif
+
+#ifdef __SPU__
+#include <spu_intrinsics.h>
+#endif
 
 //ATTRIBUTE_ALIGNED16(class) btQuadWordStorage
 //some issues under PS3 Linux with IBM 2.1 SDK, gcc compiler prevent from using aligned quadword. todo: look into this
@@ -25,10 +33,35 @@ subject to the following restrictions:
 class btQuadWordStorage
 {
 protected:
+#if defined __PPU__ || defined __SPU__
+	union {
+		vec_float4 m_Vec128;
+		struct	
+		{
+#endif
 	btScalar	m_x;
 	btScalar	m_y;
 	btScalar	m_z;
 	btScalar	m_unusedW;
+#if defined __PPU__ || defined __SPU__
+		};
+	};
+#endif
+
+public:
+#if defined __PPU__ || defined __SPU__
+
+	inline const vec_float4 get128( ) const
+	{
+		return m_Vec128;
+	}
+
+	inline vec_float4 get128( )
+	{
+		return m_Vec128;
+	}
+#endif //
+
 };
 
 

src/LinearMath/btVector3.h

@@ -47,7 +47,14 @@ public:
 
 	SIMD_FORCE_INLINE btVector3& operator+=(const btVector3& v)
 	{
+
+#ifdef __PPU__
+		m_Vec128 = vec_add( m_Vec128, v.get128() );
+#elif defined __SPU__
+		m_Vec128 = spu_add( m_Vec128, v.get128() ); 
+#else
 		m_x += v.x(); m_y += v.y(); m_z += v.z();
+#endif
 		return *this;
 	}