moved predictUnconstraintMotion to SPU

Extras/BulletMultiThreaded/SpuSampleTask/SpuSampleTask.cpp

@@ -27,10 +27,12 @@ subject to the following restrictions:
 #define spu_printf printf
 #endif
 
+#define MAX_NUM_BODIES 8192
+
 struct SampleTask_LocalStoreMemory
 {
 	ATTRIBUTE_ALIGNED16(char gLocalRigidBody [sizeof(btRigidBody)+16]);
-	ATTRIBUTE_ALIGNED16(void* gPointerArray[1024]); //at max upload 1024 pointers
+	ATTRIBUTE_ALIGNED16(void* gPointerArray[MAX_NUM_BODIES]);
 
 };
 
@@ -55,6 +57,11 @@ void processSampleTask(void* userPtr, void* lsMemory)
 			btCollisionObject** eaPtr = (btCollisionObject**)taskDesc.m_mainMemoryPtr;
 
 			int batchSize = taskDesc.m_sampleValue;
+			if (batchSize>MAX_NUM_BODIES)
+			{
+				spu_printf("SPU Error: exceed number of bodies, see MAX_NUM_BODIES in SpuSampleTask.cpp\n");
+				break;
+			}
 			int dmaArraySize = batchSize*sizeof(void*);
 
 			uint64_t ppuArrayAddress = reinterpret_cast<uint64_t>(eaPtr);
@@ -106,17 +113,82 @@ void processSampleTask(void* userPtr, void* lsMemory)
 				}
 
 			}
-
-			/*
-#ifdef __SPU__
-			spu_printf("hello SPU world\n");
-#else
-			printf("hello world\n");
-#endif
-			*/
-
 			break;
 		}
+
+
+	case CMD_SAMPLE_PREDICT_MOTION_BODIES:
+		{
+			btTransform predictedTrans;
+			btCollisionObject** eaPtr = (btCollisionObject**)taskDesc.m_mainMemoryPtr;
+
+			int batchSize = taskDesc.m_sampleValue;
+			int dmaArraySize = batchSize*sizeof(void*);
+
+			if (batchSize>MAX_NUM_BODIES)
+			{
+				spu_printf("SPU Error: exceed number of bodies, see MAX_NUM_BODIES in SpuSampleTask.cpp\n");
+				break;
+			}
+
+			uint64_t ppuArrayAddress = reinterpret_cast<uint64_t>(eaPtr);
+
+			//			spu_printf("array location is at %llx, batchSize = %d, DMA size = %d\n",ppuArrayAddress,batchSize,dmaArraySize);
+
+			if (dmaArraySize>=16)
+			{
+				cellDmaLargeGet((void*)&localMemory->gPointerArray[0], ppuArrayAddress  , dmaArraySize, DMA_TAG(1), 0, 0);	
+				cellDmaWaitTagStatusAll(DMA_MASK(1));
+			} else
+			{
+				stallingUnalignedDmaSmallGet((void*)&localMemory->gPointerArray[0], ppuArrayAddress  , dmaArraySize);
+			}
+
+
+			for ( int i=0;i<batchSize;i++)
+			{
+				///DMA rigid body
+
+				void* localPtr = &localMemory->gLocalRigidBody[0];
+				void* shortAdd = localMemory->gPointerArray[i];
+				uint64_t ppuRigidBodyAddress = reinterpret_cast<uint64_t>(shortAdd);
+
+				//	spu_printf("cellDmaGet at CMD_SAMPLE_INTEGRATE_BODIES from %llx to %llx\n",ppuRigidBodyAddress,localPtr);
+
+				int dmaBodySize = sizeof(btRigidBody);
+
+				cellDmaGet((void*)localPtr, ppuRigidBodyAddress  , dmaBodySize, DMA_TAG(1), 0, 0);	
+				cellDmaWaitTagStatusAll(DMA_MASK(1));
+
+
+				float timeStep = 1.f/60.f;
+
+				btRigidBody* body = (btRigidBody*) localPtr;//btRigidBody::upcast(colObj);
+				if (body)
+				{
+					if (!body->isStaticOrKinematicObject())
+					{
+						if (body->isActive())
+						{
+							body->integrateVelocities( timeStep);
+							//damping
+							body->applyDamping(timeStep);
+
+							body->predictIntegratedTransform(timeStep,body->getInterpolationWorldTransform());
+
+							void* ptr = (void*)localPtr;
+							cellDmaLargePut(ptr, ppuRigidBodyAddress  , dmaBodySize, DMA_TAG(1), 0, 0);
+							cellDmaWaitTagStatusAll(DMA_MASK(1));
+						}
+					}
+				}
+
+			}
+			break;
+		}
+	
+
+
 	default:
 		{
 

Extras/BulletMultiThreaded/SpuSampleTask/SpuSampleTask.h

@@ -26,7 +26,8 @@ subject to the following restrictions:
 
 enum
 {
-	CMD_SAMPLE_INTEGRATE_BODIES = 1
+	CMD_SAMPLE_INTEGRATE_BODIES = 1,
+	CMD_SAMPLE_PREDICT_MOTION_BODIES
 };