void GatherCollisionObjectAndShapeData (RaycastGatheredObjectData* gatheredObjectData, RaycastTask_LocalStoreMemory* lsMemPtr, ppu_address_t objectWrapper)
{
register int dmaSize;
register ppu_address_t dmaPpuAddress2;
/* DMA Collision object wrapper into local store */
dmaSize = sizeof(SpuCollisionObjectWrapper);
dmaPpuAddress2 = objectWrapper;
cellDmaGet(&lsMemPtr->gCollisionObjectWrapper, dmaPpuAddress2, dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
/* DMA Collision object into local store */
dmaSize = sizeof(btCollisionObject);
dmaPpuAddress2 = lsMemPtr->getCollisionObjectWrapper()->getCollisionObjectPtr();
cellDmaGet(&lsMemPtr->gColObj, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
/* Gather information about collision object and shape */
gatheredObjectData->m_worldTransform = lsMemPtr->getColObj()->getWorldTransform();
gatheredObjectData->m_collisionMargin = lsMemPtr->getCollisionObjectWrapper()->getCollisionMargin ();
gatheredObjectData->m_shapeType = lsMemPtr->getCollisionObjectWrapper()->getShapeType ();
gatheredObjectData->m_collisionShape = (ppu_address_t)lsMemPtr->getColObj()->getCollisionShape();
gatheredObjectData->m_spuCollisionShape = (void*)&lsMemPtr->gCollisionShape.collisionShape;
/* DMA shape data */
dmaCollisionShape (gatheredObjectData->m_spuCollisionShape, gatheredObjectData->m_collisionShape, 1, gatheredObjectData->m_shapeType);
cellDmaWaitTagStatusAll(DMA_MASK(1));
if (btBroadphaseProxy::isConvex (gatheredObjectData->m_shapeType))
{
btConvexInternalShape* spuConvexShape = (btConvexInternalShape*)gatheredObjectData->m_spuCollisionShape;
gatheredObjectData->m_primitiveDimensions = spuConvexShape->getImplicitShapeDimensions ();
} else {
gatheredObjectData->m_primitiveDimensions = btVector3(1.0, 1.0, 1.0);
}
}
void processDecodeSet(unsigned int uiPtr)
{
SpursSpeexTaskOutput spuOutput;
cellDmaGet(&gviSpursSpeexTaskDesc, uiPtr, sizeof(SpursSpeexTaskDesc), DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] CMD_SAMPLE_TASK_DECODESET_COMMAND\n");
if (gviSpursSpeexTaskDesc.mDebugPause)
{
snPause();
}
cellDmaLargeGet(gviSpursSpeexStateBuffer, (uint64_t)gviSpursSpeexTaskDesc.mSpeexStateBuffer, SPEEX_DECODER_STATE_BUFFER_SIZE, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
gviSpursSpeexDecodeSet(&spuOutput);
if (spuOutput.mSpeexReturnCode < 0)
{
spuDebugPrintf("SPU: failed to encode, ret = %d\n", spuOutput.mSpeexReturnCode);
}
cellDmaPut(&spuOutput, (uint64_t)gviSpursSpeexTaskDesc.mSpeexTaskOutput, sizeof(SpursSpeexTaskOutput), DMA_TAG(1),
0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
cellDmaLargePut(gviSpursSpeexStateBuffer, (uint64_t)gviSpursSpeexTaskDesc.mSpeexStateBuffer, SPEEX_DECODER_STATE_BUFFER_SIZE, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] buffer dma done\n");
}
void processDecodeInit(unsigned int uiPtr)
{
SpursSpeexTaskOutput spuOutput;
cellDmaGet(&gviSpursSpeexTaskDesc, uiPtr, sizeof(SpursSpeexTaskDesc), DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] CMD_SAMPLE_TASK_DECODE_INIT_COMMAND\n");
if (gviSpursSpeexTaskDesc.mDebugPause)
{
snPause();
}
gviSpursSpeexDecoderInitialize(&spuOutput);
if (spuOutput.mSpeexReturnCode < 0)
{
spuDebugPrintf("[Speex][SPU] failed to initialize decoder, ret = %d\n", spuOutput.mSpeexReturnCode);
}
cellDmaPut(&spuOutput, (uint64_t)gviSpursSpeexTaskDesc.mSpeexTaskOutput, sizeof(SpursSpeexTaskOutput), DMA_TAG(1),
0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
cellDmaLargePut(gviSpursSpeexStateBuffer, (uint64_t)gviSpursSpeexTaskDesc.mSpeexStateBuffer,
gviSpursSpeexTaskDesc.mSpeexStateBufferSize, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] buffer dma done\n");
}
void procesEncodeInit(unsigned int uiPtr)
{
SpursSpeexTaskOutput spuOutput;
//spuDebugPrintf("[Speex][SPU] CMD_SAMPLE_TASK_ENCODE_INIT_COMMAND\n");
cellDmaGet(&gviSpursSpeexTaskDesc, uiPtr, sizeof(SpursSpeexTaskDesc), DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
if (gviSpursSpeexTaskDesc.mDebugPause)
{
snPause();
}
gviSpursSpeexEncoderInitialize(&spuOutput);
if (spuOutput.mSpeexReturnCode < 0)
{
spuDebugPrintf("[Speex][SPU] failed to initialize encoder, ret = %d\n", spuOutput.mSpeexReturnCode);
}
//spuDebugPrintf("[Speex][SPU] done with initializing things for speex, now returning data via DMA put\n");
//printGlobalTaskDescData();
cellDmaPut(&spuOutput, (uint64_t)gviSpursSpeexTaskDesc.mSpeexTaskOutput, sizeof(SpursSpeexTaskOutput), DMA_TAG(1),
0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] task dma done\n");
cellDmaLargePut(gviSpursSpeexStateBuffer, (uint64_t)gviSpursSpeexTaskDesc.mSpeexStateBuffer, SPEEX_ENCODER_STATE_BUFFER_SIZE, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] buffer dma done\n");
}
void* cellDmaGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid)
{
#if defined (__SPU__) || defined (USE_LIBSPE2)
cellDmaGet(ls,ea,size,tag,tid,rid);
return ls;
#else
return (void*)(uint32_t)ea;
#endif
}
void gviSpursSpeexEncode(SpursSpeexTaskOutput *spuTaskOut)
{
short *inBuffer;
float *speexBuffer;
char *outBuffer;
unsigned int i;
spuTaskOut->mSpeexEncodedFrameSize = 0;
spuTaskOut->mSpeexInitialized = 1;
spuTaskOut->mSpeexSamplesPerFrame = 0;
spuTaskOut->mSpeexReturnCode = 0;
spuTaskOut->mSpeexOutBufferSize = 0;
speexBuffer = (float *)memalign(16, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(float));
inBuffer = (short *)memalign(16, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(short));
outBuffer = (char *)memalign(16, gviSpursSpeexTaskDesc.mOutputBufferSize);
memset(speexBuffer, 0, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(float));
memset(inBuffer, 0, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(short));
memset(outBuffer, 0, gviSpursSpeexTaskDesc.mOutputBufferSize);
cellDmaGet(inBuffer, (uint64_t)gviSpursSpeexTaskDesc.mInputBuffer, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(short), DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
// convert the input to floats for encoding
for(i = 0 ; i < gviSpursSpeexTaskDesc.mInputBufferSize ; i++)
speexBuffer[i] = inBuffer[i];
// (re)initialize the bits struct
speex_bits_init_buffer(&gviSpursSpeexBits,gviSpursSpeexBitsBuffer,sizeof(gviSpursSpeexBitsBuffer));
// flush the bits
speex_bits_reset(&gviSpursSpeexBits);
// encode the frame
speex_encode(gviSpursSpeexStateBuffer, speexBuffer, &gviSpursSpeexBits);
// write the bits to the output
spuTaskOut->mSpeexOutBufferSize = speex_bits_write(&gviSpursSpeexBits, (char *)outBuffer, gviSpursSpeexTaskDesc.mEncodedFrameSize);
//spuDebugPrintf("[Speex][SPU] transferring data back, output size should be: %d\n", gviSpursSpeexTaskDesc.mOutputBufferSize>16?gviSpursSpeexTaskDesc.mOutputBufferSize:16);
cellDmaPut(outBuffer, (uint64_t)gviSpursSpeexTaskDesc.mOutputBuffer, gviSpursSpeexTaskDesc.mOutputBufferSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] done transferring data back\n");
free(speexBuffer);
free(inBuffer);
free(outBuffer);
spuTaskOut->mSpeexReturnCode = 0;
}
void gviSpursSpeexDecodeAdd(SpursSpeexTaskOutput *spuTaskOut)
{
char *inBuffer;
float *speexBuffer;
short *outBuffer;
int rcode;
unsigned int i;
//spuDebugPrintf("[Speex][SPU] allocating buffers for decoding\n");
speexBuffer = (float *)memalign(16, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(float));
outBuffer = (short *)memalign(16, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(short));
inBuffer = (char *)memalign(16, gviSpursSpeexTaskDesc.mInputBufferSize);
memset(speexBuffer, 0, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(float));
memset(outBuffer, 0, gviSpursSpeexTaskDesc.mOutputBufferSize);
memset(inBuffer, 0, gviSpursSpeexTaskDesc.mInputBufferSize * sizeof(short));
//spuDebugPrintf("[Speex][SPU] done allocating, getting input data, inbuffer size: %d\n", gSpuSampleTaskDesc.mInputBufferSize);
cellDmaGet(inBuffer, (uint64_t)gviSpursSpeexTaskDesc.mInputBuffer, gviSpursSpeexTaskDesc.mInputBufferSize, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
// spuDebugPrintf("[Speex][SPU] done getting input data, preparing for speex to decode\n");
// read the data into the bits
// (re)initialize the bits struct
speex_bits_init_buffer(&gviSpursSpeexBits,gviSpursSpeexBitsBuffer,sizeof(gviSpursSpeexBitsBuffer));
speex_bits_read_from(&gviSpursSpeexBits, (char *)inBuffer, gviSpursSpeexTaskDesc.mEncodedFrameSize);
// decode it
rcode = speex_decode((void *)gviSpursSpeexStateBuffer, &gviSpursSpeexBits, speexBuffer);
assert(rcode == 0);
//spuDebugPrintf("[Speex][SPU] done with speex decode\n");
// convert the output from floats
for(i = 0 ; i < gviSpursSpeexTaskDesc.mOutputBufferSize ; i++)
outBuffer[i] = (short)speexBuffer[i];
//spuDebugPrintf("[Speex][SPU] transferring data back\n");
cellDmaPut(outBuffer, (uint64_t)gviSpursSpeexTaskDesc.mOutputBuffer, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(short), DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
//spuDebugPrintf("[Speex][SPU] done transferring data back\n");
free(speexBuffer);
free(inBuffer);
free(outBuffer);
spuTaskOut->mSpeexReturnCode = 0;
}
void
performRaycastAgainstConvex (RaycastGatheredObjectData* gatheredObjectData, const SpuRaycastTaskWorkUnit& workUnit, SpuRaycastTaskWorkUnitOut* workUnitOut, RaycastTask_LocalStoreMemory* lsMemPtr)
{
SpuVoronoiSimplexSolver simplexSolver;
btTransform rayFromTrans, rayToTrans;
rayFromTrans.setIdentity ();
rayFromTrans.setOrigin (workUnit.rayFrom);
rayToTrans.setIdentity ();
rayToTrans.setOrigin (workUnit.rayTo);
SpuCastResult result;
/* Load the vertex data if the shape is a convex hull */
/* XXX: We might be loading the shape twice */
ATTRIBUTE_ALIGNED16(char convexHullShape[sizeof(btConvexHullShape)]);
if (gatheredObjectData->m_shapeType == CONVEX_HULL_SHAPE_PROXYTYPE)
{
register int dmaSize;
register ppu_address_t dmaPpuAddress2;
dmaSize = sizeof(btConvexHullShape);
dmaPpuAddress2 = gatheredObjectData->m_collisionShape;
cellDmaGet(&convexHullShape, dmaPpuAddress2, dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
dmaConvexVertexData (&lsMemPtr->convexVertexData, (btConvexHullShape*)&convexHullShape);
cellDmaWaitTagStatusAll(DMA_MASK(2)); // dmaConvexVertexData uses dma channel 2!
lsMemPtr->convexVertexData.gSpuConvexShapePtr = gatheredObjectData->m_spuCollisionShape;
lsMemPtr->convexVertexData.gConvexPoints = &lsMemPtr->convexVertexData.g_convexPointBuffer[0];
}
/* performRaycast */
SpuSubsimplexRayCast caster (gatheredObjectData->m_spuCollisionShape, &lsMemPtr->convexVertexData, gatheredObjectData->m_shapeType, gatheredObjectData->m_collisionMargin, &simplexSolver);
bool r = caster.calcTimeOfImpact (rayFromTrans, rayToTrans, gatheredObjectData->m_worldTransform, gatheredObjectData->m_worldTransform,result);
if (r)
{
workUnitOut->hitFraction = result.m_fraction;
workUnitOut->hitNormal = result.m_normal;
}
}
void gviSpursSpeexDecodeSet(SpursSpeexTaskOutput *spuTaskOut)
{
char *inBuffer;
float *speexBuffer;
short *outBuffer;
int rcode;
unsigned int i;
speexBuffer = (float *)memalign(16, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(float));
outBuffer = (short *)memalign(16, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(short));
inBuffer = (char *)memalign(16, gviSpursSpeexTaskDesc.mInputBufferSize);
memset(speexBuffer, 0, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(float));
memset(inBuffer, 0, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(short));
memset(outBuffer, 0, gviSpursSpeexTaskDesc.mInputBufferSize);
cellDmaGet(inBuffer, (uint64_t)gviSpursSpeexTaskDesc.mInputBuffer, gviSpursSpeexTaskDesc.mInputBufferSize, DMA_TAG(1), 0,0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
// read the data into the bits
speex_bits_read_from(&gviSpursSpeexBits, (char *)inBuffer, gviSpursSpeexTaskDesc.mEncodedFrameSize);
// decode it
rcode = speex_decode((void *)gviSpursSpeexStateBuffer, &gviSpursSpeexBits, speexBuffer);
assert(rcode == 0);
// convert the output from floats
for(i = 0 ; i < gviSpursSpeexTaskDesc.mOutputBufferSize ; i++)
// Expanded to remove warnings in VS2K5
outBuffer[i] = (short)speexBuffer[i];
cellDmaPut(outBuffer, (uint64_t)gviSpursSpeexTaskDesc.mOutputBuffer, gviSpursSpeexTaskDesc.mOutputBufferSize * sizeof(short), DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
free(speexBuffer);
free(inBuffer);
free(outBuffer);
spuTaskOut->mSpeexReturnCode = 0;
}
void dmaLoadRayOutput (ppu_address_t rayOutputAddr, SpuRaycastTaskWorkUnitOut* rayOutput, uint32_t dmaTag)
{
cellDmaGet(rayOutput, rayOutputAddr, sizeof(*rayOutput), DMA_TAG(dmaTag), 0, 0);
}
//-- 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:
{
btTransform predictedTrans;
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);
// 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->isActive() && (!body->isStaticOrKinematicObject()))
{
body->predictIntegratedTransform(timeStep, predictedTrans);
body->proceedToTransform( predictedTrans);
void* ptr = (void*)localPtr;
// spu_printf("cellDmaLargePut from %llx to LS %llx\n",ptr,ppuRigidBodyAddress);
cellDmaLargePut(ptr, ppuRigidBodyAddress , dmaBodySize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
}
}
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:
{
}
};
}
