Bart/bullet3
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

More work on multi-threaded version, fixed alignment issues in DMA.

Browse Source 
BulletMultiThreaded shared implementation works now on both Win32 Threads and PS3 Cell SPU.
This commit is contained in:
ejcoumans
2007-06-16 02:06:24 +00:00
parent ca1c42f07f
commit 272a7dfac4
5 changed files with 258 additions and 161 deletions
Download Patch File Download Diff File Expand all files Collapse all files

361
Extras/BulletMultiThreaded/SpuNarrowPhaseCollisionTask/SpuGatheringCollisionTask.cpp
View File

@@ -28,6 +28,9 @@
#ifdef WIN32
#define IGNORE_ALIGNMENT 1
#define spu_printf printf
#include <stdio.h>
#endif
@@ -36,25 +39,36 @@
//int gNumConvexPoints0=0;
///Make sure no destructors are called on this memory
struct CollisionTask_LocalStoreMemory
{
ATTRIBUTE_ALIGNED16(char bufferProxy0[16]);
ATTRIBUTE_ALIGNED16(char bufferProxy1[16]);
ATTRIBUTE_ALIGNED16(btBroadphaseProxy* gProxyPtr0);
ATTRIBUTE_ALIGNED16(btBroadphaseProxy* gProxyPtr1);
ATTRIBUTE_ALIGNED16(btCollisionObject gColObj0);
ATTRIBUTE_ALIGNED16(btCollisionObject gColObj1);
DoubleBuffer<unsigned char, MIDPHASE_WORKUNIT_PAGE_SIZE> g_workUnitTaskBuffers;
btBroadphasePair gBroadphasePairs[SPU_BATCHSIZE_BROADPHASE_PAIRS];
ATTRIBUTE_ALIGNED16(btBroadphasePair gBroadphasePairs[SPU_BATCHSIZE_BROADPHASE_PAIRS]);
//SpuContactManifoldCollisionAlgorithm gSpuContactManifoldAlgo;
ATTRIBUTE_ALIGNED16(char gSpuContactManifoldAlgo[sizeof(SpuContactManifoldCollisionAlgorithm)+128]);
//ATTRIBUTE_ALIGNED16(char gSpuContactManifoldAlgo[sizeof(SpuContactManifoldCollisionAlgorithm)+128]);
SpuContactManifoldCollisionAlgorithm gSpuContactManifoldAlgo;
SpuContactManifoldCollisionAlgorithm* getlocalCollisionAlgorithm()
{
return (SpuContactManifoldCollisionAlgorithm*)&gSpuContactManifoldAlgo;
}
btPersistentManifold gPersistentManifold;
btBroadphaseProxy gProxy0;
btBroadphaseProxy gProxy1;
btCollisionObject gColObj0;
btCollisionObject gColObj1;
ATTRIBUTE_ALIGNED16(char gCollisionShape0[MAX_SHAPE_SIZE]);
ATTRIBUTE_ALIGNED16(char gCollisionShape1[MAX_SHAPE_SIZE]);
@@ -87,7 +101,8 @@ void* createCollisionLocalStoreMemory()
#elif defined(__CELLOS_LV2__)
CollisionTask_LocalStoreMemory gLocalStoreMemory;
ATTRIBUTE_ALIGNED16(CollisionTask_LocalStoreMemory gLocalStoreMemory);
void* createCollisionLocalStoreMemory()
{
return &gLocalStoreMemory;
@@ -160,8 +175,7 @@ void small_cache_read(void* buffer, uint64_t ea, size_t size)
void* ls = spe_cache_read(ea);
memcpy(buffer, ls, size);
#else
cellDmaLargeGet(buffer, ea, size, DMA_TAG(16), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(16));
stallingUnalignedDmaSmallGet(buffer,ea,size);
#endif
}
@@ -664,10 +678,12 @@ void processCollisionTask(void* userPtr, void* lsMemPtr)
CollisionTask_LocalStoreMemory* colMemPtr = (CollisionTask_LocalStoreMemory*)lsMemPtr;
CollisionTask_LocalStoreMemory& lsMem = *(colMemPtr);
spu_printf("taskDescPtr=%llx\n",taskDescPtr);
// spu_printf("taskDescPtr=%llx\n",taskDescPtr);
SpuContactResult spuContacts;
////////////////////
uint64_t dmaInPtr = taskDesc.inPtr;
unsigned int numPages = taskDesc.numPages;
unsigned int numOnLastPage = taskDesc.numOnLastPage;
@@ -679,7 +695,6 @@ void processCollisionTask(void* userPtr, void* lsMemPtr)
for (unsigned int i = 0; i < numPages; i++)
{
// wait for back buffer dma and swap buffers
unsigned char *inputPtr = lsMem.g_workUnitTaskBuffers.swapBuffers();
@@ -701,18 +716,12 @@ void processCollisionTask(void* userPtr, void* lsMemPtr)
for (j = 0; j < numOnPage; j++)
{
spu_printf("numOnPage=%d\n",numOnPage);
#ifdef DEBUG_SPU_COLLISION_DETECTION
printMidphaseInput(&wuInputs[j]);
#endif //DEBUG_SPU_COLLISION_DETECTION
int numPairs = wuInputs[j].m_endIndex - wuInputs[j].m_startIndex;
// printf("startIndex=%d, endIndex = %d\n",wuInputs[j].m_startIndex,wuInputs[j].m_endIndex);
if (numPairs)
{
@@ -728,11 +737,19 @@ void processCollisionTask(void* userPtr, void* lsMemPtr)
//for each broadphase pair, do something
btBroadphasePair& pair = lsMem.gBroadphasePairs[p];
int userInfo = int(pair.m_userInfo);
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("pair->m_userInfo = %d\n",pair.m_userInfo);
spu_printf("pair->m_algorithm = %d\n",pair.m_algorithm);
spu_printf("pair->m_pProxy0 = %d\n",pair.m_pProxy0);
spu_printf("pair->m_pProxy1 = %d\n",pair.m_pProxy1);
#endif //DEBUG_SPU_COLLISION_DETECTION
int userInfo = int(pair.m_userInfo);
if (userInfo == 2 && pair.m_algorithm && pair.m_pProxy0 && pair.m_pProxy1)
{
{
int dmaSize = sizeof(SpuContactManifoldCollisionAlgorithm);
uint64_t dmaPpuAddress2 = (uint64_t)pair.m_algorithm;
@@ -743,184 +760,216 @@ void processCollisionTask(void* userPtr, void* lsMemPtr)
SpuCollisionPairInput collisionPairInput;
collisionPairInput.m_persistentManifoldPtr = (uint64_t) lsMem.getlocalCollisionAlgorithm()->getContactManifoldPtr();
collisionPairInput.m_persistentManifoldPtr = (uint64_t) lsMem.gSpuContactManifoldAlgo.getContactManifoldPtr();
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("SPU: manifoldPtr: %llx",collisionPairInput->m_persistentManifoldPtr);
#endif //DEBUG_SPU_COLLISION_DETECTION
{
int dmaSize = sizeof(btBroadphaseProxy);
uint64_t dmaPpuAddress2 = (uint64_t)pair.m_pProxy0;
cellDmaGet(&lsMem.gProxy0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
{
int dmaSize = sizeof(btBroadphaseProxy);
uint64_t dmaPpuAddress2 = (uint64_t)pair.m_pProxy1;
cellDmaGet(&lsMem.gProxy1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
//btCollisionObject* colObj0 = (btCollisionObject*)gProxy0.m_clientObject;
//btCollisionObject* colObj1 = (btCollisionObject*)gProxy1.m_clientObject;
{
int dmaSize = sizeof(btBroadphaseProxy);
//spu_printf("dmaSize btBroadphaseProxy1 = %d\n",dmaSize);
uint64_t dmaPpuAddress2 = (uint64_t)pair.m_pProxy0;
lsMem.gProxyPtr0 = (btBroadphaseProxy*) lsMem.bufferProxy0;
//spu_printf("dmaPpuAddress2 btBroadphaseProxy1 = %llx, gProxyPtr0 = %d\n",dmaPpuAddress2,gProxyPtr0);
stallingUnalignedDmaSmallGet(lsMem.gProxyPtr0, dmaPpuAddress2 , dmaSize);
}
{
int dmaSize = sizeof(btBroadphaseProxy);
uint64_t dmaPpuAddress2 = (uint64_t)pair.m_pProxy1;
lsMem.gProxyPtr1 = (btBroadphaseProxy*) lsMem.bufferProxy1;
stallingUnalignedDmaSmallGet(lsMem.gProxyPtr1, dmaPpuAddress2 , dmaSize);
}
//btCollisionObject* colObj0 = (btCollisionObject*)gProxy0.m_clientObject;
//btCollisionObject* colObj1 = (btCollisionObject*)gProxy1.m_clientObject;
{
int dmaSize = sizeof(btCollisionObject);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gProxy0.m_clientObject;
cellDmaGet(&lsMem.gColObj0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
{
int dmaSize = sizeof(btCollisionObject);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gProxy1.m_clientObject;
cellDmaGet(&lsMem.gColObj1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
///can wait on the combined DMA_MASK, or dma on the same tag
collisionPairInput.m_shapeType0 = lsMem.getlocalCollisionAlgorithm()->getShapeType0();
collisionPairInput.m_shapeType1 = lsMem.getlocalCollisionAlgorithm()->getShapeType1();
collisionPairInput.m_collisionMargin0 = lsMem.getlocalCollisionAlgorithm()->getCollisionMargin0();
collisionPairInput.m_collisionMargin1 = lsMem.getlocalCollisionAlgorithm()->getCollisionMargin1();
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("SPU collisionPairInput->m_shapeType0 = %d\n",collisionPairInput->m_shapeType0);
spu_printf("SPU collisionPairInput->m_shapeType1 = %d\n",collisionPairInput->m_shapeType1);
#endif //DEBUG_SPU_COLLISION_DETECTION
int dmaSize = sizeof(btCollisionObject);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gProxyPtr0->m_clientObject;
cellDmaGet(&lsMem.gColObj0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
{
int dmaSize = sizeof(btCollisionObject);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gProxyPtr1->m_clientObject;
cellDmaGet(&lsMem.gColObj1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
if (1)
{
collisionPairInput.m_worldTransform0 = lsMem.gColObj0.getWorldTransform();
collisionPairInput.m_worldTransform1 = lsMem.gColObj1.getWorldTransform();
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("SPU worldTrans0.origin = (%f,%f,%f)\n",
collisionPairInput->m_worldTransform0.getOrigin().getX(),
collisionPairInput->m_worldTransform0.getOrigin().getY(),
collisionPairInput->m_worldTransform0.getOrigin().getZ());
spu_printf("SPU worldTrans1.origin = (%f,%f,%f)\n",
collisionPairInput->m_worldTransform1.getOrigin().getX(),
collisionPairInput->m_worldTransform1.getOrigin().getY(),
collisionPairInput->m_worldTransform1.getOrigin().getZ());
#endif //DEBUG_SPU_COLLISION_DETECTION
///can wait on the combined DMA_MASK, or dma on the same tag
{
int dmaSize = sizeof(btPersistentManifold);
uint64_t dmaPpuAddress2 = collisionPairInput.m_persistentManifoldPtr;
cellDmaGet(&lsMem.gPersistentManifold, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
collisionPairInput.m_shapeType0 = lsMem.gSpuContactManifoldAlgo.getShapeType0();
collisionPairInput.m_shapeType1 = lsMem.gSpuContactManifoldAlgo.getShapeType1();
collisionPairInput.m_collisionMargin0 = lsMem.gSpuContactManifoldAlgo.getCollisionMargin0();
collisionPairInput.m_collisionMargin1 = lsMem.gSpuContactManifoldAlgo.getCollisionMargin1();
if (btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType0)
&& btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType1))
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("SPU collisionPairInput->m_shapeType0 = %d\n",collisionPairInput->m_shapeType0);
spu_printf("SPU collisionPairInput->m_shapeType1 = %d\n",collisionPairInput->m_shapeType1);
#endif //DEBUG_SPU_COLLISION_DETECTION
if (1)
{
collisionPairInput.m_worldTransform0 = lsMem.gColObj0.getWorldTransform();
collisionPairInput.m_worldTransform1 = lsMem.gColObj1.getWorldTransform();
#ifdef DEBUG_SPU_COLLISION_DETECTION
spu_printf("SPU worldTrans0.origin = (%f,%f,%f)\n",
collisionPairInput->m_worldTransform0.getOrigin().getX(),
collisionPairInput->m_worldTransform0.getOrigin().getY(),
collisionPairInput->m_worldTransform0.getOrigin().getZ());
spu_printf("SPU worldTrans1.origin = (%f,%f,%f)\n",
collisionPairInput->m_worldTransform1.getOrigin().getX(),
collisionPairInput->m_worldTransform1.getOrigin().getY(),
collisionPairInput->m_worldTransform1.getOrigin().getZ());
#endif //DEBUG_SPU_COLLISION_DETECTION
{
int dmaSize = getShapeTypeSize( collisionPairInput.m_shapeType0);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj0.getCollisionShape();
cellDmaGet(lsMem.gCollisionShape0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
int dmaSize = sizeof(btPersistentManifold);
uint64_t dmaPpuAddress2 = collisionPairInput.m_persistentManifoldPtr;
if (dmaPpuAddress2 & 0x0f)
{
#ifndef IGNORE_ALIGNMENT
spu_printf("SPU ALIGNMENT ERROR\n");
spuContacts.flush();
return;
#endif
}
cellDmaGet(&lsMem.gPersistentManifold, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
if (btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType0)
&& btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType1))
{
int dmaSize = getShapeTypeSize( collisionPairInput.m_shapeType1);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj1.getCollisionShape();
cellDmaGet(lsMem.gCollisionShape1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
btConvexShape* spuConvexShape0 = (btConvexShape*)lsMem.gCollisionShape0;
btConvexShape* spuConvexShape1 = (btConvexShape*)lsMem.gCollisionShape1;
btVector3 dim0 = spuConvexShape0->getImplicitShapeDimensions();
btVector3 dim1 = spuConvexShape1->getImplicitShapeDimensions();
collisionPairInput.m_primitiveDimensions0 = dim0;
collisionPairInput.m_primitiveDimensions1 = dim1;
collisionPairInput.m_collisionShapes[0] = (uint64_t)lsMem.gColObj0.getCollisionShape();
collisionPairInput.m_collisionShapes[1] = (uint64_t)lsMem.gColObj1.getCollisionShape();
collisionPairInput.m_spuCollisionShapes[0] = spuConvexShape0;
collisionPairInput.m_spuCollisionShapes[1] = spuConvexShape1;
ProcessSpuConvexConvexCollision(&collisionPairInput,&lsMem, spuContacts);
} else
{
//a non-convex shape is involved
bool isSwapped = false;
bool handleConvexConcave = false;
if (btBroadphaseProxy::isConcave(collisionPairInput.m_shapeType0) &&
btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType1))
{
isSwapped = true;
spu_printf("SPU convex/concave swapped, unsupported!\n");
handleConvexConcave = true;
}
if (btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType0)&&
btBroadphaseProxy::isConcave(collisionPairInput.m_shapeType1))
{
handleConvexConcave = true;
}
if (handleConvexConcave && !isSwapped)
{
// spu_printf("SPU: non-convex detected\n");
{
// uint64_t dmaPpuAddress2 = (uint64_t)gProxy1.m_clientObject;
// spu_printf("SPU: gColObj1 trimesh = %llx\n",dmaPpuAddress2);
}
///dma and initialize the convex object
{
int dmaSize = getShapeTypeSize( collisionPairInput.m_shapeType0);
int dmaSize = getShapeTypeSize(collisionPairInput.m_shapeType0);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj0.getCollisionShape();
cellDmaGet(lsMem.gCollisionShape0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
///dma and initialize the concave object
{
int dmaSize = getShapeTypeSize( collisionPairInput.m_shapeType1);
int dmaSize = getShapeTypeSize(collisionPairInput.m_shapeType1);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj1.getCollisionShape();
// spu_printf("SPU: trimesh = %llx\n",dmaPpuAddress2);
if (dmaPpuAddress2 & 0x0f)
{
#ifndef IGNORE_ALIGNMENT
spu_printf("SPU ALIGNMENT ERROR2\n");
spuContacts.flush();
return;
#endif //IGNORE_ALIGNMENT
}
cellDmaGet(lsMem.gCollisionShape1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
btConvexShape* spuConvexShape0 = (btConvexShape*)lsMem.gCollisionShape0;
btBvhTriangleMeshShape* trimeshShape = (btBvhTriangleMeshShape*)lsMem.gCollisionShape1;
btConvexShape* spuConvexShape1 = (btConvexShape*)lsMem.gCollisionShape1;
btVector3 dim0 = spuConvexShape0->getImplicitShapeDimensions();
btVector3 dim1 = spuConvexShape1->getImplicitShapeDimensions();
collisionPairInput.m_primitiveDimensions0 = dim0;
collisionPairInput.m_primitiveDimensions1 = dim1;
collisionPairInput.m_collisionShapes[0] = (uint64_t)lsMem.gColObj0.getCollisionShape();
collisionPairInput.m_collisionShapes[1] = (uint64_t)lsMem.gColObj1.getCollisionShape();
collisionPairInput.m_spuCollisionShapes[0] = spuConvexShape0;
collisionPairInput.m_spuCollisionShapes[1] = trimeshShape;
ProcessConvexConcaveSpuCollision(&collisionPairInput,&lsMem,spuContacts);
collisionPairInput.m_spuCollisionShapes[1] = spuConvexShape1;
ProcessSpuConvexConvexCollision(&collisionPairInput,&lsMem,spuContacts);
} else
{
//a non-convex shape is involved
bool isSwapped = false;
bool handleConvexConcave = false;
if (btBroadphaseProxy::isConcave(collisionPairInput.m_shapeType0) &&
btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType1))
{
isSwapped = true;
spu_printf("SPU convex/concave swapped, unsupported!\n");
handleConvexConcave = true;
}
if (btBroadphaseProxy::isConvex(collisionPairInput.m_shapeType0)&&
btBroadphaseProxy::isConcave(collisionPairInput.m_shapeType1))
{
handleConvexConcave = true;
}
if (handleConvexConcave && !isSwapped)
{
// spu_printf("SPU: non-convex detected\n");
{
// uint64_t dmaPpuAddress2 = (uint64_t)gProxy1.m_clientObject;
// spu_printf("SPU: gColObj1 trimesh = %llx\n",dmaPpuAddress2);
}
///dma and initialize the convex object
{
int dmaSize = getShapeTypeSize(collisionPairInput.m_shapeType0);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj0.getCollisionShape();
cellDmaGet(lsMem.gCollisionShape0, dmaPpuAddress2 , dmaSize, DMA_TAG(1), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(1));
}
///dma and initialize the convex object
{
int dmaSize = getShapeTypeSize(collisionPairInput.m_shapeType1);
uint64_t dmaPpuAddress2 = (uint64_t)lsMem.gColObj1.getCollisionShape();
if (dmaPpuAddress2 & 0x0f)
{
#ifndef IGNORE_ALIGNMENT
spu_printf("SPU ALIGNMENT ERROR3\n");
spuContacts.flush();
return;
#endif //
}
// spu_printf("SPU: trimesh = %llx\n",dmaPpuAddress2);
cellDmaGet(lsMem.gCollisionShape1, dmaPpuAddress2 , dmaSize, DMA_TAG(2), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(2));
}
btConvexShape* spuConvexShape0 = (btConvexShape*)lsMem.gCollisionShape0;
btBvhTriangleMeshShape* trimeshShape = (btBvhTriangleMeshShape*)lsMem.gCollisionShape1;
btVector3 dim0 = spuConvexShape0->getImplicitShapeDimensions();
collisionPairInput.m_primitiveDimensions0 = dim0;
collisionPairInput.m_collisionShapes[0] = (uint64_t)lsMem.gColObj0.getCollisionShape();
collisionPairInput.m_collisionShapes[1] = (uint64_t)lsMem.gColObj1.getCollisionShape();
collisionPairInput.m_spuCollisionShapes[0] = spuConvexShape0;
collisionPairInput.m_spuCollisionShapes[1] = trimeshShape;
ProcessConvexConcaveSpuCollision(&collisionPairInput,&lsMem,spuContacts);
}
}
}
spuContacts.flush();
}
spuContacts.flush();
}
}
}
}
}
}
}
} //end for (j = 0; j < numOnPage; j++)
}// for
return;
}