static void accumulate(PvdHitType& query, Ps::Array<PvdHitType>& accumulated, const char* arrayName, Ps::Array<PvdSqHit>& dst, const SDKHitType* src, PxU32 nb, const PxQueryFilterData& fd)
{
query.mFilterFlags = fd.flags;
query.mHits = PvdReference(arrayName, dst.size(), nb);
PX_ASSERT(PxU32(-1) != nb);
for(PxU32 i=0; i<nb; i++)
dst.pushBack(PvdSqHit(src[i]));
accumulated.pushBack(query);
}
void ProfileEventHandler::reportCudaCollection(PxBufferedProfilerCallback& callback)
{
if(mCUDAEventOccurences.empty())
return;
// sort the cuda occurrences - according to eventID and startTime
Ps::sort(mCUDAEventOccurences.begin(), mCUDAEventOccurences.size(), SortCUDAProfileOccurences());
PxU16 currentEventId = mCUDAEventOccurences[0].eventId;
CUDAProfileEventOccurence currentEvent = mCUDAEventOccurences[0];
Ps::Array<CUDAProfileEventOccurence> outOccurences;
// now group the occurrences that do overlap
// put the group occurrences into out array
for (PxU32 i = 1; i < mCUDAEventOccurences.size(); i++)
{
const CUDAProfileEventOccurence& occurence = mCUDAEventOccurences[i];
// occurrences are sorted by eventId, so once it changes, we have new group occurrence
if(currentEventId != occurence.eventId)
{
outOccurences.pushBack(currentEvent);
currentEvent = occurence;
currentEventId = occurence.eventId;
}
else
{
// occurrences are sorted by start time, so we check for overlap and add occurrence or create new group
if(currentEvent.endTime >= occurence.startTime)
{
currentEvent.endTime = PxMax(currentEvent.endTime, occurence.endTime);
}
else
{
outOccurences.pushBack(currentEvent);
currentEvent = occurence;
}
}
}
outOccurences.pushBack(currentEvent);
// fire the callback with grouped occurrences
for (PxU32 i = outOccurences.size(); i--;)
{
const CUDAProfileEventOccurence& ev = outOccurences[i];
const char* name = mProfileZoneInterface->getProfileEventName(ev.eventId);
const char* profileZoneName = mProfileZoneInterface->getName();
PxBufferedProfilerEvent ce = { ev.startTime, ev.endTime, name, profileZoneName, ev.eventId, 0, 0, 0, 0 };
callback.onEvent(ce);
}
}
void DeformableMesh::generateConstraintsFromTriangles()
{
PxU32 numTriangles = mPrimitives.size() / 3;
DeformableTriEdge e;
Ps::Array<DeformableTriEdge> edges PX_DEBUG_EXP("defoMeshEdges");
edges.reserve(numTriangles * 3);
const PxU32 *i0 = mPrimitives.begin();
for (PxU32 i = 0; i < numTriangles; i++, i0 += 3)
{
const PxU32 *i1 = i0+1;
const PxU32 *i2 = i0+2;
e.set(mVertexToParticleMap[*i0], mVertexToParticleMap[*i1], mVertexToParticleMap[*i2], i);
edges.pushBack(e);
e.set(mVertexToParticleMap[*i1], mVertexToParticleMap[*i2], mVertexToParticleMap[*i0], i);
edges.pushBack(e);
e.set(mVertexToParticleMap[*i2], mVertexToParticleMap[*i0], mVertexToParticleMap[*i1], i);
edges.pushBack(e);
}
quickSortEdges(edges, 0, edges.size()-1);
DeformableConstraint constraint;
constraint.particleId[4] = -1; // only used when torn
constraint.particleId[5] = -1; // only used when torn
for(PxU32 i=0; i<edges.size(); )
{
const DeformableTriEdge &e0 = edges[i];
PxU32 p0 = constraint.particleId[0] = e0.particleId[0];
PxU32 p1 = constraint.particleId[1] = e0.particleId[1];
PxVec3 d0 = mWeldedVertices[p1] - mWeldedVertices[p0];
constraint.stretchingRestLength = d0.magnitude();
constraint.particleId[2] = e0.third;
constraint.particleId[3] = -1; // for border edges -> no bending possible
constraint.bendingRestLength = 0.0f;
constraint.flags = 0;
if (++i < edges.size()) {
const DeformableTriEdge &e1 = edges[i];
if (e0 == e1) {
constraint.particleId[2] = e0.third;
constraint.particleId[3] = e1.third;
PxVec3 d1 = mWeldedVertices[e1.third] - mWeldedVertices[e0.third];
constraint.bendingRestLength = d1.magnitude();
}
while (i < edges.size() && edges[i] == e0)
++i;
}
mConstraints.pushBack(constraint);
}
}
void InternalIndexPool::freeIndices(PxU32 num, const PxStrideIterator<const PxU32>& indexBuffer)
{
PxU32 numAllocated = mIndexCount - mFreeList.size();
if (num > numAllocated)
{
Ps::getFoundation().error(PxErrorCode::eDEBUG_WARNING, __FILE__, __LINE__,
"PxParticleExt::IndexPool::freeIndices: Provided number of indices exceeds number of actually allocated indices. Call faild.");
return;
}
#ifdef PX_CHECK
for (PxU32 i = 0; i < num; ++i)
{
if (indexBuffer[i] < mIndexCount)
continue;
Ps::getFoundation().error(PxErrorCode::eDEBUG_WARNING, __FILE__, __LINE__,
"PxParticleExt::IndexPool::freeIndices: Provided indices which where not actually allocated before. Call failed.");
return;
}
#endif
for (PxU32 i = 0; i < num; ++i)
mFreeList.pushBack(indexBuffer[i]);
}
bool Cm::CompleteBoxPruning(const PxBounds3* bounds, PxU32 nb, Ps::Array<PxU32>& pairs, const Axes& axes)
{
pairs.clear();
// Checkings
if(!nb)
return false;
// Catch axes
const PxU32 Axis0 = axes.mAxis0;
const PxU32 Axis1 = axes.mAxis1;
const PxU32 Axis2 = axes.mAxis2;
PX_UNUSED(Axis1);
PX_UNUSED(Axis2);
// Allocate some temporary data
float* PosList = reinterpret_cast<float*>(PX_ALLOC_TEMP(sizeof(float)*nb, "Cm::CompleteBoxPruning"));
// 1) Build main list using the primary axis
for(PxU32 i=0;i<nb;i++) PosList[i] = bounds[i].minimum[Axis0];
// 2) Sort the list
/*static*/ RadixSortBuffered RS; // Static for coherence
const PxU32* Sorted = RS.Sort(PosList, nb).GetRanks();
// 3) Prune the list
const PxU32* const LastSorted = &Sorted[nb];
const PxU32* RunningAddress = Sorted;
PxU32 Index0, Index1;
while(RunningAddress<LastSorted && Sorted<LastSorted)
{
Index0 = *Sorted++;
while(RunningAddress<LastSorted && PosList[*RunningAddress++]<PosList[Index0]);
const PxU32* RunningAddress2 = RunningAddress;
while(RunningAddress2<LastSorted && PosList[Index1 = *RunningAddress2++]<=bounds[Index0].maximum[Axis0])
{
if(Index0!=Index1)
{
if(bounds[Index0].intersects(bounds[Index1]))
{
pairs.pushBack(Index0);
pairs.pushBack(Index1);
}
}
}
}
PX_FREE(PosList);
return true;
}
void ProfileEventHandler::onCUDAProfileBuffer( PxU64 , PxF32 , const PxU8* cudaData, PxU32 bufLenInBytes, PxU32 bufferVersion )
{
if(bufferVersion == 1)
{
struct currentWarpProfileEvent
{
PxU16 block;
PxU8 warp;
PxU8 mpId;
PxU8 hwWarpId;
PxU8 userDataCfg;
PxU16 eventId;
PxU32 startTime;
PxU32 endTime;
};
Ps::Array<Local::OverflowRecordPair> overflowPair;
//Run through dataset. We need to be able to correct rollover, meaning one of the timers
//runs through PxU32.MAX_VALUE and resets to zero.
PxU32 numEvents = bufLenInBytes/sizeof(currentWarpProfileEvent);
const currentWarpProfileEvent* cudaEvents = reinterpret_cast<const currentWarpProfileEvent*> (cudaData);
for (PxU32 i = 0; i < numEvents; i++)
{
const currentWarpProfileEvent& cudaEvent = cudaEvents[i];
Local::OverflowRecord* record = NULL;
for (PxU32 j = 0; j < overflowPair.size(); j++)
{
if(overflowPair[j].mpId == cudaEvent.mpId)
{
record = &overflowPair[j].mOverflowRecord;
break;
}
}
if(!record)
{
Local::OverflowRecordPair pair;
pair.mpId = cudaEvent.mpId;
overflowPair.pushBack(pair);
record = &overflowPair.back().mOverflowRecord;
}
//account for overflow
PxU64 startTime = record->NextValue(cudaEvent.startTime);
PxU64 endTime = record->NextValue(cudaEvent.endTime);
CUDAProfileEventOccurence cudaEventOccurence = { cudaEvent.eventId, startTime, endTime };
mCUDAEventOccurences.pushBack(cudaEventOccurence);
}
}
}
static void collectBatchedHits(const QueryResultT* results, Ps::Array<PvdHitType>& accumulated, Ps::Array<PvdSqHit>& pvdSqHits, PxU32 nb, PxU32 startIdx, const char* arrayName)
{
for(PxU32 i=0; i<nb; i++)
{
const QueryResultT& result = results[i];
if(result.queryStatus != PxBatchQueryStatus::eSUCCESS)
continue;
PvdHitType& query = accumulated[startIdx + i];
const PxU32 nbAnyHits = result.getNbAnyHits();
if(query.mHits.mCount != nbAnyHits)
{
query.mHits = PvdReference(arrayName, pvdSqHits.size(), nbAnyHits);
for(PxU32 j=0; j<nbAnyHits; j++)
pvdSqHits.pushBack(PvdSqHit(result.getAnyHit(j)));
}
}
}
template<class Type> static void pushBackT(Ps::Array<Type>& array, const Type& item, PvdReference& ref, const char* arrayName)
{
ref = PvdReference(arrayName, array.size(), 1);
array.pushBack(item);
}
bool Cm::BipartiteBoxPruning(const PxBounds3* bounds0, PxU32 nb0, const PxBounds3* bounds1, PxU32 nb1, Ps::Array<PxU32>& pairs, const Axes& axes)
{
pairs.clear();
// Checkings
if(nb0 == 0 || nb1 == 0)
return false;
// Catch axes
PxU32 Axis0 = axes.mAxis0;
PxU32 Axis1 = axes.mAxis1;
PxU32 Axis2 = axes.mAxis2;
PX_UNUSED(Axis1);
PX_UNUSED(Axis2);
// Allocate some temporary data
float* MinPosBounds0 = reinterpret_cast<float*>(PX_ALLOC_TEMP(sizeof(float)*nb0, "Gu::BipartiteBoxPruning"));
float* MinPosBounds1 = reinterpret_cast<float*>(PX_ALLOC_TEMP(sizeof(float)*nb1, "Gu::BipartiteBoxPruning"));
// 1) Build main lists using the primary axis
for(PxU32 i=0;i<nb0;i++) MinPosBounds0[i] = bounds0[i].minimum[Axis0];
for(PxU32 i=0;i<nb1;i++) MinPosBounds1[i] = bounds1[i].minimum[Axis0];
// 2) Sort the lists
//static RadixSort RS0, RS1; // Static for coherence. Crashes on exit
RadixSortBuffered RS0, RS1; // Static for coherence.
const PxU32* Sorted0 = RS0.Sort(MinPosBounds0, nb0).GetRanks();
const PxU32* Sorted1 = RS1.Sort(MinPosBounds1, nb1).GetRanks();
// 3) Prune the lists
PxU32 Index0, Index1;
const PxU32* const LastSorted0 = &Sorted0[nb0];
const PxU32* const LastSorted1 = &Sorted1[nb1];
const PxU32* RunningAddress0 = Sorted0;
const PxU32* RunningAddress1 = Sorted1;
while(RunningAddress1<LastSorted1 && Sorted0<LastSorted0)
{
Index0 = *Sorted0++;
while(RunningAddress1<LastSorted1 && MinPosBounds1[*RunningAddress1]<MinPosBounds0[Index0]) RunningAddress1++;
const PxU32* RunningAddress2_1 = RunningAddress1;
while(RunningAddress2_1<LastSorted1 && MinPosBounds1[Index1 = *RunningAddress2_1++]<=bounds0[Index0].maximum[Axis0])
{
if(bounds0[Index0].intersects(bounds1[Index1]))
{
pairs.pushBack(Index0);
pairs.pushBack(Index1);
}
}
}
////
while(RunningAddress0<LastSorted0 && Sorted1<LastSorted1)
{
Index0 = *Sorted1++;
while(RunningAddress0<LastSorted0 && MinPosBounds0[*RunningAddress0]<=MinPosBounds1[Index0]) RunningAddress0++;
const PxU32* RunningAddress2_0 = RunningAddress0;
while(RunningAddress2_0<LastSorted0 && MinPosBounds0[Index1 = *RunningAddress2_0++]<=bounds1[Index0].maximum[Axis0])
{
if(bounds0[Index1].intersects(bounds1[Index0]))
{
pairs.pushBack(Index1);
pairs.pushBack(Index0);
}
}
}
PX_FREE(MinPosBounds1);
PX_FREE(MinPosBounds0);
return true;
}
void DeformableMesh::generateConstraintsFromTetrahedra()
{
PxU32 edgeIndices[6][2] = { {0,1}, {0,2}, {0,3}, {1,2}, {1,3}, {2,3} };
PxU32 *tetIndices;
// - tetrahedra are assumed to be unique (thus no parent code here)
PxU32 numTetrahedra = mPrimitives.size() / 4;
DeformableTetraEdge e;
Ps::Array<DeformableTetraEdge> edges PX_DEBUG_EXP("defoMeshEdges2");
edges.reserve(numTetrahedra * 6);
tetIndices = mPrimitives.begin();
PxU32 i, j;
for (i = 0; i < numTetrahedra; i++, tetIndices += 4)
{
for(j = 0; j < 6; j++)
{
PxU32 e0 = mVertexToParticleMap[tetIndices[edgeIndices[j][0]]];
PxU32 e1 = mVertexToParticleMap[tetIndices[edgeIndices[j][1]]];
e.set(e0, e1, i); edges.pushBack(e);
}
}
quickSortTetraEdges(edges, 0, edges.size()-1);
mConstraints.resize(numTetrahedra);
DeformableConstraint constraint;
DeformableTetraEdge *tetEdges[6];
tetIndices = mPrimitives.begin();
bool warningIssued = false;
for (i = 0; i < numTetrahedra; i++, tetIndices += 4)
{
for (j = 0; j < 6; j++)
{
PxU32 e0 = mVertexToParticleMap[tetIndices[edgeIndices[j][0]]];
PxU32 e1 = mVertexToParticleMap[tetIndices[edgeIndices[j][1]]];
DeformableTetraEdge goalEdge;
goalEdge.set(e0, e1, i);
tetEdges[j] = binarySearchTetraEdge(edges, goalEdge);
}
for (j = 0; j < 4; j++)
constraint.particleId[j] = mVertexToParticleMap[tetIndices[j]];
PxVec3 groundArea = (mWeldedVertices[tetIndices[1]] - mWeldedVertices[tetIndices[0]]).cross(mWeldedVertices[tetIndices[2]] - mWeldedVertices[tetIndices[0]]);
constraint.restVolume = groundArea.dot(mWeldedVertices[tetIndices[3]] - mWeldedVertices[tetIndices[0]]);
constraint.flags = 0;
if (!warningIssued && constraint.restVolume < 0.0f)
{
Ps::getFoundation().error(PxErrorCode::eDEBUG_WARNING, __FILE__, __LINE__, "Soft body mesh tetrahedron %d has illegal winding order.", i);
warningIssued = true;
}
for (j = 0; j < 6; j++)
{
PxU32 e0 = mVertexToParticleMap[tetIndices[edgeIndices[j][0]]];
PxU32 e1 = mVertexToParticleMap[tetIndices[edgeIndices[j][1]]];
PxVec3 edgeVec = mWeldedVertices[e1] - mWeldedVertices[e0];
if(tetEdges[j])
{
PX_ASSERT(tetEdges[j]->tetrahedron == i);
constraint.restEdgeLengths[j] = edgeVec.magnitude();
}
else
constraint.restEdgeLengths[j] = -edgeVec.magnitude();
}
mConstraints[i] = constraint;
}
}