Add access to overlapping pairs b3OpenCLArray.

src/Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h

@@ -35,6 +35,7 @@ public:
 	virtual b3OpenCLArray<b3SapAabb>&	getAllAabbsGPU()=0;
 	virtual b3AlignedObjectArray<b3SapAabb>&	getAllAabbsCPU()=0;
 	
+	virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU() = 0;
 	virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU() = 0;
 	virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU() = 0;
 

src/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.cpp

@@ -368,6 +368,10 @@ b3AlignedObjectArray<b3SapAabb>&	b3GpuGridBroadphase::getAllAabbsCPU()
 	return m_allAabbsCPU1;
 }
 
+b3OpenCLArray<b3Int4>& b3GpuGridBroadphase::getOverlappingPairsGPU()
+{
+	return m_gpuPairs;
+}
 b3OpenCLArray<int>& b3GpuGridBroadphase::getSmallAabbIndicesGPU()
 {
 	return m_smallAabbsMappingGPU;

src/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.h

@@ -79,6 +79,7 @@ public:
 	virtual b3OpenCLArray<b3SapAabb>&	getAllAabbsGPU();
 	virtual b3AlignedObjectArray<b3SapAabb>&	getAllAabbsCPU();
 	
+	virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU();
 	virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU();
 	virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU();
 

src/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.cpp

@@ -22,6 +22,7 @@ b3GpuParallelLinearBvh::b3GpuParallelLinearBvh(cl_context context, cl_device_id
 	
 	m_rootNodeIndex(context, queue),
 	m_maxDistanceFromRoot(context, queue),
+	m_temp(context, queue),
 	
 	m_internalNodeAabbs(context, queue),
 	m_internalNodeLeafIndexRanges(context, queue),
@@ -41,6 +42,7 @@ b3GpuParallelLinearBvh::b3GpuParallelLinearBvh(cl_context context, cl_device_id
 {
 	m_rootNodeIndex.resize(1);
 	m_maxDistanceFromRoot.resize(1);
+	m_temp.resize(1);
 	
 	//
 	const char CL_PROGRAM_PATH[] = "src/Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvh.cl";
@@ -287,14 +289,13 @@ void b3GpuParallelLinearBvh::build(const b3OpenCLArray<b3SapAabb>& worldSpaceAab
 	}
 }
 
-void b3GpuParallelLinearBvh::calculateOverlappingPairs(b3OpenCLArray<int>& out_numPairs, b3OpenCLArray<b3Int4>& out_overlappingPairs)
+void b3GpuParallelLinearBvh::calculateOverlappingPairs(b3OpenCLArray<b3Int4>& out_overlappingPairs)
 {
-	b3Assert( out_numPairs.size() == 1 );
-	
 	int maxPairs = out_overlappingPairs.size();
+	b3OpenCLArray<int>& numPairsGpu = m_temp;
 	
 	int reset = 0;
-	out_numPairs.copyFromHostPointer(&reset, 1);
+	numPairsGpu.copyFromHostPointer(&reset, 1);
 	
 	//
 	if( m_leafNodeAabbs.size() > 1 )
@@ -313,7 +314,7 @@ void b3GpuParallelLinearBvh::calculateOverlappingPairs(b3OpenCLArray<int>& out_n
 			b3BufferInfoCL( m_internalNodeLeafIndexRanges.getBufferCL() ),
 			b3BufferInfoCL( m_mortonCodesAndAabbIndicies.getBufferCL() ),
 			
-			b3BufferInfoCL( out_numPairs.getBufferCL() ),
+			b3BufferInfoCL( numPairsGpu.getBufferCL() ),
 			b3BufferInfoCL( out_overlappingPairs.getBufferCL() )
 		};
 		
@@ -338,7 +339,7 @@ void b3GpuParallelLinearBvh::calculateOverlappingPairs(b3OpenCLArray<int>& out_n
 			b3BufferInfoCL( m_leafNodeAabbs.getBufferCL() ),
 			b3BufferInfoCL( m_largeAabbs.getBufferCL() ),
 			
-			b3BufferInfoCL( out_numPairs.getBufferCL() ),
+			b3BufferInfoCL( numPairsGpu.getBufferCL() ),
 			b3BufferInfoCL( out_overlappingPairs.getBufferCL() )
 		};
 		
@@ -355,12 +356,12 @@ void b3GpuParallelLinearBvh::calculateOverlappingPairs(b3OpenCLArray<int>& out_n
 	
 	//
 	int numPairs = -1;
-	out_numPairs.copyToHostPointer(&numPairs, 1);
+	numPairsGpu.copyToHostPointer(&numPairs, 1);
 	if(numPairs > maxPairs)
 	{
 		b3Error("Error running out of pairs: numPairs = %d, maxPairs = %d.\n", numPairs, maxPairs);
 		numPairs = maxPairs;
-		out_numPairs.copyFromHostPointer(&maxPairs, 1);
+		numPairsGpu.copyFromHostPointer(&maxPairs, 1);
 	}
 	
 	out_overlappingPairs.resize(numPairs);

src/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.h

@@ -76,6 +76,7 @@ class b3GpuParallelLinearBvh
 	//1 element
 	b3OpenCLArray<int> m_rootNodeIndex;							//Most significant bit(0x80000000) is set to indicate internal node
 	b3OpenCLArray<int> m_maxDistanceFromRoot;					//Max number of internal nodes between an internal node and the root node
+	b3OpenCLArray<int> m_temp;									//Used to hold the number of pairs in calculateOverlappingPairs()
 	
 	//1 element per internal node (number_of_internal_nodes == number_of_leaves - 1)
 	b3OpenCLArray<b3SapAabb> m_internalNodeAabbs;
@@ -101,15 +102,14 @@ public:
 	b3GpuParallelLinearBvh(cl_context context, cl_device_id device, cl_command_queue queue);
 	virtual ~b3GpuParallelLinearBvh();
 	
+	///Must be called before any other function
 	void build(const b3OpenCLArray<b3SapAabb>& worldSpaceAabbs, const b3OpenCLArray<int>& smallAabbIndices, 
 				const b3OpenCLArray<int>& largeAabbIndices);
 	
-	///b3GpuParallelLinearBvh::build() must be called before this function. calculateOverlappingPairs() uses
+	///calculateOverlappingPairs() uses the worldSpaceAabbs parameter of b3GpuParallelLinearBvh::build() as the query AABBs.
-	///the worldSpaceAabbs parameter of b3GpuParallelLinearBvh::build() as the query AABBs.
-	///@param out_numPairs If number of pairs exceeds the max number of pairs, this is clamped to the max number.
 	///@param out_overlappingPairs The size() of this array is used to determine the max number of pairs.
 	///If the number of overlapping pairs is < out_overlappingPairs.size(), out_overlappingPairs is resized.
-	void calculateOverlappingPairs(b3OpenCLArray<int>& out_numPairs, b3OpenCLArray<b3Int4>& out_overlappingPairs);
+	void calculateOverlappingPairs(b3OpenCLArray<b3Int4>& out_overlappingPairs);
 	
 	///@param out_numRigidRayPairs Array of length 1; contains the number of detected ray-rigid AABB intersections;
 	///this value may be greater than out_rayRigidPairs.size() if out_rayRigidPairs is not large enough.

src/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.cpp

@@ -16,14 +16,12 @@ subject to the following restrictions:
 b3GpuParallelLinearBvhBroadphase::b3GpuParallelLinearBvhBroadphase(cl_context context, cl_device_id device, cl_command_queue queue) : 
 	m_plbvh(context, device, queue),
 	
-	m_numOverlappingPairs(context, queue),
 	m_overlappingPairsGpu(context, queue),
 	
 	m_aabbsGpu(context, queue),
 	m_smallAabbsMappingGpu(context, queue),
 	m_largeAabbsMappingGpu(context, queue)
 {
-	m_numOverlappingPairs.resize(1);
 }
 
 void b3GpuParallelLinearBvhBroadphase::createProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr, short int collisionFilterGroup, short int collisionFilterMask)
@@ -64,7 +62,7 @@ void b3GpuParallelLinearBvhBroadphase::calculateOverlappingPairs(int maxPairs)
 	
 	//
 	m_overlappingPairsGpu.resize(maxPairs);
-	m_plbvh.calculateOverlappingPairs(m_numOverlappingPairs, m_overlappingPairsGpu);
+	m_plbvh.calculateOverlappingPairs(m_overlappingPairsGpu);
 }
 void b3GpuParallelLinearBvhBroadphase::calculateOverlappingPairsHost(int maxPairs)
 {

src/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.h

@@ -22,7 +22,6 @@ class b3GpuParallelLinearBvhBroadphase : public b3GpuBroadphaseInterface
 {
 	b3GpuParallelLinearBvh m_plbvh;
 	
-	b3OpenCLArray<int> m_numOverlappingPairs;
 	b3OpenCLArray<b3Int4> m_overlappingPairsGpu;
 	
 	b3OpenCLArray<b3SapAabb> m_aabbsGpu;
@@ -51,6 +50,8 @@ public:
 
 	virtual cl_mem getAabbBufferWS() { return m_aabbsGpu.getBufferCL(); }
 	virtual b3OpenCLArray<b3SapAabb>& getAllAabbsGPU() { return m_aabbsGpu; }
+	
+	virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU() { return m_overlappingPairsGpu; }
 	virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU() { return m_smallAabbsMappingGpu; }
 	virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU() { return m_largeAabbsMappingGpu; }
 	

src/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.cpp

@@ -1308,6 +1308,10 @@ cl_mem	b3GpuSapBroadphase::getOverlappingPairBuffer()
 	return m_overlappingPairs.getBufferCL();
 }
 
+b3OpenCLArray<b3Int4>& b3GpuSapBroadphase::getOverlappingPairsGPU()
+{
+	return m_overlappingPairs;
+}
 b3OpenCLArray<int>& b3GpuSapBroadphase::getSmallAabbIndicesGPU()
 {
 	return m_smallAabbsMappingGPU;

src/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.h

@@ -143,6 +143,7 @@ public:
 	virtual int	getNumOverlap();
 	virtual cl_mem	getOverlappingPairBuffer();
 	
+	virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU();
 	virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU();
 	virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU();
 };