6 files changed, 52 insertions, 28 deletions

diff --git a/NvCloth/src/cuda/CuCloth.cpp b/NvCloth/src/cuda/CuCloth.cpp
index 3e6175b..4131b04 100644
--- a/NvCloth/src/cuda/CuCloth.cpp
+++ b/NvCloth/src/cuda/CuCloth.cpp
@@ -423,7 +423,7 @@ void CuCloth::clearParticleAccelerations()
 {
 	CuContextLock contextLock(mFactory);
 	CuDeviceVector<PxVec4>(mFactory.mContext).swap(mParticleAccelerations);
-	CuHostVector<PxVec4, CU_MEMHOSTALLOC_DEVICEMAP>::Type().swap(mParticleAccelerationsHostCopy);
+	CuHostVector<PxVec4, CU_MEMHOSTALLOC_DEVICEMAP>::Type(mFactory.mContext).swap(mParticleAccelerationsHostCopy);
 	wakeUp();
 }
 
diff --git a/NvCloth/src/cuda/CuClothData.h b/NvCloth/src/cuda/CuClothData.h
index dd836fd..5f2d7c6 100644
--- a/NvCloth/src/cuda/CuClothData.h
+++ b/NvCloth/src/cuda/CuClothData.h
@@ -32,7 +32,7 @@
 #include <foundation/Px.h>
 
 #ifndef __CUDACC__
-#include "simd.h"
+#include "Simd.h"
 #endif
 
 namespace nv
diff --git a/NvCloth/src/cuda/CuFabric.cpp b/NvCloth/src/cuda/CuFabric.cpp
index 9bc20db..957f912 100644
--- a/NvCloth/src/cuda/CuFabric.cpp
+++ b/NvCloth/src/cuda/CuFabric.cpp
@@ -31,6 +31,7 @@
 #include "CuContextLock.h"
 #include "CuFactory.h"
 #include <PsUtilities.h>
+#include <limits.h>
 
 using namespace physx;
 
diff --git a/NvCloth/src/cuda/CuPinnedAllocator.h b/NvCloth/src/cuda/CuPinnedAllocator.h
index 8b1787b..9939324 100644
--- a/NvCloth/src/cuda/CuPinnedAllocator.h
+++ b/NvCloth/src/cuda/CuPinnedAllocator.h
@@ -29,6 +29,8 @@
 
 #pragma once
 
+#include <utility>
+
 #include "CuCheckSuccess.h"
 #include "NvCloth/Allocator.h"
 
@@ -102,7 +104,7 @@ public:
 
 	void destroy(T* ptr)
 	{
-		core::unused(ptr);
+		PX_UNUSED(ptr);
 		ptr->~T();
 	}
 
@@ -122,13 +124,13 @@ bool operator!=(const CuHostAllocator<T1, Flag1>&, const CuHostAllocator<T2, Fla
 	return false;
 }
 
-//Use CuHostVectorImpl instead of physx::shdfnd::Array<T, typename CuHostAllocator<T, Flags>> 
+//Use CuHostVectorImpl instead of physx::shdfnd::Array<T, CuHostAllocator<T, Flags>> 
 //This entire class is just to make sure that the mDevicePtr from the CuHostAllocator is properly swapped together with mData
 template <typename T, unsigned Flags = 0>
-class CuHostVectorImpl : public physx::shdfnd::Array<T, typename CuHostAllocator<T, Flags>>
+class CuHostVectorImpl : public physx::shdfnd::Array<T, CuHostAllocator<T, Flags>>
 {
-	typedef physx::shdfnd::Array<T, typename CuHostAllocator<T, Flags>> Super;
-	typedef typename CuHostAllocator<T, Flags> Alloc;
+	typedef physx::shdfnd::Array<T, CuHostAllocator<T, Flags>> Super;
+	typedef CuHostAllocator<T, Flags> Alloc;
 public:
 	explicit CuHostVectorImpl(const physx::PxEMPTY v):Super(v){}
 	PX_INLINE explicit CuHostVectorImpl(const Alloc& alloc = Alloc()):Super(alloc){}
@@ -142,10 +144,10 @@ public:
 
 	PX_INLINE explicit CuHostVectorImpl(const T* first, const T* last, const Alloc& alloc = Alloc()):Super(first,last,alloc){}
 
-	void swap(physx::shdfnd::Array<T, typename CuHostAllocator<T, Flags>>& other)
+	void swap(CuHostVectorImpl<T, Flags>& other)
 	{
-		PX_ASSERT(mContext == other.mContext);
-		physx::shdfnd::swap(mDevicePtr, other.mDevicePtr);
+		NV_CLOTH_ASSERT(this->mContext == other.mContext);
+		physx::shdfnd::swap(this->mDevicePtr, other.mDevicePtr);
 		Super::swap(other);
 	}
 };
diff --git a/NvCloth/src/cuda/CuSolver.cpp b/NvCloth/src/cuda/CuSolver.cpp
index f0e328f..7ef1d32 100644
--- a/NvCloth/src/cuda/CuSolver.cpp
+++ b/NvCloth/src/cuda/CuSolver.cpp
@@ -302,6 +302,28 @@ cloth::CuSolver::~CuSolver()
 	mFactory.mSolverCount--;
 }
 
+void cloth::CuSolver::addClothAppend(Cloth* cloth)
+{
+	CuCloth& cuCloth = *static_cast<CuCloth*>(cloth);
+
+	NV_CLOTH_ASSERT(mCloths.find(&cuCloth) == mCloths.end());
+
+	mCloths.pushBack(&cuCloth);
+	// trigger update of mClothData array
+	cuCloth.notifyChanged();
+}
+
+void cloth::CuSolver::addClothUpdateData()
+{
+	CuContextLock contextLock(mFactory);
+
+	// resize containers and update kernel data
+	mClothDataHostCopy.resize(mCloths.size());
+	mClothData.resize(mCloths.size());
+	mFrameData.resize(mCloths.size());
+	updateKernelData();
+}
+
 void cloth::CuSolver::updateKernelData()
 {
 	mKernelDataHost.mClothIndex = mClothIndex.get();
@@ -326,24 +348,17 @@ struct ClothSimCostGreater
 
 void cloth::CuSolver::addCloth(Cloth* cloth)
 {
-	CuCloth& cuCloth = *static_cast<CuCloth*>(cloth);
-
-	NV_CLOTH_ASSERT(mCloths.find(&cuCloth) == mCloths.end());
-
-	mCloths.pushBack(&cuCloth);
-	// trigger update of mClothData array
-	cuCloth.notifyChanged();
-
-	// sort cloth instances by size
-	shdfnd::sort(mCloths.begin(), mCloths.size(), ClothSimCostGreater(), NonTrackingAllocator());
-
-	CuContextLock contextLock(mFactory);
+	addClothAppend(cloth);
+	addClothUpdateData();
+}
 
-	// resize containers and update kernel data
-	mClothDataHostCopy.resize(mCloths.size());
-	mClothData.resize(mCloths.size());
-	mFrameData.resize(mCloths.size());
-	updateKernelData();
+void cloth::CuSolver::addCloths(Range<Cloth*> cloths)
+{
+	for (uint32_t i = 0; i < cloths.size(); ++i)
+	{
+		addClothAppend(*(cloths.begin() + i));
+	}
+	addClothUpdateData();
 }
 
 void cloth::CuSolver::removeCloth(Cloth* cloth)
@@ -401,7 +416,8 @@ void cloth::CuSolver::endSimulation()
 
 int cloth::CuSolver::getSimulationChunkCount() const
 {
-	return 1;
+	// 0 chunks when no cloth present in the solver, 1 otherwise
+	return getNumCloths() != 0;
 }
 
 void cloth::CuSolver::beginFrame()
diff --git a/NvCloth/src/cuda/CuSolver.h b/NvCloth/src/cuda/CuSolver.h
index 0406e00..b4c6d6b 100644
--- a/NvCloth/src/cuda/CuSolver.h
+++ b/NvCloth/src/cuda/CuSolver.h
@@ -58,6 +58,7 @@ public:
 	~CuSolver();
 
 	virtual void addCloth(Cloth*) override;
+	virtual void addCloths(Range<Cloth*> cloths) override;
 	virtual void removeCloth(Cloth*) override;
 	virtual int getNumCloths() const override;
 	virtual Cloth * const * getClothList() const override;
@@ -103,6 +104,10 @@ public:
 	}
 
   private:
+	// add cloth helper functions
+	void addClothAppend(Cloth* cloth);
+	void addClothUpdateData();
+
 	void updateKernelData(); // context needs to be acquired
 
 	// simulate helper functions