path: root/PhysX_3.4/Source/LowLevel/common/src/pipeline/PxcNpMemBlockPool.cpp

//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions
// are met:
//  * Redistributions of source code must retain the above copyright
//    notice, this list of conditions and the following disclaimer.
//  * Redistributions in binary form must reproduce the above copyright
//    notice, this list of conditions and the following disclaimer in the
//    documentation and/or other materials provided with the distribution.
//  * Neither the name of NVIDIA CORPORATION nor the names of its
//    contributors may be used to endorse or promote products derived
//    from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Copyright (c) 2008-2018 NVIDIA Corporation. All rights reserved.
// Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
// Copyright (c) 2001-2004 NovodeX AG. All rights reserved.  

#include "foundation/PxPreprocessor.h"
#include "foundation/PxMath.h"
#include "PxcNpMemBlockPool.h"
#include "PsUserAllocated.h"
#include "PsInlineArray.h"
#include "PsFoundation.h"

using namespace physx;

PxcNpMemBlockPool::PxcNpMemBlockPool(PxcScratchAllocator& allocator):
  mConstraints(PX_DEBUG_EXP("PxcNpMemBlockPool::mConstraints")),
  mExceptionalConstraints(PX_DEBUG_EXP("PxcNpMemBlockPool::mExceptionalConstraints")),
  mNpCacheActiveStream(0),
  mFrictionActiveStream(0),
  mCCDCacheActiveStream(0),
  mContactIndex(0),
  mAllocatedBlocks(0),
  mMaxBlocks(0),
  mUsedBlocks(0),
  mMaxUsedBlocks(0),
  mScratchBlockAddr(0),
  mNbScratchBlocks(0),
  mScratchAllocator(allocator),
  mPeakConstraintAllocations(0),
  mConstraintAllocations(0)  
{
}

void PxcNpMemBlockPool::init(PxU32 initialBlockCount, PxU32 maxBlocks)
{
	mMaxBlocks = maxBlocks;
	mInitialBlocks = initialBlockCount;

	PxU32 reserve = PxMax<PxU32>(initialBlockCount, 64);

	mConstraints.reserve(reserve);
	mExceptionalConstraints.reserve(16);

	mFriction[0].reserve(reserve);
	mFriction[1].reserve(reserve);
	mNpCache[0].reserve(reserve);
	mNpCache[1].reserve(reserve);
	mUnused.reserve(reserve);

	setBlockCount(initialBlockCount);
}

PxU32 PxcNpMemBlockPool::getUsedBlockCount() const
{
	return mUsedBlocks;
}

PxU32 PxcNpMemBlockPool::getMaxUsedBlockCount() const
{
	return mMaxUsedBlocks;
}

PxU32 PxcNpMemBlockPool::getPeakConstraintBlockCount() const
{
	return mPeakConstraintAllocations;
}


void PxcNpMemBlockPool::setBlockCount(PxU32 blockCount)
{
	Ps::Mutex::ScopedLock lock(mLock);
	PxU32 current = getUsedBlockCount();
	for(PxU32 i=current;i<blockCount;i++)
	{
		mUnused.pushBack(reinterpret_cast<PxcNpMemBlock *>(PX_ALLOC(PxcNpMemBlock::SIZE, "PxcNpMemBlock")));
		mAllocatedBlocks++;
	}
}

void PxcNpMemBlockPool::releaseUnusedBlocks()
{
	Ps::Mutex::ScopedLock lock(mLock);
	while(mUnused.size())
	{
		PX_FREE(mUnused.popBack());
		mAllocatedBlocks--;
	}
}


PxcNpMemBlockPool::~PxcNpMemBlockPool()
{
	// swapping twice guarantees all blocks are released from the stream pairs
	swapFrictionStreams();
	swapFrictionStreams();

	swapNpCacheStreams();
	swapNpCacheStreams();

	releaseConstraintMemory();
	releaseContacts();
	releaseContacts();

	PX_ASSERT(mUsedBlocks == 0);

	flushUnused();
}

void PxcNpMemBlockPool::acquireConstraintMemory()
{
	PxU32 size;
	void* addr = mScratchAllocator.allocAll(size);
	size = size&~(PxcNpMemBlock::SIZE-1);

	PX_ASSERT(mScratchBlocks.size()==0);
	mScratchBlockAddr = reinterpret_cast<PxcNpMemBlock*>(addr);
	mNbScratchBlocks =  size/PxcNpMemBlock::SIZE;

	mScratchBlocks.resize(mNbScratchBlocks);
	for(PxU32 i=0;i<mNbScratchBlocks;i++)
		mScratchBlocks[i] = mScratchBlockAddr+i;
}

void PxcNpMemBlockPool::releaseConstraintMemory()
{
	Ps::Mutex::ScopedLock lock(mLock);

	mPeakConstraintAllocations = mConstraintAllocations = 0;
	
	while(mConstraints.size())
	{
		PxcNpMemBlock* block = mConstraints.popBack();
		if(mScratchAllocator.isScratchAddr(block))
			mScratchBlocks.pushBack(block);
		else
		{
			mUnused.pushBack(block);
			PX_ASSERT(mUsedBlocks>0);
			mUsedBlocks--;
		}
	}

	for(PxU32 i=0;i<mExceptionalConstraints.size();i++)
		PX_FREE(mExceptionalConstraints[i]);
	mExceptionalConstraints.clear();

	PX_ASSERT(mScratchBlocks.size()==mNbScratchBlocks); // check we released them all
	mScratchBlocks.clear();

	if(mScratchBlockAddr)
	{
		mScratchAllocator.free(mScratchBlockAddr);
		mScratchBlockAddr = 0;
		mNbScratchBlocks = 0;
	}
}


PxcNpMemBlock* PxcNpMemBlockPool::acquire(PxcNpMemBlockArray& trackingArray, PxU32* allocationCount, PxU32* peakAllocationCount, bool isScratchAllocation)
{
	Ps::Mutex::ScopedLock lock(mLock);
	if(allocationCount && peakAllocationCount)
	{
		*peakAllocationCount = PxMax(*allocationCount + 1, *peakAllocationCount);
		(*allocationCount)++;
	}

	// this is a bit of hack - the logic would be better placed in acquireConstraintBlock, but then we'd have to grab the mutex
	// once there to check the scratch block array and once here if we fail - or, we'd need a larger refactor to separate out
	// locking and acquisition.

	if(isScratchAllocation && mScratchBlocks.size()>0)
	{
		PxcNpMemBlock* block = mScratchBlocks.popBack();
		trackingArray.pushBack(block);
		return block;
	}

	
	if(mUnused.size())
	{
		PxcNpMemBlock* block = mUnused.popBack();
		trackingArray.pushBack(block);
		mMaxUsedBlocks = PxMax<PxU32>(mUsedBlocks+1, mMaxUsedBlocks);
		mUsedBlocks++;
		return block;
	}	


	if(mAllocatedBlocks == mMaxBlocks)
	{
#if PX_CHECKED
		Ps::getFoundation().error(PxErrorCode::eDEBUG_WARNING, __FILE__, __LINE__, 
				"Reached maximum number of allocated blocks so 16k block allocation will fail!");
#endif
		return NULL;
	}

#if PX_CHECKED
	if(mInitialBlocks)
	{
		Ps::getFoundation().error(PxErrorCode::eDEBUG_WARNING, __FILE__, __LINE__, 
			"Number of required 16k memory blocks has exceeded the initial number of blocks. Allocator is being called. Consider increasing the number of pre-allocated 16k blocks.");
	}
#endif

	// increment here so that if we hit the limit in separate threads we won't overallocated
	mAllocatedBlocks++;
	
	PxcNpMemBlock* block = reinterpret_cast<PxcNpMemBlock*>(PX_ALLOC(sizeof(PxcNpMemBlock), "PxcNpMemBlock"));

	if(block)
	{
		trackingArray.pushBack(block);
		mMaxUsedBlocks = PxMax<PxU32>(mUsedBlocks+1, mMaxUsedBlocks);
		mUsedBlocks++;
	}
	else
		mAllocatedBlocks--;

	return block;
}

PxU8* PxcNpMemBlockPool::acquireExceptionalConstraintMemory(PxU32 size)
{
	PxU8* memory = reinterpret_cast<PxU8*>(PX_ALLOC(size, "PxcNpExceptionalMemory"));
	if(memory)
	{
		Ps::Mutex::ScopedLock lock(mLock);
		mExceptionalConstraints.pushBack(memory);
	}
	return memory;
}

void PxcNpMemBlockPool::release(PxcNpMemBlockArray& deadArray, PxU32* allocationCount)
{
	Ps::Mutex::ScopedLock lock(mLock);
	PX_ASSERT(mUsedBlocks >= deadArray.size());
	mUsedBlocks -= deadArray.size();
	if(allocationCount)
	{
		*allocationCount -= deadArray.size();
	}
	while(deadArray.size())
	{
		PxcNpMemBlock* block = deadArray.popBack();
		for(PxU32 a = 0; a < mUnused.size(); ++a)
		{
			PX_ASSERT(mUnused[a] != block);
		}
		mUnused.pushBack(block);
	}
}

void PxcNpMemBlockPool::flushUnused()
{
	while(mUnused.size())
		PX_FREE(mUnused.popBack());
}


PxcNpMemBlock* PxcNpMemBlockPool::acquireConstraintBlock()
{
	// we track the scratch blocks in the constraint block array, because the code in acquireMultipleConstraintBlocks
	// assumes that acquired blocks are listed there.

	return acquire(mConstraints);
}

PxcNpMemBlock* PxcNpMemBlockPool::acquireConstraintBlock(PxcNpMemBlockArray& memBlocks)
{
	return acquire(memBlocks, &mConstraintAllocations, &mPeakConstraintAllocations, true);
}

PxcNpMemBlock* PxcNpMemBlockPool::acquireContactBlock()
{
	return acquire(mContacts[mContactIndex], NULL, NULL, true);
}


void PxcNpMemBlockPool::releaseConstraintBlocks(PxcNpMemBlockArray& memBlocks)
{
	Ps::Mutex::ScopedLock lock(mLock);
	
	while(memBlocks.size())
	{
		PxcNpMemBlock* block = memBlocks.popBack();
		if(mScratchAllocator.isScratchAddr(block))
			mScratchBlocks.pushBack(block);
		else
		{
			mUnused.pushBack(block);
			PX_ASSERT(mUsedBlocks>0);
			mUsedBlocks--;
		}
	}
}

void PxcNpMemBlockPool::releaseContacts()
{
	//releaseConstraintBlocks(mContacts);
	release(mContacts[1-mContactIndex]);
	mContactIndex = 1-mContactIndex;
}

PxcNpMemBlock* PxcNpMemBlockPool::acquireFrictionBlock()
{
	return acquire(mFriction[mFrictionActiveStream]);
}

void PxcNpMemBlockPool::swapFrictionStreams()
{
	release(mFriction[1-mFrictionActiveStream]);
	mFrictionActiveStream = 1-mFrictionActiveStream;
}

PxcNpMemBlock* PxcNpMemBlockPool::acquireNpCacheBlock()
{
	return acquire(mNpCache[mNpCacheActiveStream]);
}

void PxcNpMemBlockPool::swapNpCacheStreams()
{
	release(mNpCache[1-mNpCacheActiveStream]);
	mNpCacheActiveStream = 1-mNpCacheActiveStream;
}