RingBuffer.inl

/*
 * Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

#ifndef __MULTITHREAD_RINGBUFFER_INLINE_H_
#define __MULTITHREAD_RINGBUFFER_INLINE_H_


#include "cudaMappedMemory.h"
#include "logging.h"


// constructor
RingBuffer::RingBuffer( uint32_t flags )
{
	mFlags = flags;
	mBuffers = NULL;
	mBufferSize = 0;
	mNumBuffers = 0;
	mReadOnce = false;
	mLatestRead = 0;
	mLatestWrite = 0;
}


// destructor
RingBuffer::~RingBuffer()
{
	Free();
	
	if( mBuffers != NULL )
	{
		free(mBuffers);
		mBuffers = NULL;
	}
}


// Alloc
inline bool RingBuffer::Alloc( uint32_t numBuffers, size_t size, uint32_t flags )
{
	if( numBuffers == mNumBuffers && size <= mBufferSize && (flags & ZeroCopy) == (mFlags & ZeroCopy) )
		return true;
	
	Free();
	
	if( mBuffers != NULL && mNumBuffers != numBuffers )
	{
		free(mBuffers);
		mBuffers = NULL;
	}
	
	if( mBuffers == NULL )
	{
		const size_t bufferListSize = numBuffers * sizeof(void*);
		mBuffers = (void**)malloc(bufferListSize);
		memset(mBuffers, 0, bufferListSize);
	}
	
	for( uint32_t n=0; n < numBuffers; n++ )
	{
		if( flags & ZeroCopy )
		{
			if( !cudaAllocMapped(&mBuffers[n], size) )
			{
				LogError(LOG_CUDA "RingBuffer -- failed to allocate zero-copy buffer of %zu bytes\n", size);
				return false;
			}
		}
		else
		{
			mBuffers[n] = malloc(size);
			
			if( CUDA_FAILED(cudaMalloc(&mBuffers[n], size)) )
			{
				LogError(LOG_CUDA "RingBuffer -- failed to allocate CUDA buffer of %zu bytes\n", size);
				return false;
			}
		}
	}
		
	LogVerbose(LOG_CUDA "allocated %u ring buffers (%zu bytes each, %zu bytes total)\n", numBuffers, size, size * numBuffers);
	
	mNumBuffers = numBuffers;
	mBufferSize = size;
	mFlags     |= flags;
	
	return true;
}


// Free
inline void RingBuffer::Free()
{
	if( !mBuffers || mNumBuffers == 0 )
		return;
	
	for( uint32_t n=0; n < mNumBuffers; n++ )
	{
		if( mFlags & ZeroCopy )
			CUDA(cudaFreeHost(mBuffers[n]));
		else
			CUDA(cudaFree(mBuffers[n]));
		
		mBuffers[n] = NULL;
	}
}


// Peek
inline void* RingBuffer::Peek( uint32_t flags )
{
	flags |= mFlags;

	if( !mBuffers || mNumBuffers == 0 )
	{
		LogError(LOG_CUDA "RingBuffer::Peek() -- error, must call RingBuffer::Alloc() first\n");
		return NULL;
	}

	if( flags & Threaded )
		mMutex.Lock();

	int bufferIndex = -1;

	if( flags & Write )
		bufferIndex = (mLatestWrite + 1) % mNumBuffers;
	else if( flags & ReadLatest )
		bufferIndex = mLatestWrite;
	else if( flags & Read )
		bufferIndex = mLatestRead;
	
	if( flags & Threaded )
		mMutex.Unlock();

	if( bufferIndex < 0 )
	{
		LogError(LOG_CUDA "RingBuffer::Peek() -- error, invalid flags (must be Write or Read flags)\n");
		return NULL;
	}

	return mBuffers[bufferIndex];
}


// Next
inline void* RingBuffer::Next( uint32_t flags )
{
	flags |= mFlags;

	if( !mBuffers || mNumBuffers == 0 )
	{
		LogError(LOG_CUDA "RingBuffer::Next() -- error, must call RingBuffer::Alloc() first\n");
		return NULL;
	}

	if( flags & Threaded )
		mMutex.Lock();

	int bufferIndex = -1;

	if( flags & Write )
	{
		mLatestWrite = (mLatestWrite + 1) % mNumBuffers;
		bufferIndex  = mLatestWrite;
		mReadOnce    = false;
	}
	else if( (flags & ReadOnce) && mReadOnce )
	{
		if( flags & Threaded )
			mMutex.Unlock();

		return NULL;
	}
	else if( flags & ReadLatest )
	{
		mLatestRead = mLatestWrite;
		bufferIndex = mLatestWrite;
		mReadOnce   = true;
	}
	else if( flags & Read )
	{
		mLatestRead = (mLatestRead + 1) % mNumBuffers;
		bufferIndex = mLatestRead;
		mReadOnce   = true;
	}
	
	if( flags & Threaded )
		mMutex.Unlock();

	if( bufferIndex < 0 )
	{
		LogError(LOG_CUDA "RingBuffer::Next() -- error, invalid flags (must be Write or Read flags)\n");
		return NULL;
	}

	return mBuffers[bufferIndex];
}


// GetFlags
inline uint32_t RingBuffer::GetFlags() const
{
	return mFlags;
}
	

// SetFlags
inline void RingBuffer::SetFlags( uint32_t flags )
{
	mFlags = flags;
}
	

// SetThreaded
inline void RingBuffer::SetThreaded( bool threaded )
{
	if( threaded )
		mFlags |= Threaded;
	else
		mFlags &= ~Threaded;
}


#endif