SoCudaBufferObject Class Reference
[CUDA device]

CUDA buffer object managment class More...

#include <Inventor/devices/SoCudaBufferObject.h>

Inheritance diagram for SoCudaBufferObject:

List of all members.

Public Member Functions
	SoCudaBufferObject ()
virtual bool	setSize (size_t size)
virtual void	memcpy (SoBufferObject *sourceBufferObject, size_t destOffset=0, size_t sourceOffset=0, size_t copySize=SO_BUFFER_SIZE_ALL)
virtual void	memcpy (SoCpuBufferObject *sourceBufferObject, size_t destOffset=0, size_t sourceOffset=0, size_t copySize=SO_BUFFER_SIZE_ALL)
virtual void	memcpy (SoCudaBufferObject *sourceBufferObject, size_t destOffset=0, size_t sourceOffset=0, size_t copySize=SO_BUFFER_SIZE_ALL)
virtual void	memcpy (SoGLBufferObject *sourceBufferObject, size_t destOffset=0, size_t sourceOffset=0, size_t copySize=SO_BUFFER_SIZE_ALL)
virtual void	memset (void *value, size_t valueSize=1, size_t offset=0, size_t count=SO_BUFFER_SIZE_ALL)
virtual void	map (SoBufferObject *targetBufferObject, AccessMode accessMode, size_t startPosition=0, size_t mappingSize=SO_BUFFER_SIZE_ALL)
virtual void	map (SoCudaBufferObject *targetBufferObject, AccessMode accessMode, size_t startPosition=0, size_t mappingSize=SO_BUFFER_SIZE_ALL)
virtual void	map (SoCpuBufferObject *targetBufferObject, AccessMode accessMode, size_t startPosition=0, size_t mappingSize=SO_BUFFER_SIZE_ALL)
virtual void	map (SoGLBufferObject *targetBufferObject, AccessMode accessMode, size_t startPosition=0, size_t mappingSize=SO_BUFFER_SIZE_ALL)
virtual void	unmap (SoBufferObject *bufferObject)
virtual void	unmap (SoCudaBufferObject *bufferObject)
virtual void	unmap (SoCpuBufferObject *bufferObject)
virtual void	unmap (SoGLBufferObject *bufferObject)
virtual void *	getLocation ()
virtual SoBufferObject *	createInstance () const
virtual void	clearInstance ()

---

Detailed Description

CUDA buffer object managment class

This class provides management functions for CUDA memory buffers.

NOTES

• The SoCudaContext used for this buffer is the one active when the constructor is called.
• Most of the functions don't actually care about the context.

See related examples:

ComputeSlice, CudaMarchingCubes, LDMMandelBulb

---

Constructor & Destructor Documentation

SoCudaBufferObject::SoCudaBufferObject

(

)

Default constructor.


The SoCudaContext used for this buffer is the one active when the constructor is called.

---

Member Function Documentation

virtual void SoCudaBufferObject::clearInstance

(

)

[virtual]

Free the memory allocated by the buffer object.

Implements SoBufferObject.

virtual SoBufferObject* SoCudaBufferObject::createInstance

(

)

const [virtual]

Creates a new buffer with the same properties as the current one.


The new instance will have the same context or device properties, but no memory is allocated.

Implements SoBufferObject.

virtual void* SoCudaBufferObject::getLocation

(

)

[virtual]

Returns a CUDA pointer for this buffer object.


Be careful, the returned pointer is only valid in a CUDA device function.

Returns:

CUDA pointer of this buffer.

virtual void SoCudaBufferObject::map	(	SoGLBufferObject *	targetBufferObject,
		AccessMode	accessMode,
		size_t	startPosition = 0,
		size_t	mappingSize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Map the buffer object to a GL buffer object.


It is defined just to speed up the call when the type of the object is known.


See the general map function for more information.

virtual void SoCudaBufferObject::map	(	SoCpuBufferObject *	targetBufferObject,
		AccessMode	accessMode,
		size_t	startPosition = 0,
		size_t	mappingSize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Map the buffer object to a CPU buffer object.


It is defined just to speed up the call when the type of the object is known.


See the general map function for more information.

Implements SoBufferObject.

virtual void SoCudaBufferObject::map	(	SoCudaBufferObject *	targetBufferObject,
		AccessMode	accessMode,
		size_t	startPosition = 0,
		size_t	mappingSize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Map the buffer object to a CUDA buffer object.


It is defined just to speed up the call when the type of the object is known.


See the general map function for more information.

Implements SoBufferObject.

virtual void SoCudaBufferObject::map	(	SoBufferObject *	targetBufferObject,
		AccessMode	accessMode,
		size_t	startPosition = 0,
		size_t	mappingSize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Map the buffer object to any buffer object.


It can be called from any context, the function binds the correct context when it's called!

If the target buffer is another CUDA buffer object, the map function copies the data according to the context. It does a copy of the source buffer in a temp buffer if the two buffers don't use the same CUDA context.

Parameters:

	targetBufferObject	The target buffer object which receives the mapping from the CUDA buffer.
	accessMode	The access mode used for the mapping.
	startPosition	Offset in source buffer to map from.
	mappingSize	Number of bytes to map, if SO_BUFFER_SIZE_ALL then the whole buffer is mapped.

virtual void SoCudaBufferObject::memcpy	(	SoGLBufferObject *	sourceBufferObject,
		size_t	destOffset = 0,
		size_t	sourceOffset = 0,
		size_t	copySize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Copy data from the specified GL buffer object into this buffer object.




This is an optimized implementation. See the general memcpy method for more information.

virtual void SoCudaBufferObject::memcpy	(	SoCudaBufferObject *	sourceBufferObject,
		size_t	destOffset = 0,
		size_t	sourceOffset = 0,
		size_t	copySize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Copy data from a CUDA buffer into this CUDA buffer.

Warning:

: Source and destination overlaping is NOT supported.

This function is a specialized function for speed-up.
See the regular memcpy function for more information

Implements SoBufferObject.

virtual void SoCudaBufferObject::memcpy	(	SoCpuBufferObject *	sourceBufferObject,
		size_t	destOffset = 0,
		size_t	sourceOffset = 0,
		size_t	copySize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Copy data from a CPU buffer into the CUDA buffer.

This function is a specialized function for speed-up.
See the regular memcpy function for more information

Implements SoBufferObject.

virtual void SoCudaBufferObject::memcpy	(	SoBufferObject *	sourceBufferObject,
		size_t	destOffset = 0,
		size_t	sourceOffset = 0,
		size_t	copySize = SO_BUFFER_SIZE_ALL
	)			[virtual]

Copy data from any SoBufferObject into the CUDA memory.

Parameters:

	sourceBufferObject	The source buffer object which contains the data to copy.
	destOffset	The starting offset in the source buffer.
	sourceOffset	The starting offset in the source buffer.
	copySize	The size in bytes of the chunk of data to copy.

Warning:

: Source and destination overlaping is NOT supported.

NOTES

• Ensure that the context which was used when the constructor was called is currently bound before calling this function.