CudaComputeController Class Reference

Compute controller for launching CUDA subdivision kernels. More...

#include <cudaComputeController.h>

Public Types
typedef CudaComputeContext	ComputeContext

Public Member Functions
	CudaComputeController ()
	Constructor. More...
	~CudaComputeController ()
	Destructor. More...
template<class VERTEX_BUFFER , class VARYING_BUFFER >
void	Compute (CudaComputeContext const *context, Far::KernelBatchVector const &batches, VERTEX_BUFFER *vertexBuffer, VARYING_BUFFER *varyingBuffer, VertexBufferDescriptor const *vertexDesc=NULL, VertexBufferDescriptor const *varyingDesc=NULL)
template<class VERTEX_BUFFER >
void	Compute (CudaComputeContext const *context, Far::KernelBatchVector const &batches, VERTEX_BUFFER *vertexBuffer)
void	Synchronize ()
	Waits until all running subdivision kernels finish. More...

Protected Member Functions
void	ApplyStencilTableKernel (Far::KernelBatch const &batch, ComputeContext const *context) const
template<class VERTEX_BUFFER , class VARYING_BUFFER >
void	bind (VERTEX_BUFFER *vertexBuffer, VARYING_BUFFER *varyingBuffer, VertexBufferDescriptor const *vertexDesc, VertexBufferDescriptor const *varyingDesc)
void	unbind ()
	Unbinds any previously bound vertex and varying data buffers. More...

Friends
class	Far::KernelBatchDispatcher

Detailed Description

Compute controller for launching CUDA subdivision kernels.

CudaComputeController is a compute controller class to launch Cuda subdivision kernels. It requires CudaVertexBufferInterface as arguments of Refine function.

Controller entities execute requests from Context instances that they share common interfaces with. Controllers are attached to discrete compute devices and share the devices resources with Context entities.

Definition at line 49 of file cudaComputeController.h.

Member Typedef Documentation

typedef CudaComputeContext ComputeContext

Definition at line 51 of file cudaComputeController.h.

Constructor & Destructor Documentation

CudaComputeController

(

)

Constructor.

~CudaComputeController

(

)

Destructor.

Member Function Documentation

void ApplyStencilTableKernel ( Far::KernelBatch const &  batch, ComputeContext const *  context  ) const

protected

void bind ( VERTEX_BUFFER *  vertexBuffer, VARYING_BUFFER *  varyingBuffer, VertexBufferDescriptor const *  vertexDesc, VertexBufferDescriptor const *  varyingDesc  )

inlineprotected

Definition at line 123 of file cudaComputeController.h.

void Compute ( CudaComputeContext const *  context, Far::KernelBatchVector const &  batches, VERTEX_BUFFER *  vertexBuffer, VARYING_BUFFER *  varyingBuffer, VertexBufferDescriptor const *  vertexDesc = NULL, VertexBufferDescriptor const *  varyingDesc = NULL  )

inline

Execute subdivision kernels and apply to given vertex buffers.

Parameters

context	The CudaContext to apply refinement operations to
batches	Vector of batches of vertices organized by operative kernel
vertexBuffer	Vertex-interpolated data buffer
vertexDesc	The descriptor of vertex elements to be refined. if it's null, all primvars in the vertex buffer will be refined.
varyingBuffer	Vertex-interpolated data buffer
varyingDesc	The descriptor of varying elements to be refined. if it's null, all primvars in the vertex buffer will be refined.

Definition at line 79 of file cudaComputeController.h.

void Compute ( CudaComputeContext const *  context, Far::KernelBatchVector const &  batches, VERTEX_BUFFER *  vertexBuffer  )

inline

Execute subdivision kernels and apply to given vertex buffers.

Parameters

context	The CudaContext to apply refinement operations to
batches	Vector of batches of vertices organized by operative kernel
vertexBuffer	Vertex-interpolated data buffer

Definition at line 105 of file cudaComputeController.h.

void Synchronize

(

)

Waits until all running subdivision kernels finish.

void unbind ( )

inlineprotected

Unbinds any previously bound vertex and varying data buffers.

Definition at line 153 of file cudaComputeController.h.

Friends And Related Function Documentation

friend class Far::KernelBatchDispatcher

friend

Definition at line 117 of file cudaComputeController.h.

---

The documentation for this class was generated from the following file:

• opensubdiv/osd/cudaComputeController.h