GstCuda.CudaContext¶

Subclasses:: None

Methods¶

Inherited:: Gst.Object (27), GObject.Object (37)
Structs:: GObject.ObjectClass (5)

class	`new` (device_id)
class	`new_wrapped` (handler, device)
class	`pop` (cuda_ctx)
	`can_access_peer` (peer)
	`get_handle` ()
	`get_texture_alignment` ()
	`push` ()

Virtual Methods¶

Inherited:: Gst.Object (1), GObject.Object (7)

Properties¶

Inherited:: Gst.Object (2)

Name	Type	Flags	Short Description
`cuda-device-id`	`int`	r/w/co	Set the GPU device to use for operations
`os-handle`	`bool`	r	Whether OS specific handle is supported via virtual memory
`virtual-memory`	`bool`	r	Whether virtual memory management is supporte or not

Signals¶

Inherited:: Gst.Object (1), GObject.Object (1)

Fields¶

Inherited:: Gst.Object (1), GObject.Object (1)

Name	Type	Access	Description
object	`Gst.Object`	r

Class Details¶

class GstCuda.CudaContext(**kwargs)¶

Bases:: Gst.Object
Abstract:: No
Structure:: GstCuda.CudaContextClass

New in version 1.22.

classmethod new(device_id)¶

Parameters:: device_id (int) – device-id for creating GstCuda.CudaContext
Returns:: a new GstCuda.CudaContext or None on failure
Return type:: GstCuda.CudaContext or None

Create GstCuda.CudaContext with given device_id

New in version 1.22.

classmethod new_wrapped(handler, device)¶

Parameters:

handler (object) – A CUcontext to wrap
device (int) – A CUDevice to wrap

Returns:

A newly created GstCuda.CudaContext

Return type:

GstCuda.CudaContext or None

Note: The caller is responsible for ensuring that the CUcontext and CUdevice represented by handle and device stay alive while the returned GstCuda.CudaContext is active.

New in version 1.22.

classmethod pop(cuda_ctx)¶

Parameters:: cuda_ctx (object) –
Returns:: True if ctx was pushed without error.
Return type:: bool

Pops the current CUDA context from CPU thread

New in version 1.22.

can_access_peer(peer)¶

Parameters:: peer (GstCuda.CudaContext) – a GstCuda.CudaContext
Returns:: True if self can access peer directly
Return type:: bool

Query whether self can access any memory which belongs to peer directly.

New in version 1.22.

get_handle()¶

Returns:: the CUcontext of self
Return type:: object or None

Get CUDA device context. Caller must not modify and/or destroy returned device context.

New in version 1.22.

get_texture_alignment()¶

Returns:: the CUcontext of self
Return type:: int

Get required texture alignment by device

New in version 1.22.

push()¶

Returns:: True if self was pushed without error.
Return type:: bool

Pushes the given self onto the CPU thread’s stack of current contexts. The specified context becomes the CPU thread’s current context, so all CUDA functions that operate on the current context are affected.

New in version 1.22.

Property Details¶

GstCuda.CudaContext.props.cuda_device_id¶

Name:: cuda-device-id
Type:: int
Default Value:: 0
Flags:: READABLE, WRITABLE, CONSTRUCT_ONLY

Set the GPU device to use for operations

GstCuda.CudaContext.props.os_handle¶

Name:: os-handle
Type:: bool
Default Value:: False
Flags:: READABLE

OS handle supportability in virtual memory management

New in version 1.24.

GstCuda.CudaContext.props.virtual_memory¶

Name:: virtual-memory
Type:: bool
Default Value:: False
Flags:: READABLE

Virtual memory management supportability

New in version 1.24.