GstCuda.CudaAllocator¶

Subclasses:: GstCuda.CudaPoolAllocator

Methods¶

Inherited:: Gst.Allocator (5), Gst.Object (29), GObject.Object (37)
Structs:: GObject.ObjectClass (5)

	`alloc` (context, stream, info)
	`alloc_stream_ordered` (context, stream, info)
	`alloc_wrapped` (context, stream, info, dev_ptr, user_data, notify)
	`set_active` (active)
	`virtual_alloc` (context, stream, info, prop, granularity_flags)

Virtual Methods¶

Inherited:: Gst.Allocator (2), Gst.Object (1), GObject.Object (7)

do_set_active (active)

Properties¶

Inherited:: Gst.Object (2)

Signals¶

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

Fields¶

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

Name	Type	Access	Description
parent	`Gst.Allocator`	r

Class Details¶

class GstCuda.CudaAllocator(**kwargs)¶

Bases:: Gst.Allocator
Abstract:: No
Structure:: GstCuda.CudaAllocatorClass

A Gst.Allocator subclass for cuda memory

Added in version 1.22.

alloc(context, stream, info)¶

Parameters:

context (GstCuda.CudaContext) – a GstCuda.CudaContext
stream (GstCuda.CudaStream or None) – a GstCuda.CudaStream
info (GstVideo.VideoInfo) – a GstVideo.VideoInfo

Returns:

a newly allocated GstCuda.CudaMemory

Return type:

Gst.Memory or None

Added in version 1.22.

alloc_stream_ordered(context, stream, info)¶

Parameters:

context (GstCuda.CudaContext) – a GstCuda.CudaContext
stream (GstCuda.CudaStream) – a GstCuda.CudaStream
info (GstVideo.VideoInfo) – a GstVideo.VideoInfo

Returns:

a newly allocated GstCuda.CudaMemory

Return type:

Gst.Memory or None

Added in version 1.28.

alloc_wrapped(context, stream, info, dev_ptr, user_data, notify)¶

Parameters:

context (GstCuda.CudaContext) – a GstCuda.CudaContext
stream (GstCuda.CudaStream or None) – a GstCuda.CudaStream
info (GstVideo.VideoInfo) – a GstVideo.VideoInfo
dev_ptr (int) – a CUdeviceptr CUDA device memory
user_data (object or None) – user data
notify (GLib.DestroyNotify or None) – Called with user_data when the memory is freed

Returns:

a new Gst.Memory

Return type:

Gst.Memory

Allocates a new memory that wraps the given CUDA device memory.

info must represent actual memory layout, in other words, offset, stride and size fields of info should be matched with memory layout of dev_ptr

By default, wrapped dev_ptr will be freed at the time when Gst.Memory is freed if notify is None. Otherwise, if caller sets notify, freeing dev_ptr is callers responsibility and default GstCuda.CudaAllocator will not free it.

Added in version 1.24.

set_active(active)¶

Parameters:: active (bool) – the new active state
Returns:: True if active state of self was successfully updated.
Return type:: bool

Controls the active state of self. Default GstCuda.CudaAllocator is stateless and therefore active state is ignored, but subclass implementation (e.g., GstCuda.CudaPoolAllocator) will require explicit active state control for its internal resource management.

This method is conceptually identical to Gst.BufferPool.set_active method.

Added in version 1.24.

virtual_alloc(context, stream, info, prop, granularity_flags)¶

Parameters:

context (GstCuda.CudaContext) – a GstCuda.CudaContext
stream (GstCuda.CudaStream) – a GstCuda.CudaStream
info (GstVideo.VideoInfo) – a GstVideo.VideoInfo
prop (CudaGst.memAllocationProp) – allocation property
granularity_flags (CudaGst.memAllocationGranularity_flags) – allocation flags

Returns:

a newly allocated memory object or None if allocation is not supported

Return type:

Gst.Memory or None

Allocates new Gst.Memory object with CUDA virtual memory.

Added in version 1.24.

do_set_active(active) virtual¶

Parameters:: active (bool) – the new active state
Returns:: True if active state of allocator was successfully updated.
Return type:: bool

Controls the active state of allocator. Default GstCuda.CudaAllocator is stateless and therefore active state is ignored, but subclass implementation (e.g., GstCuda.CudaPoolAllocator) will require explicit active state control for its internal resource management.

This method is conceptually identical to Gst.BufferPool.set_active method.

Added in version 1.24.