Instead of creating new decoder instance per new sequence,
re-use configured decoder instance via cuvidReconfigureDecoder()
API. It will make output surface reusable without re-allocation.
Also, in order for application to be able to reserve higher resolution
output surface, "init-max-width" and "init-max-height" properties are
added to each decoder.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3884>
Call input resource map functions (i.e., nvEncRegisterResource,
nvEncUnregisterResource, nvEncMapInputResource, and
nvEncUnmapInputResource) only once and reuse the mapped resources,
instead of per input frame map/unmap
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3884>
Wrap mapped decoder output surface using GstCudaMemory and
output without any copy operation. Also, for application to be able to
control the number of zero-copyable output surfaces,
"num-output-surfaces" property is added.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3884>
It is really difficult for people to figure out why nvcodec has
0 features. Even the debug log is cryptic. Also make sure the errors
go to the ERROR log level, which is more likely to be enabled by
default.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3776>
NVDEC launches CUDA kernel function (ConvertNV12BLtoNV12 or so)
when CuvidMapVideoFrame() is called. Which seems to be
NVDEC's internal post-processing kernel function, maybe
to convert tiled YUV to linear YUV format or something similar.
A problem if we don't pass CUDA stream to the CuvidMapVideoFrame()
call is that the NVDEC's internel kernel function will use default CUDA stream.
Then lots of the other CUDA API calls will be blocked/serialized.
To avoid the unnecessary blocking, we should pass our own
CUDA stream object to the CuvidMapVideoFrame() call
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3605>
Rewriting GstCudaConverter object, since the old implementation was not
well organized and it's hard to add new features.
Moreover, the conversion operations were not very optimized.
Major change of this implementation:
* Remove redundant intermediate conversion operations such as
any RGB -> ARGB(64) conversion or any YUV -> Y444 (or 16bits Y444).
That's not required most of cases. The only required case is
converting 24bits (such as RGB/BGR) packed format to 32bits format
because CUDA texture object does not support sampling 24bits format
* Use normalized sample fetching (i.e., [0, 1] range float value)
and also normalized coordinates system for CUDA texture.
It's consistent with the other graphics APIs such as Direct3D
and OpenGL, that makes sampling operations much easier.
* Support a kind of viewport and adopt math for colorspace conversion
from GstD3D11 implementation
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>
GstCudaConverter object can do colorspace conversion and scale at once.
Adding new element "cudaconvertscale" to do that, this can
save unnecessary GPU operation if colorspace conversion and
rescale is required for given input stream format.
Most of codes are taken from d3d11convert element
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>
Handle d3d11 device context in set_context() method with
additional device compatibility check so that only NVIDIA GPU
associated d3d11 device can be configured in the element.
And clear old d3d11 device per set_info() for d3d11 device to be
updated as well.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3018>
Adding nvautogpu{h264,h265}enc class which will accept upstream logical
GPU device object (GstCudaContext or GstD3D11Device) instead of
using pre-assigned GPU instance.
If upstream logical GPU device object is not NVENC compatible
(e.g., D3D11 device of non-NVIDIA GPU) or it's system memory,
then user specified "cuda-device-id" or "adapter-luid" property
will be used for GPU device selection.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2666>
GstCudaMemory supports CPU access via CUDA pinned host memory already
and it would show faster memory transfer performance between
GPU and CPU than copying from/to normal system memory.
If downstream supports video meta, we can passthrough CUDA memory.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2690>
