User can get the required buffer size by using buffer pool config.
Since d3d11 implementation is a candidate for public library in the future,
we need to hide everything from header as much as possible.
Note that the total size of allocated d3d11 texture memory by GPU is not
controllable factor. It depends on hardware specific alignment/padding
requirement. So, GstD3D11 implementation updates actual buffer size
by allocating D3D11 texture, since there's no way to get CPU accessible
memory size without allocating real D3D11 texture.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2482>
Some GPUs support BGRA format and it will be converted to subsampled
YUV format by GPU internally. Disable this implicit conversion
since the conversion parameters such as input/output colorimetry
are not exposed nor it's written in bitstream (e.g., VUI).
We prefer explicit conversion via our conversion elements.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2410>
When B-frame is enabled, encoder seems to adjust PTS of encoded sample
by using frame duration.
For instance, one observed timestamp pattern by using B-frame enabled
and 30fps stream is:
* Frame-1: MF pts 0:00.033333300 MF dts 0:00.000000000
* Frame-2: MF pts 0:00.133333300 MF dts 0:00.033333300
* Frame-3: MF pts 0:00.066666600 MF dts 0:00.066666600
* Frame-4: MF pts 0:00.099999900 MF dts 0:00.100000000
We can notice that the amount of PTS shift is frame duration and
Frame-4 exhibits PTS < DTS.
To compensate shifted timestamp, we should
calculate the timestamp offset and re-calculate DTS correspondingly.
Otherwise, total timeline of output stream will be shifted, and that
can cause time sync issue.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2354>
Initial support for d3d11 texture so that encoder can copy
upstream d3d11 texture into encoder's own texture pool without
downloading memory.
This implementation requires MFTEnum2() API for creating
MFT (Media Foundation Transform) object for specific GPU but
the API is Windows 10 desktop only. So UWP is not target
of this change.
See also https://docs.microsoft.com/en-us/windows/win32/api/mfapi/nf-mfapi-mftenum2
Note that, for MF plugin to be able to support old OS versions
without breakage, this commit will load MFTEnum2() symbol
by using g_module_open()
Summary of required system environment:
- Needs Windows 10 (probably at least RS 1 update)
- GPU should support ExtendedNV12SharedTextureSupported feature
- Desktop application only (UWP is not supported yet)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1903>
The behavior for zero AVEncMPVGOPSize value would be
varying depending on GPU vendor implementation and some
GPU will produce keyframe only once at the beginning of encoding.
That's unlikely expected result for users.
To make this property behave consistently among various GPUs,
this commit will change default value of "gop-size" property to -1
which means "auto". When "gop-size" is unspecified, then
mfvideoenc will calculate GOP size based on framerate
like that of our x264enc implementation.
See also
https://docs.microsoft.com/en-us/windows/win32/directshow/avencmpvgopsize-property
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1911>
Unlike software MFT (Media Foundation Transform) which is synchronous
in terms of processing input and output data, hardware MFT works
in asynchronous mode. output data might not be available right after
we pushed one input data into MFT.
Note that async MFT will fire two events, one is "METransformNeedInput"
which happens when MFT can accept more input data,
and the other is "METransformHaveOutput", that's for signaling
there's pending data which can be outputted immediately.
To listen the events, we can wait synchronously via
IMFMediaEventGenerator::GetEvent() or make use of IMFAsyncCallback
object which is asynchronous way and the event will be notified
from Media Foundation's internal worker queue thread.
To handle such asynchronous operation, previous working flow was
as follows (IMFMediaEventGenerator::GetEvent() was used for now)
- Check if there is pending output data and push the data toward downstream.
- Pulling events (from streaming thread) until there's at least
one pending "METransformNeedInput" event
- Then, push one data into MFT from streaming thread
- Check if there is pending "METransformHaveOutput" again.
If there is, push new output data to downstream
(unlikely there is pending output data at this moment)
Above flow was processed from upstream streaming thread. That means
even if there's available output data, it could be outputted later
when the next buffer is pushed from upstream streaming thread.
It would introduce at least one frame latency in case of live stream.
To reduce such latency, this commit modifies the flow to be fully
asynchronous like hardware MFT was designed and to be able to
output encoded data whenever it's available. More specifically,
IMFAsyncCallback object will be used for handling
"METransformNeedInput" and "METransformHaveOutput" events from
Media Foundation's internal thread, and new output data will be
also outputted from the Media Foundation's thread.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1520>
Add custom IMFMediaBuffer and IMF2DBuffer implementation in order to
keep track of lifecycle of Media Foundation memory object.
By this new implementation, we can pass raw memory of upstream buffer
to Media Foundation without copy.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1518>
