We've been doing retry with 1ms sleep if DecoderBeginFrame()
returned E_PENDING which means application should call
DecoderBeginFrame() again because GPU is busy.
The 1ms sleep() during retry would result in usually about 15ms delay
in reality because of bad clock precision on Windows.
To improve throughput performance, this commit will enable
high precision clock only for NVIDIA platform since
DecoderBeginFrame() call on the other GPU vendors seems to
succeed without retry.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2099>
In a case of a scr different from 0, after a seek,
the src_segment.stop has been updated with the duration
not including the base_time (scr). The segment position
needs to be tested upon segment.stop + base_time (scr)
to check for an EOS.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2048>
After the VA filter creation, when changing the element's state from NULL
to READY, immediatly checks for any filter operation requested by the user.
If any, the passthrough mode is disabled early, so there's no need for a
future renegotiation.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2094>
Useful when having a service that runs a GStreamer pipeline
or application in Google Cloud to avoid storing the inputs
and outputs in the running container or service. For example
when analyzing a video from a Google Cloud Storage bucket
and extracting images or converting the video and then uploading
the results into another Google Cloud Storage bucket.
- gssrc allows to read from a file located in Google Cloud
Storage and it supports seeking.
- gssink allows to write to a file located in Google Cloud
Storage. There are 2 modes, one similar to multifilesink and
the other similar to filesink.
Example:
gst-launch-1.0 gssrc location=gs://mybucket/videos/sample.mp4 ! decodebin ! glimagesink
gst-launch-1.0 playbin uri=gs://mybucket/videos/sample.mp4
gst-launch-1.0 videotestsrc num-buffers=5 ! pngenc ! gssink object-name="img/img%05d.png" bucket-name="mybucket" next-file=buffer
gst-launch-1.0 filesrc location=sample.mp4 ! gssink object-name="videos/video.mp4" bucket-name="mybucket" next-file=none
When running locally simply set GOOGLE_APPLICATION_CREDENTIALS. But
when running in Google Cloud Run or Google Cloud Engine, just set the
"service-account-email" property on each element.
Closes#1264
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1369>
In order to always insert a PCR packet right on time, we need to
check whether one is needed when outputting any packet, not only
a packet for the PCR stream. Most of the PCR packets will remain
data-carrying packets, but as a last resort we may insert stuffing
packets on the PCR stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2073>
Change PCR / SI scheduling so that instead of checking if
the current PCR is larger than the next target time, instead
check if the PCR of the next packet would be too late, so PCR
and SI are always scheduled earlier than the target, not later.
There are still cases where PCR can be written too late though,
because we don't check before each output packet.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2073>
Prior to that, cccombiner's behaviour was essentially that of
a funnel: it strictly looked at input timestamps to associate
together video and caption buffers.
This patch instead exposes a "schedule" property, with a default
of TRUE, to control whether caption buffers should be smoothly
scheduled, in order to have exactly one per output video buffer.
This can involve rewriting input captions, for example when the
input is CDP sequence counters are rewritten, time codes are dropped
and potentially re-injected if the input video frame had a time code
meta.
Caption buffers may also get split up in order to assign captions to
the correct field when the input is interlaced.
This can also imply that the input will drift from synchronization,
when there isn't enough padding in the input stream to catch up. In
that case the element will start dropping old caption buffers once
the number of buffers in its internal queue reaches a certain limit
(configurable).
The property is exposed so that existing users of cccombiner can
revert back to the original behaviour, but should eventually be
removed, as that behaviour was simply inadequate.
This commit also disallows changing the input caption type, as
this would needlessly complicate implementation, and removes
the corresponding test.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2076>
When we transform the caps from the sink to src, or vice versa, the
"caps" passed to us may only contain parts of the features. Which
makes our vpp lose some feature in caps and get a negotiation error.
The correct way should be:
Cleaning the format and resolution of that caps, but adding all VA,
DMA features to it, making it a full feature caps. Then, clipping it
with the pad template.
fixes: #1551
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2081>
For reverse playback, we are always copying decoded
frame to downstream buffer. So the pool size can be
and need to be large enough.
In case that forward playback, however, we need to restrict
the max pool size for performance reason. Otherwise decoder
will keep copying decoded texture to downstream buffer pool
if decoding is faster than downstream throughput
performance and also there are queue element between them.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2083>
Decoder might be able to copy decoded texture to the other buffer pool
during playback depending on context. In that case, copied one
has no D3D11_BIND_DECODER bind flag.
If we used ID3D11VideoProcessor previously for decoder texture,
and incoming texture supports ID3D11VideoProcessor as well even if it has no
D3D11_BIND_DECODER flag (having D3D11_BIND_RENDER_TARGET for example),
allow zero-copying instead of using our fallback texture.
Frequent conversion tool change (between ID3D11VideoProcessor and generic shader)
might result in inconsistent image quality.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2084>
... instead of QueryInterface-ing per elements. Note that
ID3D11VideoDevice and ID3D11VideoContext objects might not be available
if device doesn't support video interface.
So GstD3D11Device object will create those objects only when requested.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
Direct3D11 objects are COM, and most COM C APIs are verbose
(C++ is a little better). So, by using C++ APIs, we can make code
shorter and more readable.
Moreover, "ComPtr" helper class (which is C++ only) can be
utilized, that is very helpful for avoiding error-prone COM refcounting
issue/leak.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2077>
