Delaying the pool re-negotiation untill we push all decoded (and queued)
frames downstream. Otherwise for the multi-resolution videos, the
GstVideoVideoMemory will be having wrong resolution and which leads
to nasty behaviours, especially when using software renderers.
sample media file: RAP_B_Bossen_1.bin
case explained:
The first SPS Nal will report resoultion of 448x256 and having crop rectangles to
get the final resoultion 416x240.
Starting from 25 th frame, the resolution will change to 416x240. But parser
elements won't report this since the effective croped resolution is same in
both cases. Here the core libgstvaapi will detect this through it's internal
parsing and do all context/pool destory/reset stuffs. Also it will notify this
change to plugins in advance. But if the plugin try to do re-negotiaion of pool
immediately, this will not sync with the resolution of already decoded and queued
frames and which will lead to failure in gst_video_frame_map() in downstream(if we use the
software renderer). So we have to delay the pool renegotiation in vaapidecode,
untill we push all decoded frames downstream.
https://bugzilla.gnome.org/show_bug.cgi?id=753914
Because the decoder uses the thread from handle_frame() to decode a frame,
the src pad task creates an unsolveable AB-BA deadlock between
handle_frame() waiting for a free surface and decode_loop() pushing
decoded frames out.
Instead, have handle_frame() take responsibility for pushing surfaces,
and remove the deadlock completely. If you need a separate thread
downstream, you can insert a queue between vaapidecode and its downstream
to get one.
Another justification for the single thread implementation is,
there are two many point of locking in gstreamer-vaapi's current
implementation which can lead to deadlocks.
https://bugzilla.gnome.org/show_bug.cgi?id=742605
Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Signed-off-by: Víctor Manuel Jáquez Leal <victorx.jaquez@intel.com>
Signed-off-by: Sreerenj Balachandran <sreerenj.balachandran@intel.com>
Currently the src caps are set immediately after the sink caps are set, but in
that moment the pipeline might not fully constructed and the video sink has
not negotiated its supported caps and features. As a consequence, in many cases
of playback, the least optimized caps feature is forced. This is partially the
responsible of bug #744039.
Also, vaapidecode doesn't attend the reconfigure events from downstream,
which is a problem too, since the video sink can be changed with different
caps features.
This patch delays the src caps, setting them until the first frame arrives to
the decoder, assuming until that very moment the whole pipeline is already
negotiated. Particularly, it checks if the src pad needs to be reconfigured,
as a consequence of a reconfiguration event from downstream.
A key part of this patch is the new GstVaapiCapsFeature
GST_VAAPI_CAPS_FEATURE_NOT_NEGOTIATED, which is returned when the src pad
doesn't have a peer yet. Also, for a better report of the caps allowed
through the src pad and its peer, this patch uses gst_pad_get_allowed_caps()
instead of gst_pad_peer_query_caps() when looking for the preferred feature.
v3: move the input_state unref to close(), since videodecoder resets at
some events such as navigation.
v4: a) the state_changed() callback replaces the input_state if the media
changed, so this case is also handled.
b) since the parameter ref_state in gst_vaapidecode_update_src_caps() is
always the input_state, the parameter were removed.
c) there were a lot of repeated code handling the input_state, so I
refactored it with the function gst_vaapi_decode_input_state_replace().
https://bugzilla.gnome.org/show_bug.cgi?id=744618
Signed-off-by: Víctor Manuel Jáquez Leal <victorx.jaquez@intel.com>
Signed-off-by: Sreerenj Balachandran <sreerenj.balachandran@intel.com>
Introduce a new GstVaapiPluginBase object that will contain all common
data structures and perform all common tasks. First step is to have a
single place to hold VA displays.
While we are at it, also make sure to store and subsequently release
the appropriate debug category for the subclasses.
Review all interactions between the main video decoder stream thread
and the decode task to derive a correct sequence of operations for
decoding. Also avoid extra atomic operations that become implicit under
the GstVideoDecoder stream lock.
vaapidecode used to wait up to one second past the expected time of
presentation for the last decoded frame. This is not realistic in
practice when it comes to video pause/resume. Changed behaviour to
unconditionnally wait for a free VA surface prior to continuing the
decoding. The decode task will continue pushing the output frames to
the downstream element while also reporting errors at the same time
to the main thread.
https://bugzilla.gnome.org/show_bug.cgi?id=707108
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
If the allocation meta GST_VIDEO_GL_TEXTURE_UPLOAD_META_API_TYPE is
requested, and more specifically under a GLX configuration, then add
the GstVideoGLTextureUploadMeta to the output buffer.
https://bugzilla.gnome.org/show_bug.cgi?id=703236
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
If there is no frame delimiter at the end of the stream, e.g. no
end-of-stream or end-of-sequence marker, and that the current frame
was fully parsed correctly, then assume that last frame is complete
and submit it to the decoder.
https://bugzilla.gnome.org/show_bug.cgi?id=705123
Signed-off-by: Guangxin.Xu <Guangxin.Xu@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Don't return static caps that don't mean anything for the underlying codecs
that are actually supported for decoding. i.e. always allocate a VA display
and retrieve the exact set of HW decoders available. That VA display may be
re-used later on during negotiation through GstVideoContext "prepare-context".
This fixes fallback to SW decoding if no HW decoder is available.
Make vaapidecode derive from the standard GstVideoDecoder base element
class. This simplifies the code to the strict minimum for the decoder
element and makes it easier to port to GStreamer 1.x API.
Signed-off-by: Sreerenj Balachandran <sreerenj.balachandran@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
The use of heap allocated GMutex/GCond is deprecated. Instead place them
inside the structure they are locking.
These changes switch to use g_mutex_init/g_cond_init rather than the heap
allocation functions.
Because we cannot test for a NULL pointer for the GMutex/GCond we must
initialise inside the GObject _init function and clear inside the _finalize
which is guaranteed to only be called once and after the object is no longer
in use.
Previously, vaapidecode would wait up to one second until a free surface
is available, or it aborts decoding. Now, vaapidecode waits until the
last decoded surface was to be presented, plus one second. Besides, end
times are now expressed relative to the monotonic clock.
GStreamer codecparsers-based decoders are the only supported decoders now.
Though, FFmpeg decoders are still available in gstreamer-vaapi 0.3.x series.
Rationale: playbin2 links all elements at run-time. Once vaapidecode
is created and a NEWSEGMENT event arrives, downstream element may not
be ready yet. So, delay this event until next element is chained in,
otherwise basesink could output "Received buffer without a new-segment.
Assuming timestamps start from 0".
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Otherwise, the decoder would always create its own X display instead
of probing it from the downstream element, which is not reliable.
e.g. DISPLAY is not :0 or when running on Wayland.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
