Don't create temporary GstBuffers for all decoder units, even if they
are lightweight "sub-buffers", since it is not really necessary to keep
the buffer data around.
Implement GstVaapiDecoder.start_frame() and end_frame() semantics so
that to create new VA context earlier and submit VA pictures to the
HW for decoding as soon as possible. i.e. don't wait for the next
frame to start decoding the previous one.
Use GstVaapiDpb interface instead of maintaining our own prev and next
picture pointers. While doing so, try to derive a sensible POC value.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Avoid usage of goto. Simplify decode_step() process to first accumulate all
pending buffers into the GstAdapter, and then parse and decode units from
that input adapter. Stop the process once a frame is fully decoded or an
error occurred.
Make sure we always have a free surface left to use for decoding the
current frame. This means that decode_step() has to return once a frame
gets decoded. If the current adapter contains more buffers with valid
frames, they will get parsed and decoded on subsequent iterations.
Keep only one DPB interface and rename gst_vaapi_dpb2_get_references()
to gst_vaapi_dpb_get_neighbours() so that to retrieve pictures in DPB
around the specified picture POC.
Move GstVaapiDpbMpeg2 API to a more generic version that could also be
useful to other decoders that require 2 reference pictures, e.g. VC-1.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Fix support for global-alpha subpictures. The previous changes brought
the ability to check for GstVideoOverlayRectangle changes by comparing
the underlying pixel buffer pointers. If sequence number and pixel data
did not change, then this is an indication that only the global-alpha
value changed. Now, try to update the underlying VA subpicture global-alpha
value.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Don't re-upload VA subpicture if only the render rectangle changed.
Rather deassociate the subpicture and re-associate it with the new
render rectangle.
A GstVideoOverlayRectangle is created whenever the underlying pixels data
change. However, when global-alpha is supported, it is possible to re-use
the same GstVideoOverlayRectangle but with a change to the global-alpha
value. This process causes a change of sequence number, so we can no longer
check for that.
Still, if sequence numbers did not change, then there was no change in
global-alpha either. So, we need a way to compare the underlying GstBuffer
pointers. There is no API to retrieve the original pixels buffer from
a GstVideoOverlayRectangle. So, we use the following heuristics:
1. Use gst_video_overlay_rectangle_get_pixels_unscaled_argb() with the same
format flags from which the GstVideoOverlayRectangle was created. This
will work if there was no prior consumer of the GstVideoOverlayRectangle
with alternate (non-"native") format flags.
2. In overlay_rectangle_has_changed_pixels(), we have to use the same
gst_video_overlay_rectangle_get_pixels_unscaled_argb() function but
with flags that match the subpicture. This is needed to cope with
platforms that don't support global-alpha in HW, so the gst-video
layer takes care of that and fixes this up with a possibly new
GstBuffer, and hence pixels data (or) in-place by caching the current
global-alpha value applied. So we have to determine the rectangle
was previously used, based on what previous flags were used to
retrieve the ARGB pixels buffer.
We previously assumed that an overlay composition changed if the number
of overlay rectangles in there actually changed, or that the rectangle
was updated, and thus its seqnum was also updated.
Now, we can cope with cases where the GstVideoOverlayComposition grew
by one or a few more overlay rectangles, and the initial overlay rectangles
are kept as is.
Create the GPtrArray once in the _init() function and destroy it only
in the _finalize() function. Then use overlay_clear() to remove all
subpicture associations for intermediate updates, don't recreate the
GPtrArray.
Make GstVaapiOverlayRectangle a reference counted object. Also make
sure that overlay_rectangle_new() actually creates and associates the
VA subpicture.
Handle global-alpha from GstVideoOverlayComposition API. Likewise,
the same code path could also work for premultiplied-alpha but this
was not tested.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Add the necessary helpers in GstVaapiDisplay to determine whether subpictures
with global alpha are supported or not. Also add accessors in GstVaapiSubpicture
to address this feature.
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Add premultiplied-alpha and global-alpha feature flags, along with converters
between VA-API and gstreamer-vaapi definitions. Another round of helpers is
also necessary for GstVideoOverlayComposition API.
Use GPOINTER_TO_SIZE() instead of GPOINTER_TO_UINT() while manipulating
pointers. The latter is meant to be 32-bit only, not uintptr_t like size.
Only a gsize can hold all bits of a pointer.
Thanks to Ouping Zhang for spotting this error.
Heuristic: if the second start-code is available, check whether that
one marks the start of a new frame because e.g. this is a sequence
or picture header. This doesn't save much, since we already cache the
results.
Accelerate scan for start codes by skipping up to 3 bytes per iteration.
A start code prefix is defined by the following bytes: 00 00 01. Thus,
for any group of 3 bytes (xx yy zz), we have the following possible cases:
1. If zz != 1, this cannot be a start code, then skip 3 bytes;
2. If yy != 0, this cannot be a start code, then skip 2 bytes;
3. If xx != 0 or zz != 1, this cannot be a start code, then skip 1 byte;
4. xx == 00, yy == 00, zz == 1, we have match!
This algorithm requires to peek bytes from the adapter. This increases the
amount of bytes copied to a temporary buffer, but this process is much faster
than scanning for all the bytes and using shift/masks. So, overall, this is
a win.
Move parsing back to decoding step, but keep functions separate for now.
This is needed for future optimizations that may introduce some meta data
for parsed info attached to codec frames.
Optimize pre-allocation of decoder units, thus avoiding un-necessary
memory reallocations. The heuristic used is that we could have around
one slice unit per macroblock line.
Use a GArray to hold decoder units in a frame, instead of a single-linked
list. This makes 'append' calls faster, but not that much. At least, this
makes things clearer.
Allocate decoder unit earlier in the main parse() function and don't
delegate this task to derived classes. The ultimate purpose is to get
rid of dynamic allocation of decoder units.
The SPS, PPS and slice headers are not fully zero-initialized in the
codecparsers/ library. Rather, the standard upstream behaviour is to
initialize only certain syntax elements with some inferred values if
they are not present in the bitstream.
At the gstreamer-vaapi decoder level, we need to further initialize
certain syntax elements with some sensible default values so that to
not complicate VA drivers that just pass those verbatim to the HW,
and also avoid an memset() of the whole decoder unit.
Signed-off-by: Sreerenj Balachandran <sreerenj.balachandran@intel.com>
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Make GstVaapiVideoBuffer a simple wrapper for video meta. This buffer is
no longer necessary but for compatibility with GStreamer 0.10 APIs or users
expecting a GstSurfaceBuffer like Clutter.
Use PTS value computed by the decoder, which could also be derived from
the GstVideoCodecFrame PTS. This makes it possible to fix up the PTS if
the original one was miscomputed or only represented a DTS instead.
