5 years agovideobuffer: wrap video meta into a surface buffer.
Gwenole Beauchesne [Sat, 5 Jan 2013 16:37:13 +0000 (17:37 +0100)]
videobuffer: wrap video meta into a surface buffer.

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.

5 years agovideobuffer: add video meta information.
Gwenole Beauchesne [Sat, 5 Jan 2013 07:31:24 +0000 (08:31 +0100)]
videobuffer: add video meta information.

Add new GstVaapiVideoMeta object that holds all information needed to
convey gst-vaapi specific data as a GstBuffer.

5 years agovaapidecode: fix calculation of the time-out value.
Gwenole Beauchesne [Thu, 3 Jan 2013 12:10:33 +0000 (13:10 +0100)]
vaapidecode: fix calculation of the time-out value.

Fix calculation of the time-out value for cases where no VA surface is
available for decoding. In this case, we need to wait until downstream
sink consumed at least one surface. The time-out was miscalculated as
it was always set to <current-time> + one second, which is not suitable
for streams with larger gaps.

5 years agodecoder: always use the calculated presentation timestamp.
Gwenole Beauchesne [Thu, 3 Jan 2013 12:05:47 +0000 (13:05 +0100)]
decoder: always use the calculated presentation timestamp.

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.

5 years agoh264: don't create sub-buffer for slice data.
Gwenole Beauchesne [Wed, 2 Jan 2013 16:33:15 +0000 (17:33 +0100)]
h264: don't create sub-buffer for slice data.

5 years agodecoder: create new context when encoded resolution changes.
Gwenole Beauchesne [Thu, 3 Jan 2013 10:16:44 +0000 (11:16 +0100)]
decoder: create new context when encoded resolution changes.

Create a new VA context if the encoded surface size changes because we
need to keep the underlying surface pool until the last one was released.
Otherwise, either of the following cases could have happened: (i) release
a VA surface to an inexistent pool, or (ii) release VA surface to an
existing surface pool, but with different size.

5 years agompeg2: don't create sub-buffer for slice data.
Gwenole Beauchesne [Wed, 2 Jan 2013 16:23:53 +0000 (17:23 +0100)]
mpeg2: don't create sub-buffer for slice data.

Avoid creating a GstBuffer for slice data. Rather, directly use the codec
frame input buffer data. This is possible because the codec frame is valid
until end_frame() where we submit the VA buffers for decoding. Anyway, the
slice data buffer is copied into the VA buffer when it is created.

5 years agompeg2: minor clean-ups.
Gwenole Beauchesne [Wed, 2 Jan 2013 13:45:50 +0000 (14:45 +0100)]
mpeg2: minor clean-ups.

Drop explicit initialization of most fields that are implicitly set to
zero. Remove some useless checks for NULL pointers.

5 years agompeg2: optimize scan for the second start code.
Gwenole Beauchesne [Wed, 2 Jan 2013 13:18:31 +0000 (14:18 +0100)]
mpeg2: optimize scan for the second start code.

Optimize scan for the second start code, on the next parse() call so that
to avoid scanning again earlier bytes where we didn't find any start code.

5 years agompeg2: use sequence_display_extension() to compute PAR.
Gwenole Beauchesne [Wed, 2 Jan 2013 13:10:20 +0000 (14:10 +0100)]
mpeg2: use sequence_display_extension() to compute PAR.

Also compute pixel-aspect-ratio from sequence_display_extension(),
should it exist in the bitstream.

5 years agompeg2: handle sequence_display_extension().
Gwenole Beauchesne [Wed, 2 Jan 2013 13:02:29 +0000 (14:02 +0100)]
mpeg2: handle sequence_display_extension().

5 years agompeg2: implement {start,end}_frame() hooks.
Gwenole Beauchesne [Thu, 27 Dec 2012 14:18:55 +0000 (15:18 +0100)]
mpeg2: implement {start,end}_frame() hooks.

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.

5 years agompeg2: parse slice() header earlier.
Gwenole Beauchesne [Thu, 27 Dec 2012 13:54:29 +0000 (14:54 +0100)]
mpeg2: parse slice() header earlier.

Parse slice() header and first macroblock position earlier in _parse()
function instead of waiting for the _decode() stage. This doesn't change
anything but readability.

5 years agompeg2: add codec specific decoder unit.
Gwenole Beauchesne [Thu, 27 Dec 2012 13:41:04 +0000 (14:41 +0100)]
mpeg2: add codec specific decoder unit.

Introduce new GstVaapiDecoderUnitMpeg2 object, which holds the standard
GstMpegVideoPacket and additional parsed header info. Besides, we now
parse as early as in the _parse() function so that to avoid un-necessary
creation of sub-buffers in _decode() for video packets that are not slices.

5 years agodecoder: introduce lists of units to decode before/after frame.
Gwenole Beauchesne [Thu, 27 Dec 2012 17:52:43 +0000 (18:52 +0100)]
decoder: introduce lists of units to decode before/after frame.

Theory of operations: all units marked as "slice" are moved to the "units"
list. Since this list only contains slice data units, the prev_slice pointer
was removed. Besides, we now maintain two extra lists of units to be decoded
before or after slice data units.

In particular, all units in the "pre_units" list will be decoded before
GstVaapiDecoder::start_frame() is called and units in the "post_units"
list will be decoded after GstVaapiDecoder::end_frame() is called.

5 years agodecoder: drop useless checks for codec objects.
Gwenole Beauchesne [Wed, 2 Jan 2013 15:06:18 +0000 (16:06 +0100)]
decoder: drop useless checks for codec objects.

Codec objects are used internally only and they are bound to be created
with a valid GstVaapiDecoder object.

5 years agovaapidecode: use GST_ERROR to print error messages.
Gwenole Beauchesne [Thu, 27 Dec 2012 09:35:45 +0000 (10:35 +0100)]
vaapidecode: use GST_ERROR to print error messages.

5 years agovaapidecode: avoid double release of frame on error.
Gwenole Beauchesne [Thu, 27 Dec 2012 08:55:14 +0000 (09:55 +0100)]
vaapidecode: avoid double release of frame on error.

Don't call gst_video_decoder_drop_frame() if gst_video_decoder_finish_frame()
was already called before and it returned an error. In that case, we were
releasing the frame again, thus leading to a "double-free" condition.

5 years agoAdd videoutils submodule for GstVideoDecoder APIs.
Gwenole Beauchesne [Fri, 21 Dec 2012 13:29:01 +0000 (14:29 +0100)]
Add videoutils submodule for GstVideoDecoder APIs.

5 years agoconfigure: check for GstVideoDecoder API.
Gwenole Beauchesne [Tue, 18 Dec 2012 15:36:01 +0000 (16:36 +0100)]
configure: check for GstVideoDecoder API.

GstVideoDecoder API is part of an unreleased GStreamer 0.10 stack. In particular,
this is only available in git 0.10 branch or GStreamer >= 1.0 stack. Interested
parties may either use upstream git 0.10 branch or backport the necessary support
for GstVideoDecoder API, thus including helper tools like GstVideoCodecFrame et al.

5 years agodocs: remove obsolete gst_vaapi_surface_proxy_get_type().
Gwenole Beauchesne [Tue, 18 Dec 2012 15:21:31 +0000 (16:21 +0100)]
docs: remove obsolete gst_vaapi_surface_proxy_get_type().

GstVaapiSurfaceProxy is no longer based on the GType system.

5 years agodocs: fix entries for GstVaapiSurfaceProxy.
Gwenole Beauchesne [Tue, 18 Dec 2012 15:17:22 +0000 (16:17 +0100)]
docs: fix entries for GstVaapiSurfaceProxy.

5 years agoNEWS: updates.
Gwenole Beauchesne [Tue, 18 Dec 2012 14:29:58 +0000 (15:29 +0100)]
NEWS: updates.

5 years agoBump library major version.
Gwenole Beauchesne [Tue, 18 Dec 2012 14:15:52 +0000 (15:15 +0100)]
Bump library major version.

Increase library major so that to cope with API/ABI incompatible changes
since 0.4.x series and avoid user issues.

5 years agosurfaceproxy: minor clean-ups.
Gwenole Beauchesne [Thu, 13 Dec 2012 15:02:52 +0000 (16:02 +0100)]
surfaceproxy: minor clean-ups.

5 years agosurfaceproxy: drop accessors to obsolete attributes.
Gwenole Beauchesne [Thu, 13 Dec 2012 14:51:24 +0000 (15:51 +0100)]
surfaceproxy: drop accessors to obsolete attributes.

Make GstVaapiSurfaceProxy only a thin wrapper around a VA context and a
VA surface. i.e. drop any other attribute like timestamp, duration,
interlaced or top-field-first.

5 years agodecoder: maintain decoded frames as GstVideoCodecFrame objects.
Gwenole Beauchesne [Thu, 13 Dec 2012 14:34:10 +0000 (15:34 +0100)]
decoder: maintain decoded frames as GstVideoCodecFrame objects.

Maintain decoded surfaces as GstVideoCodecFrame objects instead of
GstVaapiSurfaceProxy objects. The latter will tend to be reduced to
the strict minimum: a context and a surface.

5 years agovaapidecode: output all decoded frames as soon as possible.
Gwenole Beauchesne [Thu, 13 Dec 2012 13:30:18 +0000 (14:30 +0100)]
vaapidecode: output all decoded frames as soon as possible.

Make sure to push all decoded frames downstream as soon as possible.
This makes sure we don't need to wait for a new frame to be ready to
be decoded before receiving new decoded frames.

This also separates the decode process and the output process. The latter
could be moved to a specific GstTask later on.

5 years agodecoder: add gst_vaapi_decoder_get_frame() API.
Gwenole Beauchesne [Thu, 13 Dec 2012 13:27:18 +0000 (14:27 +0100)]
decoder: add gst_vaapi_decoder_get_frame() API.

Add new gst_vaapi_decoder_get_frame() function meant to be used with
gst_vaapi_decoder_decode(). The purpose is to return the next decoded
frame as a GstVideoCodecFrame and the associated GstVaapiSurfaceProxy
as the user-data object.

5 years agovaapipostproc: use GstBuffer flags for TFF.
Gwenole Beauchesne [Thu, 13 Dec 2012 14:47:27 +0000 (15:47 +0100)]
vaapipostproc: use GstBuffer flags for TFF.

Determine whether the buffer represents the top-field only by checking for
the GST_VIDEO_BUFFER_TFF flag instead of relying on the GstVaapiSurfaceProxy
flag. Also trust "interlaced" caps to determine whether the input frame
is interleaved or not.

5 years agovaapipostproc: handle video sub-buffers.
Gwenole Beauchesne [Thu, 13 Dec 2012 12:27:33 +0000 (13:27 +0100)]
vaapipostproc: handle video sub-buffers.

Intermediate elements may produce a sub-buffer from a valid GstVaapiVideoBuffer
for non raw YUV cases. Make sure vaapipostproc now understands those buffers.

5 years agoh264: optimize initialization process of decoder units.
Gwenole Beauchesne [Tue, 18 Dec 2012 13:57:36 +0000 (14:57 +0100)]
h264: optimize initialization process of decoder units.

Decoder units were zero-initialized, including the SPS/PPS/slice headers.
The latter don't require zero-initialization since the codecparsers/ lib
will do so for key variables already. This is not a great value per se but
at least it makes it possible to check whether the default initialization
decisions made in the codecparsers/ lib were right or not.

This can be reverted if this exposes too many issues.

5 years agoh264: minor clean-ups.
Gwenole Beauchesne [Thu, 13 Dec 2012 10:48:06 +0000 (11:48 +0100)]
h264: minor clean-ups.

Drop explicit initialization of most fields that are implicitly set to
zero. Drop helper macros for casting to GstVaapiPictureH264 or
GstVaapiFrameStore. Also remove some useless checks for NULL pointers.

5 years agoh264: drop GstVaapiSliceH264 object.
Gwenole Beauchesne [Fri, 7 Dec 2012 16:45:03 +0000 (17:45 +0100)]
h264: drop GstVaapiSliceH264 object.

Use standard GstVaapiSlice object from now on since we already have
parsed and recorded the slice headers (GstH264SliceHdr decode units).

5 years agoh264: detect new pictures from decode-units.
Gwenole Beauchesne [Thu, 13 Dec 2012 09:47:25 +0000 (10:47 +0100)]
h264: detect new pictures from decode-units.

Update is_new_picture() to cope with GstVaapiDecoderUnitH264, instead
of assuming frame boundaries when first_mb_in_slice is zero.

5 years agoh264: implement {start,end}_frame() hooks.
Gwenole Beauchesne [Thu, 13 Dec 2012 09:21:46 +0000 (10:21 +0100)]
h264: implement {start,end}_frame() hooks.

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.

5 years agoh264: optimize scan for the second start code.
Gwenole Beauchesne [Wed, 12 Dec 2012 17:33:52 +0000 (18:33 +0100)]
h264: optimize scan for the second start code.

Optimize scan for the second start code, on the next parse() call so that
to avoid scanning again earlier bytes where we didn't find any start code.

5 years agoh264: add codec specific decoder unit.
Gwenole Beauchesne [Thu, 6 Dec 2012 16:25:01 +0000 (17:25 +0100)]
h264: add codec specific decoder unit.

Introduce new GstVaapiDecoderUnitH264 object, which holds the standard
NAL unit header (GstH264NalUnit) and additional parsed header info.
Besides, we now parse headers as early as in the _parse() function so
that to avoid un-necessary creation of sub-buffers in _decode() for
NAL units that are not slices.

This is a performance win by ~+1.1% only.

5 years agovaapisink: handle sub video-buffers.
Gwenole Beauchesne [Tue, 4 Dec 2012 10:01:42 +0000 (11:01 +0100)]
vaapisink: handle sub video-buffers.

Intermediate elements may produce a sub-buffer from a valid GstVaapiVideoBuffer
for non raw YUV cases. Make sure vaapisink now understands those buffers.

5 years agovaapidecode: use gst_vaapi_decoder_get_codec_state().
Sreerenj Balachandran [Wed, 12 Dec 2012 14:22:32 +0000 (15:22 +0100)]
vaapidecode: use gst_vaapi_decoder_get_codec_state().

Directly use the GstVideoCodecState associated with the VA decoder
instead of parsing caps again.

Signed-off-by: Sreerenj Balachandran <>
Signed-off-by: Gwenole Beauchesne <>
5 years agovaapidecode: use more standard helpers.
Gwenole Beauchesne [Tue, 4 Dec 2012 13:53:15 +0000 (14:53 +0100)]
vaapidecode: use more standard helpers.

Use g_clear_object() [glib >= 2.28] and gst_caps_replace() helper functions
in more places.

5 years agovaapidecode: move to GstVideoDecoder base class.
Sreerenj Balachandran [Tue, 4 Dec 2012 13:45:29 +0000 (14:45 +0100)]
vaapidecode: move to GstVideoDecoder base class.

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 <>
Signed-off-by: Gwenole Beauchesne <>
5 years agojpeg: initial port to new GstVaapiDecoder API
Gwenole Beauchesne [Thu, 6 Dec 2012 13:02:25 +0000 (14:02 +0100)]
jpeg: initial port to new GstVaapiDecoder API

5 years agovc1: initial port to new GstVaapiDecoder API
Gwenole Beauchesne [Thu, 6 Dec 2012 13:02:21 +0000 (14:02 +0100)]
vc1: initial port to new GstVaapiDecoder API

5 years agoh264: initial port to new GstVaapiDecoder API
Gwenole Beauchesne [Thu, 6 Dec 2012 13:02:17 +0000 (14:02 +0100)]
h264: initial port to new GstVaapiDecoder API

5 years agompeg4: initial port to new GstVaapiDecoder API
Gwenole Beauchesne [Mon, 17 Dec 2012 17:47:20 +0000 (09:47 -0800)]
mpeg4: initial port to new GstVaapiDecoder API

5 years agompeg2: initial port to new GstVaapiDecoder API.
Gwenole Beauchesne [Thu, 6 Dec 2012 13:01:46 +0000 (14:01 +0100)]
mpeg2: initial port to new GstVaapiDecoder API.

5 years agodecoder: use GstVideoCodecState.
Sreerenj Balachandran [Wed, 12 Dec 2012 14:09:21 +0000 (15:09 +0100)]
decoder: use GstVideoCodecState.

Use standard GstVideoCodecState throughout GstVaapiDecoder and expose
it with a new gst_vaapi_decoder_get_codec_state() function. This makes
it possible to drop picture size (width, height) information, framerate
(fps_n, fps_d) information, pixel aspect ratio (par_n, par_d) information,
and interlace mode (is_interlaced field).

This is a new API with backwards compatibility maintained. In particular,
gst_vaapi_decoder_get_caps() is still available.

Signed-off-by: Sreerenj Balachandran <>
Signed-off-by: Gwenole Beauchesne <>
5 years agodecoder: update gst_vaapi_decoder_get_surface() semantics.
Gwenole Beauchesne [Wed, 12 Dec 2012 12:44:07 +0000 (13:44 +0100)]
decoder: update gst_vaapi_decoder_get_surface() semantics.

Align gst_vaapi_decoder_get_surface() semantics with the rest of the
API. That is, return a GstVaapiDecoderStatus and the decoded surface
as a handle to GstVaapiSurfaceProxy in parameter.

This is an API/ABI change.

5 years agodecoder: use standard helper functions.
Gwenole Beauchesne [Fri, 7 Dec 2012 15:40:44 +0000 (16:40 +0100)]
decoder: use standard helper functions.

Use g_clear_object(), gst_buffer_replace() and gst_caps_replace()
whenever necessary.

5 years agodecoder: expose new parse/decode API.
Gwenole Beauchesne [Thu, 29 Nov 2012 14:06:00 +0000 (15:06 +0100)]
decoder: expose new parse/decode API.

Introduce new decoding process whereby a GstVideoCodecFrame is created
first. Next, input stream buffers are accumulated into a GstAdapter,
that is then passed to the _parse() function. The GstVaapiDecoder object
accumulates all parsed units and when a complete frame or field is
detected, that GstVideoCodecFrame is passed to the _decode() function.

Ultimately, the caller receives a GstVaapiSurfaceProxy if decoding
process was successful.

5 years agodecoder: add {start,end}_frame() hooks.
Gwenole Beauchesne [Thu, 13 Dec 2012 09:20:35 +0000 (10:20 +0100)]
decoder: add {start,end}_frame() hooks.

The start_frame() hook is called prior to traversing all decode-units
for decoding. The unit argument represents the first slice in the frame.
Some codecs (e.g. H.264) need to wait for the first slice in order to
determine the actual VA context parameters.

5 years agodecoder: add new GstVaapiDecoder API.
Gwenole Beauchesne [Thu, 6 Dec 2012 12:57:42 +0000 (13:57 +0100)]
decoder: add new GstVaapiDecoder API.

Split decoding process into two steps: (i) parse incoming bitstreams
into simple decoder-units until the frame or field is complete; and
(ii) decode the whole frame or field at once.

This is an ABI change.

5 years agodecoder: add new "decoder-frame" object.
Gwenole Beauchesne [Wed, 5 Dec 2012 09:51:41 +0000 (10:51 +0100)]
decoder: add new "decoder-frame" object.

Introduce a new GstVaapiDecoderFrame that is just a list of decoder units
(GstVaapiDecoderUnit objects) that constitute a frame. This object is just
an extension to GstVideoCodecFrame for VA decoder purposes. It is available
as the user-data member element.

This is a libgstvaapi internal object.

5 years agodecoder: add new "decoder-unit" object.
Gwenole Beauchesne [Thu, 6 Dec 2012 08:44:01 +0000 (09:44 +0100)]
decoder: add new "decoder-unit" object.

Introduce GstVaapiDecoderUnit which represents a fragment of the source
stream to be decoded. For instance, a decode-unit will be a NAL unit for
H.264 streams, an EBDU for VC-1 streams, and a video packet for MPEG-2

This is a libgstvaapi internal object.

5 years agoPort GstVaapiFrameStore to GstVaapiMiniObject.
Gwenole Beauchesne [Mon, 3 Dec 2012 13:09:01 +0000 (14:09 +0100)]
Port GstVaapiFrameStore to GstVaapiMiniObject.

5 years agoPort codec objects to GstVaapiMiniObject.
Gwenole Beauchesne [Mon, 3 Dec 2012 10:19:08 +0000 (11:19 +0100)]
Port codec objects to GstVaapiMiniObject.

5 years agosurfaceproxy: port to GstVaapiMiniObject.
Gwenole Beauchesne [Mon, 3 Dec 2012 12:46:28 +0000 (13:46 +0100)]
surfaceproxy: port to GstVaapiMiniObject.

GstVaapiSurfaceProxy does not use any particular functionality from
GObject. Actually, it only needs a basic object type with reference

This is an API and ABI change.

5 years agoAdd GstVaapiMiniObject.
Gwenole Beauchesne [Fri, 30 Nov 2012 16:25:07 +0000 (17:25 +0100)]
Add GstVaapiMiniObject.

Introduce a new reference counted object that is very lightweight and
also provides flags and user-data functionalities. Initialization and
finalization times are reduced by up to a factor 5x vs GstMiniObject
from GStreamer 0.10 stack.

This is a libgstvaapi internal object.

5 years agotests: add test for MPEG-4:2 decoding.
Gwenole Beauchesne [Mon, 17 Dec 2012 10:51:17 +0000 (02:51 -0800)]
tests: add test for MPEG-4:2 decoding.

5 years agoh264: initialize VA context before allocating the first slice.
Gwenole Beauchesne [Mon, 17 Dec 2012 12:42:29 +0000 (04:42 -0800)]
h264: initialize VA context before allocating the first slice.

Fix decode_slice() to ensure a VA context exists prior to creating a
new GstVaapiSliceH264, which invokes vaCreateBuffer() with some VA
context ID. i.e. the latter was not initialized, thus causing failures
on Cedar Trail for example.

5 years agoconfigure: install plugin elements in GST_PLUGIN_PATH, if set.
Zhao Halley [Wed, 5 Dec 2012 01:15:32 +0000 (09:15 +0800)]
configure: install plugin elements in GST_PLUGIN_PATH, if set.

If GST_PLUGIN_PATH environment variable exists and points to a valid
directory, then use it as the system installation path for gst-vaapi
plugin elements.

Signed-off-by: Gwenole Beauchesne <>
5 years agoconfigure: downgrade glib required version to 2.28.
Gwenole Beauchesne [Mon, 17 Dec 2012 13:27:56 +0000 (14:27 +0100)]
configure: downgrade glib required version to 2.28.

5 years agolibs: fix compatibility with glib 2.28.
Gwenole Beauchesne [Mon, 17 Dec 2012 08:41:24 +0000 (09:41 +0100)]
libs: fix compatibility with glib 2.28.

Always prefer non deprecated APIs by default and provide compatibility
glue for older glib versions when necessary.

5 years agolibs: use glib >= 2.32 semantics for mutexes.
Gwenole Beauchesne [Mon, 17 Dec 2012 09:10:55 +0000 (10:10 +0100)]
libs: use glib >= 2.32 semantics for mutexes.

Use glib >= 2.32 semantics for GMutex and GRecMutex wrt. initialization
and termination. Basically, the new mutex objects can be used as static
mutex objects from the deprecated APIs, e.g. GStaticMutex and GStaticRecMutex.

5 years agolibs: only export gst_vaapi_*() symbols.
Gwenole Beauchesne [Mon, 17 Dec 2012 12:15:53 +0000 (04:15 -0800)]
libs: only export gst_vaapi_*() symbols.

This fixes symbol clashes between the gst-vaapi built-in codecparsers/
library and the system-provided one, mainly used by videoparses/. Now,
only symbols with the gst_vaapi_* prefix will be exported, if they are
not marked as "hidden" to libgstvaapi.

5 years agovaapiupload: reset direct-rendering to zero when changing caps.
Gwenole Beauchesne [Tue, 20 Nov 2012 17:21:41 +0000 (18:21 +0100)]
vaapiupload: reset direct-rendering to zero when changing caps.

Make sure to reset direct-rendering flag to zero when caps are changed,
and only derive it to one when the next checks succeed.

5 years agovaapiupload: fix sink caps to report the supported set of YUV caps.
Gwenole Beauchesne [Tue, 20 Nov 2012 13:42:24 +0000 (14:42 +0100)]
vaapiupload: fix sink caps to report the supported set of YUV caps.

Try to allocate the GstVaapiUploader helper object prior to listing the
supported image formats. Otherwise, only a single generic caps is output
with no particular pixel format referenced in there.

5 years agovaapiupload: use new GstVaapiUploader helper.
Zhao Halley [Tue, 20 Nov 2012 13:32:40 +0000 (14:32 +0100)]
vaapiupload: use new GstVaapiUploader helper.

Use GstVaapiUploader helper that automatically handles direct rendering
mode, thus making the "direct-rendering" property obsolete and hence it
is now removed.

The "direct-rendering" level 2, i.e. exposing VA surface buffers, was never
really well supported and it could actually trigger degraded performance.

Signed-off-by: Gwenole Beauchesne <>
5 years agovaapisink: compute and expose the supported set of YUV caps.
Gwenole Beauchesne [Tue, 20 Nov 2012 14:50:56 +0000 (15:50 +0100)]
vaapisink: compute and expose the supported set of YUV caps.

Make vaapisink expose only the set of supported caps for raw YUV buffers.

Add gst_vaapi_uploader_get_caps() helper function to determine the set
of supported YUV caps as source (for images). This function actually
tries to zero and upload each image to a 64x64 test surface. Of course,
this relies on VA drivers to not claim success if vaPutImage() is not
correctly supported.

5 years agovaapisink: add support for raw YUV buffers.
Gwenole Beauchesne [Tue, 20 Nov 2012 13:28:55 +0000 (14:28 +0100)]
vaapisink: add support for raw YUV buffers.

Add new GstVaapiUploader helper to upload raw YUV buffers to VA surfaces.
It is up to the caller to negotiate source caps (for images) and output
caps (for surfaces). gst_vaapi_uploader_has_direct_rendering() is available
to help decide between the creation of a GstVaapiVideoBuffer or a regular
GstBuffer on sink pads.

Signed-off-by: Zhao Halley <>
Signed-off-by: Gwenole Beauchesne <>
5 years agoimage: fix GstVaapiImage map and unmap.
Gwenole Beauchesne [Tue, 20 Nov 2012 13:36:29 +0000 (14:36 +0100)]
image: fix GstVaapiImage map and unmap.

Fix gst_vaapi_image_map() to return TRUE and the GstVaapiImageRaw
structure correctly filled in if the image was already mapped.
Likewise, make gst_vaapi_image_unmap() return TRUE if the image
was already unmapped.

5 years agovideobuffer: fix memory leak for surface and image.
Wind Yuan [Tue, 30 Oct 2012 05:15:45 +0000 (13:15 +0800)]
videobuffer: fix memory leak for surface and image.

Fix reference leak of surface and image in GstVaapiVideoBuffer wrapper,
thus resulting on actual memory leak of GstVaapiImage when using them
for downloads/uploads from VA surfaces and more specifically surfaces
when the pipeline is shutdown. i.e. vaTerminate() was never called
because the resources were not unreferenced, and thus not deallocated
in the end.

Signed-off-by: Gwenole Beauchesne <>
5 years agoNEWS: updates.
Gwenole Beauchesne [Mon, 19 Nov 2012 09:04:52 +0000 (10:04 +0100)]
NEWS: updates.

5 years agoh264: fix picture size in macroblocks.
Gwenole Beauchesne [Fri, 16 Nov 2012 17:00:10 +0000 (18:00 +0100)]
h264: fix picture size in macroblocks.

The picture size signalled by sps->{width,height} is the actual size with
cropping applied, not the original size derived from pic_width_in_mbs_minus1
and pic_height_in_map_units_minus1. VA driver expects that original size,

There is another issue pending: frame cropping information needs to be
taken care of.

5 years agocodecparsers: always build parserutils first.
Gwenole Beauchesne [Fri, 16 Nov 2012 15:18:52 +0000 (16:18 +0100)]
codecparsers: always build parserutils first.

Fix commit 18245b4 so that to link and build parserutils.[ch] first.
This is needed since that's the common dependency for actual codec
parsers (gstvc1parser.c for instance).

5 years agocodecparsers: always build the VC-1 parser library.
Gwenole Beauchesne [Thu, 15 Nov 2012 16:50:45 +0000 (17:50 +0100)]
codecparsers: always build the VC-1 parser library.

... this is useful to make sure pixel-aspect-ratio and framerate
information are correctly parsed since we have no means to detect
that at configure time.

5 years agompeg2: fix PAR calculation from commit bd11bae.
Sreerenj Balachandran [Thu, 8 Nov 2012 09:40:47 +0000 (11:40 +0200)]
mpeg2: fix PAR calculation from commit bd11bae.

Invoke gst_mpeg_video_finalise_mpeg2_sequence_header() to get the
correct PAR values. While doing so, require a newer version of the
bitstream parser library.

Note: it may be necessary to also parse the Sequence_Display_Extension()

Signed-off-by: Sreerenj Balachandran <>
Signed-off-by: Gwenole Beauchesne <>
5 years agoFix build with the GNU gold linker.
Gwenole Beauchesne [Thu, 15 Nov 2012 14:00:43 +0000 (15:00 +0100)]
Fix build with the GNU gold linker.

In particular, fix libgstvaapi-glx DSO dependencies to include libgstbase
and libgstvideo libs, e.g. for gst_video_buffer_get_overlay_composition().

5 years agowayland: port to 1.0 version of the protocol.
Rob Bradford [Fri, 2 Nov 2012 18:18:37 +0000 (18:18 +0000)]
wayland: port to 1.0 version of the protocol.

This patch updates to relect the 1.0 version of the protocol. The main
changes are the switch to wl_registry for global object notifications
and the way that the event queue and file descriptor is processed.

Signed-off-by: Gwenole Beauchesne <>
5 years agoh264: fix incorrect integration of previous commit (4d31e1e).
Gwenole Beauchesne [Wed, 14 Nov 2012 18:22:13 +0000 (19:22 +0100)]
h264: fix incorrect integration of previous commit (4d31e1e).

git am got confused somehow, though the end result doesn't change at
all since we require both SPS and PPS to be parsed prior to decoding
the first slice.

5 years agoh264: start decoding slices after first SPS/PPS activation.
Gwenole Beauchesne [Wed, 14 Nov 2012 17:40:47 +0000 (18:40 +0100)]
h264: start decoding slices after first SPS/PPS activation.

Only start decoding slices when at least one SPS and PPS got activated.
This fixes cases when a source represents a substream of another stream
and no SPS and PPS was inserted before the first slice of the generated

5 years agoh264: fix VAPictureParameterBufferH264.ReferenceFrames[] construction.
Gwenole Beauchesne [Wed, 14 Nov 2012 13:25:34 +0000 (14:25 +0100)]
h264: fix VAPictureParameterBufferH264.ReferenceFrames[] construction.

... for interlaced streams. The short_ref[] and long_ref[] arrays may
contain up to 32 fields but VA ReferenceFrames[] array expects up to
16 reference frames, thus including both fields.

5 years agoh264: fix interlaced stream decoding with MMCO.
Gwenole Beauchesne [Wed, 14 Nov 2012 09:27:12 +0000 (10:27 +0100)]
h264: fix interlaced stream decoding with MMCO.

Fix decoding of interlaced streams when adaptive_ref_pic_marking_mode_flag
is equal to 1, i.e. when memory management control operations are used. In
particular, when field_pic_flag is set to 0, the new reference flags shall
be applied to both fields.

5 years agoh264: add initial support for interlaced streams.
Gwenole Beauchesne [Tue, 13 Nov 2012 16:14:39 +0000 (17:14 +0100)]
h264: add initial support for interlaced streams.

Decoded frames are only output when they are complete, i.e. when both
fields are decoded. This also means that the "interlaced" caps is not
propagated to vaapipostproc or vaapisink elements. Another limitation
is that interlaced bitstreams with MMCO are unlikely to work.

5 years agoh264: split remove_reference_at() into finer units.
Gwenole Beauchesne [Tue, 13 Nov 2012 15:35:30 +0000 (16:35 +0100)]
h264: split remove_reference_at() into finer units.

Split remove_reference_at() into a function that actually removes the
specified entry from the short-term or long-term reference picture array,
and a function that sets reference flags to the desired value, possibly
zero. The latters marks the picture as "unused for reference".

5 years agodecoder: fix gst_vaapi_picture_new_field() object type.
Gwenole Beauchesne [Tue, 23 Oct 2012 12:04:22 +0000 (14:04 +0200)]
decoder: fix gst_vaapi_picture_new_field() object type.

Fix gst_vaapi_picture_new_field() to preserve the original picture type.
e.g. gst_vaapi_picture_new_field() with a GstVaapiPictureH264 argument
shall generate a GstVaapiPictureH264 object.

5 years agoh264: add picture structure for reference picture marking process.
Gwenole Beauchesne [Tue, 13 Nov 2012 13:04:31 +0000 (14:04 +0100)]
h264: add picture structure for reference picture marking process.

Introduce new `structure' field to the H.264 specific picture structure
so that to simplify the reference picture marking process. That local
picture structure is derived from the original picture structure, as
defined by the syntax elements field_pic_flag and bottom_field_flag.

5 years agoh264: introduce new frame store structure.
Gwenole Beauchesne [Fri, 2 Nov 2012 14:14:58 +0000 (15:14 +0100)]
h264: introduce new frame store structure.

The frame store represents a Decoded Picture Buffer entry, which can
hold up to two fields. So far, the frame store is only used to hold
full frames.

5 years agocodecparsers: update to gst-vaapi-rebased commit 73d6aab.
Gwenole Beauchesne [Tue, 13 Nov 2012 09:10:31 +0000 (10:10 +0100)]
codecparsers: update to gst-vaapi-rebased commit 73d6aab.

73d6aab h264: fix rbsp_more_data() implementation
25d04cf h264: fix error code for invalid size parsed in SPS
84798e5 fix FSF address

5 years agoh264: minor clean-ups.
Gwenole Beauchesne [Wed, 31 Oct 2012 15:37:14 +0000 (16:37 +0100)]
h264: minor clean-ups.

Move DPB flush up if the current picture to decode is an IDR. Besides,
don't bother to check for IDR pictures in dpb_add() function since an
explicit DPB flush was already performed in this case.

5 years agoh264: simplify reference picture marking process.
Gwenole Beauchesne [Wed, 31 Oct 2012 13:24:09 +0000 (14:24 +0100)]
h264: simplify reference picture marking process.

... to build the short_ref[] and long_ref[] lists from the DPB, instead
of maintaining them separately. This avoids refs/unrefs while making it
possible to generate the list based on the actual picture structure.

This also ensures that the list of generated ReferenceFrames[] actually
matches what reference frames are available in the DPB. i.e. short_ref[]
and long_ref[] entries are implied from the DPB, so there is no risk of
having "dangling" references.

5 years agoh264: introduce per-field POC in GstVaapiPictureH264.
Gwenole Beauchesne [Wed, 31 Oct 2012 10:52:03 +0000 (11:52 +0100)]
h264: introduce per-field POC in GstVaapiPictureH264.

Use the POC member available in the GstVaapiPicture base class and
get rid of the dependency on the local VAPictureH264 TopFieldOrderCnt
and BottomFieldOrderCnt. Rather, use a simple field_poc[] array
initialized to INT_MAX, so that to simplify picture POC calculation
for non frame pictures.

5 years agoh264: introduce GST_VAAPI_PICTURE_{SHORT,LONG}_TERM_REFERENCE flags.
Gwenole Beauchesne [Wed, 31 Oct 2012 10:45:14 +0000 (11:45 +0100)]

Further get rid of GstVaapiPictureH264-local VAPictureH264.flags for
reference bits, thus simplifying the reference picture marking process
to only track a single set of reference flags. Also introduce a new
long_term_frame_idx member.

5 years agoh264: introduce GST_VAAPI_PICTURE_FLAG_IDR flag.
Gwenole Beauchesne [Wed, 31 Oct 2012 10:33:40 +0000 (11:33 +0100)]
h264: introduce GST_VAAPI_PICTURE_FLAG_IDR flag.

5 years agoh264: fill in GstVaapiPicture structure.
Gwenole Beauchesne [Wed, 31 Oct 2012 09:56:15 +0000 (10:56 +0100)]
h264: fill in GstVaapiPicture structure.

... and get rid of local VAPictureH264.flags fields in GstVaapiPictureH264.

5 years agoh264: add vaapi_fill_picture() helper.
Gwenole Beauchesne [Wed, 31 Oct 2012 10:07:48 +0000 (11:07 +0100)]
h264: add vaapi_fill_picture() helper.

Add vaapi_fill_picture() helper function to convert GstVaapiPictureH264
to VAPictureH264 structure. This is preparatory work to get rid of the
local VAPictureH264 member in GstVaapiPictureH264.

5 years agoh264: fix activation order of picture and sequence parameters.
Gwenole Beauchesne [Fri, 26 Oct 2012 14:12:05 +0000 (16:12 +0200)]
h264: fix activation order of picture and sequence parameters.

Delay ensure_context() until we actually need a VA context for allocating
new VA surfaces, and then GstVaapiPictures, but also when a real activation
of a new picture parameter set occurs, thus also implying an activation
of the related sequence parameter set.

The most important thing was to drop the global pps and sps pointers since
they may not have matched the currently activated picture parameter or
sequence parameter sets at the specified decode point.

Anoter positive side-effect is that this cleans up all occurrences of
decode_current_picture() to only keep those useful in decode_picture(),
before a new picture is allocated, or in decode_sequence_end() when
an end-of-stream or end-of-sequence condition occurred.

5 years agoh264: fix scaling list generation.
Gwenole Beauchesne [Fri, 26 Oct 2012 11:17:43 +0000 (13:17 +0200)]
h264: fix scaling list generation.

... aka fix regression from efaab79. In particular, ScalingList8x8[]
array was partially copied to the VAIQMatrixBufferH264. While we are
at it, also improve bounds checking and avoid copying 8x8 scaling
lists if transform_8x8_mode_flag is set to 0.

5 years agoh264: fix detection of picture boundaries.
Gwenole Beauchesne [Wed, 24 Oct 2012 16:23:09 +0000 (18:23 +0200)]
h264: fix detection of picture boundaries.

Strictly follow the standard ( to detect the first VCL NAL
unit of a primary coded picture.