There is no reason to use a temporary buffer as destination
for the new macroblock before copying it into its proper place.
(Originally, this has been added in commit
b68ab2609c due to concerns about
copying from GPU memory.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Only set the ScratchpadContext and let the users that need it
(i.e. encoders) set the MotionEstContext stuff themselves.
Also add an explicit pointer to ScratchpadContext to point
to the allocated buffer so that none of the other scratchpad
pointers is singled out as being used for the allocations.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
These pointers point to the same buffers, so one can just
use a union for them and avoid synchronising one of them.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Given that MPVPictures are already directly shared between threads
in case of frame-threaded decoding, one can simply use it to
pass decoding progress information between threads. This allows
to avoid one level of indirection; it also means avoids allocations
(of the ThreadFrameProgress structure) in case of frame-threading
and indeed makes ff_thread_release_ext_buffer() decoder-only
(actually, H.264-decoder-only).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Up until now, an initialized MpegEncContext had an array of
MPVPictures (way more than were ever needed) and the MPVPicture*
contained in the MPVWorkPictures as well as the input_picture
and reordered_input_picture arrays (for the encoder) pointed
into this array. Several of the pointers could point to the
same slot and because there was no reference counting involved,
one had to check for aliasing before unreferencing.
Furthermore, given that these pointers were not ownership pointers
the pointers were often simply reset without unreferencing
the slot (happened e.g. for the RV30 and RV40 decoders) or
there were moved without resetting the src pointer (happened
for the encoders where the entries in the input_picture
and reordered_input_picture arrays were not reset).
Instead actually releasing these pictures was performed by looping
over the whole array and checking which one of the entries needed
to be kept. Given that the array had way too many slots (36),
this meant that more than 30 MPVPictures have been unnecessarily
unreferenced in every ff_mpv_frame_start(); something similar
happened for the encoder.
This commit changes this by making the MPVPictures refcounted
via the RefStruct API. The MPVPictures itself are part of a pool
so that this does not entail constant allocations; instead,
the amount of allocations actually goes down, because the
earlier code used such a large array of MPVPictures (36 entries) and
allocated an AVFrame for every one of these on every
ff_mpv_common_init(). In fact, the pool is only freed when closing
the codec, so that reinitializations don't lead to new allocations
(this avoids having to sync the pool in update_thread_context).
Making MPVPictures refcounted also has another key benefit:
It makes it possible to directly share them across threads
(when using frame-threaded decoding), eliminating ugly code
with underlying av_frame_ref()'s; sharing these pictures
can't fail any more.
The pool is allocated in ff_mpv_decode_init() for decoders,
which therefore can fail now. This and the fact that the pool
is not unreferenced in ff_mpv_common_end() also necessitated
to mark several mpegvideo-decoders with the FF_CODEC_CAP_INIT_CLEANUP
flag.
*: This also means that there is no good reason any more for
ff_mpv_common_frame_size_change() to exist.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
There are two types of MPVPictures: Three (cur_pic, last_pic, next_pic)
that are directly part of MpegEncContext and an array of MPVPictures
that are separately allocated and are mostly accessed via pointers
(cur|last|next)_pic_ptr; they are also used to store AVFrames in the
encoder (necessary due to B-frames). As the name implies, each of the
former is directly associated with one of the _ptr pointers:
They actually share the same underlying buffers, but the ones
that are part of the context can have their data pointers offset
and their linesize doubled for field pictures.
Up until now, each of these had their own references; in particular,
there was an underlying av_frame_ref() to sync cur_pic and cur_pic_ptr
etc. This is wasteful.
This commit changes this relationship: cur_pic, last_pic and next_pic
now become MPVWorkPictures; this structure does not have an AVFrame
at all any more, but only the cached values of data and linesize.
It also contains a pointer to the corresponding MPVPicture, establishing
a more natural relationsship between the two.
This already means that creating the context-pictures from the pointers
can no longer fail.
What has not been changed is the fact that the MPVPicture* pointers
are not ownership pointers and that the MPVPictures are part of an
array of MPVPictures that is owned by a single AVCodecContext.
Doing so will be done in a latter commit.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
ff_alloc_picture() currently does two things: It checks the
consistency of the linesize (which should not be necessary, but is)
and it allocates certain buffers. (It does not actually allocate
the picture buffers, so its name is misleading.)
This commit splits it into two separate functions. The rationale
for this is that for the encoders, every picture needs its linesizes
checked, but not every picture needs these extra buffers.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This avoids an indirection and is in preparation for removing
the AVFrame from MpegEncContext.(cur|last|next)_pic altogether.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
No mpegvideo based codec supports alpha.
While just at it, also make the define shorter.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It involves less allocations and therefore has less
potential errors to be checked. One consequence thereof
is that updating the picture tables can no longer fail.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This avoids constant allocations+frees and will also allow
to simply switch to the RefStruct API, thereby avoiding
the overhead of the AVBuffer API.
It also simplifies the code, because it removes the "needs_realloc"
field: It was added in 435c0b87d2,
before the introduction of the AVBuffer API: given that these buffers
may be used by different threads, they were not freed immediately
and instead were marked as being freed later by setting needs_realloc.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The mpegvideo-based codecs currently require the linesize to be
constant (except when the frame dimensions change); one reason
for this is that certain scratch buffers whose size depend on
linesize are only allocated once and are presumed to be correctly
sized if the pointers are != NULL.
This commit changes this by storing the actual linesize these
buffers belong to and reallocating the buffers if it does not
suffice. This is not enough to actually support changing linesizes,
but it is a start. And it is a prerequisite for the next patch.
Also don't emit an error message in case the source ctx's
edge_emu_buffer is unset in ff_mpeg_update_thread_context().
It need not be an error at all; e.g. it is a perfectly normal
state in case a hardware acceleration is used as the scratch
buffers are not allocated in this case (it is easy to run into
this issue with MPEG-4) or if the src context was not initialized
at all (e.g. because the first packet contained garbage).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is unnecessary since the removal of non-thread-safe callbacks
in e0786a8eeb. Since then, the
AVCodecContext has only been used as logcontext.
Removing ff_thread_release_buffer() allowed to remove AVCodecContext*
parameters from several other functions (not only unref functions,
but also e.g. ff_h264_ref_picture() which calls ff_h264_unref_picture()
on error).
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Since at least commit c954cf1e1b
(adding ff_encode_alloc_frame()), a large part of ff_alloc_picture()
is completely separate for the two callers. Move the caller-specific
parts out to the callers.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
ff_alloc_picture() performs two tasks: a) In most instances,
it allocates frame buffers and b) it allocates certain
auxiliary buffers.
The exception to a) is the case when the encoder can reuse
user-supplied frames. And for these frames the auxiliary buffers
are unused, because this frame will never be used as current_picture
(and therefore also not as next_picture or last_picture);
see select_input_picture().
This means that we can simply avoid calling ff_alloc_picture()
with user-supplied frames at all.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Avoids allocations and therefore error checks: Syncing
hwaccel_picture_private across threads can't fail any more.
Also gets rid of an unnecessary pointer in structures and
in the parameter list of ff_hwaccel_frame_priv_alloc().
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Reviewed-by: Lynne <dev@lynne.ee>
Tested-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
These fields are only ever set by the encoder for the current picture
and for no other picture. So only one set of these values needs to
exist, so move them to MpegEncContext.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
These tables are only used by encoders and only for the current picture;
ergo they need not be put into the picture at all, but rather into
the encoder's context. They also don't need to be refcounted,
because there is only one owner.
In contrast to this, the earlier code refcounts them which
incurs unnecessary overhead. These references are not unreferenced
in ff_mpeg_unref_picture() (they are kept in order to have something
like a buffer pool), so that several buffers are kept at the same
time, although only one is needed, thereby wasting memory.
The code also propagates references to other pictures not part of
the pictures array (namely the copy of the current/next/last picture
in the MpegEncContext which get references of their own). These
references are not unreferenced in ff_mpeg_unref_picture() (the
buffers are probably kept in order to have something like a pool),
yet if the current picture is a B-frame, it gets unreferenced
at the end of ff_mpv_encode_picture() and its slot in the picture
array will therefore be reused the next time; but the copy of the
current picture also still has its references and therefore
these buffers will be made duplicated in order to make them writable
in the next call to ff_mpv_encode_picture(). This is of course
unnecessary.
Finally, ff_find_unused_picture() is supposed to just return
any unused picture and the code is supposed to work with it;
yet for the vsynth*-mpeg4-adap tests the result depends upon
the content of these buffers; given that this patchset
changes the content of these buffers (the initial content is now
the state of these buffers after encoding the last frame;
before this patch the buffers used came from the last picture
that occupied the same slot in the picture array) their ref-files
needed to be changed. This points to a bug somewhere (if one removes
the initialization, one gets uninitialized reads in
adaptive_quantization in ratecontrol.c).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The majority of frame-threaded decoders (mainly the intra-only)
need exactly one part of ThreadFrame: The AVFrame. They don't
need the owners nor the progress, yet they had to use it because
ff_thread_(get|release)_buffer() requires it.
This commit changes this and makes these functions work with ordinary
AVFrames; the decoders that need the extra fields for progress
use ff_thread_(get|release)_ext_buffer() which work exactly
as ff_thread_(get|release)_buffer() used to do.
This also avoids some unnecessary allocations of progress AVBuffers,
namely for H.264 and HEVC film grain frames: These frames are not
used for synchronization and therefore don't need a ThreadFrame.
Also move the ThreadFrame structure as well as ff_thread_ref_frame()
to threadframe.h, the header for frame-threaded decoders with
inter-frame dependencies.
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Also use said function in mpegvideo.c and mpegvideo_enc.c;
and make ff_free_picture_tables() static as it isn't needed anymore
outside of mpegpicture.c.
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The current size is AV_NUM_DATA_POINTERS (i.e. eight).
This number is chosen in order to minimize the amount of allocations
for AVFrame.extended_(data|buf) for audio; it is meaningless
for video for which four is sufficient. So decrease this array
in order to minimize what is copied in ff_mpeg_ref_picture()
and at the places that copy a whole MpegEncContext.
Also do the same for snowenc.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Do it only when requested with the AV_CODEC_EXPORT_DATA_VIDEO_ENC_PARAMS
flag.
Drop previous code using the long-deprecated AV_FRAME_DATA_QP_TABLE*
API. Temporarily disable fate-filter-pp, fate-filter-pp7,
fate-filter-spp. They will be reenabled once these filters are converted
in following commits.
These field are difficult to interpret, and are provided by a single
encoder (mpegvideoenc). In general they do not belong to a structure
containing raw data only, so remove them from AVFrame.
Mpegvideoenc now uses a private field in Picture for its internal
computations.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
* commit '7061bf0925':
mpegvideo: Move Picture-related functions to a separate file
Conflicts:
libavcodec/mpegvideo.c
libavcodec/mpegvideo.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
