Many of the fields of MpegEncContext (which is also used by decoders)
are actually only used by encoders. Therefore this commit adds
a new encoder-only structure and moves all of the encoder-only
fields to it except for those which require more explicit
synchronisation between the main slice context and the other
slice contexts. This synchronisation is currently mainly provided
by ff_update_thread_context() which simply copies most of
the main slice context over the other slice contexts. Fields
which are moved to the new MPVEncContext no longer participate
in this (which is desired, because it is horrible and for the
fields b) below wasteful) which means that some fields can only
be moved when explicit synchronisation code is added in later commits.
More explicitly, this commit moves the following fields:
a) Fields not copied by ff_update_duplicate_context():
dct_error_sum and dct_count; the former does not need synchronisation,
the latter is synchronised in merge_context_after_encode().
b) Fields which do not change after initialisation (these fields
could also be put into MPVMainEncContext at the cost of
an indirection to access them): lambda_table, adaptive_quant,
{luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp,
{p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table,
[pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff,
{inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields,
the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info,
mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function
pointers, mpv_flags, quantizer_noise_shaping,
frame_reconstruction_bitfield, error_rate and intra_penalty.
c) Fields which are already (re)set explicitly: The PutBitContexts
pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics
fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir,
esc_pos (reset when writing the header).
d) Fields which are only used by encoders not supporting slice
threading for which synchronisation doesn't matter: esc3_level_length
and the remaining mb_info fields.
e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD
is set (which implies trellis) and even then it is only used for
non-intra blocks. For these blocks dct_quantize_trellis_c() either
sets coded_score[n] or returns a last_non_zero value of -1
in which case coded_score will be reset in encode_mb_internal().
Therefore no old values are ever used.
The MotionEstContext has not been moved yet.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This avoids relying on ff_update_duplicate_context() to copy
these fields to the slice contexts.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Namely after the main slice context has already been initialized,
so that this initialized state is directly copied to the newly
created slice contexts without relying on it being copied
by ff_update_duplicate_context(). This is in preparation for further
commits.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
ff_MPV_encode_init_x86() has been renamed to ff_dct_encode_init_x86()
in 6b33e91899 in FFmpeg; libav renamed
it to ff_mpv_encode_init_x86() in commit
6b33e91899. This hasn't been noticed
in c1df467d73 when merging.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
last_pict_type, last_non_b_pict_type and last_lambda_for
are only used by the encoder's main thread.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This is in preparation for adding a special slice context for
the encoders and moving all the encoder-specific fields to it.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(Checking in the same way for intra_only is not straightforward,
because at least p_mv_table is written to even in intra_only
mode.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Also allocate them jointly (thereby reducing the number of base
arrays considerably) and already set them on all slice contexts;
don't rely on ff_update_duplicate_context().
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
dct_error_sum and me.map are allocated per slice-context
and therefore their allocation has not been moved to
mpegvideoenc.c in 0154fb43e3.
This commit changes this by allocating them jointly and
moving said allocations to mpegvideo_enc.c like the other
encoder-only buffers.
The buffers are suitably aligned to ensure that no false
sharing occurs.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Several fields are not used by any generic code and can therefore
be moved to more specialized contexts.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Switch to using the length+symbol table combination that is
also used by the decoder and remove MVTable as well as
the old tables altogether.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Up until now, the decoder uses VLCs to determine whether
this is an escape element or an index in two separate tables
giving the x and y components of the motion vector.
Given that both these components fit into an int16_t and therefore
into a VLCElem's symbol one can simply use a symbol table
to avoid the latter lookups.
This necessitated to combine the two tables for x and y components.
While just at it, also switch to ff_vlc_init_tables_from_lengths().
This will allow to get rid of the code tables (about 4KiB) once
the encoder has been ported, too.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is a pointer to const to allow the slice threads to inspect
values without modifying them; also make it a simple cast
for codecs that don't support slice threading.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It gets rid of a switch (with accompanying CONFIG_*_ENCODER checks);
for MJPEG, it even lets one perform the check for whether one is
really encoding or only recording the macroblock once during init.
Furthermore, the switch actually contained lots of dead code --
it is compiled three times (for different pixel formats: 420, 422, 444),
yet most encoders only support 420. The approach used here automatically
fixes this.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Will avoid forward declarations later.
While just at it, forward ff_mpv_encode_init() errors.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The header is always written by the main thread, so it is
not surprising that header_bits can be moved to MPVMainEncContext
as well.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
avpriv_slicethread_free() is one of our functions that takes
a pointer to a pointer and resets the pointer when done.
It is legal for such functions to be passed a pointer to a NULL
pointer, yet passing a NULL pointer would be insane and should
not be tolerated.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This is no longer necessary given that the MPEG-1 encoder
now adapts the size of the last slice to ensure that
no slice header has an invalid slice_start_code.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The valid values for slice_start_code are 0x1..0xAF, which implies
that one can't start a slice with row nb > 174 (zero-based).
This problem can be encountered with files of sufficient height
(more than 2800 pixels) either by using the slice option or by
imposing an RTP payload limit.
Fix this by making the last slice start on the maximum allowed
slice row if necessary and divide the first 174 rows to the remaining
slices. This will impede parallelism both in the decoder and encoder,
but that is unavoidable.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Everything except setting the dc_scale_tables can be done
before calling ff_mpv_encode_init(); setting the dc_scale_tables
unfortunately still has to be performed via a (now inlined)
callback performed by ff_mpv_encode_init() as it relies
on intra_dc_precision having been sanitized.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
