FFmpeg/libavcodec/mpegpicture.c

259 lines
7.8 KiB
C
Raw Normal View History

/*
* Mpeg video formats-related picture management functions
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/avassert.h"
#include "libavutil/common.h"
#include "libavutil/mem.h"
#include "libavutil/pixdesc.h"
#include "libavutil/imgutils.h"
#include "avcodec.h"
#include "mpegpicture.h"
#include "refstruct.h"
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
static void mpv_pic_reset(FFRefStructOpaque unused, void *obj)
{
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
MPVPicture *pic = obj;
av_frame_unref(pic->f);
ff_thread_progress_reset(&pic->progress);
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
ff_refstruct_unref(&pic->hwaccel_picture_private);
ff_refstruct_unref(&pic->mbskip_table);
ff_refstruct_unref(&pic->qscale_table_base);
ff_refstruct_unref(&pic->mb_type_base);
for (int i = 0; i < 2; i++) {
ff_refstruct_unref(&pic->motion_val_base[i]);
ff_refstruct_unref(&pic->ref_index[i]);
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
pic->motion_val[i] = NULL;
}
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
pic->mb_type = NULL;
pic->qscale_table = NULL;
pic->mb_stride =
pic->mb_width =
pic->mb_height = 0;
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
pic->dummy = 0;
pic->field_picture = 0;
pic->b_frame_score = 0;
pic->reference = 0;
pic->shared = 0;
pic->display_picture_number = 0;
pic->coded_picture_number = 0;
}
static int av_cold mpv_pic_init(FFRefStructOpaque opaque, void *obj)
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
{
MPVPicture *pic = obj;
int ret, init_progress = (uintptr_t)opaque.nc;
ret = ff_thread_progress_init(&pic->progress, init_progress);
if (ret < 0)
return ret;
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
pic->f = av_frame_alloc();
if (!pic->f)
return AVERROR(ENOMEM);
return 0;
}
static void av_cold mpv_pic_free(FFRefStructOpaque unused, void *obj)
{
MPVPicture *pic = obj;
ff_thread_progress_destroy(&pic->progress);
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
av_frame_free(&pic->f);
}
av_cold FFRefStructPool *ff_mpv_alloc_pic_pool(int init_progress)
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
{
return ff_refstruct_pool_alloc_ext(sizeof(MPVPicture),
FF_REFSTRUCT_POOL_FLAG_FREE_ON_INIT_ERROR,
(void*)(uintptr_t)init_progress,
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
mpv_pic_init, mpv_pic_reset, mpv_pic_free, NULL);
}
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
void ff_mpv_unref_picture(MPVWorkPicture *pic)
{
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
ff_refstruct_unref(&pic->ptr);
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
memset(pic, 0, sizeof(*pic));
}
static void set_workpic_from_pic(MPVWorkPicture *wpic, const MPVPicture *pic)
{
for (int i = 0; i < MPV_MAX_PLANES; i++) {
wpic->data[i] = pic->f->data[i];
wpic->linesize[i] = pic->f->linesize[i];
}
wpic->qscale_table = pic->qscale_table;
wpic->mb_type = pic->mb_type;
wpic->mbskip_table = pic->mbskip_table;
for (int i = 0; i < 2; i++) {
wpic->motion_val[i] = pic->motion_val[i];
wpic->ref_index[i] = pic->ref_index[i];
}
wpic->reference = pic->reference;
}
void ff_mpv_replace_picture(MPVWorkPicture *dst, const MPVWorkPicture *src)
{
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
av_assert1(dst != src);
ff_refstruct_replace(&dst->ptr, src->ptr);
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
memcpy(dst, src, sizeof(*dst));
}
void ff_mpv_workpic_from_pic(MPVWorkPicture *wpic, MPVPicture *pic)
{
avcodec/mpegpicture: Make MPVPicture refcounted 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>
2023-10-15 19:34:43 +00:00
ff_refstruct_replace(&wpic->ptr, pic);
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
if (!pic) {
memset(wpic, 0, sizeof(*wpic));
return;
}
set_workpic_from_pic(wpic, pic);
}
int ff_mpv_framesize_alloc(AVCodecContext *avctx,
ScratchpadContext *sc, int linesize)
{
# define EMU_EDGE_HEIGHT (4 * 70)
int linesizeabs = FFABS(linesize);
int alloc_size = FFALIGN(linesizeabs + 64, 32);
if (linesizeabs <= sc->linesize)
return 0;
if (avctx->hwaccel)
return 0;
if (linesizeabs < 24) {
av_log(avctx, AV_LOG_ERROR, "Image too small, temporary buffers cannot function\n");
return AVERROR_PATCHWELCOME;
}
if (av_image_check_size2(alloc_size, EMU_EDGE_HEIGHT, avctx->max_pixels, AV_PIX_FMT_NONE, 0, avctx) < 0)
return AVERROR(ENOMEM);
av_freep(&sc->edge_emu_buffer);
av_freep(&sc->scratchpad_buf);
// edge emu needs blocksize + filter length - 1
// (= 17x17 for halfpel / 21x21 for H.264)
// VC-1 computes luma and chroma simultaneously and needs 19X19 + 9x9
// at uvlinesize. It supports only YUV420 so 24x24 is enough
// linesize * interlaced * MBsize
// we also use this buffer for encoding in encode_mb_internal() needig an additional 32 lines
if (!FF_ALLOCZ_TYPED_ARRAY(sc->edge_emu_buffer, alloc_size * EMU_EDGE_HEIGHT) ||
!FF_ALLOCZ_TYPED_ARRAY(sc->scratchpad_buf, alloc_size * 4 * 16 * 2)) {
sc->linesize = 0;
av_freep(&sc->edge_emu_buffer);
return AVERROR(ENOMEM);
}
sc->linesize = linesizeabs;
sc->obmc_scratchpad = sc->scratchpad_buf + 16;
return 0;
}
int ff_mpv_pic_check_linesize(void *logctx, const AVFrame *f,
ptrdiff_t *linesizep, ptrdiff_t *uvlinesizep)
{
ptrdiff_t linesize = *linesizep, uvlinesize = *uvlinesizep;
if ((linesize && linesize != f->linesize[0]) ||
(uvlinesize && uvlinesize != f->linesize[1])) {
av_log(logctx, AV_LOG_ERROR, "Stride change unsupported: "
"linesize=%"PTRDIFF_SPECIFIER"/%d uvlinesize=%"PTRDIFF_SPECIFIER"/%d)\n",
linesize, f->linesize[0],
uvlinesize, f->linesize[1]);
return AVERROR_PATCHWELCOME;
}
if (av_pix_fmt_count_planes(f->format) > 2 &&
f->linesize[1] != f->linesize[2]) {
av_log(logctx, AV_LOG_ERROR, "uv stride mismatch unsupported\n");
return AVERROR_PATCHWELCOME;
}
*linesizep = f->linesize[0];
*uvlinesizep = f->linesize[1];
return 0;
}
static int alloc_picture_tables(BufferPoolContext *pools, MPVPicture *pic,
int mb_height)
{
#define GET_BUFFER(name, buf_suffix, idx_suffix) do { \
pic->name ## buf_suffix idx_suffix = ff_refstruct_pool_get(pools->name ## _pool); \
if (!pic->name ## buf_suffix idx_suffix) \
return AVERROR(ENOMEM); \
} while (0)
GET_BUFFER(qscale_table, _base,);
GET_BUFFER(mb_type, _base,);
if (pools->motion_val_pool) {
if (pools->mbskip_table_pool)
GET_BUFFER(mbskip_table,,);
for (int i = 0; i < 2; i++) {
GET_BUFFER(ref_index,, [i]);
GET_BUFFER(motion_val, _base, [i]);
pic->motion_val[i] = pic->motion_val_base[i] + 4;
}
}
#undef GET_BUFFER
pic->mb_width = pools->alloc_mb_width;
pic->mb_height = mb_height;
pic->mb_stride = pools->alloc_mb_stride;
pic->qscale_table = pic->qscale_table_base + 2 * pic->mb_stride + 1;
pic->mb_type = pic->mb_type_base + 2 * pic->mb_stride + 1;
return 0;
}
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
int ff_mpv_alloc_pic_accessories(AVCodecContext *avctx, MPVWorkPicture *wpic,
ScratchpadContext *sc,
BufferPoolContext *pools, int mb_height)
{
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
MPVPicture *pic = wpic->ptr;
int ret;
ret = ff_mpv_framesize_alloc(avctx, sc, pic->f->linesize[0]);
if (ret < 0)
goto fail;
ret = alloc_picture_tables(pools, pic, mb_height);
if (ret < 0)
goto fail;
avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic 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>
2023-10-08 10:25:07 +00:00
set_workpic_from_pic(wpic, pic);
return 0;
fail:
av_log(avctx, AV_LOG_ERROR, "Error allocating picture accessories.\n");
return ret;
}