From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
To: ffmpeg-devel@ffmpeg.org
Cc: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Subject: [FFmpeg-devel] [PATCH 42/57] avcodec/mpegpicture: Make MPVPicture refcounted
Date: Mon, 29 Apr 2024 23:14:23 +0200
Message-ID: <AS8P250MB0744B0601994778005D45B848F1B2@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM> (raw)
In-Reply-To: <GV1SPRMB0021759DDCCEFEB3B9609B808F1B2@GV1SPRMB0021.EURP250.PROD.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>
---
libavcodec/flvdec.c | 3 +-
libavcodec/h261dec.c | 5 +-
libavcodec/h263dec.c | 14 ++--
libavcodec/intelh263dec.c | 3 +-
libavcodec/mjpegenc.c | 2 +-
libavcodec/mpeg12dec.c | 11 ++-
libavcodec/mpeg4videodec.c | 3 +-
libavcodec/mpegpicture.c | 160 ++++++++++++-------------------------
libavcodec/mpegpicture.h | 14 ++--
libavcodec/mpegvideo.c | 21 +----
libavcodec/mpegvideo.h | 2 +-
libavcodec/mpegvideo_dec.c | 72 +++++------------
libavcodec/mpegvideo_enc.c | 74 +++++++++--------
libavcodec/mpegvideodec.h | 4 +-
libavcodec/msmpeg4dec.c | 12 ++-
libavcodec/rv10.c | 8 +-
libavcodec/rv34.c | 6 +-
libavcodec/svq1enc.c | 3 -
libavcodec/vc1dec.c | 8 +-
19 files changed, 176 insertions(+), 249 deletions(-)
diff --git a/libavcodec/flvdec.c b/libavcodec/flvdec.c
index 1bb1b12917..f4bfd99417 100644
--- a/libavcodec/flvdec.c
+++ b/libavcodec/flvdec.c
@@ -121,6 +121,7 @@ const FFCodec ff_flv_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.p.max_lowres = 3,
};
diff --git a/libavcodec/h261dec.c b/libavcodec/h261dec.c
index d8d0dcf3cf..392f1aef1d 100644
--- a/libavcodec/h261dec.c
+++ b/libavcodec/h261dec.c
@@ -84,10 +84,13 @@ static av_cold int h261_decode_init(AVCodecContext *avctx)
static AVOnce init_static_once = AV_ONCE_INIT;
H261DecContext *const h = avctx->priv_data;
MpegEncContext *const s = &h->s;
+ int ret;
s->private_ctx = &h->common;
// set defaults
- ff_mpv_decode_init(s, avctx);
+ ret = ff_mpv_decode_init(s, avctx);
+ if (ret < 0)
+ return ret;
s->out_format = FMT_H261;
s->low_delay = 1;
diff --git a/libavcodec/h263dec.c b/libavcodec/h263dec.c
index b8db4ffc98..b9762be9c9 100644
--- a/libavcodec/h263dec.c
+++ b/libavcodec/h263dec.c
@@ -95,7 +95,9 @@ av_cold int ff_h263_decode_init(AVCodecContext *avctx)
s->out_format = FMT_H263;
// set defaults
- ff_mpv_decode_init(s, avctx);
+ ret = ff_mpv_decode_init(s, avctx);
+ if (ret < 0)
+ return ret;
s->quant_precision = 5;
s->decode_mb = ff_h263_decode_mb;
@@ -427,7 +429,7 @@ int ff_h263_decode_frame(AVCodecContext *avctx, AVFrame *pict,
if (s->low_delay == 0 && s->next_pic.ptr) {
if ((ret = av_frame_ref(pict, s->next_pic.ptr->f)) < 0)
return ret;
- s->next_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->next_pic);
*got_frame = 1;
} else if (s->skipped_last_frame && s->cur_pic.ptr) {
@@ -439,7 +441,7 @@ int ff_h263_decode_frame(AVCodecContext *avctx, AVFrame *pict,
* returned picture would be reused */
if ((ret = ff_decode_frame_props(avctx, pict)) < 0)
return ret;
- s->cur_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->cur_pic);
*got_frame = 1;
}
@@ -698,7 +700,8 @@ const FFCodec ff_h263_decoder = {
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1 |
AV_CODEC_CAP_DELAY,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.flush = ff_mpeg_flush,
.p.max_lowres = 3,
.hw_configs = h263_hw_config_list,
@@ -715,7 +718,8 @@ const FFCodec ff_h263p_decoder = {
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1 |
AV_CODEC_CAP_DELAY,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.flush = ff_mpeg_flush,
.p.max_lowres = 3,
.hw_configs = h263_hw_config_list,
diff --git a/libavcodec/intelh263dec.c b/libavcodec/intelh263dec.c
index d4051c36f1..4efae7938c 100644
--- a/libavcodec/intelh263dec.c
+++ b/libavcodec/intelh263dec.c
@@ -135,5 +135,6 @@ const FFCodec ff_h263i_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
};
diff --git a/libavcodec/mjpegenc.c b/libavcodec/mjpegenc.c
index b6de50edce..9d4c3a4f41 100644
--- a/libavcodec/mjpegenc.c
+++ b/libavcodec/mjpegenc.c
@@ -80,7 +80,7 @@ static av_cold void init_uni_ac_vlc(const uint8_t huff_size_ac[256],
static void mjpeg_encode_picture_header(MpegEncContext *s)
{
- ff_mjpeg_encode_picture_header(s->avctx, &s->pb, s->picture->f, s->mjpeg_ctx,
+ ff_mjpeg_encode_picture_header(s->avctx, &s->pb, s->cur_pic.ptr->f, s->mjpeg_ctx,
s->intra_scantable.permutated, 0,
s->intra_matrix, s->chroma_intra_matrix,
s->slice_context_count > 1);
diff --git a/libavcodec/mpeg12dec.c b/libavcodec/mpeg12dec.c
index cf8493ba51..0d5540fd2f 100644
--- a/libavcodec/mpeg12dec.c
+++ b/libavcodec/mpeg12dec.c
@@ -793,13 +793,16 @@ static av_cold int mpeg_decode_init(AVCodecContext *avctx)
{
Mpeg1Context *s = avctx->priv_data;
MpegEncContext *s2 = &s->mpeg_enc_ctx;
+ int ret;
s2->out_format = FMT_MPEG1;
if ( avctx->codec_tag != AV_RL32("VCR2")
&& avctx->codec_tag != AV_RL32("BW10"))
avctx->coded_width = avctx->coded_height = 0; // do not trust dimensions from input
- ff_mpv_decode_init(s2, avctx);
+ ret = ff_mpv_decode_init(s2, avctx);
+ if (ret < 0)
+ return ret;
ff_mpeg12_init_vlcs();
@@ -2529,7 +2532,7 @@ static int mpeg_decode_frame(AVCodecContext *avctx, AVFrame *picture,
if (ret < 0)
return ret;
- s2->next_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s2->next_pic);
*got_output = 1;
}
@@ -2552,14 +2555,14 @@ static int mpeg_decode_frame(AVCodecContext *avctx, AVFrame *picture,
}
s->extradata_decoded = 1;
if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE)) {
- s2->cur_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s2->cur_pic);
return ret;
}
}
ret = decode_chunks(avctx, picture, got_output, buf, buf_size);
if (ret<0 || *got_output) {
- s2->cur_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s2->cur_pic);
if (s->timecode_frame_start != -1 && *got_output) {
char tcbuf[AV_TIMECODE_STR_SIZE];
diff --git a/libavcodec/mpeg4videodec.c b/libavcodec/mpeg4videodec.c
index 58896250ce..b07d016a8b 100644
--- a/libavcodec/mpeg4videodec.c
+++ b/libavcodec/mpeg4videodec.c
@@ -3860,7 +3860,8 @@ const FFCodec ff_mpeg4_decoder = {
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1 |
AV_CODEC_CAP_DELAY | AV_CODEC_CAP_FRAME_THREADS,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.flush = ff_mpeg_flush,
.p.max_lowres = 3,
.p.profiles = NULL_IF_CONFIG_SMALL(ff_mpeg4_video_profiles),
diff --git a/libavcodec/mpegpicture.c b/libavcodec/mpegpicture.c
index 84d9209c4c..caba4ea2e9 100644
--- a/libavcodec/mpegpicture.c
+++ b/libavcodec/mpegpicture.c
@@ -30,8 +30,14 @@
#include "refstruct.h"
#include "threadframe.h"
-static void av_noinline free_picture_tables(MPVPicture *pic)
+static void mpv_pic_reset(FFRefStructOpaque unused, void *obj)
{
+ MPVPicture *pic = obj;
+
+ ff_thread_release_ext_buffer(&pic->tf);
+
+ 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);
@@ -39,16 +45,54 @@ static void av_noinline free_picture_tables(MPVPicture *pic)
for (int i = 0; i < 2; i++) {
ff_refstruct_unref(&pic->motion_val_base[i]);
ff_refstruct_unref(&pic->ref_index[i]);
+
+ pic->motion_val[i] = NULL;
}
+ pic->mb_type = NULL;
+ pic->qscale_table = NULL;
+
+ pic->mb_stride =
pic->mb_width =
pic->mb_height = 0;
+
+ pic->dummy = 0;
+ pic->field_picture = 0;
+ pic->interlaced = 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 unused, void *obj)
+{
+ MPVPicture *pic = obj;
+
+ pic->f = av_frame_alloc();
+ if (!pic->f)
+ return AVERROR(ENOMEM);
+ pic->tf.f = pic->f;
+ return 0;
+}
+
+static void av_cold mpv_pic_free(FFRefStructOpaque unused, void *obj)
+{
+ MPVPicture *pic = obj;
+
+ av_frame_free(&pic->f);
+}
+
+av_cold FFRefStructPool *ff_mpv_alloc_pic_pool(void)
+{
+ return ff_refstruct_pool_alloc_ext(sizeof(MPVPicture), 0, NULL,
+ mpv_pic_init, mpv_pic_reset, mpv_pic_free, NULL);
}
void ff_mpv_unref_picture(MPVWorkPicture *pic)
{
- if (pic->ptr)
- ff_mpeg_unref_picture(pic->ptr);
+ ff_refstruct_unref(&pic->ptr);
memset(pic, 0, sizeof(*pic));
}
@@ -71,16 +115,18 @@ static void set_workpic_from_pic(MPVWorkPicture *wpic, const MPVPicture *pic)
void ff_mpv_replace_picture(MPVWorkPicture *dst, const MPVWorkPicture *src)
{
+ av_assert1(dst != src);
+ ff_refstruct_replace(&dst->ptr, src->ptr);
memcpy(dst, src, sizeof(*dst));
}
void ff_mpv_workpic_from_pic(MPVWorkPicture *wpic, MPVPicture *pic)
{
+ ff_refstruct_replace(&wpic->ptr, pic);
if (!pic) {
memset(wpic, 0, sizeof(*wpic));
return;
}
- wpic->ptr = pic;
set_workpic_from_pic(wpic, pic);
}
@@ -211,109 +257,3 @@ fail:
av_log(avctx, AV_LOG_ERROR, "Error allocating picture accessories.\n");
return ret;
}
-
-/**
- * Deallocate a picture; frees the picture tables in case they
- * need to be reallocated anyway.
- */
-void ff_mpeg_unref_picture(MPVPicture *pic)
-{
- pic->tf.f = pic->f;
- ff_thread_release_ext_buffer(&pic->tf);
-
- ff_refstruct_unref(&pic->hwaccel_picture_private);
-
- free_picture_tables(pic);
-
- pic->dummy = 0;
-
- pic->field_picture = 0;
- pic->interlaced = 0;
- pic->b_frame_score = 0;
- pic->reference = 0;
- pic->shared = 0;
- pic->display_picture_number = 0;
- pic->coded_picture_number = 0;
-}
-
-static void update_picture_tables(MPVPicture *dst, const MPVPicture *src)
-{
- ff_refstruct_replace(&dst->mbskip_table, src->mbskip_table);
- ff_refstruct_replace(&dst->qscale_table_base, src->qscale_table_base);
- ff_refstruct_replace(&dst->mb_type_base, src->mb_type_base);
- for (int i = 0; i < 2; i++) {
- ff_refstruct_replace(&dst->motion_val_base[i], src->motion_val_base[i]);
- ff_refstruct_replace(&dst->ref_index[i], src->ref_index[i]);
- }
-
- dst->qscale_table = src->qscale_table;
- dst->mb_type = src->mb_type;
- for (int i = 0; i < 2; i++)
- dst->motion_val[i] = src->motion_val[i];
-
- dst->mb_width = src->mb_width;
- dst->mb_height = src->mb_height;
- dst->mb_stride = src->mb_stride;
-}
-
-int ff_mpeg_ref_picture(MPVPicture *dst, MPVPicture *src)
-{
- int ret;
-
- av_assert0(!dst->f->buf[0]);
- av_assert0(src->f->buf[0]);
-
- src->tf.f = src->f;
- dst->tf.f = dst->f;
- ret = ff_thread_ref_frame(&dst->tf, &src->tf);
- if (ret < 0)
- goto fail;
-
- update_picture_tables(dst, src);
-
- ff_refstruct_replace(&dst->hwaccel_picture_private,
- src->hwaccel_picture_private);
-
- dst->dummy = src->dummy;
- dst->field_picture = src->field_picture;
- dst->b_frame_score = src->b_frame_score;
- dst->reference = src->reference;
- dst->shared = src->shared;
- dst->display_picture_number = src->display_picture_number;
- dst->coded_picture_number = src->coded_picture_number;
- dst->interlaced = src->interlaced;
-
- return 0;
-fail:
- ff_mpeg_unref_picture(dst);
- return ret;
-}
-
-int ff_find_unused_picture(AVCodecContext *avctx, MPVPicture *picture, int shared)
-{
- for (int i = 0; i < MAX_PICTURE_COUNT; i++)
- if (!picture[i].f->buf[0])
- return i;
-
- av_log(avctx, AV_LOG_FATAL,
- "Internal error, picture buffer overflow\n");
- /* We could return -1, but the codec would crash trying to draw into a
- * non-existing frame anyway. This is safer than waiting for a random crash.
- * Also the return of this is never useful, an encoder must only allocate
- * as much as allowed in the specification. This has no relationship to how
- * much libavcodec could allocate (and MAX_PICTURE_COUNT is always large
- * enough for such valid streams).
- * Plus, a decoder has to check stream validity and remove frames if too
- * many reference frames are around. Waiting for "OOM" is not correct at
- * all. Similarly, missing reference frames have to be replaced by
- * interpolated/MC frames, anything else is a bug in the codec ...
- */
- abort();
- return -1;
-}
-
-void av_cold ff_mpv_picture_free(MPVPicture *pic)
-{
- ff_mpeg_unref_picture(pic);
- av_frame_free(&pic->f);
-}
diff --git a/libavcodec/mpegpicture.h b/libavcodec/mpegpicture.h
index 774c5c8841..63cf092806 100644
--- a/libavcodec/mpegpicture.h
+++ b/libavcodec/mpegpicture.h
@@ -29,7 +29,6 @@
#include "threadframe.h"
#define MPV_MAX_PLANES 3
-#define MAX_PICTURE_COUNT 36
#define EDGE_WIDTH 16
typedef struct ScratchpadContext {
@@ -95,7 +94,7 @@ typedef struct MPVWorkPicture {
uint8_t *data[MPV_MAX_PLANES];
ptrdiff_t linesize[MPV_MAX_PLANES];
- MPVPicture *ptr;
+ MPVPicture *ptr; ///< RefStruct reference
int8_t *qscale_table;
@@ -110,6 +109,11 @@ typedef struct MPVWorkPicture {
int reference;
} MPVWorkPicture;
+/**
+ * Allocate a pool of MPVPictures.
+ */
+struct FFRefStructPool *ff_mpv_alloc_pic_pool(void);
+
/**
* Allocate an MPVPicture's accessories (but not the AVFrame's buffer itself)
* and set the MPVWorkPicture's fields.
@@ -130,14 +134,8 @@ int ff_mpv_pic_check_linesize(void *logctx, const struct AVFrame *f,
int ff_mpeg_framesize_alloc(AVCodecContext *avctx, MotionEstContext *me,
ScratchpadContext *sc, int linesize);
-int ff_mpeg_ref_picture(MPVPicture *dst, MPVPicture *src);
void ff_mpv_unref_picture(MPVWorkPicture *pic);
void ff_mpv_workpic_from_pic(MPVWorkPicture *wpic, MPVPicture *pic);
void ff_mpv_replace_picture(MPVWorkPicture *dst, const MPVWorkPicture *src);
-void ff_mpeg_unref_picture(MPVPicture *picture);
-
-void ff_mpv_picture_free(MPVPicture *pic);
-
-int ff_find_unused_picture(AVCodecContext *avctx, MPVPicture *picture, int shared);
#endif /* AVCODEC_MPEGPICTURE_H */
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
index a4bea7273a..dc7fee2ac7 100644
--- a/libavcodec/mpegvideo.c
+++ b/libavcodec/mpegvideo.c
@@ -698,7 +698,6 @@ static void clear_context(MpegEncContext *s)
s->bitstream_buffer = NULL;
s->allocated_bitstream_buffer_size = 0;
- s->picture = NULL;
s->p_field_mv_table_base = NULL;
for (int i = 0; i < 2; i++)
for (int j = 0; j < 2; j++)
@@ -723,10 +722,10 @@ static void clear_context(MpegEncContext *s)
*/
av_cold int ff_mpv_common_init(MpegEncContext *s)
{
- int i, ret;
int nb_slices = (HAVE_THREADS &&
s->avctx->active_thread_type & FF_THREAD_SLICE) ?
s->avctx->thread_count : 1;
+ int ret;
clear_context(s);
@@ -752,14 +751,6 @@ av_cold int ff_mpv_common_init(MpegEncContext *s)
if (ret)
return ret;
- if (!FF_ALLOCZ_TYPED_ARRAY(s->picture, MAX_PICTURE_COUNT))
- return AVERROR(ENOMEM);
- for (i = 0; i < MAX_PICTURE_COUNT; i++) {
- s->picture[i].f = av_frame_alloc();
- if (!s->picture[i].f)
- goto fail_nomem;
- }
-
if ((ret = ff_mpv_init_context_frame(s)))
goto fail;
@@ -786,8 +777,6 @@ av_cold int ff_mpv_common_init(MpegEncContext *s)
// }
return 0;
- fail_nomem:
- ret = AVERROR(ENOMEM);
fail:
ff_mpv_common_end(s);
return ret;
@@ -827,11 +816,9 @@ void ff_mpv_common_end(MpegEncContext *s)
av_freep(&s->bitstream_buffer);
s->allocated_bitstream_buffer_size = 0;
- if (s->picture) {
- for (int i = 0; i < MAX_PICTURE_COUNT; i++)
- ff_mpv_picture_free(&s->picture[i]);
- }
- av_freep(&s->picture);
+ ff_mpv_unref_picture(&s->last_pic);
+ ff_mpv_unref_picture(&s->cur_pic);
+ ff_mpv_unref_picture(&s->next_pic);
s->context_initialized = 0;
s->context_reinit = 0;
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
index 75120f2082..6c30e4119d 100644
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -128,7 +128,7 @@ typedef struct MpegEncContext {
int mb_num; ///< number of MBs of a picture
ptrdiff_t linesize; ///< line size, in bytes, may be different from width
ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
- MPVPicture *picture; ///< main picture buffer
+ struct FFRefStructPool *picture_pool; ///< Pool for MPVPictures
MPVPicture **input_picture;///< next pictures on display order for encoding
MPVPicture **reordered_input_picture; ///< pointer to the next pictures in coded order for encoding
diff --git a/libavcodec/mpegvideo_dec.c b/libavcodec/mpegvideo_dec.c
index c8952f4831..d596f94df3 100644
--- a/libavcodec/mpegvideo_dec.c
+++ b/libavcodec/mpegvideo_dec.c
@@ -38,11 +38,12 @@
#include "mpegvideo.h"
#include "mpegvideodec.h"
#include "mpeg4videodec.h"
+#include "refstruct.h"
#include "thread.h"
#include "threadframe.h"
#include "wmv2dec.h"
-void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx)
+int ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx)
{
ff_mpv_common_defaults(s);
@@ -57,6 +58,14 @@ void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx)
ff_mpv_idct_init(s);
ff_h264chroma_init(&s->h264chroma, 8); //for lowres
+
+ if (!s->picture_pool && // VC-1 can call this multiple times
+ ff_thread_sync_ref(avctx, offsetof(MpegEncContext, picture_pool))) {
+ s->picture_pool = ff_mpv_alloc_pic_pool();
+ if (!s->picture_pool)
+ return AVERROR(ENOMEM);
+ }
+ return 0;
}
int ff_mpeg_update_thread_context(AVCodecContext *dst,
@@ -103,26 +112,9 @@ int ff_mpeg_update_thread_context(AVCodecContext *dst,
s->coded_picture_number = s1->coded_picture_number;
s->picture_number = s1->picture_number;
- av_assert0(!s->picture || s->picture != s1->picture);
- if (s->picture)
- for (int i = 0; i < MAX_PICTURE_COUNT; i++) {
- ff_mpeg_unref_picture(&s->picture[i]);
- if (s1->picture && s1->picture[i].f->buf[0] &&
- (ret = ff_mpeg_ref_picture(&s->picture[i], &s1->picture[i])) < 0)
- return ret;
- }
-
-#define UPDATE_PICTURE(pic)\
-do {\
- if (s->picture && s1->picture && s1->pic.ptr && s1->pic.ptr->f->buf[0]) {\
- ff_mpv_workpic_from_pic(&s->pic, &s->picture[s1->pic.ptr - s1->picture]);\
- } else\
- ff_mpv_unref_picture(&s->pic);\
-} while (0)
-
- UPDATE_PICTURE(cur_pic);
- UPDATE_PICTURE(last_pic);
- UPDATE_PICTURE(next_pic);
+ ff_mpv_replace_picture(&s->cur_pic, &s1->cur_pic);
+ ff_mpv_replace_picture(&s->last_pic, &s1->last_pic);
+ ff_mpv_replace_picture(&s->next_pic, &s1->next_pic);
s->linesize = s1->linesize;
s->uvlinesize = s1->uvlinesize;
@@ -177,6 +169,7 @@ int ff_mpv_decode_close(AVCodecContext *avctx)
{
MpegEncContext *s = avctx->priv_data;
+ ff_refstruct_pool_uninit(&s->picture_pool);
ff_mpv_common_end(s);
return 0;
}
@@ -190,9 +183,9 @@ int ff_mpv_common_frame_size_change(MpegEncContext *s)
ff_mpv_free_context_frame(s);
- s->last_pic.ptr =
- s->next_pic.ptr =
- s->cur_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->last_pic);
+ ff_mpv_unref_picture(&s->next_pic);
+ ff_mpv_unref_picture(&s->cur_pic);
if ((s->width || s->height) &&
(err = av_image_check_size(s->width, s->height, 0, s->avctx)) < 0)
@@ -228,14 +221,12 @@ int ff_mpv_common_frame_size_change(MpegEncContext *s)
static int alloc_picture(MpegEncContext *s, MPVWorkPicture *dst, int reference)
{
AVCodecContext *avctx = s->avctx;
- int idx = ff_find_unused_picture(s->avctx, s->picture, 0);
- MPVPicture *pic;
+ MPVPicture *pic = ff_refstruct_pool_get(s->picture_pool);
int ret;
- if (idx < 0)
- return idx;
+ if (!pic)
+ return AVERROR(ENOMEM);
- pic = &s->picture[idx];
dst->ptr = pic;
pic->tf.f = pic->f;
@@ -368,22 +359,7 @@ int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx)
return AVERROR_BUG;
}
- /* mark & release old frames */
- if (s->pict_type != AV_PICTURE_TYPE_B && s->last_pic.ptr &&
- s->last_pic.ptr != s->next_pic.ptr &&
- s->last_pic.ptr->f->buf[0]) {
- ff_mpeg_unref_picture(s->last_pic.ptr);
- }
-
- /* release non reference/forgotten frames */
- for (int i = 0; i < MAX_PICTURE_COUNT; i++) {
- if (!s->picture[i].reference ||
- (&s->picture[i] != s->last_pic.ptr &&
- &s->picture[i] != s->next_pic.ptr)) {
- ff_mpeg_unref_picture(&s->picture[i]);
- }
- }
-
+ ff_mpv_unref_picture(&s->cur_pic);
ret = alloc_picture(s, &s->cur_pic,
s->pict_type != AV_PICTURE_TYPE_B && !s->droppable);
if (ret < 0)
@@ -495,12 +471,6 @@ void ff_mpeg_flush(AVCodecContext *avctx)
{
MpegEncContext *const s = avctx->priv_data;
- if (!s->picture)
- return;
-
- for (int i = 0; i < MAX_PICTURE_COUNT; i++)
- ff_mpeg_unref_picture(&s->picture[i]);
-
ff_mpv_unref_picture(&s->cur_pic);
ff_mpv_unref_picture(&s->last_pic);
ff_mpv_unref_picture(&s->next_pic);
diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c
index ce827712af..e65a30f96f 100644
--- a/libavcodec/mpegvideo_enc.c
+++ b/libavcodec/mpegvideo_enc.c
@@ -75,6 +75,7 @@
#include "wmv2enc.h"
#include "rv10enc.h"
#include "packet_internal.h"
+#include "refstruct.h"
#include <limits.h>
#include "sp5x.h"
@@ -821,7 +822,8 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
!FF_ALLOCZ_TYPED_ARRAY(s->q_inter_matrix16, 32) ||
!FF_ALLOCZ_TYPED_ARRAY(s->input_picture, MAX_B_FRAMES + 1) ||
!FF_ALLOCZ_TYPED_ARRAY(s->reordered_input_picture, MAX_B_FRAMES + 1) ||
- !(s->new_pic = av_frame_alloc()))
+ !(s->new_pic = av_frame_alloc()) ||
+ !(s->picture_pool = ff_mpv_alloc_pic_pool()))
return AVERROR(ENOMEM);
/* Allocate MV tables; the MV and MB tables will be copied
@@ -992,7 +994,14 @@ av_cold int ff_mpv_encode_end(AVCodecContext *avctx)
ff_rate_control_uninit(s);
ff_mpv_common_end(s);
+ ff_refstruct_pool_uninit(&s->picture_pool);
+ if (s->input_picture && s->reordered_input_picture) {
+ for (int i = 0; i < MAX_B_FRAMES + 1; i++) {
+ ff_refstruct_unref(&s->input_picture[i]);
+ ff_refstruct_unref(&s->reordered_input_picture[i]);
+ }
+ }
for (i = 0; i < FF_ARRAY_ELEMS(s->tmp_frames); i++)
av_frame_free(&s->tmp_frames[i]);
@@ -1131,12 +1140,14 @@ static int load_input_picture(MpegEncContext *s, const AVFrame *pic_arg)
{
MPVPicture *pic = NULL;
int64_t pts;
- int i, display_picture_number = 0, ret;
+ int display_picture_number = 0, ret;
int encoding_delay = s->max_b_frames ? s->max_b_frames
: (s->low_delay ? 0 : 1);
int flush_offset = 1;
int direct = 1;
+ av_assert1(!s->input_picture[0]);
+
if (pic_arg) {
pts = pic_arg->pts;
display_picture_number = s->input_picture_number++;
@@ -1182,16 +1193,13 @@ static int load_input_picture(MpegEncContext *s, const AVFrame *pic_arg)
ff_dlog(s->avctx, "%d %d %"PTRDIFF_SPECIFIER" %"PTRDIFF_SPECIFIER"\n", pic_arg->linesize[0],
pic_arg->linesize[1], s->linesize, s->uvlinesize);
- i = ff_find_unused_picture(s->avctx, s->picture, direct);
- if (i < 0)
- return i;
-
- pic = &s->picture[i];
- pic->reference = 3;
+ pic = ff_refstruct_pool_get(s->picture_pool);
+ if (!pic)
+ return AVERROR(ENOMEM);
if (direct) {
if ((ret = av_frame_ref(pic->f, pic_arg)) < 0)
- return ret;
+ goto fail;
pic->shared = 1;
} else {
ret = prepare_picture(s, pic->f, pic_arg);
@@ -1241,17 +1249,17 @@ static int load_input_picture(MpegEncContext *s, const AVFrame *pic_arg)
pic->display_picture_number = display_picture_number;
pic->f->pts = pts; // we set this here to avoid modifying pic_arg
- } else {
- /* Flushing: When we have not received enough input frames,
- * ensure s->input_picture[0] contains the first picture */
+ } else if (!s->reordered_input_picture[1]) {
+ /* Flushing: When the above check is true, the encoder is about to run
+ * out of frames to encode. Check if there are input_pictures left;
+ * if so, ensure s->input_picture[0] contains the first picture.
+ * A flush_offset != 1 will only happen if we did not receive enough
+ * input frames. */
for (flush_offset = 0; flush_offset < encoding_delay + 1; flush_offset++)
if (s->input_picture[flush_offset])
break;
- if (flush_offset <= 1)
- flush_offset = 1;
- else
- encoding_delay = encoding_delay - flush_offset + 1;
+ encoding_delay -= flush_offset - 1;
}
/* shift buffer entries */
@@ -1262,7 +1270,7 @@ static int load_input_picture(MpegEncContext *s, const AVFrame *pic_arg)
return 0;
fail:
- ff_mpeg_unref_picture(pic);
+ ff_refstruct_unref(&pic);
return ret;
}
@@ -1475,8 +1483,10 @@ fail:
/**
* Determines whether an input picture is discarded or not
* and if not determines the length of the next chain of B frames
- * and puts these pictures (including the P frame) into
+ * and moves these pictures (including the P frame) into
* reordered_input_picture.
+ * input_picture[0] is always NULL when exiting this function, even on error;
+ * reordered_input_picture[0] is always NULL when exiting this function on error.
*/
static int set_bframe_chain_length(MpegEncContext *s)
{
@@ -1490,7 +1500,7 @@ static int set_bframe_chain_length(MpegEncContext *s)
s->next_pic.ptr &&
skip_check(s, s->input_picture[0], s->next_pic.ptr)) {
// FIXME check that the gop check above is +-1 correct
- ff_mpeg_unref_picture(s->input_picture[0]);
+ ff_refstruct_unref(&s->input_picture[0]);
ff_vbv_update(s, 0);
@@ -1501,6 +1511,7 @@ static int set_bframe_chain_length(MpegEncContext *s)
if (/*s->picture_in_gop_number >= s->gop_size ||*/
!s->next_pic.ptr || s->intra_only) {
s->reordered_input_picture[0] = s->input_picture[0];
+ s->input_picture[0] = NULL;
s->reordered_input_picture[0]->f->pict_type = AV_PICTURE_TYPE_I;
s->reordered_input_picture[0]->coded_picture_number =
s->coded_picture_number++;
@@ -1555,7 +1566,7 @@ static int set_bframe_chain_length(MpegEncContext *s)
} else if (s->b_frame_strategy == 2) {
b_frames = estimate_best_b_count(s);
if (b_frames < 0) {
- ff_mpeg_unref_picture(s->input_picture[0]);
+ ff_refstruct_unref(&s->input_picture[0]);
return b_frames;
}
}
@@ -1589,12 +1600,14 @@ static int set_bframe_chain_length(MpegEncContext *s)
b_frames--;
s->reordered_input_picture[0] = s->input_picture[b_frames];
+ s->input_picture[b_frames] = NULL;
if (s->reordered_input_picture[0]->f->pict_type != AV_PICTURE_TYPE_I)
s->reordered_input_picture[0]->f->pict_type = AV_PICTURE_TYPE_P;
s->reordered_input_picture[0]->coded_picture_number =
s->coded_picture_number++;
for (int i = 0; i < b_frames; i++) {
s->reordered_input_picture[i + 1] = s->input_picture[i];
+ s->input_picture[i] = NULL;
s->reordered_input_picture[i + 1]->f->pict_type =
AV_PICTURE_TYPE_B;
s->reordered_input_picture[i + 1]->coded_picture_number =
@@ -1609,11 +1622,14 @@ static int select_input_picture(MpegEncContext *s)
{
int ret;
+ av_assert1(!s->reordered_input_picture[0]);
+
for (int i = 1; i <= MAX_B_FRAMES; i++)
s->reordered_input_picture[i - 1] = s->reordered_input_picture[i];
s->reordered_input_picture[MAX_B_FRAMES] = NULL;
ret = set_bframe_chain_length(s);
+ av_assert1(!s->input_picture[0]);
if (ret < 0)
return ret;
@@ -1643,6 +1659,7 @@ static int select_input_picture(MpegEncContext *s)
}
}
s->cur_pic.ptr = s->reordered_input_picture[0];
+ s->reordered_input_picture[0] = NULL;
av_assert1(s->mb_width == s->buffer_pools.alloc_mb_width);
av_assert1(s->mb_height == s->buffer_pools.alloc_mb_height);
av_assert1(s->mb_stride == s->buffer_pools.alloc_mb_stride);
@@ -1657,7 +1674,7 @@ static int select_input_picture(MpegEncContext *s)
}
return 0;
fail:
- ff_mpeg_unref_picture(s->reordered_input_picture[0]);
+ ff_refstruct_unref(&s->reordered_input_picture[0]);
return ret;
}
@@ -1720,13 +1737,6 @@ static void update_noise_reduction(MpegEncContext *s)
static void frame_start(MpegEncContext *s)
{
- /* mark & release old frames */
- if (s->pict_type != AV_PICTURE_TYPE_B && s->last_pic.ptr &&
- s->last_pic.ptr != s->next_pic.ptr &&
- s->last_pic.ptr->f->buf[0]) {
- ff_mpv_unref_picture(&s->last_pic);
- }
-
s->cur_pic.ptr->f->pict_type = s->pict_type;
if (s->pict_type != AV_PICTURE_TYPE_B) {
@@ -1747,6 +1757,8 @@ int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
int i, stuffing_count, ret;
int context_count = s->slice_context_count;
+ ff_mpv_unref_picture(&s->cur_pic);
+
s->vbv_ignore_qmax = 0;
s->picture_in_gop_number++;
@@ -1973,11 +1985,7 @@ vbv_retry:
s->frame_bits = 0;
}
- /* release non-reference frames */
- for (i = 0; i < MAX_PICTURE_COUNT; i++) {
- if (!s->picture[i].reference)
- ff_mpeg_unref_picture(&s->picture[i]);
- }
+ ff_mpv_unref_picture(&s->cur_pic);
av_assert1((s->frame_bits & 7) == 0);
diff --git a/libavcodec/mpegvideodec.h b/libavcodec/mpegvideodec.h
index 35e9081d2c..6100364715 100644
--- a/libavcodec/mpegvideodec.h
+++ b/libavcodec/mpegvideodec.h
@@ -44,8 +44,10 @@
* Initialize the given MpegEncContext for decoding.
* the changed fields will not depend upon
* the prior state of the MpegEncContext.
+ *
+ * Also initialize the picture pool.
*/
-void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx);
+int ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx);
int ff_mpv_common_frame_size_change(MpegEncContext *s);
diff --git a/libavcodec/msmpeg4dec.c b/libavcodec/msmpeg4dec.c
index 20d735a152..4143c46c15 100644
--- a/libavcodec/msmpeg4dec.c
+++ b/libavcodec/msmpeg4dec.c
@@ -852,7 +852,8 @@ const FFCodec ff_msmpeg4v1_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.p.max_lowres = 3,
};
@@ -866,7 +867,8 @@ const FFCodec ff_msmpeg4v2_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.p.max_lowres = 3,
};
@@ -880,7 +882,8 @@ const FFCodec ff_msmpeg4v3_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.p.max_lowres = 3,
};
@@ -894,6 +897,7 @@ const FFCodec ff_wmv1_decoder = {
FF_CODEC_DECODE_CB(ff_h263_decode_frame),
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DRAW_HORIZ_BAND | AV_CODEC_CAP_DR1,
- .caps_internal = FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP |
+ FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM,
.p.max_lowres = 3,
};
diff --git a/libavcodec/rv10.c b/libavcodec/rv10.c
index 6ece6a5a25..3dcee0a065 100644
--- a/libavcodec/rv10.c
+++ b/libavcodec/rv10.c
@@ -364,7 +364,9 @@ static av_cold int rv10_decode_init(AVCodecContext *avctx)
avctx->coded_height, 0, avctx)) < 0)
return ret;
- ff_mpv_decode_init(s, avctx);
+ ret = ff_mpv_decode_init(s, avctx);
+ if (ret < 0)
+ return ret;
s->out_format = FMT_H263;
@@ -645,7 +647,7 @@ static int rv10_decode_frame(AVCodecContext *avctx, AVFrame *pict,
}
// so we can detect if frame_end was not called (find some nicer solution...)
- s->cur_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->cur_pic);
}
return avpkt->size;
@@ -662,6 +664,7 @@ const FFCodec ff_rv10_decoder = {
.close = ff_mpv_decode_close,
.p.capabilities = AV_CODEC_CAP_DR1,
.p.max_lowres = 3,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
};
const FFCodec ff_rv20_decoder = {
@@ -676,4 +679,5 @@ const FFCodec ff_rv20_decoder = {
.p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY,
.flush = ff_mpeg_flush,
.p.max_lowres = 3,
+ .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
};
diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c
index 90296d7de3..94d5ee30df 100644
--- a/libavcodec/rv34.c
+++ b/libavcodec/rv34.c
@@ -1508,7 +1508,9 @@ av_cold int ff_rv34_decode_init(AVCodecContext *avctx)
MpegEncContext *s = &r->s;
int ret;
- ff_mpv_decode_init(s, avctx);
+ ret = ff_mpv_decode_init(s, avctx);
+ if (ret < 0)
+ return ret;
s->out_format = FMT_H263;
avctx->pix_fmt = AV_PIX_FMT_YUV420P;
@@ -1630,7 +1632,7 @@ int ff_rv34_decode_frame(AVCodecContext *avctx, AVFrame *pict,
if (s->next_pic.ptr) {
if ((ret = av_frame_ref(pict, s->next_pic.ptr->f)) < 0)
return ret;
- s->next_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->next_pic);
*got_picture_ptr = 1;
}
diff --git a/libavcodec/svq1enc.c b/libavcodec/svq1enc.c
index 9631fa243d..9c9be8c6b3 100644
--- a/libavcodec/svq1enc.c
+++ b/libavcodec/svq1enc.c
@@ -60,7 +60,6 @@ typedef struct SVQ1EncContext {
* else, the idea is to make the motion estimation eventually independent
* of MpegEncContext, so this will be removed then. */
MpegEncContext m;
- MPVPicture cur_pic, last_pic;
AVCodecContext *avctx;
MECmpContext mecc;
HpelDSPContext hdsp;
@@ -327,8 +326,6 @@ static int svq1_encode_plane(SVQ1EncContext *s, int plane,
if (s->pict_type == AV_PICTURE_TYPE_P) {
s->m.avctx = s->avctx;
- s->m.cur_pic.ptr = &s->cur_pic;
- s->m.last_pic.ptr = &s->last_pic;
s->m.last_pic.data[0] = ref_plane;
s->m.linesize =
s->m.last_pic.linesize[0] =
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
index ea4f015acd..0fac163c4f 100644
--- a/libavcodec/vc1dec.c
+++ b/libavcodec/vc1dec.c
@@ -461,7 +461,9 @@ av_cold int ff_vc1_decode_init(AVCodecContext *avctx)
if (ret < 0)
return ret;
- ff_mpv_decode_init(s, avctx);
+ ret = ff_mpv_decode_init(s, avctx);
+ if (ret < 0)
+ return ret;
avctx->pix_fmt = vc1_get_format(avctx);
@@ -846,7 +848,7 @@ static int vc1_decode_frame(AVCodecContext *avctx, AVFrame *pict,
if (s->low_delay == 0 && s->next_pic.ptr) {
if ((ret = av_frame_ref(pict, s->next_pic.ptr->f)) < 0)
return ret;
- s->next_pic.ptr = NULL;
+ ff_mpv_unref_picture(&s->next_pic);
*got_frame = 1;
}
@@ -997,7 +999,7 @@ static int vc1_decode_frame(AVCodecContext *avctx, AVFrame *pict,
if (s->context_initialized &&
(s->width != avctx->coded_width ||
s->height != avctx->coded_height)) {
- ff_vc1_decode_end(avctx);
+ vc1_decode_reset(avctx);
}
if (!s->context_initialized) {
--
2.40.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2024-04-29 21:19 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-04-29 0:28 [FFmpeg-devel] [PATCH 01/14] avcodec/get_buffer: Remove redundant check Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 02/14] avcodec/mpegpicture: Store linesize in ScratchpadContext Andreas Rheinhardt
2024-04-29 15:09 ` [FFmpeg-devel] [PATCH v2 2/14] " Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 03/14] avcodec/mpegvideo_dec: Sync linesize and uvlinesize between threads Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 04/14] avcodec/mpegvideo_dec: Factor allocating dummy frames out Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 05/14] avcodec/mpegpicture: Mark dummy frames as such Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 06/14] avcodec/mpeg12dec: Allocate dummy frames for non-I fields Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 07/14] avcodec/mpegvideo_motion: Remove dead checks for existence of reference Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 08/14] avcodec/mpegvideo_motion: Optimize check away Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 09/14] " Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 10/14] avcodec/mpegvideo_motion: Avoid constant function argument Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 11/14] avcodec/msmpeg4enc: Only calculate coded_cbp when used Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 12/14] avcodec/mpegvideo: Only allocate coded_block when needed Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 13/14] avcodec/mpegvideo: Don't reset coded_block unnecessarily Andreas Rheinhardt
2024-04-29 0:35 ` [FFmpeg-devel] [PATCH 14/14] avcodec/mpegvideo: Only allocate cbp_table, pred_dir_table when needed Andreas Rheinhardt
2024-04-29 21:13 ` [FFmpeg-devel] [PATCH 15/57] avcodec/mpegpicture: Always reset motion val buffer Andreas Rheinhardt
2024-04-29 21:13 ` [FFmpeg-devel] [PATCH 16/57] avcodec/mpegvideo: Redo aligning mb_height for VC-1 Andreas Rheinhardt
2024-04-29 21:13 ` [FFmpeg-devel] [PATCH 17/57] avcodec/mpegvideo, mpegpicture: Add buffer pool Andreas Rheinhardt
2024-04-30 20:40 ` Michael Niedermayer
2024-04-30 20:58 ` James Almer
2024-04-30 21:07 ` Andreas Rheinhardt
2024-04-30 21:06 ` Andreas Rheinhardt
2024-04-29 21:13 ` [FFmpeg-devel] [PATCH 18/57] avcodec/mpegpicture: Reindent after the previous commit Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 19/57] avcodec/mpegpicture: Use RefStruct-pool API Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 20/57] avcodec/mpegvideo: Shorten variable names Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 21/57] avcodec/mpegpicture: Reduce value of MAX_PLANES define Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 22/57] avcodec/mpegpicture: Cache AVFrame.data and linesize values Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 23/57] avcodec/rv30, rv34, rv40: Avoid indirection Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 24/57] avcodec/mpegvideo: Add const where appropriate Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 25/57] avcodec/vc1_pred: Remove unused function parameter Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 26/57] avcodec/mpegpicture: Improve error messages and code Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 27/57] avcodec/mpegpicture: Split ff_alloc_picture() into check and alloc part Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 28/57] avcodec/mpegvideo_enc: Pass AVFrame*, not Picture* to alloc_picture() Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 29/57] avcodec/mpegvideo_enc: Move copying properties " Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 30/57] avcodec/mpegpicture: Rename Picture->MPVPicture Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 31/57] avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic Andreas Rheinhardt
2024-04-30 18:57 ` Michael Niedermayer
2024-04-30 19:07 ` Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 32/57] avcodec/error_resilience: Deduplicate cleanup code Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 33/57] avcodec/mpegvideo_enc: Factor setting length of B frame chain out Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 34/57] avcodec/mpegvideo_enc: Return early when getting length of B frame chain Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 35/57] avcodec/mpegvideo_enc: Reindentation Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 36/57] avcodec/mpeg12dec: Don't initialize inter tables for IPU Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 37/57] avcodec/mpeg12dec: Only initialize IDCT " Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 38/57] avcodec/mpeg12dec: Remove write-only assignment Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 39/57] avcodec/mpeg12dec: Set out_format only once Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 40/57] avformat/riff: Declare VCR2 to be MPEG-2 Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 41/57] avcodec/mpegvideo_dec: Add close function for mpegvideo-decoders Andreas Rheinhardt
2024-04-29 21:14 ` Andreas Rheinhardt [this message]
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 43/57] avcodec/mpeg4videoenc: Avoid branch for writing stuffing Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 44/57] avcodec/mpeg4videoenc: Simplify writing startcodes Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 45/57] avcodec/mpegpicture: Use ThreadProgress instead of ThreadFrame API Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 46/57] avcodec/mpegpicture: Avoid loop and branch when setting motion_val Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 47/57] avcodec/mpegpicture: Use union for b_scratchpad and rd_scratchpad Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 48/57] avcodec/mpegpicture: Avoid MotionEstContext in ff_mpeg_framesize_alloc() Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 49/57] avcodec/mpegvideo_enc: Unify initializing PutBitContexts Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 50/57] avcodec/mpeg12enc: Simplify writing startcodes Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 51/57] avcodec/mpegvideo_dec: Simplify check for "does pic exist?" Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 52/57] avcodec/mpegvideo_dec: Don't sync encoder-only coded_picture_number Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 53/57] avcodec/mpeg12dec: Pass Mpeg1Context* in mpeg_field_start() Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 54/57] avcodec/mpeg12dec: Don't initialize inter_scantable Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 55/57] avcodec/mpegvideo: Remove pblocks Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 56/57] avcodec/mpegvideo: Use enum for msmpeg4_version Andreas Rheinhardt
2024-04-29 21:14 ` [FFmpeg-devel] [PATCH 57/57] avcodec/ituh263enc: Remove redundant check Andreas Rheinhardt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=AS8P250MB0744B0601994778005D45B848F1B2@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM \
--to=andreas.rheinhardt@outlook.com \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git