From 77ab1578003a88587c83ef9d8cc52ca837e6337c Mon Sep 17 00:00:00 2001
From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Date: Sun, 2 Mar 2025 02:55:04 +0100
Subject: [PATCH 23/77] avcodec/mpegvideo: Move sequence-level properties to
MPVMainEncContext
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
libavcodec/mpeg12enc.c | 14 ++++---
libavcodec/mpeg12enc.h | 4 +-
libavcodec/mpegvideo.h | 15 -------
libavcodec/mpegvideo_enc.c | 86 +++++++++++++++++++-------------------
libavcodec/mpegvideoenc.h | 16 +++++++
libavcodec/ratecontrol.c | 4 +-
6 files changed, 71 insertions(+), 68 deletions(-)
diff --git a/libavcodec/mpeg12enc.c b/libavcodec/mpeg12enc.c
index d84d628563..ed2c63086f 100644
--- a/libavcodec/mpeg12enc.c
+++ b/libavcodec/mpeg12enc.c
@@ -144,9 +144,9 @@ static void put_header(MpegEncContext *s, uint32_t header)
}
/* put sequence header if needed */
-static void mpeg1_encode_sequence_header(MpegEncContext *s)
+static void mpeg1_encode_sequence_header(MPEG12EncContext *mpeg12)
{
- MPEG12EncContext *const mpeg12 = (MPEG12EncContext*)s;
+ MpegEncContext *const s = &mpeg12->mpeg.s;
unsigned int vbv_buffer_size, fps, v;
int constraint_parameter_flag;
AVRational framerate = ff_mpeg12_frame_rate_tab[mpeg12->frame_rate_index];
@@ -299,7 +299,7 @@ static void mpeg1_encode_sequence_header(MpegEncContext *s)
put_bits(&s->pb, 6, (uint32_t)((time_code / fps) % 60));
put_bits(&s->pb, 6, (uint32_t)((time_code % fps)));
put_bits(&s->pb, 1, !!(s->avctx->flags & AV_CODEC_FLAG_CLOSED_GOP) ||
- s->intra_only || !mpeg12->gop_picture_number);
+ mpeg12->mpeg.intra_only || !mpeg12->gop_picture_number);
put_bits(&s->pb, 1, 0); // broken link
}
@@ -333,11 +333,13 @@ void ff_mpeg1_encode_slice_header(MpegEncContext *s)
put_bits(&s->pb, 1, 0);
}
-void ff_mpeg1_encode_picture_header(MpegEncContext *s)
+void ff_mpeg1_encode_picture_header(MPVMainEncContext *const m)
{
- MPEG12EncContext *const mpeg12 = (MPEG12EncContext*)s;
+ MPEG12EncContext *const mpeg12 = (MPEG12EncContext*)m;
+ MpegEncContext *const s = &m->s;
const AVFrameSideData *side_data;
- mpeg1_encode_sequence_header(s);
+
+ mpeg1_encode_sequence_header(mpeg12);
/* MPEG-1 picture header */
put_header(s, PICTURE_START_CODE);
diff --git a/libavcodec/mpeg12enc.h b/libavcodec/mpeg12enc.h
index 908971905d..fa1504177a 100644
--- a/libavcodec/mpeg12enc.h
+++ b/libavcodec/mpeg12enc.h
@@ -24,10 +24,10 @@
#include <stdint.h>
-#include "mpegvideo.h"
+#include "mpegvideoenc.h"
#include "mpegvideodata.h"
-void ff_mpeg1_encode_picture_header(MpegEncContext *s);
+void ff_mpeg1_encode_picture_header(MPVMainEncContext *m);
void ff_mpeg1_encode_mb(MpegEncContext *s, int16_t block[8][64],
int motion_x, int motion_y);
void ff_mpeg1_encode_slice_header(MpegEncContext *s);
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
index b532088730..3259c1117c 100644
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -91,8 +91,6 @@ typedef struct MpegEncContext {
void *private_ctx;
/* the following parameters must be initialized before encoding */
int width, height;///< picture size. must be a multiple of 16
- int gop_size;
- int intra_only; ///< if true, only intra pictures are generated
int64_t bit_rate; ///< wanted bit rate
enum OutputFormat out_format; ///< output format
int h263_pred; ///< use MPEG-4/H.263 ac/dc predictions
@@ -114,10 +112,7 @@ typedef struct MpegEncContext {
/* sequence parameters */
int context_initialized;
- int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
- int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
int picture_number; //FIXME remove, unclear definition
- int picture_in_gop_number; ///< 0-> first pic in gop, ...
int mb_width, mb_height; ///< number of MBs horizontally & vertically
int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
@@ -131,16 +126,6 @@ typedef struct MpegEncContext {
BufferPoolContext buffer_pools;
- int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_send_frame()
- /**
- * pts difference between the first and second input frame, used for
- * calculating dts of the first frame when there's a delay */
- int64_t dts_delta;
- /**
- * reordered pts to be used as dts for the next output frame when there's
- * a delay */
- int64_t reordered_pts;
-
/** bit output */
PutBitContext pb;
diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c
index 0a9c18ff79..cc3c7a745f 100644
--- a/libavcodec/mpegvideo_enc.c
+++ b/libavcodec/mpegvideo_enc.c
@@ -291,9 +291,6 @@ static av_cold void mpv_encode_defaults(MpegEncContext *s)
s->y_dc_scale_table =
s->c_dc_scale_table = ff_mpeg1_dc_scale_table;
}
-
- s->input_picture_number = 0;
- s->picture_in_gop_number = 0;
}
av_cold void ff_dct_encode_init(MpegEncContext *s)
@@ -354,9 +351,10 @@ static av_cold int me_cmp_init(MPVMainEncContext *const m, AVCodecContext *avctx
}
#define ALLOCZ_ARRAYS(p, mult, numb) ((p) = av_calloc(numb, mult * sizeof(*(p))))
-static av_cold int init_matrices(MpegEncContext *s, AVCodecContext *avctx)
+static av_cold int init_matrices(MPVMainEncContext *const m, AVCodecContext *avctx)
{
- const int nb_matrices = 1 + (s->out_format == FMT_MJPEG) + !s->intra_only;
+ MpegEncContext *const s = &m->s;
+ const int nb_matrices = 1 + (s->out_format == FMT_MJPEG) + !m->intra_only;
const uint16_t *intra_matrix, *inter_matrix;
int ret;
@@ -368,14 +366,14 @@ static av_cold int init_matrices(MpegEncContext *s, AVCodecContext *avctx)
s->q_chroma_intra_matrix = s->q_intra_matrix + 32;
s->q_chroma_intra_matrix16 = s->q_intra_matrix16 + 32;
// No need to set q_inter_matrix
- av_assert1(s->intra_only);
+ av_assert1(m->intra_only);
// intra_matrix, chroma_intra_matrix will be set later for MJPEG.
return 0;
} else {
s->q_chroma_intra_matrix = s->q_intra_matrix;
s->q_chroma_intra_matrix16 = s->q_intra_matrix16;
}
- if (!s->intra_only) {
+ if (!m->intra_only) {
s->q_inter_matrix = s->q_intra_matrix + 32;
s->q_inter_matrix16 = s->q_intra_matrix16 + 32;
}
@@ -460,7 +458,7 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
avctx->gop_size, 600);
avctx->gop_size = 600;
}
- s->gop_size = avctx->gop_size;
+ m->gop_size = avctx->gop_size;
s->avctx = avctx;
if (avctx->max_b_frames > MPVENC_MAX_B_FRAMES) {
av_log(avctx, AV_LOG_ERROR, "Too many B-frames requested, maximum "
@@ -499,13 +497,13 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
av_log(avctx, AV_LOG_ERROR, "intra dc precision too large\n");
return AVERROR(EINVAL);
}
- s->user_specified_pts = AV_NOPTS_VALUE;
+ m->user_specified_pts = AV_NOPTS_VALUE;
- if (s->gop_size <= 1) {
- s->intra_only = 1;
- s->gop_size = 12;
+ if (m->gop_size <= 1) {
+ m->intra_only = 1;
+ m->gop_size = 12;
} else {
- s->intra_only = 0;
+ m->intra_only = 0;
}
/* Fixed QSCALE */
@@ -753,14 +751,14 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
case AV_CODEC_ID_MJPEG:
case AV_CODEC_ID_AMV:
s->out_format = FMT_MJPEG;
- s->intra_only = 1; /* force intra only for jpeg */
+ m->intra_only = 1; /* force intra only for jpeg */
avctx->delay = 0;
s->low_delay = 1;
break;
#endif
case AV_CODEC_ID_SPEEDHQ:
s->out_format = FMT_SPEEDHQ;
- s->intra_only = 1; /* force intra only for SHQ */
+ m->intra_only = 1; /* force intra only for SHQ */
avctx->delay = 0;
s->low_delay = 1;
break;
@@ -885,7 +883,7 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
s->frame_reconstruction_bitfield = (1 << AV_PICTURE_TYPE_I) |
(1 << AV_PICTURE_TYPE_P) |
(1 << AV_PICTURE_TYPE_B);
- } else if (!s->intra_only) {
+ } else if (!m->intra_only) {
s->frame_reconstruction_bitfield = (1 << AV_PICTURE_TYPE_I) |
(1 << AV_PICTURE_TYPE_P);
} else {
@@ -916,7 +914,7 @@ av_cold int ff_mpv_encode_init(AVCodecContext *avctx)
!(s->picture_pool = ff_mpv_alloc_pic_pool(0)))
return AVERROR(ENOMEM);
- ret = init_matrices(s, avctx);
+ ret = init_matrices(m, avctx);
if (ret < 0)
return ret;
@@ -1259,11 +1257,11 @@ static int load_input_picture(MPVMainEncContext *const m, const AVFrame *pic_arg
if (pic_arg) {
pts = pic_arg->pts;
- display_picture_number = s->input_picture_number++;
+ display_picture_number = m->input_picture_number++;
if (pts != AV_NOPTS_VALUE) {
- if (s->user_specified_pts != AV_NOPTS_VALUE) {
- int64_t last = s->user_specified_pts;
+ if (m->user_specified_pts != AV_NOPTS_VALUE) {
+ int64_t last = m->user_specified_pts;
if (pts <= last) {
av_log(s->avctx, AV_LOG_ERROR,
@@ -1273,13 +1271,13 @@ static int load_input_picture(MPVMainEncContext *const m, const AVFrame *pic_arg
}
if (!s->low_delay && display_picture_number == 1)
- s->dts_delta = pts - last;
+ m->dts_delta = pts - last;
}
- s->user_specified_pts = pts;
+ m->user_specified_pts = pts;
} else {
- if (s->user_specified_pts != AV_NOPTS_VALUE) {
- s->user_specified_pts =
- pts = s->user_specified_pts + 1;
+ if (m->user_specified_pts != AV_NOPTS_VALUE) {
+ m->user_specified_pts =
+ pts = m->user_specified_pts + 1;
av_log(s->avctx, AV_LOG_INFO,
"Warning: AVFrame.pts=? trying to guess (%"PRId64")\n",
pts);
@@ -1612,7 +1610,7 @@ static int set_bframe_chain_length(MPVMainEncContext *const m)
/* set next picture type & ordering */
if (m->frame_skip_threshold || m->frame_skip_factor) {
- if (s->picture_in_gop_number < s->gop_size &&
+ if (m->picture_in_gop_number < m->gop_size &&
s->next_pic.ptr &&
skip_check(m, s->input_picture[0], s->next_pic.ptr)) {
// FIXME check that the gop check above is +-1 correct
@@ -1624,13 +1622,13 @@ static int set_bframe_chain_length(MPVMainEncContext *const m)
}
}
- if (/*s->picture_in_gop_number >= s->gop_size ||*/
- !s->next_pic.ptr || s->intra_only) {
+ if (/* m->picture_in_gop_number >= m->gop_size || */
+ !s->next_pic.ptr || m->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++;
+ m->coded_picture_number++;
} else {
int b_frames = 0;
@@ -1700,10 +1698,10 @@ static int set_bframe_chain_length(MPVMainEncContext *const m)
"warning, too many B-frames in a row\n");
}
- if (s->picture_in_gop_number + b_frames >= s->gop_size) {
+ if (m->picture_in_gop_number + b_frames >= m->gop_size) {
if ((s->mpv_flags & FF_MPV_FLAG_STRICT_GOP) &&
- s->gop_size > s->picture_in_gop_number) {
- b_frames = s->gop_size - s->picture_in_gop_number - 1;
+ m->gop_size > m->picture_in_gop_number) {
+ b_frames = m->gop_size - m->picture_in_gop_number - 1;
} else {
if (s->avctx->flags & AV_CODEC_FLAG_CLOSED_GOP)
b_frames = 0;
@@ -1720,14 +1718,14 @@ static int set_bframe_chain_length(MPVMainEncContext *const m)
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++;
+ m->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 =
- s->coded_picture_number++;
+ m->coded_picture_number++;
}
}
@@ -1794,11 +1792,13 @@ fail:
return ret;
}
-static void frame_end(MpegEncContext *s)
+static void frame_end(MPVMainEncContext *const m)
{
+ MpegEncContext *const s = &m->s;
+
if (s->unrestricted_mv &&
s->cur_pic.reference &&
- !s->intra_only) {
+ !m->intra_only) {
int hshift = s->chroma_x_shift;
int vshift = s->chroma_y_shift;
s->mpvencdsp.draw_edges(s->cur_pic.data[0],
@@ -1878,7 +1878,7 @@ int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
s->vbv_ignore_qmax = 0;
- s->picture_in_gop_number++;
+ m->picture_in_gop_number++;
ret = load_input_picture(m, pic_arg);
if (ret < 0)
@@ -1923,7 +1923,7 @@ vbv_retry:
if (ret < 0)
return -1;
- frame_end(s);
+ frame_end(m);
if ((CONFIG_MJPEG_ENCODER || CONFIG_AMV_ENCODER) && s->out_format == FMT_MJPEG)
ff_mjpeg_encode_picture_trailer(&s->pb, s->header_bits);
@@ -2066,10 +2066,10 @@ vbv_retry:
pkt->duration = s->cur_pic.ptr->f->duration;
if (!s->low_delay && s->pict_type != AV_PICTURE_TYPE_B) {
if (!s->cur_pic.ptr->coded_picture_number)
- pkt->dts = pkt->pts - s->dts_delta;
+ pkt->dts = pkt->pts - m->dts_delta;
else
- pkt->dts = s->reordered_pts;
- s->reordered_pts = pkt->pts;
+ pkt->dts = m->reordered_pts;
+ m->reordered_pts = pkt->pts;
} else
pkt->dts = pkt->pts;
@@ -3882,7 +3882,7 @@ static int encode_picture(MPVMainEncContext *const m, const AVPacket *pkt)
s->cur_pic.ptr->f->pict_type = s->pict_type;
if (s->cur_pic.ptr->f->flags & AV_FRAME_FLAG_KEY)
- s->picture_in_gop_number=0;
+ m->picture_in_gop_number = 0;
s->mb_x = s->mb_y = 0;
s->last_bits= put_bits_count(&s->pb);
@@ -3925,7 +3925,7 @@ static int encode_picture(MPVMainEncContext *const m, const AVPacket *pkt)
break;
case FMT_MPEG1:
if (CONFIG_MPEG1VIDEO_ENCODER || CONFIG_MPEG2VIDEO_ENCODER)
- ff_mpeg1_encode_picture_header(s);
+ ff_mpeg1_encode_picture_header(m);
break;
default:
av_assert0(0);
diff --git a/libavcodec/mpegvideoenc.h b/libavcodec/mpegvideoenc.h
index 9f0d63c31e..a76d17f63e 100644
--- a/libavcodec/mpegvideoenc.h
+++ b/libavcodec/mpegvideoenc.h
@@ -39,6 +39,22 @@
typedef struct MPVMainEncContext {
MpegEncContext s; ///< The main slicecontext
+ int intra_only; ///< if true, only intra pictures are generated
+ int gop_size;
+ int picture_in_gop_number; ///< 0-> first pic in gop, ...
+ int input_picture_number; ///< used to set pic->display_picture_number
+ int coded_picture_number; ///< used to set pic->coded_picture_number
+
+ int64_t user_specified_pts; ///< last non-zero pts from user-supplied AVFrame
+ /**
+ * pts difference between the first and second input frame, used for
+ * calculating dts of the first frame when there's a delay */
+ int64_t dts_delta;
+ /**
+ * reordered pts to be used as dts for the next output frame when there's
+ * a delay */
+ int64_t reordered_pts;
+
/// temporary frames used by b_frame_strategy = 2
AVFrame *tmp_frames[MPVENC_MAX_B_FRAMES + 2];
int b_frame_strategy;
diff --git a/libavcodec/ratecontrol.c b/libavcodec/ratecontrol.c
index ebc3a30fc1..4aa358a4e5 100644
--- a/libavcodec/ratecontrol.c
+++ b/libavcodec/ratecontrol.c
@@ -660,7 +660,7 @@ av_cold int ff_rate_control_init(MPVMainEncContext *const m)
double bits = rcc->initial_cplx * (i / 10000.0 + 1.0) * s->mb_num;
RateControlEntry rce;
- if (i % ((s->gop_size + 3) / 4) == 0)
+ if (i % ((m->gop_size + 3) / 4) == 0)
rce.pict_type = AV_PICTURE_TYPE_I;
else if (i % (s->max_b_frames + 1))
rce.pict_type = AV_PICTURE_TYPE_B;
@@ -1033,7 +1033,7 @@ float ff_rate_estimate_qscale(MPVMainEncContext *const m, int dry_run)
av_assert0(q > 0.0);
// FIXME type dependent blur like in 2-pass
- if (pict_type == AV_PICTURE_TYPE_P || s->intra_only) {
+ if (pict_type == AV_PICTURE_TYPE_P || m->intra_only) {
rcc->short_term_qsum *= a->qblur;
rcc->short_term_qcount *= a->qblur;
--
2.45.2