1
0
mirror of https://github.com/mpv-player/mpv synced 2024-12-27 01:22:30 +00:00
mpv/audio/aframe.c
wm4 b9d351f02a Implement backwards playback
See manpage additions. This is a huge hack. You can bet there are shit
tons of bugs. It's literally forcing square pegs into round holes.
Hopefully, the manpage wall of text makes it clear enough that the whole
shit can easily crash and burn. (Although it shouldn't literally crash.
That would be a bug. It possibly _could_ start a fire by entering some
sort of endless loop, not a literal one, just something where it tries
to do work without making progress.)

(Some obvious bugs I simply ignored for this initial version, but
there's a number of potential bugs I can't even imagine. Normal playback
should remain completely unaffected, though.)

How this works is also described in the manpage. Basically, we demux in
reverse, then we decode in reverse, then we render in reverse.

The decoding part is the simplest: just reorder the decoder output. This
weirdly integrates with the timeline/ordered chapter code, which also
has special requirements on feeding the packets to the decoder in a
non-straightforward way (it doesn't conflict, although a bugmessmass
breaks correct slicing of segments, so EDL/ordered chapter playback is
broken in backward direction).

Backward demuxing is pretty involved. In theory, it could be much
easier: simply iterating the usual demuxer output backward. But this
just doesn't fit into our code, so there's a cthulhu nightmare of shit.
To be specific, each stream (audio, video) is reversed separately. At
least this means we can do backward playback within cached content (for
example, you could play backwards in a live stream; on that note, it
disables prefetching, which would lead to losing new live video, but
this could be avoided).

The fuckmess also meant that I didn't bother trying to support
subtitles. Subtitles are a problem because they're "sparse" streams.
They need to be "passively" demuxed: you don't try to read a subtitle
packet, you demux audio and video, and then look whether there was a
subtitle packet. This means to get subtitles for a time range, you need
to know that you demuxed video and audio over this range, which becomes
pretty messy when you demux audio and video backwards separately.

Backward display is the most weird (and potentially buggy) part. To
avoid that we need to touch a LOT of timing code, we negate all
timestamps. The basic idea is that due to the navigation, all
comparisons and subtractions of timestamps keep working, and you don't
need to touch every single of them to "reverse" them.

E.g.:

    bool before = pts_a < pts_b;

would need to be:

    bool before = forward
        ? pts_a < pts_b
        : pts_a > pts_b;

or:

    bool before = pts_a * dir < pts_b * dir;

or if you, as it's implemented now, just do this after decoding:

    pts_a *= dir;
    pts_b *= dir;

and then in the normal timing/renderer code:

    bool before = pts_a < pts_b;

Consequently, we don't need many changes in the latter code. But some
assumptions inhererently true for forward playback may have been broken
anyway. What is mainly needed is fixing places where values are passed
between positive and negative "domains". For example, seeking and
timestamp user display always uses positive timestamps. The main mess is
that it's not obvious which domain a given variable should or does use.

Well, in my tests with a single file, it suddenly started to work when I
did this. I'm honestly surprised that it did, and that I didn't have to
change a single line in the timing code past decoder (just something
minor to make external/cached text subtitles display). I committed it
immediately while avoiding thinking about it. But there really likely
are subtle problems of all sorts.

As far as I'm aware, gstreamer also supports backward playback. When I
looked at this years ago, I couldn't find a way to actually try this,
and I didn't revisit it now. Back then I also read talk slides from the
person who implemented it, and I'm not sure if and which ideas I might
have taken from it. It's possible that the timestamp reversal is
inspired by it, but I didn't check. (I think it claimed that it could
avoid large changes by changing a sign?)

VapourSynth has some sort of reverse function, which provides a backward
view on a video. The function itself is trivial to implement, as
VapourSynth aims to provide random access to video by frame numbers (so
you just request decreasing frame numbers). From what I remember, it
wasn't exactly fluid, but it worked. It's implemented by creating an
index, and seeking to the target on demand, and a bunch of caching. mpv
could use it, but it would either require using VapourSynth as demuxer
and decoder for everything, or replacing the current file every time
something is supposed to be played backwards.

FFmpeg's libavfilter has reversal filters for audio and video. These
require buffering the entire media data of the file, and don't really
fit into mpv's architecture. It could be used by playing a libavfilter
graph that also demuxes, but that's like VapourSynth but worse.
2019-09-19 20:37:04 +02:00

633 lines
19 KiB
C

/*
* This file is part of mpv.
*
* mpv 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.
*
* mpv 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 mpv. If not, see <http://www.gnu.org/licenses/>.
*/
#include <libavutil/frame.h>
#include <libavutil/mem.h>
#include "common/common.h"
#include "chmap.h"
#include "fmt-conversion.h"
#include "format.h"
#include "aframe.h"
struct mp_aframe {
AVFrame *av_frame;
// We support channel layouts different from AVFrame channel masks
struct mp_chmap chmap;
// We support spdif formats, which are allocated as AV_SAMPLE_FMT_S16.
int format;
double pts;
double speed;
};
struct avframe_opaque {
double speed;
};
static void free_frame(void *ptr)
{
struct mp_aframe *frame = ptr;
av_frame_free(&frame->av_frame);
}
struct mp_aframe *mp_aframe_create(void)
{
struct mp_aframe *frame = talloc_zero(NULL, struct mp_aframe);
frame->av_frame = av_frame_alloc();
if (!frame->av_frame)
abort();
talloc_set_destructor(frame, free_frame);
mp_aframe_reset(frame);
return frame;
}
struct mp_aframe *mp_aframe_new_ref(struct mp_aframe *frame)
{
if (!frame)
return NULL;
struct mp_aframe *dst = mp_aframe_create();
dst->chmap = frame->chmap;
dst->format = frame->format;
dst->pts = frame->pts;
dst->speed = frame->speed;
if (mp_aframe_is_allocated(frame)) {
if (av_frame_ref(dst->av_frame, frame->av_frame) < 0)
abort();
} else {
// av_frame_ref() would fail.
mp_aframe_config_copy(dst, frame);
}
return dst;
}
// Revert to state after mp_aframe_create().
void mp_aframe_reset(struct mp_aframe *frame)
{
av_frame_unref(frame->av_frame);
frame->chmap.num = 0;
frame->format = 0;
frame->pts = MP_NOPTS_VALUE;
frame->speed = 1.0;
}
// Remove all actual audio data and leave only the metadata.
void mp_aframe_unref_data(struct mp_aframe *frame)
{
// In a fucked up way, this is less complex than just unreffing the data.
struct mp_aframe *tmp = mp_aframe_create();
MPSWAP(struct mp_aframe, *tmp, *frame);
mp_aframe_reset(frame);
mp_aframe_config_copy(frame, tmp);
talloc_free(tmp);
}
// Return a new reference to the data in av_frame. av_frame itself is not
// touched. Returns NULL if not representable, or if input is NULL.
// Does not copy the timestamps.
struct mp_aframe *mp_aframe_from_avframe(struct AVFrame *av_frame)
{
if (!av_frame || av_frame->width > 0 || av_frame->height > 0)
return NULL;
int format = af_from_avformat(av_frame->format);
if (!format && av_frame->format != AV_SAMPLE_FMT_NONE)
return NULL;
struct mp_aframe *frame = mp_aframe_create();
// This also takes care of forcing refcounting.
if (av_frame_ref(frame->av_frame, av_frame) < 0)
abort();
frame->format = format;
mp_chmap_from_lavc(&frame->chmap, frame->av_frame->channel_layout);
#if LIBAVUTIL_VERSION_MICRO >= 100
// FFmpeg being a stupid POS again
if (frame->chmap.num != frame->av_frame->channels)
mp_chmap_from_channels(&frame->chmap, av_frame->channels);
#endif
if (av_frame->opaque_ref) {
struct avframe_opaque *op = (void *)av_frame->opaque_ref->data;
frame->speed = op->speed;
}
return frame;
}
// Return a new reference to the data in frame. Returns NULL is not
// representable (), or if input is NULL.
// Does not copy the timestamps.
struct AVFrame *mp_aframe_to_avframe(struct mp_aframe *frame)
{
if (!frame)
return NULL;
if (af_to_avformat(frame->format) != frame->av_frame->format)
return NULL;
if (!mp_chmap_is_lavc(&frame->chmap))
return NULL;
if (!frame->av_frame->opaque_ref && frame->speed != 1.0) {
frame->av_frame->opaque_ref =
av_buffer_alloc(sizeof(struct avframe_opaque));
if (!frame->av_frame->opaque_ref)
return NULL;
struct avframe_opaque *op = (void *)frame->av_frame->opaque_ref->data;
op->speed = frame->speed;
}
return av_frame_clone(frame->av_frame);
}
struct AVFrame *mp_aframe_to_avframe_and_unref(struct mp_aframe *frame)
{
AVFrame *av = mp_aframe_to_avframe(frame);
talloc_free(frame);
return av;
}
// You must not use this.
struct AVFrame *mp_aframe_get_raw_avframe(struct mp_aframe *frame)
{
return frame->av_frame;
}
// Return whether it has associated audio data. (If not, metadata only.)
bool mp_aframe_is_allocated(struct mp_aframe *frame)
{
return frame->av_frame->buf[0] || frame->av_frame->extended_data[0];
}
// Clear dst, and then copy the configuration to it.
void mp_aframe_config_copy(struct mp_aframe *dst, struct mp_aframe *src)
{
mp_aframe_reset(dst);
dst->chmap = src->chmap;
dst->format = src->format;
mp_aframe_copy_attributes(dst, src);
dst->av_frame->sample_rate = src->av_frame->sample_rate;
dst->av_frame->format = src->av_frame->format;
dst->av_frame->channel_layout = src->av_frame->channel_layout;
#if LIBAVUTIL_VERSION_MICRO >= 100
// FFmpeg being a stupid POS again
dst->av_frame->channels = src->av_frame->channels;
#endif
}
// Copy "soft" attributes from src to dst, excluding things which affect
// frame allocation and organization.
void mp_aframe_copy_attributes(struct mp_aframe *dst, struct mp_aframe *src)
{
dst->pts = src->pts;
dst->speed = src->speed;
int rate = dst->av_frame->sample_rate;
if (av_frame_copy_props(dst->av_frame, src->av_frame) < 0)
abort();
dst->av_frame->sample_rate = rate;
}
// Return whether a and b use the same physical audio format. Extra metadata
// such as PTS, per-frame signalling, and AVFrame side data is not compared.
bool mp_aframe_config_equals(struct mp_aframe *a, struct mp_aframe *b)
{
struct mp_chmap ca = {0}, cb = {0};
mp_aframe_get_chmap(a, &ca);
mp_aframe_get_chmap(b, &cb);
return mp_chmap_equals(&ca, &cb) &&
mp_aframe_get_rate(a) == mp_aframe_get_rate(b) &&
mp_aframe_get_format(a) == mp_aframe_get_format(b);
}
// Return whether all required format fields have been set.
bool mp_aframe_config_is_valid(struct mp_aframe *frame)
{
return frame->format && frame->chmap.num && frame->av_frame->sample_rate;
}
// Return the pointer to the first sample for each plane. The pointers stay
// valid until the next call that mutates frame somehow. You must not write to
// the audio data. Returns NULL if no frame allocated.
uint8_t **mp_aframe_get_data_ro(struct mp_aframe *frame)
{
return mp_aframe_is_allocated(frame) ? frame->av_frame->extended_data : NULL;
}
// Like mp_aframe_get_data_ro(), but you can write to the audio data.
// Additionally, it will return NULL if copy-on-write fails.
uint8_t **mp_aframe_get_data_rw(struct mp_aframe *frame)
{
if (!mp_aframe_is_allocated(frame))
return NULL;
if (av_frame_make_writable(frame->av_frame) < 0)
return NULL;
return frame->av_frame->extended_data;
}
int mp_aframe_get_format(struct mp_aframe *frame)
{
return frame->format;
}
bool mp_aframe_get_chmap(struct mp_aframe *frame, struct mp_chmap *out)
{
if (!mp_chmap_is_valid(&frame->chmap))
return false;
*out = frame->chmap;
return true;
}
int mp_aframe_get_channels(struct mp_aframe *frame)
{
return frame->chmap.num;
}
int mp_aframe_get_rate(struct mp_aframe *frame)
{
return frame->av_frame->sample_rate;
}
int mp_aframe_get_size(struct mp_aframe *frame)
{
return frame->av_frame->nb_samples;
}
double mp_aframe_get_pts(struct mp_aframe *frame)
{
return frame->pts;
}
bool mp_aframe_set_format(struct mp_aframe *frame, int format)
{
if (mp_aframe_is_allocated(frame))
return false;
enum AVSampleFormat av_format = af_to_avformat(format);
if (av_format == AV_SAMPLE_FMT_NONE && format) {
if (!af_fmt_is_spdif(format))
return false;
av_format = AV_SAMPLE_FMT_S16;
}
frame->format = format;
frame->av_frame->format = av_format;
return true;
}
bool mp_aframe_set_chmap(struct mp_aframe *frame, struct mp_chmap *in)
{
if (!mp_chmap_is_valid(in) && !mp_chmap_is_empty(in))
return false;
if (mp_aframe_is_allocated(frame) && in->num != frame->chmap.num)
return false;
uint64_t lavc_layout = mp_chmap_to_lavc_unchecked(in);
if (!lavc_layout && in->num)
return false;
frame->chmap = *in;
frame->av_frame->channel_layout = lavc_layout;
#if LIBAVUTIL_VERSION_MICRO >= 100
// FFmpeg being a stupid POS again
frame->av_frame->channels = frame->chmap.num;
#endif
return true;
}
bool mp_aframe_set_rate(struct mp_aframe *frame, int rate)
{
if (rate < 1 || rate > 10000000)
return false;
frame->av_frame->sample_rate = rate;
return true;
}
bool mp_aframe_set_size(struct mp_aframe *frame, int samples)
{
if (!mp_aframe_is_allocated(frame) || mp_aframe_get_size(frame) < samples)
return false;
frame->av_frame->nb_samples = MPMAX(samples, 0);
return true;
}
void mp_aframe_set_pts(struct mp_aframe *frame, double pts)
{
frame->pts = pts;
}
// Set a speed factor. This is multiplied with the sample rate to get the
// "effective" samplerate (mp_aframe_get_effective_rate()), which will be used
// to do PTS calculations. If speed!=1.0, the PTS values always refer to the
// original PTS (before changing speed), and if you want reasonably continuous
// PTS between frames, you need to use the effective samplerate.
void mp_aframe_set_speed(struct mp_aframe *frame, double factor)
{
frame->speed = factor;
}
// Adjust current speed factor.
void mp_aframe_mul_speed(struct mp_aframe *frame, double factor)
{
frame->speed *= factor;
}
double mp_aframe_get_speed(struct mp_aframe *frame)
{
return frame->speed;
}
// Matters for speed changed frames (such as a frame which has been resampled
// to play at a different speed).
// Return the sample rate at which the frame would have to be played to result
// in the same duration as the original frame before the speed change.
// This is used for A/V sync.
double mp_aframe_get_effective_rate(struct mp_aframe *frame)
{
return mp_aframe_get_rate(frame) / frame->speed;
}
// Return number of data pointers.
int mp_aframe_get_planes(struct mp_aframe *frame)
{
return af_fmt_is_planar(mp_aframe_get_format(frame))
? mp_aframe_get_channels(frame) : 1;
}
// Return number of bytes between 2 consecutive samples on the same plane.
size_t mp_aframe_get_sstride(struct mp_aframe *frame)
{
int format = mp_aframe_get_format(frame);
return af_fmt_to_bytes(format) *
(af_fmt_is_planar(format) ? 1 : mp_aframe_get_channels(frame));
}
// Return total number of samples on each plane.
int mp_aframe_get_total_plane_samples(struct mp_aframe *frame)
{
return frame->av_frame->nb_samples *
(af_fmt_is_planar(mp_aframe_get_format(frame))
? 1 : mp_aframe_get_channels(frame));
}
char *mp_aframe_format_str_buf(char *buf, size_t buf_size, struct mp_aframe *fmt)
{
char ch[128];
mp_chmap_to_str_buf(ch, sizeof(ch), &fmt->chmap);
char *hr_ch = mp_chmap_to_str_hr(&fmt->chmap);
if (strcmp(hr_ch, ch) != 0)
mp_snprintf_cat(ch, sizeof(ch), " (%s)", hr_ch);
snprintf(buf, buf_size, "%dHz %s %dch %s", fmt->av_frame->sample_rate,
ch, fmt->chmap.num, af_fmt_to_str(fmt->format));
return buf;
}
// Set data to the audio after the given number of samples (i.e. slice it).
void mp_aframe_skip_samples(struct mp_aframe *f, int samples)
{
assert(samples >= 0 && samples <= mp_aframe_get_size(f));
int num_planes = mp_aframe_get_planes(f);
size_t sstride = mp_aframe_get_sstride(f);
for (int n = 0; n < num_planes; n++)
f->av_frame->extended_data[n] += samples * sstride;
f->av_frame->nb_samples -= samples;
if (f->pts != MP_NOPTS_VALUE)
f->pts += samples / mp_aframe_get_effective_rate(f);
}
// Return the timestamp of the sample just after the end of this frame.
double mp_aframe_end_pts(struct mp_aframe *f)
{
double rate = mp_aframe_get_effective_rate(f);
if (f->pts == MP_NOPTS_VALUE || rate <= 0)
return MP_NOPTS_VALUE;
return f->pts + f->av_frame->nb_samples / rate;
}
// Return the duration in seconds of the frame (0 if invalid).
double mp_aframe_duration(struct mp_aframe *f)
{
double rate = mp_aframe_get_effective_rate(f);
if (rate <= 0)
return 0;
return f->av_frame->nb_samples / rate;
}
// Clip the given frame to the given timestamp range. Adjusts the frame size
// and timestamp.
// Refuses to change spdif frames.
void mp_aframe_clip_timestamps(struct mp_aframe *f, double start, double end)
{
double f_end = mp_aframe_end_pts(f);
double rate = mp_aframe_get_effective_rate(f);
if (f_end == MP_NOPTS_VALUE)
return;
if (af_fmt_is_spdif(mp_aframe_get_format(f)))
return;
if (end != MP_NOPTS_VALUE) {
if (f_end >= end) {
if (f->pts >= end) {
f->av_frame->nb_samples = 0;
} else {
int new = (end - f->pts) * rate;
f->av_frame->nb_samples = MPCLAMP(new, 0, f->av_frame->nb_samples);
}
}
}
if (start != MP_NOPTS_VALUE) {
if (f->pts < start) {
if (f_end <= start) {
f->av_frame->nb_samples = 0;
f->pts = f_end;
} else {
int skip = (start - f->pts) * rate;
skip = MPCLAMP(skip, 0, f->av_frame->nb_samples);
mp_aframe_skip_samples(f, skip);
}
}
}
}
bool mp_aframe_copy_samples(struct mp_aframe *dst, int dst_offset,
struct mp_aframe *src, int src_offset,
int samples)
{
if (!mp_aframe_config_equals(dst, src))
return false;
if (mp_aframe_get_size(dst) < dst_offset + samples ||
mp_aframe_get_size(src) < src_offset + samples)
return false;
uint8_t **s = mp_aframe_get_data_ro(src);
uint8_t **d = mp_aframe_get_data_rw(dst);
if (!s || !d)
return false;
int planes = mp_aframe_get_planes(dst);
size_t sstride = mp_aframe_get_sstride(dst);
for (int n = 0; n < planes; n++) {
memcpy(d[n] + dst_offset * sstride, s[n] + src_offset * sstride,
samples * sstride);
}
return true;
}
bool mp_aframe_set_silence(struct mp_aframe *f, int offset, int samples)
{
if (mp_aframe_get_size(f) < offset + samples)
return false;
int format = mp_aframe_get_format(f);
uint8_t **d = mp_aframe_get_data_rw(f);
if (!d)
return false;
int planes = mp_aframe_get_planes(f);
size_t sstride = mp_aframe_get_sstride(f);
for (int n = 0; n < planes; n++)
af_fill_silence(d[n] + offset * sstride, samples * sstride, format);
return true;
}
bool mp_aframe_reverse(struct mp_aframe *f)
{
int format = mp_aframe_get_format(f);
size_t bps = af_fmt_to_bytes(format);
if (!af_fmt_is_pcm(format) || bps > 16)
return false;
uint8_t **d = mp_aframe_get_data_rw(f);
if (!d)
return false;
int planes = mp_aframe_get_planes(f);
int samples = mp_aframe_get_size(f);
int channels = mp_aframe_get_channels(f);
size_t sstride = mp_aframe_get_sstride(f);
int plane_samples = channels;
if (af_fmt_is_planar(format))
plane_samples = 1;
for (int p = 0; p < planes; p++) {
for (int n = 0; n < samples / 2; n++) {
int s1_offset = n * sstride;
int s2_offset = (samples - 1 - n) * sstride;
for (int c = 0; c < plane_samples; c++) {
// Nobody said it'd be fast.
char tmp[16];
uint8_t *s1 = d[p] + s1_offset + c * bps;
uint8_t *s2 = d[p] + s2_offset + c * bps;
memcpy(tmp, s2, bps);
memcpy(s2, s1, bps);
memcpy(s1, tmp, bps);
}
}
}
return true;
}
int mp_aframe_approx_byte_size(struct mp_aframe *frame)
{
// God damn, AVFrame is too fucking annoying. Just go with the size that
// allocating a new frame would use.
int planes = mp_aframe_get_planes(frame);
size_t sstride = mp_aframe_get_sstride(frame);
int samples = frame->av_frame->nb_samples;
int plane_size = MP_ALIGN_UP(sstride * MPMAX(samples, 1), 32);
return plane_size * planes + sizeof(*frame);
}
struct mp_aframe_pool {
AVBufferPool *avpool;
int element_size;
};
struct mp_aframe_pool *mp_aframe_pool_create(void *ta_parent)
{
return talloc_zero(ta_parent, struct mp_aframe_pool);
}
static void mp_aframe_pool_destructor(void *p)
{
struct mp_aframe_pool *pool = p;
av_buffer_pool_uninit(&pool->avpool);
}
// Like mp_aframe_allocate(), but use the pool to allocate data.
int mp_aframe_pool_allocate(struct mp_aframe_pool *pool, struct mp_aframe *frame,
int samples)
{
int planes = mp_aframe_get_planes(frame);
size_t sstride = mp_aframe_get_sstride(frame);
int plane_size = MP_ALIGN_UP(sstride * MPMAX(samples, 1), 32);
int size = plane_size * planes;
if (size <= 0 || mp_aframe_is_allocated(frame))
return -1;
if (!pool->avpool || size > pool->element_size) {
size_t alloc = ta_calc_prealloc_elems(size);
if (alloc >= INT_MAX)
return -1;
av_buffer_pool_uninit(&pool->avpool);
pool->element_size = alloc;
pool->avpool = av_buffer_pool_init(pool->element_size, NULL);
if (!pool->avpool)
return -1;
talloc_set_destructor(pool, mp_aframe_pool_destructor);
}
// Yes, you have to do all this shit manually.
// At least it's less stupid than av_frame_get_buffer(), which just wipes
// the entire frame struct on error for no reason.
AVFrame *av_frame = frame->av_frame;
if (av_frame->extended_data != av_frame->data)
av_freep(&av_frame->extended_data); // sigh
av_frame->extended_data =
av_mallocz_array(planes, sizeof(av_frame->extended_data[0]));
if (!av_frame->extended_data)
abort();
av_frame->buf[0] = av_buffer_pool_get(pool->avpool);
if (!av_frame->buf[0])
return -1;
av_frame->linesize[0] = samples * sstride;
for (int n = 0; n < planes; n++)
av_frame->extended_data[n] = av_frame->buf[0]->data + n * plane_size;
av_frame->nb_samples = samples;
return 0;
}