mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-23 12:43:46 +02:00
lavfi/buffersink: move to the new design.
This commit is contained in:
parent
32c59a115d
commit
912969a33e
@ -23,12 +23,10 @@
|
||||
* buffer sink
|
||||
*/
|
||||
|
||||
#include "libavutil/audio_fifo.h"
|
||||
#include "libavutil/avassert.h"
|
||||
#include "libavutil/channel_layout.h"
|
||||
#include "libavutil/common.h"
|
||||
#include "libavutil/internal.h"
|
||||
#include "libavutil/mathematics.h"
|
||||
#include "libavutil/opt.h"
|
||||
|
||||
#define FF_INTERNAL_FIELDS 1
|
||||
@ -37,11 +35,11 @@
|
||||
#include "audio.h"
|
||||
#include "avfilter.h"
|
||||
#include "buffersink.h"
|
||||
#include "filters.h"
|
||||
#include "internal.h"
|
||||
|
||||
typedef struct BufferSinkContext {
|
||||
const AVClass *class;
|
||||
AVFifoBuffer *fifo; ///< FIFO buffer of video frame references
|
||||
unsigned warning_limit;
|
||||
|
||||
/* only used for video */
|
||||
@ -59,182 +57,74 @@ typedef struct BufferSinkContext {
|
||||
int *sample_rates; ///< list of accepted sample rates, terminated by -1
|
||||
int sample_rates_size;
|
||||
|
||||
/* only used for compat API */
|
||||
AVAudioFifo *audio_fifo; ///< FIFO for audio samples
|
||||
int64_t next_pts; ///< interpolating audio pts
|
||||
AVFrame *peeked_frame;
|
||||
} BufferSinkContext;
|
||||
|
||||
#define NB_ITEMS(list) (list ## _size / sizeof(*list))
|
||||
#define FIFO_INIT_SIZE 8
|
||||
#define FIFO_INIT_ELEMENT_SIZE sizeof(void *)
|
||||
|
||||
static av_cold void uninit(AVFilterContext *ctx)
|
||||
{
|
||||
BufferSinkContext *sink = ctx->priv;
|
||||
AVFrame *frame;
|
||||
|
||||
if (sink->audio_fifo)
|
||||
av_audio_fifo_free(sink->audio_fifo);
|
||||
|
||||
if (sink->fifo) {
|
||||
while (av_fifo_size(sink->fifo) >= FIFO_INIT_ELEMENT_SIZE) {
|
||||
av_fifo_generic_read(sink->fifo, &frame, sizeof(frame), NULL);
|
||||
av_frame_free(&frame);
|
||||
}
|
||||
av_fifo_freep(&sink->fifo);
|
||||
}
|
||||
}
|
||||
|
||||
static int add_buffer_ref(AVFilterContext *ctx, AVFrame *ref)
|
||||
{
|
||||
BufferSinkContext *buf = ctx->priv;
|
||||
|
||||
if (av_fifo_space(buf->fifo) < FIFO_INIT_ELEMENT_SIZE) {
|
||||
/* realloc fifo size */
|
||||
if (av_fifo_realloc2(buf->fifo, av_fifo_size(buf->fifo) * 2) < 0) {
|
||||
av_log(ctx, AV_LOG_ERROR,
|
||||
"Cannot buffer more frames. Consume some available frames "
|
||||
"before adding new ones.\n");
|
||||
return AVERROR(ENOMEM);
|
||||
}
|
||||
}
|
||||
|
||||
/* cache frame */
|
||||
av_fifo_generic_write(buf->fifo, &ref, FIFO_INIT_ELEMENT_SIZE, NULL);
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int filter_frame(AVFilterLink *link, AVFrame *frame)
|
||||
{
|
||||
AVFilterContext *ctx = link->dst;
|
||||
BufferSinkContext *buf = link->dst->priv;
|
||||
int ret;
|
||||
|
||||
if ((ret = add_buffer_ref(ctx, frame)) < 0)
|
||||
return ret;
|
||||
if (buf->warning_limit &&
|
||||
av_fifo_size(buf->fifo) / FIFO_INIT_ELEMENT_SIZE >= buf->warning_limit) {
|
||||
av_log(ctx, AV_LOG_WARNING,
|
||||
"%d buffers queued in %s, something may be wrong.\n",
|
||||
buf->warning_limit,
|
||||
(char *)av_x_if_null(ctx->name, ctx->filter->name));
|
||||
buf->warning_limit *= 10;
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
int attribute_align_arg av_buffersink_get_frame(AVFilterContext *ctx, AVFrame *frame)
|
||||
{
|
||||
return av_buffersink_get_frame_flags(ctx, frame, 0);
|
||||
}
|
||||
|
||||
int attribute_align_arg av_buffersink_get_frame_flags(AVFilterContext *ctx, AVFrame *frame, int flags)
|
||||
static int return_or_keep_frame(BufferSinkContext *buf, AVFrame *out, AVFrame *in, int flags)
|
||||
{
|
||||
if ((flags & AV_BUFFERSINK_FLAG_PEEK)) {
|
||||
buf->peeked_frame = in;
|
||||
return out ? av_frame_ref(out, in) : 0;
|
||||
} else {
|
||||
av_assert1(out);
|
||||
buf->peeked_frame = NULL;
|
||||
av_frame_move_ref(out, in);
|
||||
av_frame_free(&in);
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
static int get_frame_internal(AVFilterContext *ctx, AVFrame *frame, int flags, int samples)
|
||||
{
|
||||
BufferSinkContext *buf = ctx->priv;
|
||||
AVFilterLink *inlink = ctx->inputs[0];
|
||||
int peek_in_framequeue = 0, ret;
|
||||
int64_t frame_count;
|
||||
int status, ret;
|
||||
AVFrame *cur_frame;
|
||||
int64_t pts;
|
||||
|
||||
/* no picref available, fetch it from the filterchain */
|
||||
while (!av_fifo_size(buf->fifo)) {
|
||||
/* if peek_in_framequeue is true later, then ff_request_frame() and
|
||||
the ff_filter_graph_run_once() loop will take a frame from it and
|
||||
move it to the internal fifo, ending the global loop */
|
||||
av_assert0(!peek_in_framequeue);
|
||||
if (inlink->status_out)
|
||||
return inlink->status_out;
|
||||
peek_in_framequeue = ff_framequeue_queued_frames(&inlink->fifo) &&
|
||||
ff_framequeue_queued_samples(&inlink->fifo) >= inlink->min_samples;
|
||||
if ((flags & AV_BUFFERSINK_FLAG_NO_REQUEST) && !peek_in_framequeue)
|
||||
return AVERROR(EAGAIN);
|
||||
if ((ret = ff_request_frame(inlink)) < 0)
|
||||
if (buf->peeked_frame)
|
||||
return return_or_keep_frame(buf, frame, buf->peeked_frame, flags);
|
||||
|
||||
while (1) {
|
||||
ret = samples ? ff_inlink_consume_samples(inlink, samples, samples, &cur_frame) :
|
||||
ff_inlink_consume_frame(inlink, &cur_frame);
|
||||
if (ret < 0) {
|
||||
return ret;
|
||||
frame_count = inlink->frame_count_out;
|
||||
while (frame_count == inlink->frame_count_out) {
|
||||
} else if (ret) {
|
||||
/* TODO return the frame instead of copying it */
|
||||
return return_or_keep_frame(buf, frame, cur_frame, flags);
|
||||
} else if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
|
||||
return status;
|
||||
} else if ((flags & AV_BUFFERSINK_FLAG_NO_REQUEST)) {
|
||||
return AVERROR(EAGAIN);
|
||||
} else if (inlink->frame_wanted_out) {
|
||||
ret = ff_filter_graph_run_once(ctx->graph);
|
||||
if (ret < 0)
|
||||
return ret;
|
||||
} else {
|
||||
ff_inlink_request_frame(inlink);
|
||||
}
|
||||
}
|
||||
|
||||
if (flags & AV_BUFFERSINK_FLAG_PEEK) {
|
||||
cur_frame = *((AVFrame **)av_fifo_peek2(buf->fifo, 0));
|
||||
if ((ret = av_frame_ref(frame, cur_frame)) < 0)
|
||||
return ret;
|
||||
} else {
|
||||
av_fifo_generic_read(buf->fifo, &cur_frame, sizeof(cur_frame), NULL);
|
||||
av_frame_move_ref(frame, cur_frame);
|
||||
av_frame_free(&cur_frame);
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int read_from_fifo(AVFilterContext *ctx, AVFrame *frame,
|
||||
int nb_samples)
|
||||
int attribute_align_arg av_buffersink_get_frame_flags(AVFilterContext *ctx, AVFrame *frame, int flags)
|
||||
{
|
||||
BufferSinkContext *s = ctx->priv;
|
||||
AVFilterLink *link = ctx->inputs[0];
|
||||
AVFrame *tmp;
|
||||
|
||||
if (!(tmp = ff_get_audio_buffer(link, nb_samples)))
|
||||
return AVERROR(ENOMEM);
|
||||
av_audio_fifo_read(s->audio_fifo, (void**)tmp->extended_data, nb_samples);
|
||||
|
||||
tmp->pts = s->next_pts;
|
||||
if (s->next_pts != AV_NOPTS_VALUE)
|
||||
s->next_pts += av_rescale_q(nb_samples, (AVRational){1, link->sample_rate},
|
||||
link->time_base);
|
||||
|
||||
av_frame_move_ref(frame, tmp);
|
||||
av_frame_free(&tmp);
|
||||
|
||||
return 0;
|
||||
return get_frame_internal(ctx, frame, flags, ctx->inputs[0]->min_samples);
|
||||
}
|
||||
|
||||
int attribute_align_arg av_buffersink_get_samples(AVFilterContext *ctx,
|
||||
AVFrame *frame, int nb_samples)
|
||||
{
|
||||
BufferSinkContext *s = ctx->priv;
|
||||
AVFilterLink *link = ctx->inputs[0];
|
||||
AVFrame *cur_frame;
|
||||
int ret = 0;
|
||||
|
||||
if (!s->audio_fifo) {
|
||||
int nb_channels = link->channels;
|
||||
if (!(s->audio_fifo = av_audio_fifo_alloc(link->format, nb_channels, nb_samples)))
|
||||
return AVERROR(ENOMEM);
|
||||
}
|
||||
|
||||
while (ret >= 0) {
|
||||
if (av_audio_fifo_size(s->audio_fifo) >= nb_samples)
|
||||
return read_from_fifo(ctx, frame, nb_samples);
|
||||
|
||||
if (!(cur_frame = av_frame_alloc()))
|
||||
return AVERROR(ENOMEM);
|
||||
ret = av_buffersink_get_frame_flags(ctx, cur_frame, 0);
|
||||
if (ret == AVERROR_EOF && av_audio_fifo_size(s->audio_fifo)) {
|
||||
av_frame_free(&cur_frame);
|
||||
return read_from_fifo(ctx, frame, av_audio_fifo_size(s->audio_fifo));
|
||||
} else if (ret < 0) {
|
||||
av_frame_free(&cur_frame);
|
||||
return ret;
|
||||
}
|
||||
|
||||
if (cur_frame->pts != AV_NOPTS_VALUE) {
|
||||
s->next_pts = cur_frame->pts -
|
||||
av_rescale_q(av_audio_fifo_size(s->audio_fifo),
|
||||
(AVRational){ 1, link->sample_rate },
|
||||
link->time_base);
|
||||
}
|
||||
|
||||
ret = av_audio_fifo_write(s->audio_fifo, (void**)cur_frame->extended_data,
|
||||
cur_frame->nb_samples);
|
||||
av_frame_free(&cur_frame);
|
||||
}
|
||||
|
||||
return ret;
|
||||
return get_frame_internal(ctx, frame, 0, nb_samples);
|
||||
}
|
||||
|
||||
AVBufferSinkParams *av_buffersink_params_alloc(void)
|
||||
@ -261,13 +151,24 @@ static av_cold int common_init(AVFilterContext *ctx)
|
||||
{
|
||||
BufferSinkContext *buf = ctx->priv;
|
||||
|
||||
buf->fifo = av_fifo_alloc_array(FIFO_INIT_SIZE, FIFO_INIT_ELEMENT_SIZE);
|
||||
if (!buf->fifo) {
|
||||
av_log(ctx, AV_LOG_ERROR, "Failed to allocate fifo\n");
|
||||
return AVERROR(ENOMEM);
|
||||
}
|
||||
buf->warning_limit = 100;
|
||||
buf->next_pts = AV_NOPTS_VALUE;
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int activate(AVFilterContext *ctx)
|
||||
{
|
||||
BufferSinkContext *buf = ctx->priv;
|
||||
|
||||
if (buf->warning_limit &&
|
||||
ff_framequeue_queued_frames(&ctx->inputs[0]->fifo) >= buf->warning_limit) {
|
||||
av_log(ctx, AV_LOG_WARNING,
|
||||
"%d buffers queued in %s, something may be wrong.\n",
|
||||
buf->warning_limit,
|
||||
(char *)av_x_if_null(ctx->name, ctx->filter->name));
|
||||
buf->warning_limit *= 10;
|
||||
}
|
||||
|
||||
/* The frame is queued, the rest is up to get_frame_internal */
|
||||
return 0;
|
||||
}
|
||||
|
||||
@ -281,7 +182,7 @@ void av_buffersink_set_frame_size(AVFilterContext *ctx, unsigned frame_size)
|
||||
|
||||
#define MAKE_AVFILTERLINK_ACCESSOR(type, field) \
|
||||
type av_buffersink_get_##field(const AVFilterContext *ctx) { \
|
||||
av_assert0(ctx->filter->uninit == uninit); \
|
||||
av_assert0(ctx->filter->activate == activate); \
|
||||
return ctx->inputs[0]->field; \
|
||||
}
|
||||
|
||||
@ -437,7 +338,6 @@ static const AVFilterPad avfilter_vsink_buffer_inputs[] = {
|
||||
{
|
||||
.name = "default",
|
||||
.type = AVMEDIA_TYPE_VIDEO,
|
||||
.filter_frame = filter_frame,
|
||||
},
|
||||
{ NULL }
|
||||
};
|
||||
@ -448,9 +348,9 @@ AVFilter ff_vsink_buffer = {
|
||||
.priv_size = sizeof(BufferSinkContext),
|
||||
.priv_class = &buffersink_class,
|
||||
.init_opaque = vsink_init,
|
||||
.uninit = uninit,
|
||||
|
||||
.query_formats = vsink_query_formats,
|
||||
.activate = activate,
|
||||
.inputs = avfilter_vsink_buffer_inputs,
|
||||
.outputs = NULL,
|
||||
};
|
||||
@ -459,7 +359,6 @@ static const AVFilterPad avfilter_asink_abuffer_inputs[] = {
|
||||
{
|
||||
.name = "default",
|
||||
.type = AVMEDIA_TYPE_AUDIO,
|
||||
.filter_frame = filter_frame,
|
||||
},
|
||||
{ NULL }
|
||||
};
|
||||
@ -470,9 +369,9 @@ AVFilter ff_asink_abuffer = {
|
||||
.priv_class = &abuffersink_class,
|
||||
.priv_size = sizeof(BufferSinkContext),
|
||||
.init_opaque = asink_init,
|
||||
.uninit = uninit,
|
||||
|
||||
.query_formats = asink_query_formats,
|
||||
.activate = activate,
|
||||
.inputs = avfilter_asink_abuffer_inputs,
|
||||
.outputs = NULL,
|
||||
};
|
||||
|
Loading…
Reference in New Issue
Block a user