1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-18 03:19:31 +02:00
FFmpeg/libavcodec/frame_thread_encoder.c

319 lines
11 KiB
C
Raw Normal View History

/*
* Copyright (c) 2012 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <stdatomic.h>
#include "frame_thread_encoder.h"
#include "libavutil/avassert.h"
#include "libavutil/imgutils.h"
#include "libavutil/opt.h"
#include "libavutil/thread.h"
#include "avcodec.h"
#include "internal.h"
#include "thread.h"
#define MAX_THREADS 64
/* There can be as many as MAX_THREADS + 1 outstanding tasks.
* An additional + 1 is needed so that one can distinguish
* the case of zero and MAX_THREADS + 1 outstanding tasks modulo
* the number of buffers. */
#define BUFFER_SIZE (MAX_THREADS + 2)
typedef struct{
AVFrame *indata;
AVPacket *outdata;
int return_code;
int finished;
} Task;
typedef struct{
AVCodecContext *parent_avctx;
pthread_mutex_t buffer_mutex;
pthread_mutex_t task_fifo_mutex; /* Used to guard (next_)task_index */
pthread_cond_t task_fifo_cond;
unsigned max_tasks;
Task tasks[BUFFER_SIZE];
pthread_mutex_t finished_task_mutex; /* Guards tasks[i].finished */
pthread_cond_t finished_task_cond;
unsigned next_task_index;
unsigned task_index;
unsigned finished_task_index;
pthread_t worker[MAX_THREADS];
atomic_int exit;
} ThreadContext;
static void * attribute_align_arg worker(void *v){
AVCodecContext *avctx = v;
ThreadContext *c = avctx->internal->frame_thread_encoder;
while (!atomic_load(&c->exit)) {
int got_packet = 0, ret;
AVPacket *pkt;
AVFrame *frame;
Task *task;
unsigned task_index;
pthread_mutex_lock(&c->task_fifo_mutex);
while (c->next_task_index == c->task_index || atomic_load(&c->exit)) {
if (atomic_load(&c->exit)) {
pthread_mutex_unlock(&c->task_fifo_mutex);
goto end;
}
pthread_cond_wait(&c->task_fifo_cond, &c->task_fifo_mutex);
}
task_index = c->next_task_index;
c->next_task_index = (c->next_task_index + 1) % c->max_tasks;
pthread_mutex_unlock(&c->task_fifo_mutex);
/* The main thread ensures that any two outstanding tasks have
* different indices, ergo each worker thread owns its element
* of c->tasks with the exception of finished, which is shared
* with the main thread and guarded by finished_task_mutex. */
task = &c->tasks[task_index];
frame = task->indata;
pkt = task->outdata;
ret = avctx->codec->encode2(avctx, pkt, frame, &got_packet);
if(got_packet) {
int ret2 = av_packet_make_refcounted(pkt);
if (ret >= 0 && ret2 < 0)
ret = ret2;
pkt->pts = pkt->dts = frame->pts;
} else {
pkt->data = NULL;
pkt->size = 0;
}
pthread_mutex_lock(&c->buffer_mutex);
av_frame_unref(frame);
pthread_mutex_unlock(&c->buffer_mutex);
pthread_mutex_lock(&c->finished_task_mutex);
task->return_code = ret;
task->finished = 1;
pthread_cond_signal(&c->finished_task_cond);
pthread_mutex_unlock(&c->finished_task_mutex);
}
end:
pthread_mutex_lock(&c->buffer_mutex);
avcodec_close(avctx);
pthread_mutex_unlock(&c->buffer_mutex);
av_freep(&avctx);
return NULL;
}
int ff_frame_thread_encoder_init(AVCodecContext *avctx, AVDictionary *options){
int i=0;
ThreadContext *c;
if( !(avctx->thread_type & FF_THREAD_FRAME)
|| !(avctx->codec->capabilities & AV_CODEC_CAP_FRAME_THREADS))
return 0;
if( !avctx->thread_count
&& avctx->codec_id == AV_CODEC_ID_MJPEG
&& !(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
av_log(avctx, AV_LOG_DEBUG,
"Forcing thread count to 1 for MJPEG encoding, use -thread_type slice "
"or a constant quantizer if you want to use multiple cpu cores\n");
avctx->thread_count = 1;
}
if( avctx->thread_count > 1
&& avctx->codec_id == AV_CODEC_ID_MJPEG
&& !(avctx->flags & AV_CODEC_FLAG_QSCALE))
av_log(avctx, AV_LOG_WARNING,
"MJPEG CBR encoding works badly with frame multi-threading, consider "
"using -threads 1, -thread_type slice or a constant quantizer.\n");
if (avctx->codec_id == AV_CODEC_ID_HUFFYUV ||
avctx->codec_id == AV_CODEC_ID_FFVHUFF) {
int warn = 0;
int context_model = 0;
AVDictionaryEntry *con = av_dict_get(options, "context", NULL, AV_DICT_MATCH_CASE);
if (con && con->value)
context_model = atoi(con->value);
if (avctx->flags & AV_CODEC_FLAG_PASS1)
warn = 1;
else if(context_model > 0) {
AVDictionaryEntry *t = av_dict_get(options, "non_deterministic",
NULL, AV_DICT_MATCH_CASE);
warn = !t || !t->value || !atoi(t->value) ? 1 : 0;
}
// huffyuv does not support these with multiple frame threads currently
if (warn) {
av_log(avctx, AV_LOG_WARNING,
"Forcing thread count to 1 for huffyuv encoding with first pass or context 1\n");
avctx->thread_count = 1;
}
}
if(!avctx->thread_count) {
avctx->thread_count = av_cpu_count();
avctx->thread_count = FFMIN(avctx->thread_count, MAX_THREADS);
}
if(avctx->thread_count <= 1)
return 0;
if(avctx->thread_count > MAX_THREADS)
return AVERROR(EINVAL);
av_assert0(!avctx->internal->frame_thread_encoder);
c = avctx->internal->frame_thread_encoder = av_mallocz(sizeof(ThreadContext));
if(!c)
return AVERROR(ENOMEM);
c->parent_avctx = avctx;
pthread_mutex_init(&c->task_fifo_mutex, NULL);
pthread_mutex_init(&c->finished_task_mutex, NULL);
pthread_mutex_init(&c->buffer_mutex, NULL);
pthread_cond_init(&c->task_fifo_cond, NULL);
pthread_cond_init(&c->finished_task_cond, NULL);
atomic_init(&c->exit, 0);
c->max_tasks = avctx->thread_count + 2;
for (unsigned i = 0; i < c->max_tasks; i++) {
if (!(c->tasks[i].indata = av_frame_alloc()) ||
!(c->tasks[i].outdata = av_packet_alloc()))
goto fail;
}
for(i=0; i<avctx->thread_count ; i++){
AVDictionary *tmp = NULL;
int ret;
void *tmpv;
AVCodecContext *thread_avctx = avcodec_alloc_context3(avctx->codec);
if(!thread_avctx)
goto fail;
tmpv = thread_avctx->priv_data;
*thread_avctx = *avctx;
ret = av_opt_copy(thread_avctx, avctx);
if (ret < 0)
goto fail;
thread_avctx->priv_data = tmpv;
thread_avctx->internal = NULL;
if (avctx->codec->priv_class) {
int ret = av_opt_copy(thread_avctx->priv_data, avctx->priv_data);
if (ret < 0)
goto fail;
} else if (avctx->codec->priv_data_size) {
memcpy(thread_avctx->priv_data, avctx->priv_data, avctx->codec->priv_data_size);
}
thread_avctx->thread_count = 1;
thread_avctx->active_thread_type &= ~FF_THREAD_FRAME;
av_dict_copy(&tmp, options, 0);
av_dict_set(&tmp, "threads", "1", 0);
if(avcodec_open2(thread_avctx, avctx->codec, &tmp) < 0) {
av_dict_free(&tmp);
goto fail;
}
av_dict_free(&tmp);
av_assert0(!thread_avctx->internal->frame_thread_encoder);
thread_avctx->internal->frame_thread_encoder = c;
if(pthread_create(&c->worker[i], NULL, worker, thread_avctx)) {
goto fail;
}
}
avctx->active_thread_type = FF_THREAD_FRAME;
return 0;
fail:
avctx->thread_count = i;
av_log(avctx, AV_LOG_ERROR, "ff_frame_thread_encoder_init failed\n");
ff_frame_thread_encoder_free(avctx);
return -1;
}
void ff_frame_thread_encoder_free(AVCodecContext *avctx){
int i;
ThreadContext *c= avctx->internal->frame_thread_encoder;
pthread_mutex_lock(&c->task_fifo_mutex);
atomic_store(&c->exit, 1);
pthread_cond_broadcast(&c->task_fifo_cond);
pthread_mutex_unlock(&c->task_fifo_mutex);
for (i=0; i<avctx->thread_count; i++) {
pthread_join(c->worker[i], NULL);
}
for (unsigned i = 0; i < c->max_tasks; i++) {
av_frame_free(&c->tasks[i].indata);
av_packet_free(&c->tasks[i].outdata);
}
pthread_mutex_destroy(&c->task_fifo_mutex);
pthread_mutex_destroy(&c->finished_task_mutex);
pthread_mutex_destroy(&c->buffer_mutex);
pthread_cond_destroy(&c->task_fifo_cond);
pthread_cond_destroy(&c->finished_task_cond);
av_freep(&avctx->internal->frame_thread_encoder);
}
int ff_thread_video_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
AVFrame *frame, int *got_packet_ptr)
{
ThreadContext *c = avctx->internal->frame_thread_encoder;
Task *outtask;
av_assert1(!*got_packet_ptr);
if(frame){
av_frame_move_ref(c->tasks[c->task_index].indata, frame);
pthread_mutex_lock(&c->task_fifo_mutex);
c->task_index = (c->task_index + 1) % c->max_tasks;
pthread_cond_signal(&c->task_fifo_cond);
pthread_mutex_unlock(&c->task_fifo_mutex);
}
outtask = &c->tasks[c->finished_task_index];
pthread_mutex_lock(&c->finished_task_mutex);
/* The access to task_index in the following code is ok,
* because it is only ever changed by the main thread. */
if (c->task_index == c->finished_task_index ||
(frame && !outtask->finished &&
(c->task_index - c->finished_task_index + c->max_tasks) % c->max_tasks <= avctx->thread_count)) {
pthread_mutex_unlock(&c->finished_task_mutex);
return 0;
}
while (!outtask->finished) {
pthread_cond_wait(&c->finished_task_cond, &c->finished_task_mutex);
}
pthread_mutex_unlock(&c->finished_task_mutex);
/* We now own outtask completely: No worker thread touches it any more,
* because there is no outstanding task with this index. */
outtask->finished = 0;
av_packet_move_ref(pkt, outtask->outdata);
if(pkt->data)
*got_packet_ptr = 1;
c->finished_task_index = (c->finished_task_index + 1) % c->max_tasks;
return outtask->return_code;
}