1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00
FFmpeg/libavcodec/executor.c
Nuo Mi 846fbc395b avcodec/vvc: simplify priority logical to improve performance for 4K/8K
For 4K/8K video processing, it's possible to have over 1,000 tasks pending on the executor.
In such cases, O(n) and O(log(n)) insertion times are too costly.
Reducing this to O(1) will significantly decrease the time spent in critical sections

clip                                                        | before | after  | delta
------------------------------------------------------------|--------|--------|-------
VVC_HDR_UHDTV2_OpenGOP_7680x4320_50fps_HLG10.bit            |    24  |   27   |  12.5%
VVC_HDR_UHDTV2_OpenGOP_7680x4320_50fps_HLG10_HighBitrate.bit|    12  |   17   |  41.7%
tears_of_steel_4k_8M_8bit_2000.vvc                          |    34  |  102   | 200.0%
VVC_UHDTV1_OpenGOP_3840x2160_60fps_HLG10.bit                |   126  |  128   |   1.6%
RitualDance_1920x1080_60_10_420_37_RA.266                   |   350  |  378   |   8.0%
NovosobornayaSquare_1920x1080.bin                           |   341  |  369   |   8.2%
Tango2_3840x2160_60_10_420_27_LD.266                        |    69  |   70   |   1.4%
RitualDance_1920x1080_60_10_420_32_LD.266                   |   243  |  259   |   6.6%
Chimera_8bit_1080P_1000_frames.vvc                          |   420  |  392   |  -6.7%
BQTerrace_1920x1080_60_10_420_22_RA.vvc                     |   148  |  144   |  -2.7%
2024-10-04 21:58:42 +08:00

234 lines
5.5 KiB
C

/*
* Copyright (C) 2024 Nuo Mi
*
* 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 "config.h"
#include <stdbool.h>
#include "libavutil/mem.h"
#include "libavutil/thread.h"
#include "executor.h"
#if !HAVE_THREADS
#define ExecutorThread char
#define executor_thread_create(t, a, s, ar) 0
#define executor_thread_join(t, r) do {} while(0)
#else
#define ExecutorThread pthread_t
#define executor_thread_create(t, a, s, ar) pthread_create(t, a, s, ar)
#define executor_thread_join(t, r) pthread_join(t, r)
#endif //!HAVE_THREADS
typedef struct ThreadInfo {
FFExecutor *e;
ExecutorThread thread;
} ThreadInfo;
typedef struct Queue {
FFTask *head;
FFTask *tail;
} Queue;
struct FFExecutor {
FFTaskCallbacks cb;
int thread_count;
bool recursive;
ThreadInfo *threads;
uint8_t *local_contexts;
AVMutex lock;
AVCond cond;
int die;
Queue *q;
};
static FFTask* remove_task(Queue *q)
{
FFTask *t = q->head;
if (t) {
q->head = t->next;
t->next = NULL;
if (!q->head)
q->tail = NULL;
}
return t;
}
static void add_task(Queue *q, FFTask *t)
{
t->next = NULL;
if (!q->head)
q->tail = q->head = t;
else
q->tail = q->tail->next = t;
}
static int run_one_task(FFExecutor *e, void *lc)
{
FFTaskCallbacks *cb = &e->cb;
FFTask *t = NULL;
for (int i = 0; i < e->cb.priorities && !t; i++)
t = remove_task(e->q + i);
if (t) {
if (e->thread_count > 0)
ff_mutex_unlock(&e->lock);
cb->run(t, lc, cb->user_data);
if (e->thread_count > 0)
ff_mutex_lock(&e->lock);
return 1;
}
return 0;
}
#if HAVE_THREADS
static void *executor_worker_task(void *data)
{
ThreadInfo *ti = (ThreadInfo*)data;
FFExecutor *e = ti->e;
void *lc = e->local_contexts + (ti - e->threads) * e->cb.local_context_size;
ff_mutex_lock(&e->lock);
while (1) {
if (e->die) break;
if (!run_one_task(e, lc)) {
//no task in one loop
ff_cond_wait(&e->cond, &e->lock);
}
}
ff_mutex_unlock(&e->lock);
return NULL;
}
#endif
static void executor_free(FFExecutor *e, const int has_lock, const int has_cond)
{
if (e->thread_count) {
//signal die
ff_mutex_lock(&e->lock);
e->die = 1;
ff_cond_broadcast(&e->cond);
ff_mutex_unlock(&e->lock);
for (int i = 0; i < e->thread_count; i++)
executor_thread_join(e->threads[i].thread, NULL);
}
if (has_cond)
ff_cond_destroy(&e->cond);
if (has_lock)
ff_mutex_destroy(&e->lock);
av_free(e->threads);
av_free(e->q);
av_free(e->local_contexts);
av_free(e);
}
FFExecutor* ff_executor_alloc(const FFTaskCallbacks *cb, int thread_count)
{
FFExecutor *e;
int has_lock = 0, has_cond = 0;
if (!cb || !cb->user_data || !cb->run || !cb->priorities)
return NULL;
e = av_mallocz(sizeof(*e));
if (!e)
return NULL;
e->cb = *cb;
e->local_contexts = av_calloc(FFMAX(thread_count, 1), e->cb.local_context_size);
if (!e->local_contexts)
goto free_executor;
e->q = av_calloc(e->cb.priorities, sizeof(Queue));
if (!e->q)
goto free_executor;
e->threads = av_calloc(FFMAX(thread_count, 1), sizeof(*e->threads));
if (!e->threads)
goto free_executor;
if (!thread_count)
return e;
has_lock = !ff_mutex_init(&e->lock, NULL);
has_cond = !ff_cond_init(&e->cond, NULL);
if (!has_lock || !has_cond)
goto free_executor;
for (/* nothing */; e->thread_count < thread_count; e->thread_count++) {
ThreadInfo *ti = e->threads + e->thread_count;
ti->e = e;
if (executor_thread_create(&ti->thread, NULL, executor_worker_task, ti))
goto free_executor;
}
return e;
free_executor:
executor_free(e, has_lock, has_cond);
return NULL;
}
void ff_executor_free(FFExecutor **executor)
{
int thread_count;
if (!executor || !*executor)
return;
thread_count = (*executor)->thread_count;
executor_free(*executor, thread_count, thread_count);
*executor = NULL;
}
void ff_executor_execute(FFExecutor *e, FFTask *t)
{
if (e->thread_count)
ff_mutex_lock(&e->lock);
if (t)
add_task(e->q + t->priority % e->cb.priorities, t);
if (e->thread_count) {
ff_cond_signal(&e->cond);
ff_mutex_unlock(&e->lock);
}
if (!e->thread_count || !HAVE_THREADS) {
if (e->recursive)
return;
e->recursive = true;
// We are running in a single-threaded environment, so we must handle all tasks ourselves
while (run_one_task(e, e->local_contexts))
/* nothing */;
e->recursive = false;
}
}