Commit 399cec2f authored by Michael Niedermayer's avatar Michael Niedermayer

count > thread_count for execute()

Originally committed as revision 2842 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent 43465395
......@@ -17,7 +17,7 @@ extern "C" {
#define FFMPEG_VERSION_INT 0x000408
#define FFMPEG_VERSION "0.4.8"
#define LIBAVCODEC_BUILD 4706
#define LIBAVCODEC_BUILD 4707
#define LIBAVCODEC_VERSION_INT FFMPEG_VERSION_INT
#define LIBAVCODEC_VERSION FFMPEG_VERSION
......@@ -1547,7 +1547,7 @@ typedef struct AVCodecContext {
* the codec may call this to execute several independant things. it will return only after
* finishing all tasks, the user may replace this with some multithreaded implementation, the
* default implementation will execute the parts serially
* @param count the number of functions this will be identical to thread_count if possible
* @param count the number of things to execute
* - encoding: set by lavc, user can override
* - decoding: set by lavc, user can override
*/
......
......@@ -24,30 +24,57 @@
#include "avcodec.h"
#include "common.h"
typedef struct JobContext{
sem_t available_sem;
int assigned;
int (*func)(AVCodecContext *c, void *arg);
void *arg;
int ret;
}JobContext;
typedef struct ThreadContext{
typedef struct WorkerContext{
AVCodecContext *avctx;
pthread_t thread;
int start_index;
sem_t work_sem;
sem_t done_sem;
int (*func)(AVCodecContext *c, void *arg);
void *arg;
int ret;
}WorkerContext;
typedef struct ThreadContext{
WorkerContext *worker;
JobContext *job;
int job_count;
int allocated_job_count;
}ThreadContext;
static void * thread_func(void *v){
ThreadContext *c= v;
WorkerContext *w= v;
ThreadContext *c= w->avctx->thread_opaque;
int i;
for(;;){
//printf("thread_func %X enter wait\n", (int)v); fflush(stdout);
sem_wait(&c->work_sem);
//printf("thread_func %X after wait (func=%X)\n", (int)v, (int)c->func); fflush(stdout);
if(c->func)
c->ret= c->func(c->avctx, c->arg);
else
return NULL;
//printf("thread_func %X signal complete\n", (int)v); fflush(stdout);
sem_post(&c->done_sem);
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X enter wait\n", (int)v);
sem_wait(&w->work_sem);
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X after wait\n", (int)v);
if(c->job_count == 0)
break;
for(i=0; i<c->job_count; i++){
int index= (i + w->start_index) % c->job_count;
JobContext *j= &c->job[index];
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X first check of %d\n", (int)v, index);
if(j->assigned) continue; //unsynced check, if != 0 it is already given to another worker, it never becomes available before the next execute() call so this should be safe
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X second check of %d\n", (int)v, index);
if(sem_trywait(&j->available_sem) == 0){
j->assigned=1;
j->ret= j->func(w->avctx, j->arg);
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X done %d\n", (int)v, index);
}
}
//av_log(w->avctx, AV_LOG_DEBUG, "thread_func %X complete\n", (int)v);
sem_post(&w->done_sem);
}
return NULL;
......@@ -59,73 +86,108 @@ static void * thread_func(void *v){
*/
void avcodec_thread_free(AVCodecContext *s){
ThreadContext *c= s->thread_opaque;
int i;
int i, val;
for(i=0; i<s->thread_count; i++){
int val;
for(i=0; i<c->allocated_job_count; i++){
sem_getvalue(&c->job[i].available_sem, &val); assert(val == 0);
sem_destroy(&c->job[i].available_sem);
}
sem_getvalue(&c[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c[i].done_sem, &val); assert(val == 0);
c->job_count= 0;
for(i=0; i<s->thread_count; i++){
sem_getvalue(&c->worker[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c->worker[i].done_sem, &val); assert(val == 0);
c[i].func= NULL;
sem_post(&c[i].work_sem);
pthread_join(c[i].thread, NULL);
sem_destroy(&c[i].work_sem);
sem_destroy(&c[i].done_sem);
sem_post(&c->worker[i].work_sem);
pthread_join(c->worker[i].thread, NULL);
sem_destroy(&c->worker[i].work_sem);
sem_destroy(&c->worker[i].done_sem);
}
av_freep(&c->job);
av_freep(&c->worker);
av_freep(&s->thread_opaque);
}
int avcodec_thread_execute(AVCodecContext *s, int (*func)(AVCodecContext *c2, void *arg2),void **arg, int *ret, int count){
int avcodec_thread_execute(AVCodecContext *s, int (*func)(AVCodecContext *c2, void *arg2),void **arg, int *ret, int job_count){
ThreadContext *c= s->thread_opaque;
int i, val;
assert(s == c->avctx);
assert(count <= s->thread_count);
if(job_count > c->allocated_job_count){
c->job= av_realloc(c->job, job_count*sizeof(JobContext));
for(i=c->allocated_job_count; i<job_count; i++){
memset(&c->job[i], 0, sizeof(JobContext));
c->allocated_job_count++;
if(sem_init(&c->job[i].available_sem, 0, 0))
return -1;
}
}
c->job_count= job_count;
/* note, we can be certain that this is not called with the same AVCodecContext by different threads at the same time */
for(i=0; i<count; i++){
sem_getvalue(&c[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c[i].done_sem, &val); assert(val == 0);
for(i=0; i<job_count; i++){
sem_getvalue(&c->job[i].available_sem, &val); assert(val == 0);
c[i].arg= arg[i];
c[i].func= func;
c[i].ret= 12345;
sem_post(&c[i].work_sem);
c->job[i].arg= arg[i];
c->job[i].func= func;
c->job[i].ret= 12345;
c->job[i].assigned= 0;
sem_post(&c->job[i].available_sem);
}
for(i=0; i<count; i++){
sem_wait(&c[i].done_sem);
sem_getvalue(&c[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c[i].done_sem, &val); assert(val == 0);
for(i=0; i<s->thread_count && i<job_count; i++){
sem_getvalue(&c->worker[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c->worker[i].done_sem, &val); assert(val == 0);
c[i].func= NULL;
if(ret) ret[i]= c[i].ret;
c->worker[i].start_index= (i + job_count/2)/job_count;
//av_log(s, AV_LOG_DEBUG, "start worker %d\n", i);
sem_post(&c->worker[i].work_sem);
}
for(i=0; i<s->thread_count && i<job_count; i++){
//av_log(s, AV_LOG_DEBUG, "wait for worker %d\n", i);
sem_wait(&c->worker[i].done_sem);
sem_getvalue(&c->worker[i].work_sem, &val); assert(val == 0);
sem_getvalue(&c->worker[i].done_sem, &val); assert(val == 0);
}
for(i=0; i<job_count; i++){
sem_getvalue(&c->job[i].available_sem, &val); assert(val == 0);
c->job[i].func= NULL;
if(ret) ret[i]= c->job[i].ret;
}
return 0;
}
int avcodec_thread_init(AVCodecContext *s, int thread_count){
int i;
ThreadContext *c;
WorkerContext *worker;
s->thread_count= thread_count;
assert(!s->thread_opaque);
c= av_mallocz(sizeof(ThreadContext)*thread_count);
c= av_mallocz(sizeof(ThreadContext));
worker= av_mallocz(sizeof(WorkerContext)*thread_count);
s->thread_opaque= c;
c->worker= worker;
for(i=0; i<thread_count; i++){
//printf("init semaphors %d\n", i); fflush(stdout);
c[i].avctx= s;
if(sem_init(&c[i].work_sem, 0, 0))
worker[i].avctx= s;
if(sem_init(&worker[i].work_sem, 0, 0))
goto fail;
if(sem_init(&c[i].done_sem, 0, 0))
if(sem_init(&worker[i].done_sem, 0, 0))
goto fail;
//printf("create thread %d\n", i); fflush(stdout);
if(pthread_create(&c[i].thread, NULL, thread_func, &c[i]))
if(pthread_create(&worker[i].thread, NULL, thread_func, &worker[i]))
goto fail;
}
//printf("init done\n"); fflush(stdout);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment