* to Michael Niedermayer <michaelni@gmx.at> for writing initial
* implementation.
*
- * This file is part of Libav.
+ * This file is part of FFmpeg.
*
- * Libav is free software; you can redistribute it and/or
+ * 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.
*
- * Libav is distributed in the hope that it will be useful,
+ * 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 Libav; if not, write to the Free Software
+ * 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"
+
+#if HAVE_SCHED_GETAFFINITY
+#define _GNU_SOURCE
+#include <sched.h>
+#endif
+#if HAVE_GETPROCESSAFFINITYMASK
+#include <windows.h>
+#endif
+#if HAVE_SYSCTL
+#if HAVE_SYS_PARAM_H
+#include <sys/param.h>
+#endif
+#include <sys/types.h>
+#include <sys/param.h>
+#include <sys/sysctl.h>
+#endif
+#if HAVE_SYSCONF
+#include <unistd.h>
+#endif
+
#include "avcodec.h"
#include "internal.h"
#include "thread.h"
#include <pthread.h>
#elif HAVE_W32THREADS
#include "w32pthreads.h"
+#elif HAVE_OS2THREADS
+#include "os2threads.h"
#endif
typedef int (action_func)(AVCodecContext *c, void *arg);
int die; ///< Set when threads should exit.
} FrameThreadContext;
+
+/* H264 slice threading seems to be buggy with more than 16 threads,
+ * limit the number of threads to 16 for automatic detection */
+#define MAX_AUTO_THREADS 16
+
+static int get_logical_cpus(AVCodecContext *avctx)
+{
+ int ret, nb_cpus = 1;
+#if HAVE_SCHED_GETAFFINITY && defined(CPU_COUNT)
+ cpu_set_t cpuset;
+
+ CPU_ZERO(&cpuset);
+
+ ret = sched_getaffinity(0, sizeof(cpuset), &cpuset);
+ if (!ret) {
+ nb_cpus = CPU_COUNT(&cpuset);
+ }
+#elif HAVE_GETPROCESSAFFINITYMASK
+ DWORD_PTR proc_aff, sys_aff;
+ ret = GetProcessAffinityMask(GetCurrentProcess(), &proc_aff, &sys_aff);
+ if (ret)
+ nb_cpus = av_popcount64(proc_aff);
+#elif HAVE_SYSCTL && defined(HW_NCPU)
+ int mib[2] = { CTL_HW, HW_NCPU };
+ size_t len = sizeof(nb_cpus);
+
+ ret = sysctl(mib, 2, &nb_cpus, &len, NULL, 0);
+ if (ret == -1)
+ nb_cpus = 0;
+#elif HAVE_SYSCONF && defined(_SC_NPROC_ONLN)
+ nb_cpus = sysconf(_SC_NPROC_ONLN);
+#elif HAVE_SYSCONF && defined(_SC_NPROCESSORS_ONLN)
+ nb_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+#endif
+ av_log(avctx, AV_LOG_DEBUG, "detected %d logical cores\n", nb_cpus);
+
+ if (avctx->height)
+ nb_cpus = FFMIN(nb_cpus, (avctx->height+15)/16);
+
+ return nb_cpus;
+}
+
+
static void* attribute_align_arg worker(void *v)
{
AVCodecContext *avctx = v;
ThreadContext *c;
int thread_count = avctx->thread_count;
+ if (!thread_count) {
+ int nb_cpus = get_logical_cpus(avctx);
+ // use number of cores + 1 as thread count if there is more than one
+ if (nb_cpus > 1)
+ thread_count = avctx->thread_count = FFMIN(nb_cpus + 1, MAX_AUTO_THREADS);
+ else
+ thread_count = avctx->thread_count = 1;
+ }
+
if (thread_count <= 1) {
avctx->active_thread_type = 0;
return 0;
AVCodec *codec = avctx->codec;
while (1) {
+ int i;
if (p->state == STATE_INPUT_READY && !fctx->die) {
pthread_mutex_lock(&p->mutex);
while (p->state == STATE_INPUT_READY && !fctx->die)
if (fctx->die) break;
- if (!codec->update_thread_context && avctx->thread_safe_callbacks)
+ if (!codec->update_thread_context && (avctx->thread_safe_callbacks || avctx->get_buffer == avcodec_default_get_buffer))
ff_thread_finish_setup(avctx);
pthread_mutex_lock(&p->mutex);
p->state = STATE_INPUT_READY;
pthread_mutex_lock(&p->progress_mutex);
+ for (i = 0; i < MAX_BUFFERS; i++)
+ if (p->progress_used[i]) {
+ p->progress[i][0] = INT_MAX;
+ p->progress[i][1] = INT_MAX;
+ }
+ pthread_cond_broadcast(&p->progress_cond);
pthread_cond_signal(&p->output_cond);
pthread_mutex_unlock(&p->progress_mutex);
dst->has_b_frames = src->has_b_frames;
dst->idct_algo = src->idct_algo;
- dst->slice_count = src->slice_count;
dst->bits_per_coded_sample = src->bits_per_coded_sample;
dst->sample_aspect_ratio = src->sample_aspect_ratio;
}
if (for_user) {
+ dst->delay = src->thread_count - 1;
dst->coded_frame = src->coded_frame;
} else {
if (dst->codec->update_thread_context)
*
* @param dst The destination context.
* @param src The source context.
+ * @return 0 on success, negative error code on failure
*/
-static void update_context_from_user(AVCodecContext *dst, AVCodecContext *src)
+static int update_context_from_user(AVCodecContext *dst, AVCodecContext *src)
{
#define copy_fields(s, e) memcpy(&dst->s, &src->s, (char*)&dst->e - (char*)&dst->s);
dst->flags = src->flags;
dst->frame_number = src->frame_number;
dst->reordered_opaque = src->reordered_opaque;
+ dst->thread_safe_callbacks = src->thread_safe_callbacks;
+
+ if (src->slice_count && src->slice_offset) {
+ if (dst->slice_count < src->slice_count) {
+ int *tmp = av_realloc(dst->slice_offset, src->slice_count *
+ sizeof(*dst->slice_offset));
+ if (!tmp) {
+ av_free(dst->slice_offset);
+ return AVERROR(ENOMEM);
+ }
+ dst->slice_offset = tmp;
+ }
+ memcpy(dst->slice_offset, src->slice_offset,
+ src->slice_count * sizeof(*dst->slice_offset));
+ }
+ dst->slice_count = src->slice_count;
+ return 0;
#undef copy_fields
}
*/
p = &fctx->threads[fctx->next_decoding];
- update_context_from_user(p->avctx, avctx);
+ err = update_context_from_user(p->avctx, avctx);
+ if (err) return err;
err = submit_packet(p, avpkt);
if (err) return err;
*picture = p->frame;
*got_picture_ptr = p->got_frame;
picture->pkt_dts = p->avpkt.dts;
+ picture->sample_aspect_ratio = avctx->sample_aspect_ratio;
+ picture->width = avctx->width;
+ picture->height = avctx->height;
+ picture->format = avctx->pix_fmt;
/*
* A later call with avkpt->size == 0 may loop over all threads,
if (!(avctx->active_thread_type&FF_THREAD_FRAME)) return;
+ if(p->state == STATE_SETUP_FINISHED){
+ av_log(avctx, AV_LOG_WARNING, "Multiple ff_thread_finish_setup() calls\n");
+ }
+
pthread_mutex_lock(&p->progress_mutex);
p->state = STATE_SETUP_FINISHED;
pthread_cond_broadcast(&p->progress_cond);
pthread_cond_wait(&p->output_cond, &p->progress_mutex);
pthread_mutex_unlock(&p->progress_mutex);
}
+ p->got_frame = 0;
}
}
if (p->thread_init)
pthread_join(p->thread, NULL);
+ p->thread_init=0;
if (codec->close)
codec->close(p->avctx);
if (i) {
av_freep(&p->avctx->priv_data);
av_freep(&p->avctx->internal);
+ av_freep(&p->avctx->slice_offset);
}
av_freep(&p->avctx);
FrameThreadContext *fctx;
int i, err = 0;
+ if (!thread_count) {
+ int nb_cpus = get_logical_cpus(avctx);
+ if ((avctx->debug & (FF_DEBUG_VIS_QP | FF_DEBUG_VIS_MB_TYPE)) || avctx->debug_mv)
+ nb_cpus = 1;
+ // use number of cores + 1 as thread count if there is more than one
+ if (nb_cpus > 1)
+ thread_count = avctx->thread_count = FFMIN(nb_cpus + 1, MAX_AUTO_THREADS);
+ else
+ thread_count = avctx->thread_count = 1;
+ }
+
if (thread_count <= 1) {
avctx->active_thread_type = 0;
return 0;
err = AVERROR(ENOMEM);
goto error;
}
- *(copy->internal) = *(src->internal);
+ *copy->internal = *src->internal;
copy->internal->is_copy = 1;
if (codec->init_thread_copy)
if (err) goto error;
- if (!pthread_create(&p->thread, NULL, frame_worker_thread, p))
- p->thread_init = 1;
+ p->thread_init= !pthread_create(&p->thread, NULL, frame_worker_thread, p);
+ if(!p->thread_init)
+ goto error;
}
return 0;
f->owner = avctx;
+ ff_init_buffer_info(avctx, f);
+
if (!(avctx->active_thread_type&FF_THREAD_FRAME)) {
f->thread_opaque = NULL;
return avctx->get_buffer(avctx, f);
}
if (p->state != STATE_SETTING_UP &&
- (avctx->codec->update_thread_context || !avctx->thread_safe_callbacks)) {
+ (avctx->codec->update_thread_context || (!avctx->thread_safe_callbacks &&
+ avctx->get_buffer != avcodec_default_get_buffer))) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() cannot be called after ff_thread_finish_setup()\n");
return -1;
}
p->requested_frame = f;
p->state = STATE_GET_BUFFER;
pthread_mutex_lock(&p->progress_mutex);
- pthread_cond_signal(&p->progress_cond);
+ pthread_cond_broadcast(&p->progress_cond);
while (p->state != STATE_SETTING_UP)
pthread_cond_wait(&p->progress_cond, &p->progress_mutex);
} else if (avctx->codec->capabilities & CODEC_CAP_SLICE_THREADS &&
avctx->thread_type & FF_THREAD_SLICE) {
avctx->active_thread_type = FF_THREAD_SLICE;
+ } else if (!(avctx->codec->capabilities & CODEC_CAP_AUTO_THREADS)) {
+ avctx->thread_count = 1;
+ avctx->active_thread_type = 0;
}
}
projects / ffmpeg / blobdiff