]> git.sesse.net Git - ffmpeg/blobdiff - ffplay.c
dirac: enable diracdsp_mmx
[ffmpeg] / ffplay.c
index 29e1120d8849a4b7778fb0111489731d2ce5362e..5a32d4e5dcbd4a71ff8c4dd04f9ddce6217026db 100644 (file)
--- a/ffplay.c
+++ b/ffplay.c
@@ -38,6 +38,7 @@
 #include "libavcodec/audioconvert.h"
 #include "libavutil/opt.h"
 #include "libavcodec/avfft.h"
+#include "libswresample/swresample.h"
 
 #if CONFIG_AVFILTER
 # include "libavfilter/avcodec.h"
@@ -70,8 +71,6 @@ const int program_birth_year = 2003;
 /* no AV correction is done if too big error */
 #define AV_NOSYNC_THRESHOLD 10.0
 
-#define FRAME_SKIP_FACTOR 0.05
-
 /* maximum audio speed change to get correct sync */
 #define SAMPLE_CORRECTION_PERCENT_MAX 10
 
@@ -97,12 +96,13 @@ typedef struct PacketQueue {
 
 typedef struct VideoPicture {
     double pts;                                  ///<presentation time stamp for this picture
-    double target_clock;                         ///<av_gettime() time at which this should be displayed ideally
     double duration;                             ///<expected duration of the frame
     int64_t pos;                                 ///<byte position in file
+    int skip;
     SDL_Overlay *bmp;
     int width, height; /* source height & width */
     int allocated;
+    int reallocate;
     enum PixelFormat pix_fmt;
 
 #if CONFIG_AVFILTER
@@ -152,9 +152,9 @@ typedef struct VideoState {
     PacketQueue audioq;
     int audio_hw_buf_size;
     /* samples output by the codec. we reserve more space for avsync
-       compensation */
-    DECLARE_ALIGNED(16,uint8_t,audio_buf1)[(AVCODEC_MAX_AUDIO_FRAME_SIZE * 3) / 2];
-    DECLARE_ALIGNED(16,uint8_t,audio_buf2)[(AVCODEC_MAX_AUDIO_FRAME_SIZE * 3) / 2];
+       compensation, resampling and format conversion */
+    DECLARE_ALIGNED(16,uint8_t,audio_buf1)[AVCODEC_MAX_AUDIO_FRAME_SIZE * 4];
+    DECLARE_ALIGNED(16,uint8_t,audio_buf2)[AVCODEC_MAX_AUDIO_FRAME_SIZE * 4];
     uint8_t *audio_buf;
     unsigned int audio_buf_size; /* in bytes */
     int audio_buf_index; /* in bytes */
@@ -162,9 +162,18 @@ typedef struct VideoState {
     AVPacket audio_pkt_temp;
     AVPacket audio_pkt;
     enum AVSampleFormat audio_src_fmt;
-    AVAudioConvert *reformat_ctx;
+    enum AVSampleFormat audio_tgt_fmt;
+    int audio_src_channels;
+    int audio_tgt_channels;
+    int64_t audio_src_channel_layout;
+    int64_t audio_tgt_channel_layout;
+    int audio_src_freq;
+    int audio_tgt_freq;
+    struct SwrContext *swr_ctx;
     double audio_current_pts;
     double audio_current_pts_drift;
+    int frame_drops_early;
+    int frame_drops_late;
 
     enum ShowMode {
         SHOW_MODE_NONE = -1, SHOW_MODE_VIDEO = 0, SHOW_MODE_WAVES, SHOW_MODE_RDFT, SHOW_MODE_NB
@@ -189,7 +198,11 @@ typedef struct VideoState {
 
     double frame_timer;
     double frame_last_pts;
-    double frame_last_delay;
+    double frame_last_duration;
+    double frame_last_dropped_pts;
+    double frame_last_returned_time;
+    double frame_last_filter_delay;
+    int64_t frame_last_dropped_pos;
     double video_clock;                          ///<pts of last decoded frame / predicted pts of next decoded frame
     int video_stream;
     AVStream *video_st;
@@ -213,8 +226,6 @@ typedef struct VideoState {
     AVFilterContext *out_video_filter;          ///<the last filter in the video chain
 #endif
 
-    float skip_frames;
-    float skip_frames_index;
     int refresh;
 } VideoState;
 
@@ -258,6 +269,9 @@ static int exit_on_mousedown;
 static int loop=1;
 static int framedrop=-1;
 static enum ShowMode show_mode = SHOW_MODE_NONE;
+static const char *audio_codec_name;
+static const char *subtitle_codec_name;
+static const char *video_codec_name;
 
 static int rdftspeed=20;
 #if CONFIG_AVFILTER
@@ -729,7 +743,7 @@ static void video_audio_display(VideoState *s)
     nb_freq= 1<<(rdft_bits-1);
 
     /* compute display index : center on currently output samples */
-    channels = s->audio_st->codec->channels;
+    channels = s->audio_tgt_channels;
     nb_display_channels = channels;
     if (!s->paused) {
         int data_used= s->show_mode == SHOW_MODE_WAVES ? s->width : (2*nb_freq);
@@ -741,7 +755,7 @@ static void video_audio_display(VideoState *s)
            the last buffer computation */
         if (audio_callback_time) {
             time_diff = av_gettime() - audio_callback_time;
-            delay -= (time_diff * s->audio_st->codec->sample_rate) / 1000000;
+            delay -= (time_diff * s->audio_tgt_freq) / 1000000;
         }
 
         delay += 2*data_used;
@@ -936,13 +950,7 @@ static int video_open(VideoState *is){
     if(screen && is->width == screen->w && screen->w == w
        && is->height== screen->h && screen->h == h)
         return 0;
-
-#ifndef __APPLE__
     screen = SDL_SetVideoMode(w, h, 0, flags);
-#else
-    /* setting bits_per_pixel = 0 or 32 causes blank video on OS X */
-    screen = SDL_SetVideoMode(w, h, 24, flags);
-#endif
     if (!screen) {
         fprintf(stderr, "SDL: could not set video mode - exiting\n");
         do_exit(is);
@@ -1060,19 +1068,9 @@ static void stream_toggle_pause(VideoState *is)
     is->paused = !is->paused;
 }
 
-static double compute_target_time(double frame_current_pts, VideoState *is)
+static double compute_target_delay(double delay, VideoState *is)
 {
-    double delay, sync_threshold, diff;
-
-    /* compute nominal delay */
-    delay = frame_current_pts - is->frame_last_pts;
-    if (delay <= 0 || delay >= 10.0) {
-        /* if incorrect delay, use previous one */
-        delay = is->frame_last_delay;
-    } else {
-        is->frame_last_delay = delay;
-    }
-    is->frame_last_pts = frame_current_pts;
+    double sync_threshold, diff;
 
     /* update delay to follow master synchronisation source */
     if (((is->av_sync_type == AV_SYNC_AUDIO_MASTER && is->audio_st) ||
@@ -1092,12 +1090,31 @@ static double compute_target_time(double frame_current_pts, VideoState *is)
                 delay = 2 * delay;
         }
     }
-    is->frame_timer += delay;
 
-    av_dlog(NULL, "video: delay=%0.3f pts=%0.3f A-V=%f\n",
-            delay, frame_current_pts, -diff);
+    av_dlog(NULL, "video: delay=%0.3f A-V=%f\n",
+            delay, -diff);
 
-    return is->frame_timer;
+    return delay;
+}
+
+static void pictq_next_picture(VideoState *is) {
+    /* update queue size and signal for next picture */
+    if (++is->pictq_rindex == VIDEO_PICTURE_QUEUE_SIZE)
+        is->pictq_rindex = 0;
+
+    SDL_LockMutex(is->pictq_mutex);
+    is->pictq_size--;
+    SDL_CondSignal(is->pictq_cond);
+    SDL_UnlockMutex(is->pictq_mutex);
+}
+
+static void update_video_pts(VideoState *is, double pts, int64_t pos) {
+    double time = av_gettime() / 1000000.0;
+    /* update current video pts */
+    is->video_current_pts = pts;
+    is->video_current_pts_drift = is->video_current_pts - time;
+    is->video_current_pos = pos;
+    is->frame_last_pts = pts;
 }
 
 /* called to display each frame */
@@ -1105,43 +1122,60 @@ static void video_refresh(void *opaque)
 {
     VideoState *is = opaque;
     VideoPicture *vp;
+    double time;
 
     SubPicture *sp, *sp2;
 
     if (is->video_st) {
 retry:
         if (is->pictq_size == 0) {
+            SDL_LockMutex(is->pictq_mutex);
+            if (is->frame_last_dropped_pts != AV_NOPTS_VALUE && is->frame_last_dropped_pts > is->frame_last_pts) {
+                update_video_pts(is, is->frame_last_dropped_pts, is->frame_last_dropped_pos);
+                is->frame_last_dropped_pts = AV_NOPTS_VALUE;
+            }
+            SDL_UnlockMutex(is->pictq_mutex);
             //nothing to do, no picture to display in the que
         } else {
-            double time= av_gettime()/1000000.0;
-            double next_target;
+            double last_duration, duration, delay;
             /* dequeue the picture */
             vp = &is->pictq[is->pictq_rindex];
 
-            if(time < vp->target_clock)
+            if (vp->skip) {
+                pictq_next_picture(is);
+                goto retry;
+            }
+
+            /* compute nominal last_duration */
+            last_duration = vp->pts - is->frame_last_pts;
+            if (last_duration > 0 && last_duration < 10.0) {
+                /* if duration of the last frame was sane, update last_duration in video state */
+                is->frame_last_duration = last_duration;
+            }
+            delay = compute_target_delay(is->frame_last_duration, is);
+
+            time= av_gettime()/1000000.0;
+            if(time < is->frame_timer + delay)
                 return;
-            /* update current video pts */
-            is->video_current_pts = vp->pts;
-            is->video_current_pts_drift = is->video_current_pts - time;
-            is->video_current_pos = vp->pos;
-            if(is->pictq_size > 1){
-                VideoPicture *nextvp= &is->pictq[(is->pictq_rindex+1)%VIDEO_PICTURE_QUEUE_SIZE];
-                assert(nextvp->target_clock >= vp->target_clock);
-                next_target= nextvp->target_clock;
-            }else{
-                next_target= vp->target_clock + vp->duration;
+
+            if (delay > 0)
+                is->frame_timer += delay * FFMAX(1, floor((time-is->frame_timer) / delay));
+
+            SDL_LockMutex(is->pictq_mutex);
+            update_video_pts(is, vp->pts, vp->pos);
+            SDL_UnlockMutex(is->pictq_mutex);
+
+            if(is->pictq_size > 1) {
+                 VideoPicture *nextvp= &is->pictq[(is->pictq_rindex+1)%VIDEO_PICTURE_QUEUE_SIZE];
+                 duration = nextvp->pts - vp->pts; // More accurate this way, 1/time_base is often not reflecting FPS
+            } else {
+                 duration = vp->duration;
             }
-            if((framedrop>0 || (framedrop && is->audio_st)) && time > next_target){
-                is->skip_frames *= 1.0 + FRAME_SKIP_FACTOR;
+
+            if((framedrop>0 || (framedrop && is->audio_st)) && time > is->frame_timer + duration){
                 if(is->pictq_size > 1){
-                    /* update queue size and signal for next picture */
-                    if (++is->pictq_rindex == VIDEO_PICTURE_QUEUE_SIZE)
-                        is->pictq_rindex = 0;
-
-                    SDL_LockMutex(is->pictq_mutex);
-                    is->pictq_size--;
-                    SDL_CondSignal(is->pictq_cond);
-                    SDL_UnlockMutex(is->pictq_mutex);
+                    is->frame_drops_late++;
+                    pictq_next_picture(is);
                     goto retry;
                 }
             }
@@ -1194,14 +1228,7 @@ retry:
             if (!display_disable)
                 video_display(is);
 
-            /* update queue size and signal for next picture */
-            if (++is->pictq_rindex == VIDEO_PICTURE_QUEUE_SIZE)
-                is->pictq_rindex = 0;
-
-            SDL_LockMutex(is->pictq_mutex);
-            is->pictq_size--;
-            SDL_CondSignal(is->pictq_cond);
-            SDL_UnlockMutex(is->pictq_mutex);
+            pictq_next_picture(is);
         }
     } else if (is->audio_st) {
         /* draw the next audio frame */
@@ -1233,10 +1260,10 @@ retry:
             av_diff = 0;
             if (is->audio_st && is->video_st)
                 av_diff = get_audio_clock(is) - get_video_clock(is);
-            printf("%7.2f A-V:%7.3f s:%3.1f aq=%5dKB vq=%5dKB sq=%5dB f=%"PRId64"/%"PRId64"   \r",
+            printf("%7.2f A-V:%7.3f fd=%4d aq=%5dKB vq=%5dKB sq=%5dB f=%"PRId64"/%"PRId64"   \r",
                    get_master_clock(is),
                    av_diff,
-                   FFMAX(is->skip_frames-1, 0),
+                   is->frame_drops_early + is->frame_drops_late,
                    aqsize / 1024,
                    vqsize / 1024,
                    sqsize,
@@ -1320,9 +1347,6 @@ static int queue_picture(VideoState *is, AVFrame *src_frame, double pts1, int64_
     /* wait until we have space to put a new picture */
     SDL_LockMutex(is->pictq_mutex);
 
-    if(is->pictq_size>=VIDEO_PICTURE_QUEUE_SIZE && !is->refresh)
-        is->skip_frames= FFMAX(1.0 - FRAME_SKIP_FACTOR, is->skip_frames * (1.0-FRAME_SKIP_FACTOR));
-
     while (is->pictq_size >= VIDEO_PICTURE_QUEUE_SIZE &&
            !is->videoq.abort_request) {
         SDL_CondWait(is->pictq_cond, is->pictq_mutex);
@@ -1337,7 +1361,7 @@ static int queue_picture(VideoState *is, AVFrame *src_frame, double pts1, int64_
     vp->duration = frame_delay;
 
     /* alloc or resize hardware picture buffer */
-    if (!vp->bmp ||
+    if (!vp->bmp || vp->reallocate ||
 #if CONFIG_AVFILTER
         vp->width  != is->out_video_filter->inputs[0]->w ||
         vp->height != is->out_video_filter->inputs[0]->h) {
@@ -1347,7 +1371,8 @@ static int queue_picture(VideoState *is, AVFrame *src_frame, double pts1, int64_
 #endif
         SDL_Event event;
 
-        vp->allocated = 0;
+        vp->allocated  = 0;
+        vp->reallocate = 0;
 
         /* the allocation must be done in the main thread to avoid
            locking problems */
@@ -1414,13 +1439,12 @@ static int queue_picture(VideoState *is, AVFrame *src_frame, double pts1, int64_
 
         vp->pts = pts;
         vp->pos = pos;
+        vp->skip = 0;
 
         /* now we can update the picture count */
         if (++is->pictq_windex == VIDEO_PICTURE_QUEUE_SIZE)
             is->pictq_windex = 0;
         SDL_LockMutex(is->pictq_mutex);
-        vp->target_clock= compute_target_time(vp->pts, is);
-
         is->pictq_size++;
         SDL_UnlockMutex(is->pictq_mutex);
     }
@@ -1440,25 +1464,26 @@ static int get_video_frame(VideoState *is, AVFrame *frame, int64_t *pts, AVPacke
         SDL_LockMutex(is->pictq_mutex);
         //Make sure there are no long delay timers (ideally we should just flush the que but thats harder)
         for (i = 0; i < VIDEO_PICTURE_QUEUE_SIZE; i++) {
-            is->pictq[i].target_clock= 0;
+            is->pictq[i].skip = 1;
         }
         while (is->pictq_size && !is->videoq.abort_request) {
             SDL_CondWait(is->pictq_cond, is->pictq_mutex);
         }
         is->video_current_pos = -1;
-        SDL_UnlockMutex(is->pictq_mutex);
-
         is->frame_last_pts = AV_NOPTS_VALUE;
-        is->frame_last_delay = 0;
+        is->frame_last_duration = 0;
         is->frame_timer = (double)av_gettime() / 1000000.0;
-        is->skip_frames = 1;
-        is->skip_frames_index = 0;
+        is->frame_last_dropped_pts = AV_NOPTS_VALUE;
+        SDL_UnlockMutex(is->pictq_mutex);
+
         return 0;
     }
 
     avcodec_decode_video2(is->video_st->codec, frame, &got_picture, pkt);
 
     if (got_picture) {
+        int ret = 1;
+
         if (decoder_reorder_pts == -1) {
             *pts = frame->best_effort_timestamp;
         } else if (decoder_reorder_pts) {
@@ -1471,12 +1496,29 @@ static int get_video_frame(VideoState *is, AVFrame *frame, int64_t *pts, AVPacke
             *pts = 0;
         }
 
-        is->skip_frames_index += 1;
-        if(is->skip_frames_index >= is->skip_frames){
-            is->skip_frames_index -= FFMAX(is->skip_frames, 1.0);
-            return 1;
+        if (((is->av_sync_type == AV_SYNC_AUDIO_MASTER && is->audio_st) || is->av_sync_type == AV_SYNC_EXTERNAL_CLOCK) &&
+             (framedrop>0 || (framedrop && is->audio_st))) {
+            SDL_LockMutex(is->pictq_mutex);
+            if (is->frame_last_pts != AV_NOPTS_VALUE && *pts) {
+                double clockdiff = get_video_clock(is) - get_master_clock(is);
+                double dpts = av_q2d(is->video_st->time_base) * *pts;
+                double ptsdiff = dpts - is->frame_last_pts;
+                if (fabs(clockdiff) < AV_NOSYNC_THRESHOLD &&
+                     ptsdiff > 0 && ptsdiff < AV_NOSYNC_THRESHOLD &&
+                     clockdiff + ptsdiff - is->frame_last_filter_delay < 0) {
+                    is->frame_last_dropped_pos = pkt->pos;
+                    is->frame_last_dropped_pts = dpts;
+                    is->frame_drops_early++;
+                    ret = 0;
+                }
+            }
+            SDL_UnlockMutex(is->pictq_mutex);
         }
 
+        if (ret)
+            is->frame_last_returned_time = av_gettime() / 1000000.0;
+
+        return ret;
     }
     return 0;
 }
@@ -1800,6 +1842,10 @@ static int video_thread(void *arg)
 
         if (ret < 0) goto the_end;
 
+        is->frame_last_filter_delay = av_gettime() / 1000000.0 - is->frame_last_returned_time;
+        if (fabs(is->frame_last_filter_delay) > AV_NOSYNC_THRESHOLD / 10.0)
+            is->frame_last_filter_delay = 0;
+
 #if CONFIG_AVFILTER
         if (!picref)
             continue;
@@ -1919,7 +1965,7 @@ static int synchronize_audio(VideoState *is, short *samples,
     int n, samples_size;
     double ref_clock;
 
-    n = 2 * is->audio_st->codec->channels;
+    n = av_get_bytes_per_sample(is->audio_tgt_fmt) * is->audio_tgt_channels;
     samples_size = samples_size1;
 
     /* if not master, then we try to remove or add samples to correct the clock */
@@ -1941,15 +1987,15 @@ static int synchronize_audio(VideoState *is, short *samples,
                 avg_diff = is->audio_diff_cum * (1.0 - is->audio_diff_avg_coef);
 
                 if (fabs(avg_diff) >= is->audio_diff_threshold) {
-                    wanted_size = samples_size + ((int)(diff * is->audio_st->codec->sample_rate) * n);
+                    wanted_size = samples_size + ((int)(diff * is->audio_tgt_freq) * n);
                     nb_samples = samples_size / n;
 
                     min_size = ((nb_samples * (100 - SAMPLE_CORRECTION_PERCENT_MAX)) / 100) * n;
                     max_size = ((nb_samples * (100 + SAMPLE_CORRECTION_PERCENT_MAX)) / 100) * n;
                     if (wanted_size < min_size)
                         wanted_size = min_size;
-                    else if (wanted_size > max_size)
-                        wanted_size = max_size;
+                    else if (wanted_size > FFMIN3(max_size, sizeof(is->audio_buf1), sizeof(is->audio_buf2)))
+                        wanted_size = FFMIN3(max_size, sizeof(is->audio_buf1), sizeof(is->audio_buf2));
 
                     /* add or remove samples to correction the synchro */
                     if (wanted_size < samples_size) {
@@ -1992,12 +2038,18 @@ static int audio_decode_frame(VideoState *is, double *pts_ptr)
     AVPacket *pkt_temp = &is->audio_pkt_temp;
     AVPacket *pkt = &is->audio_pkt;
     AVCodecContext *dec= is->audio_st->codec;
-    int n, len1, data_size;
+    int len1, len2, data_size, resampled_data_size;
+    int64_t dec_channel_layout;
     double pts;
+    int new_packet = 0;
+    int flush_complete = 0;
 
     for(;;) {
         /* NOTE: the audio packet can contain several frames */
-        while (pkt_temp->size > 0) {
+        while (pkt_temp->size > 0 || (!pkt_temp->data && new_packet)) {
+            if (flush_complete)
+                break;
+            new_packet = 0;
             data_size = sizeof(is->audio_buf1);
             len1 = avcodec_decode_audio3(dec,
                                         (int16_t *)is->audio_buf1, &data_size,
@@ -2010,47 +2062,62 @@ static int audio_decode_frame(VideoState *is, double *pts_ptr)
 
             pkt_temp->data += len1;
             pkt_temp->size -= len1;
-            if (data_size <= 0)
+
+            if (data_size <= 0) {
+                /* stop sending empty packets if the decoder is finished */
+                if (!pkt_temp->data && dec->codec->capabilities & CODEC_CAP_DELAY)
+                    flush_complete = 1;
                 continue;
+            }
 
-            if (dec->sample_fmt != is->audio_src_fmt) {
-                if (is->reformat_ctx)
-                    av_audio_convert_free(is->reformat_ctx);
-                is->reformat_ctx= av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
-                                                         dec->sample_fmt, 1, NULL, 0);
-                if (!is->reformat_ctx) {
-                    fprintf(stderr, "Cannot convert %s sample format to %s sample format\n",
+            dec_channel_layout = (dec->channel_layout && dec->channels == av_get_channel_layout_nb_channels(dec->channel_layout)) ? dec->channel_layout : av_get_default_channel_layout(dec->channels);
+
+            if (dec->sample_fmt != is->audio_src_fmt || dec_channel_layout != is->audio_src_channel_layout || dec->sample_rate != is->audio_src_freq) {
+                if (is->swr_ctx)
+                    swr_free(&is->swr_ctx);
+                is->swr_ctx = swr_alloc2(NULL, is->audio_tgt_channel_layout, is->audio_tgt_fmt, is->audio_tgt_freq,
+                                               dec_channel_layout,          dec->sample_fmt,   dec->sample_rate,
+                                               0, NULL);
+                if (!is->swr_ctx || swr_init(is->swr_ctx) < 0) {
+                    fprintf(stderr, "Cannot create sample rate converter for conversion of %d Hz %s %d channels to %d Hz %s %d channels!\n",
+                        dec->sample_rate,
                         av_get_sample_fmt_name(dec->sample_fmt),
-                        av_get_sample_fmt_name(AV_SAMPLE_FMT_S16));
-                        break;
+                        dec->channels,
+                        is->audio_tgt_freq,
+                        av_get_sample_fmt_name(is->audio_tgt_fmt),
+                        is->audio_tgt_channels);
+                    break;
                 }
-                is->audio_src_fmt= dec->sample_fmt;
+                is->audio_src_channel_layout = dec_channel_layout;
+                is->audio_src_channels = dec->channels;
+                is->audio_src_freq = dec->sample_rate;
+                is->audio_src_fmt = dec->sample_fmt;
             }
 
-            if (is->reformat_ctx) {
-                const void *ibuf[6]= {is->audio_buf1};
-                void *obuf[6]= {is->audio_buf2};
-                int istride[6]= {av_get_bytes_per_sample(dec->sample_fmt)};
-                int ostride[6]= {2};
-                int len= data_size/istride[0];
-                if (av_audio_convert(is->reformat_ctx, obuf, ostride, ibuf, istride, len)<0) {
-                    printf("av_audio_convert() failed\n");
+            resampled_data_size = data_size;
+            if (is->swr_ctx) {
+                const uint8_t *in[] = {is->audio_buf1};
+                uint8_t *out[] = {is->audio_buf2};
+                len2 = swr_convert(is->swr_ctx, out, sizeof(is->audio_buf2) / is->audio_tgt_channels / av_get_bytes_per_sample(is->audio_tgt_fmt),
+                                                in, data_size / dec->channels / av_get_bytes_per_sample(dec->sample_fmt));
+                if (len2 < 0) {
+                    fprintf(stderr, "audio_resample() failed\n");
                     break;
                 }
-                is->audio_buf= is->audio_buf2;
-                /* FIXME: existing code assume that data_size equals framesize*channels*2
-                          remove this legacy cruft */
-                data_size= len*2;
-            }else{
+                if (len2 == sizeof(is->audio_buf2) / is->audio_tgt_channels / av_get_bytes_per_sample(is->audio_tgt_fmt)) {
+                    fprintf(stderr, "warning: audio buffer is probably too small\n");
+                    swr_init(is->swr_ctx);
+                }
+                is->audio_buf = is->audio_buf2;
+                resampled_data_size = len2 * is->audio_tgt_channels * av_get_bytes_per_sample(is->audio_tgt_fmt);
+            } else {
                 is->audio_buf= is->audio_buf1;
             }
 
             /* if no pts, then compute it */
             pts = is->audio_clock;
             *pts_ptr = pts;
-            n = 2 * dec->channels;
-            is->audio_clock += (double)data_size /
-                (double)(n * dec->sample_rate);
+            is->audio_clock += (double)data_size / (dec->channels * dec->sample_rate * av_get_bytes_per_sample(dec->sample_fmt));
 #ifdef DEBUG
             {
                 static double last_clock;
@@ -2060,7 +2127,7 @@ static int audio_decode_frame(VideoState *is, double *pts_ptr)
                 last_clock = is->audio_clock;
             }
 #endif
-            return data_size;
+            return resampled_data_size;
         }
 
         /* free the current packet */
@@ -2072,12 +2139,11 @@ static int audio_decode_frame(VideoState *is, double *pts_ptr)
         }
 
         /* read next packet */
-        if (packet_queue_get(&is->audioq, pkt, 1) < 0)
+        if ((new_packet = packet_queue_get(&is->audioq, pkt, 1)) < 0)
             return -1;
-        if(pkt->data == flush_pkt.data){
+
+        if (pkt->data == flush_pkt.data)
             avcodec_flush_buffers(dec);
-            continue;
-        }
 
         pkt_temp->data = pkt->data;
         pkt_temp->size = pkt->size;
@@ -2105,7 +2171,7 @@ static void sdl_audio_callback(void *opaque, Uint8 *stream, int len)
            if (audio_size < 0) {
                 /* if error, just output silence */
                is->audio_buf = is->audio_buf1;
-               is->audio_buf_size = 1024;
+               is->audio_buf_size = 256 * is->audio_tgt_channels * av_get_bytes_per_sample(is->audio_tgt_fmt);
                memset(is->audio_buf, 0, is->audio_buf_size);
            } else {
                if (is->show_mode != SHOW_MODE_VIDEO)
@@ -2124,8 +2190,7 @@ static void sdl_audio_callback(void *opaque, Uint8 *stream, int len)
         stream += len1;
         is->audio_buf_index += len1;
     }
-    bytes_per_sec = is->audio_st->codec->sample_rate *
-            2 * is->audio_st->codec->channels;
+    bytes_per_sec = is->audio_tgt_freq * is->audio_tgt_channels * av_get_bytes_per_sample(is->audio_tgt_fmt);
     is->audio_write_buf_size = is->audio_buf_size - is->audio_buf_index;
     /* Let's assume the audio driver that is used by SDL has two periods. */
     is->audio_current_pts = is->audio_clock - (double)(2 * is->audio_hw_buf_size + is->audio_write_buf_size) / bytes_per_sec;
@@ -2141,6 +2206,7 @@ static int stream_component_open(VideoState *is, int stream_index)
     SDL_AudioSpec wanted_spec, spec;
     AVDictionary *opts;
     AVDictionaryEntry *t = NULL;
+    int64_t wanted_channel_layout = 0;
 
     if (stream_index < 0 || stream_index >= ic->nb_streams)
         return -1;
@@ -2148,16 +2214,12 @@ static int stream_component_open(VideoState *is, int stream_index)
 
     opts = filter_codec_opts(codec_opts, avctx->codec_id, ic, ic->streams[stream_index]);
 
-    /* prepare audio output */
-    if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
-        if (avctx->channels > 0) {
-            avctx->request_channels = FFMIN(2, avctx->channels);
-        } else {
-            avctx->request_channels = 2;
-        }
-    }
-
     codec = avcodec_find_decoder(avctx->codec_id);
+    switch(avctx->codec_type){
+        case AVMEDIA_TYPE_AUDIO   : if(audio_codec_name   ) codec= avcodec_find_decoder_by_name(   audio_codec_name); break;
+        case AVMEDIA_TYPE_SUBTITLE: if(subtitle_codec_name) codec= avcodec_find_decoder_by_name(subtitle_codec_name); break;
+        case AVMEDIA_TYPE_VIDEO   : if(video_codec_name   ) codec= avcodec_find_decoder_by_name(   video_codec_name); break;
+    }
     if (!codec)
         return -1;
 
@@ -2175,6 +2237,17 @@ static int stream_component_open(VideoState *is, int stream_index)
     if(codec->capabilities & CODEC_CAP_DR1)
         avctx->flags |= CODEC_FLAG_EMU_EDGE;
 
+    if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
+        wanted_channel_layout = (avctx->channel_layout && avctx->channels == av_get_channel_layout_nb_channels(avctx->channels)) ? avctx->channel_layout : av_get_default_channel_layout(avctx->channels);
+        wanted_channel_layout &= ~AV_CH_LAYOUT_STEREO_DOWNMIX;
+        wanted_spec.channels = av_get_channel_layout_nb_channels(wanted_channel_layout);
+        wanted_spec.freq = avctx->sample_rate;
+        if (wanted_spec.freq <= 0 || wanted_spec.channels <= 0) {
+            fprintf(stderr, "Invalid sample rate or channel count!\n");
+            return -1;
+        }
+    }
+
     if (!codec ||
         avcodec_open2(avctx, codec, &opts) < 0)
         return -1;
@@ -2185,13 +2258,7 @@ static int stream_component_open(VideoState *is, int stream_index)
 
     /* prepare audio output */
     if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
-        if(avctx->sample_rate <= 0 || avctx->channels <= 0){
-            fprintf(stderr, "Invalid sample rate or channel count\n");
-            return -1;
-        }
-        wanted_spec.freq = avctx->sample_rate;
         wanted_spec.format = AUDIO_S16SYS;
-        wanted_spec.channels = avctx->channels;
         wanted_spec.silence = 0;
         wanted_spec.samples = SDL_AUDIO_BUFFER_SIZE;
         wanted_spec.callback = sdl_audio_callback;
@@ -2201,7 +2268,21 @@ static int stream_component_open(VideoState *is, int stream_index)
             return -1;
         }
         is->audio_hw_buf_size = spec.size;
-        is->audio_src_fmt= AV_SAMPLE_FMT_S16;
+        if (spec.format != AUDIO_S16SYS) {
+            fprintf(stderr, "SDL advised audio format %d is not supported!\n", spec.format);
+            return -1;
+        }
+        if (spec.channels != wanted_spec.channels) {
+            wanted_channel_layout = av_get_default_channel_layout(spec.channels);
+            if (!wanted_channel_layout) {
+                fprintf(stderr, "SDL advised channel count %d is not supported!\n", spec.channels);
+                return -1;
+            }
+        }
+        is->audio_src_fmt = is->audio_tgt_fmt = AV_SAMPLE_FMT_S16;
+        is->audio_src_freq = is->audio_tgt_freq = spec.freq;
+        is->audio_src_channel_layout = is->audio_tgt_channel_layout = wanted_channel_layout;
+        is->audio_src_channels = is->audio_tgt_channels = spec.channels;
     }
 
     ic->streams[stream_index]->discard = AVDISCARD_DEFAULT;
@@ -2217,7 +2298,7 @@ static int stream_component_open(VideoState *is, int stream_index)
         is->audio_diff_avg_count = 0;
         /* since we do not have a precise anough audio fifo fullness,
            we correct audio sync only if larger than this threshold */
-        is->audio_diff_threshold = 2.0 * SDL_AUDIO_BUFFER_SIZE / avctx->sample_rate;
+        is->audio_diff_threshold = 2.0 * SDL_AUDIO_BUFFER_SIZE / wanted_spec.freq;
 
         memset(&is->audio_pkt, 0, sizeof(is->audio_pkt));
         packet_queue_init(&is->audioq);
@@ -2259,9 +2340,14 @@ static void stream_component_close(VideoState *is, int stream_index)
         SDL_CloseAudio();
 
         packet_queue_end(&is->audioq);
-        if (is->reformat_ctx)
-            av_audio_convert_free(is->reformat_ctx);
-        is->reformat_ctx = NULL;
+        if (is->swr_ctx)
+            swr_free(&is->swr_ctx);
+        av_free_packet(&is->audio_pkt);
+
+        if (is->rdft) {
+            av_rdft_end(is->rdft);
+            av_freep(&is->rdft_data);
+        }
         break;
     case AVMEDIA_TYPE_VIDEO:
         packet_queue_abort(&is->videoq);
@@ -2362,6 +2448,8 @@ static int read_thread(void *arg)
     if(genpts)
         ic->flags |= AVFMT_FLAG_GENPTS;
 
+    av_dict_set(&codec_opts, "request_channels", "2", 0);
+
     opts = setup_find_stream_info_opts(ic, codec_opts);
     orig_nb_streams = ic->nb_streams;
 
@@ -2508,6 +2596,14 @@ static int read_thread(void *arg)
                 pkt->stream_index= is->video_stream;
                 packet_queue_put(&is->videoq, pkt);
             }
+            if (is->audio_stream >= 0 &&
+                is->audio_st->codec->codec->capabilities & CODEC_CAP_DELAY) {
+                av_init_packet(pkt);
+                pkt->data = NULL;
+                pkt->size = 0;
+                pkt->stream_index = is->audio_stream;
+                packet_queue_put(&is->audioq, pkt);
+            }
             SDL_Delay(10);
             if(is->audioq.size + is->videoq.size + is->subtitleq.size ==0){
                 if(loop!=1 && (!loop || --loop)){
@@ -2659,6 +2755,12 @@ static void stream_cycle_channel(VideoState *is, int codec_type)
 static void toggle_full_screen(VideoState *is)
 {
     is_full_screen = !is_full_screen;
+#if defined(__APPLE__) && SDL_VERSION_ATLEAST(1, 2, 14)
+    /* OSX needs to reallocate the SDL overlays */
+    for (int i = 0; i < VIDEO_PICTURE_QUEUE_SIZE; i++) {
+        is->pictq[i].reallocate = 1;
+    }
+#endif
     video_open(is);
 }
 
@@ -2906,6 +3008,16 @@ static void opt_input_file(void *optctx, const char *filename)
     input_filename = filename;
 }
 
+static int opt_codec(void *o, const char *opt, const char *arg)
+{
+    switch(opt[strlen(opt)-1]){
+    case 'a' :    audio_codec_name = arg; break;
+    case 's' : subtitle_codec_name = arg; break;
+    case 'v' :    video_codec_name = arg; break;
+    }
+    return 0;
+}
+
 static int dummy;
 
 static const OptionDef options[] = {
@@ -2951,6 +3063,7 @@ static const OptionDef options[] = {
     { "showmode", HAS_ARG, {(void*)opt_show_mode}, "select show mode (0 = video, 1 = waves, 2 = RDFT)", "mode" },
     { "default", HAS_ARG | OPT_AUDIO | OPT_VIDEO | OPT_EXPERT, {(void*)opt_default}, "generic catch all option", "" },
     { "i", OPT_BOOL, {(void *)&dummy}, "read specified file", "input_file"},
+    { "codec", HAS_ARG | OPT_FUNC2, {(void*)opt_codec}, "force decoder", "decoder" },
     { NULL, },
 };
 
@@ -2963,7 +3076,6 @@ static void show_usage(void)
 
 static int opt_help(const char *opt, const char *arg)
 {
-    const AVClass *class;
     av_log_set_callback(log_callback_help);
     show_usage();
     show_help_options(options, "Main options:\n",
@@ -2971,18 +3083,10 @@ static int opt_help(const char *opt, const char *arg)
     show_help_options(options, "\nAdvanced options:\n",
                       OPT_EXPERT, OPT_EXPERT);
     printf("\n");
-    class = avcodec_get_class();
-    av_opt_show2(&class, NULL,
-                 AV_OPT_FLAG_DECODING_PARAM, 0);
-    printf("\n");
-    class = avformat_get_class();
-    av_opt_show2(&class, NULL,
-                 AV_OPT_FLAG_DECODING_PARAM, 0);
+    show_help_children(avcodec_get_class(), AV_OPT_FLAG_DECODING_PARAM);
+    show_help_children(avformat_get_class(), AV_OPT_FLAG_DECODING_PARAM);
 #if !CONFIG_AVFILTER
-    printf("\n");
-    class = sws_get_class();
-    av_opt_show2(&class, NULL,
-                 AV_OPT_FLAG_ENCODING_PARAM, 0);
+    show_help_children(sws_get_class(), AV_OPT_FLAG_ENCODING_PARAM);
 #endif
     printf("\nWhile playing:\n"
            "q, ESC              quit\n"
@@ -3026,6 +3130,7 @@ int main(int argc, char **argv)
     VideoState *is;
 
     av_log_set_flags(AV_LOG_SKIP_REPEATED);
+    parse_loglevel(argc, argv, options);
 
     /* register all codecs, demux and protocols */
     avcodec_register_all();