Merge commit '07fd0a22192805d56c635eb294dc26b0a54ae325'

* commit '07fd0a22192805d56c635eb294dc26b0a54ae325':
  avconv: add infrastructure for using hwaccels

Conflicts:
	ffmpeg.c
	ffmpeg.h
	ffmpeg_filter.c
	ffmpeg_opt.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

5 files changed, 179 insertions, 2 deletions

diff --git a/doc/ffmpeg.texi b/doc/ffmpeg.texi
index af4cf9fb25..1a2a801770 100644
--- a/doc/ffmpeg.texi
+++ b/doc/ffmpeg.texi
@@ -621,6 +621,33 @@ would be more efficient.
 @item -copyinkf[:@var{stream_specifier}] (@emph{output,per-stream})
 When doing stream copy, copy also non-key frames found at the
 beginning.
+
+@item -hwaccel[:@var{stream_specifier}] @var{hwaccel} (@emph{input,per-stream})
+Use hardware acceleration to decode the matching stream(s). The allowed values
+of @var{hwaccel} are:
+@table @option
+@item none
+Do not use any hardware acceleration (the default).
+
+@item auto
+Automatically select the hardware acceleration method.
+@end table
+
+This option has no effect if the selected hwaccel is not available or not
+supported by the chosen decoder.
+
+Note that most acceleration methods are intended for playback and will not be
+faster than software decoding on modern CPUs. Additionally, @command{ffmpeg}
+will usually need to copy the decoded frames from the GPU memory into the system
+memory, resulting in further performance loss. This option is thus mainly
+useful for testing.
+
+@item -hwaccel_device[:@var{stream_specifier}] @var{hwaccel_device} (@emph{input,per-stream})
+Select a device to use for hardware acceleration.
+
+This option only makes sense when the @option{-hwaccel} option is also
+specified. Its exact meaning depends on the specific hardware acceleration
+method chosen.
 @end table
 
 @section Audio Options
diff --git a/ffmpeg.c b/ffmpeg.c
index de361fff3d..26cbbe3894 100644
--- a/ffmpeg.c
+++ b/ffmpeg.c
@@ -489,6 +489,7 @@ static void ffmpeg_cleanup(int ret)
         avsubtitle_free(&input_streams[i]->prev_sub.subtitle);
         av_frame_free(&input_streams[i]->sub2video.frame);
         av_freep(&input_streams[i]->filters);
+        av_freep(&input_streams[i]->hwaccel_device);
         av_freep(&input_streams[i]);
     }
 
@@ -1707,6 +1708,13 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
     if(ist->top_field_first>=0)
         decoded_frame->top_field_first = ist->top_field_first;
 
+    if (ist->hwaccel_retrieve_data && decoded_frame->format == ist->hwaccel_pix_fmt) {
+        err = ist->hwaccel_retrieve_data(ist->st->codec, decoded_frame);
+        if (err < 0)
+            goto fail;
+    }
+    ist->hwaccel_retrieved_pix_fmt = decoded_frame->format;
+
     best_effort_timestamp= av_frame_get_best_effort_timestamp(decoded_frame);
     if(best_effort_timestamp != AV_NOPTS_VALUE)
         ist->next_pts = ist->pts = av_rescale_q(decoded_frame->pts = best_effort_timestamp, ist->st->time_base, AV_TIME_BASE_Q);
@@ -1771,6 +1779,7 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
         }
     }
 
+fail:
     av_frame_unref(ist->filter_frame);
     av_frame_unref(decoded_frame);
     return err < 0 ? err : ret;
@@ -1982,6 +1991,63 @@ static void print_sdp(void)
     av_freep(&avc);
 }
 
+static const HWAccel *get_hwaccel(enum AVPixelFormat pix_fmt)
+{
+    int i;
+    for (i = 0; hwaccels[i].name; i++)
+        if (hwaccels[i].pix_fmt == pix_fmt)
+            return &hwaccels[i];
+    return NULL;
+}
+
+static enum AVPixelFormat get_format(AVCodecContext *s, const enum AVPixelFormat *pix_fmts)
+{
+    InputStream *ist = s->opaque;
+    const enum AVPixelFormat *p;
+    int ret;
+
+    for (p = pix_fmts; *p != -1; p++) {
+        const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(*p);
+        const HWAccel *hwaccel;
+
+        if (!(desc->flags & AV_PIX_FMT_FLAG_HWACCEL))
+            break;
+
+        hwaccel = get_hwaccel(*p);
+        if (!hwaccel ||
+            (ist->active_hwaccel_id && ist->active_hwaccel_id != hwaccel->id) ||
+            (ist->hwaccel_id != HWACCEL_AUTO && ist->hwaccel_id != hwaccel->id))
+            continue;
+
+        ret = hwaccel->init(s);
+        if (ret < 0) {
+            if (ist->hwaccel_id == hwaccel->id) {
+                av_log(NULL, AV_LOG_FATAL,
+                       "%s hwaccel requested for input stream #%d:%d, "
+                       "but cannot be initialized.\n", hwaccel->name,
+                       ist->file_index, ist->st->index);
+                exit_program(1);
+            }
+            continue;
+        }
+        ist->active_hwaccel_id = hwaccel->id;
+        ist->hwaccel_pix_fmt   = *p;
+        break;
+    }
+
+    return *p;
+}
+
+static int get_buffer(AVCodecContext *s, AVFrame *frame, int flags)
+{
+    InputStream *ist = s->opaque;
+
+    if (ist->hwaccel_get_buffer && frame->format == ist->hwaccel_pix_fmt)
+        return ist->hwaccel_get_buffer(s, frame, flags);
+
+    return avcodec_default_get_buffer2(s, frame, flags);
+}
+
 static int init_input_stream(int ist_index, char *error, int error_len)
 {
     int ret;
@@ -1995,6 +2061,11 @@ static int init_input_stream(int ist_index, char *error, int error_len)
             return AVERROR(EINVAL);
         }
 
+        ist->st->codec->opaque      = ist;
+        ist->st->codec->get_format  = get_format;
+        ist->st->codec->get_buffer2 = get_buffer;
+        ist->st->codec->thread_safe_callbacks = 1;
+
         av_opt_set_int(ist->st->codec, "refcounted_frames", 1, 0);
 
         if (!av_dict_get(ist->opts, "threads", NULL, 0))
@@ -3326,6 +3397,8 @@ static int transcode(void)
         ist = input_streams[i];
         if (ist->decoding_needed) {
             avcodec_close(ist->st->codec);
+            if (ist->hwaccel_uninit)
+                ist->hwaccel_uninit(ist->st->codec);
         }
     }
 
diff --git a/ffmpeg.h b/ffmpeg.h
index 9f6bd96b1f..ff296fc854 100644
--- a/ffmpeg.h
+++ b/ffmpeg.h
@@ -56,6 +56,18 @@
 
 #define MAX_STREAMS 1024    /* arbitrary sanity check value */
 
+enum HWAccelID {
+    HWACCEL_NONE = 0,
+    HWACCEL_AUTO,
+};
+
+typedef struct HWAccel {
+    const char *name;
+    int (*init)(AVCodecContext *s);
+    enum HWAccelID id;
+    enum AVPixelFormat pix_fmt;
+} HWAccel;
+
 /* select an input stream for an output stream */
 typedef struct StreamMap {
     int disabled;           /* 1 is this mapping is disabled by a negative map */
@@ -100,6 +112,10 @@ typedef struct OptionsContext {
     int        nb_ts_scale;
     SpecifierOpt *dump_attachment;
     int        nb_dump_attachment;
+    SpecifierOpt *hwaccels;
+    int        nb_hwaccels;
+    SpecifierOpt *hwaccel_devices;
+    int        nb_hwaccel_devices;
 
     /* output options */
     StreamMap *stream_maps;
@@ -275,6 +291,19 @@ typedef struct InputStream {
     int        nb_filters;
 
     int reinit_filters;
+
+    /* hwaccel options */
+    enum HWAccelID hwaccel_id;
+    char  *hwaccel_device;
+
+    /* hwaccel context */
+    enum HWAccelID active_hwaccel_id;
+    void  *hwaccel_ctx;
+    void (*hwaccel_uninit)(AVCodecContext *s);
+    int  (*hwaccel_get_buffer)(AVCodecContext *s, AVFrame *frame, int flags);
+    int  (*hwaccel_retrieve_data)(AVCodecContext *s, AVFrame *frame);
+    enum AVPixelFormat hwaccel_pix_fmt;
+    enum AVPixelFormat hwaccel_retrieved_pix_fmt;
 } InputStream;
 
 typedef struct InputFile {
@@ -431,6 +460,8 @@ extern float max_error_rate;
 extern const AVIOInterruptCB int_cb;
 
 extern const OptionDef options[];
+extern const HWAccel hwaccels[];
+
 
 void term_init(void);
 void term_exit(void);
diff --git a/ffmpeg_filter.c b/ffmpeg_filter.c
index 461cc900fe..ad97f0dd81 100644
--- a/ffmpeg_filter.c
+++ b/ffmpeg_filter.c
@@ -654,7 +654,8 @@ static int configure_input_video_filter(FilterGraph *fg, InputFilter *ifilter,
     av_bprintf(&args,
              "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d:"
              "pixel_aspect=%d/%d:sws_param=flags=%d", ist->resample_width,
-             ist->resample_height, ist->resample_pix_fmt,
+             ist->resample_height,
+             ist->hwaccel_retrieve_data ? ist->hwaccel_retrieved_pix_fmt : ist->resample_pix_fmt,
              tb.num, tb.den, sar.num, sar.den,
              SWS_BILINEAR + ((ist->st->codec->flags&CODEC_FLAG_BITEXACT) ? SWS_BITEXACT:0));
     if (fr.num && fr.den)
diff --git a/ffmpeg_opt.c b/ffmpeg_opt.c
index 8da1ce43a6..8e01c685d1 100644
--- a/ffmpeg_opt.c
+++ b/ffmpeg_opt.c
@@ -62,6 +62,11 @@
             outvar = o->name[i].u.type;\
     }\
 }
+
+const HWAccel hwaccels[] = {
+    { 0 },
+};
+
 char *vstats_filename;
 
 float audio_drift_threshold = 0.1;
@@ -557,7 +562,7 @@ static void add_input_streams(OptionsContext *o, AVFormatContext *ic)
         AVStream *st = ic->streams[i];
         AVCodecContext *dec = st->codec;
         InputStream *ist = av_mallocz(sizeof(*ist));
-        char *framerate = NULL;
+        char *framerate = NULL, *hwaccel = NULL, *hwaccel_device = NULL;
 
         if (!ist)
             exit_program(1);
@@ -612,6 +617,40 @@ static void add_input_streams(OptionsContext *o, AVFormatContext *ic)
             ist->top_field_first = -1;
             MATCH_PER_STREAM_OPT(top_field_first, i, ist->top_field_first, ic, st);
 
+            MATCH_PER_STREAM_OPT(hwaccels, str, hwaccel, ic, st);
+            if (hwaccel) {
+                if (!strcmp(hwaccel, "none"))
+                    ist->hwaccel_id = HWACCEL_NONE;
+                else if (!strcmp(hwaccel, "auto"))
+                    ist->hwaccel_id = HWACCEL_AUTO;
+                else {
+                    int i;
+                    for (i = 0; hwaccels[i].name; i++) {
+                        if (!strcmp(hwaccels[i].name, hwaccel)) {
+                            ist->hwaccel_id = hwaccels[i].id;
+                            break;
+                        }
+                    }
+
+                    if (!ist->hwaccel_id) {
+                        av_log(NULL, AV_LOG_FATAL, "Unrecognized hwaccel: %s.\n",
+                               hwaccel);
+                        av_log(NULL, AV_LOG_FATAL, "Supported hwaccels: ");
+                        for (i = 0; hwaccels[i].name; i++)
+                            av_log(NULL, AV_LOG_FATAL, "%s ", hwaccels[i].name);
+                        av_log(NULL, AV_LOG_FATAL, "\n");
+                        exit_program(1);
+                    }
+                }
+            }
+
+            MATCH_PER_STREAM_OPT(hwaccel_devices, str, hwaccel_device, ic, st);
+            if (hwaccel_device) {
+                ist->hwaccel_device = av_strdup(hwaccel_device);
+                if (!ist->hwaccel_device)
+                    exit_program(1);
+            }
+
             break;
         case AVMEDIA_TYPE_AUDIO:
             ist->guess_layout_max = INT_MAX;
@@ -2835,6 +2874,12 @@ const OptionDef options[] = {
         "force key frames at specified timestamps", "timestamps" },
     { "b",            OPT_VIDEO | HAS_ARG | OPT_PERFILE | OPT_OUTPUT,            { .func_arg = opt_bitrate },
         "video bitrate (please use -b:v)", "bitrate" },
+    { "hwaccel",          OPT_VIDEO | OPT_STRING | HAS_ARG | OPT_EXPERT |
+                          OPT_SPEC | OPT_INPUT,                                  { .off = OFFSET(hwaccels) },
+        "use HW accelerated decoding", "hwaccel name" },
+    { "hwaccel_device",   OPT_VIDEO | OPT_STRING | HAS_ARG | OPT_EXPERT |
+                          OPT_SPEC | OPT_INPUT,                                  { .off = OFFSET(hwaccel_devices) },
+        "select a device for HW acceleration" "devicename" },
 
     /* audio options */
     { "aframes",        OPT_AUDIO | HAS_ARG  | OPT_PERFILE | OPT_OUTPUT,           { .func_arg = opt_audio_frames },