寫在前面

前幾篇文章大概分析到了數(shù)據(jù)的讀取罐柳，接下來(lái)就該解碼和播放了张吉。那么ijkplayer解碼和播放又是怎么做的呢肮蛹？

解碼線程

從上一篇文章我們可以看到蔗崎，ijkplayer的音頻解碼線程的入口函數(shù)是audio_thread()缓苛，那么我們就跟蹤到audio_thread()/ff_ffplayer.c函數(shù)里面：

static int audio_thread(void *arg)
{
//...
    do {
        ffp_audio_statistic_l(ffp);
        if ((got_frame = decoder_decode_frame(ffp, &is->auddec, frame, NULL)) < 0)
            goto the_end;
            //...
            while ((ret = av_buffersink_get_frame_flags(is->out_audio_filter, frame, 0)) >= 0) {
    //...
                if (!(af = frame_queue_peek_writable(&is->sampq)))
                    goto the_end;

        //...

                av_frame_move_ref(af->frame, frame);
                frame_queue_push(&is->sampq);
//...
        }
    } while (ret >= 0 || ret == AVERROR(EAGAIN) || ret == AVERROR_EOF);
 the_end:
//...
    av_frame_free(&frame);
    return ret;
}

從上面代碼可以看出，一開(kāi)始就進(jìn)入循環(huán)冬耿，然后調(diào)用decoder_decode_frame()進(jìn)行解碼亦镶，解碼后的幀存放到frame中缤骨，然后調(diào)用frame_queue_peek_writable()判斷是否能把剛剛解碼的frame寫入is->sampq中绊起，因?yàn)?code>is->sampq是音頻解碼幀列表虱歪，然而播放線程直接從這里面讀取數(shù)據(jù)笋鄙，然后播放出來(lái)局装。最后 av_frame_move_ref(af->frame, frame);把frame放入到sampq相應(yīng)位置。由于前面af = frame_queue_peek_writable(&is->sampq)哆姻，af就是指向這一幀frame應(yīng)該放的位置的指針矛缨，所以直接把值賦值給它的結(jié)構(gòu)體里面的frame就行了灵妨。

然后frame_queue_push(&is->sampq);里面是一個(gè)喚醒線程的操作泌霍，如查音頻播放線程因?yàn)閟ampq隊(duì)列為空而阻塞朱转，這里可以喚醒它藤为。

在decoder_decode_frame()里面是調(diào)用傳進(jìn)去的codec的codec->decode()方法解碼缅疟。

在frame_queue_peek_writable()里面會(huì)判斷sampq隊(duì)列是否滿了存淫，如果沒(méi)位置放我們的frame的話纫雁，會(huì)調(diào)用pthread_cond_wait()方法阻塞隊(duì)列轧邪。如果有位置放frame的話忌愚，就會(huì)返回frame應(yīng)該放置的位置的地址硕糊。

解碼線程大概就結(jié)束了檬某。

播放流程

之前在初始化的時(shí)候恢恼，有個(gè)地方還沒(méi)分析到场斑，那就是在ijkmp_android_create()/ijkplayer_android.c里面：

IjkMediaPlayer *ijkmp_android_create(int(*msg_loop)(void*))
{
    IjkMediaPlayer *mp = ijkmp_create(msg_loop);

    mp->ffplayer->vout = SDL_VoutAndroid_CreateForAndroidSurface();
    if (!mp->ffplayer->vout)
        goto fail;

    mp->ffplayer->pipeline = ffpipeline_create_from_android(mp->ffplayer);

    ffpipeline_set_vout(mp->ffplayer->pipeline, mp->ffplayer->vout);

    return mp;

}

在ffpipeline_create_from_android()里面有一句

pipeline->func_open_audio_output  = func_open_audio_output;

接著我們看看func_open_audio_output()/ffpipeline_android.c:

static SDL_Aout *func_open_audio_output(IJKFF_Pipeline *pipeline, FFPlayer *ffp)
{
    SDL_Aout *aout = NULL;
    if (ffp->opensles) {
        aout = SDL_AoutAndroid_CreateForOpenSLES();
    } else {
        aout = SDL_AoutAndroid_CreateForAudioTrack();
    }
    if (aout)
        SDL_AoutSetStereoVolume(aout, pipeline->opaque->left_volume, pipeline->opaque->right_volume);
    return aout;
}

從上面可以看出，音頻播放也分為:opensles,audiotrack,然后我們就來(lái)看看audiotrack吧：

SDL_Aout *SDL_AoutAndroid_CreateForAudioTrack()
{
    SDL_Aout *aout = SDL_Aout_CreateInternal(sizeof(SDL_Aout_Opaque));
    if (!aout)
        return NULL;

    SDL_Aout_Opaque *opaque = aout->opaque;
    opaque->wakeup_cond  = SDL_CreateCond();
    opaque->wakeup_mutex = SDL_CreateMutex();
    opaque->speed        = 1.0f;

    aout->opaque_class = &g_audiotrack_class;
    aout->free_l       = aout_free_l;
    aout->open_audio   = aout_open_audio;
    aout->pause_audio  = aout_pause_audio;
    aout->flush_audio  = aout_flush_audio;
    aout->set_volume   = aout_set_volume;
    aout->close_audio  = aout_close_audio;
    aout->func_get_audio_session_id = aout_get_audio_session_id;
    aout->func_set_playback_rate    = func_set_playback_rate;

    return aout;
}

我們要分析音頻播放對(duì)吧？在前面數(shù)據(jù)讀取線程我們分析到了在stream_component_open()里面會(huì)調(diào)用aout->open_audio(aout, pause_on);爽柒，相信大家還記得吧浩村？我們現(xiàn)在看看在初始化的時(shí)候:

    aout->open_audio   = aout_open_audio;

所以在前面stream_component_open()里面相當(dāng)于直接調(diào)用了aout_open_audio()/ijksdl_aout_android_audiotrack.c心墅。當(dāng)然要是我們?cè)诓シ牌骼锩嬗玫?code>opensles怎燥，程序流程是差不多的铐姚，大家有興趣的話可以下去看看隐绵。

上面接著會(huì)調(diào)用aout_open_audio_n()/ijksdl_aout_android_audiotrack.c,然后：

SDL_CreateThreadEx(&opaque->_audio_tid, aout_thread, aout, "ff_aout_android");

這里創(chuàng)建的線程就是播放線程依许。

接著我們看看入口函數(shù)aout_thread,在這個(gè)函數(shù)內(nèi)部會(huì)調(diào)用aout_thread_n()/ijksdl_aout_android_audiotrack.c:

static int aout_thread_n(JNIEnv *env, SDL_Aout *aout)
{
  SDL_Aout_Opaque *opaque = aout->opaque;
    SDL_Android_AudioTrack *atrack = opaque->atrack;
    SDL_AudioCallback audio_cblk = opaque->spec.callback;
    void *userdata = opaque->spec.userdata;
    uint8_t *buffer = opaque->buffer;
  //...
    if (!opaque->abort_request && !opaque->pause_on)
        SDL_Android_AudioTrack_play(env, atrack);

    while (!opaque->abort_request) {
        SDL_LockMutex(opaque->wakeup_mutex);
        if (!opaque->abort_request && opaque->pause_on) {
            SDL_Android_AudioTrack_pause(env, atrack);
            while (!opaque->abort_request && opaque->pause_on) {
                SDL_CondWaitTimeout(opaque->wakeup_cond, opaque->wakeup_mutex, 1000);
            }
            if (!opaque->abort_request && !opaque->pause_on)
                SDL_Android_AudioTrack_play(env, atrack);
        }
        if (opaque->need_flush) {
            opaque->need_flush = 0;
            SDL_Android_AudioTrack_flush(env, atrack);
        }
        if (opaque->need_set_volume) {
            opaque->need_set_volume = 0;
            SDL_Android_AudioTrack_set_volume(env, atrack, opaque->left_volume, opaque->right_volume);
        }
        if (opaque->speed_changed) {
            opaque->speed_changed = 0;
            if (J4A_GetSystemAndroidApiLevel(env) >= 23) {
                SDL_Android_AudioTrack_setSpeed(env, atrack, opaque->speed);
            }
        }
        SDL_UnlockMutex(opaque->wakeup_mutex);

        audio_cblk(userdata, buffer, copy_size);
        if (opaque->need_flush) {
            SDL_Android_AudioTrack_flush(env, atrack);
            opaque->need_flush = false;
        }

        if (opaque->need_flush) {
            opaque->need_flush = 0;
            SDL_Android_AudioTrack_flush(env, atrack);
        } else {
            int written = SDL_Android_AudioTrack_write(env, atrack, buffer, copy_size);
            if (written != copy_size) {
                ALOGW("AudioTrack: not all data copied %d/%d", (int)written, (int)copy_size);
            }
        }

        // TODO: 1 if callback return -1 or 0
    }

這個(gè)函數(shù)的開(kāi)始有很多SDL_Android_AudioTrack_set_xxx()，主要是設(shè)置播放器相關(guān)的配置悬襟，比如播放速度古胆，聲音大小等。

接著就是audio_cblk()夭谤，不知道大家還記得在上一節(jié)我說(shuō)過(guò)在stream_component_open()的里面調(diào)用的audio_open()朗儒，會(huì)有這么一句代碼:

    wanted_spec.callback = sdl_audio_callback;

現(xiàn)在這里排上用場(chǎng)了醉锄。

在上面aout_thread_n()里調(diào)用的audio_cblk(),實(shí)際上就是調(diào)用的opaque->spec.callback,其實(shí)就是調(diào)用到sdl_audio_callback()這個(gè)函數(shù)來(lái)了。

然后繼續(xù)分析:

static void sdl_audio_callback(void *opaque, Uint8 *stream, int len)
{
//...
    while (len > 0) {
        if (is->audio_buf_index >= is->audio_buf_size) {
           audio_size = audio_decode_frame(is);
           if (audio_size < 0) {
                /* 發(fā)生錯(cuò)誤烟勋，就輸出silence */
           //...
           } else {
               if (is->show_mode != SHOW_MODE_VIDEO)
                   update_sample_display(is, (int16_t *)is->audio_buf, audio_size);
               is->audio_buf_size = audio_size;
           }
           is->audio_buf_index = 0;
        }
        len1 = is->audio_buf_size - is->audio_buf_index;
        if (len1 > len)
            len1 = len;
        if (!is->muted && is->audio_buf && is->audio_volume == SDL_MIX_MAXVOLUME)
            memcpy(stream, (uint8_t *)is->audio_buf + is->audio_buf_index, len1);
        else {
           //...
        }
        len -= len1;
        stream += len1;
        is->audio_buf_index += len1;
    }
    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. */
  //...
}

保留了部分相對(duì)重要代碼卵惦，其中重要代碼有:

audio_size = audio_decode_frame(ffp);
memcpy(stream, (uint8_t *)is->audio_buf + is->audio_buf_index, len1);

我們繼續(xù)看audio_decode_frame()沮尿，還好作者有注釋畜疾，根據(jù)注釋提取重要代碼:

/**
 * Decode one audio frame and return its uncompressed size.
 *
 * The processed audio frame is decoded, converted if required, and
 * stored in is->audio_buf, with size in bytes given by the return
 * value.
 */
static int audio_decode_frame(FFPlayer *ffp)｛
    af = frame_queue_peek_readable(&is->sampq)
    is->audio_buf = af->frame->data[0];
｝

從上面代碼看，這里主要是判斷解碼后的is->sampq是否為空届慈，其中和解碼的時(shí)候一樣，如果為空(解碼時(shí)放入is->sampq判斷是否滿)臊泌，如果為空渠概，就阻塞(還記得解碼的時(shí)候播揪，每向is->sampq放入一frame猪狈，就喚醒線程么雇庙？)疆前，否則返回隊(duì)列的第一個(gè)frame竹椒。然后賦值給ffp->is->audio_buf碾牌。

接著返回到上面sdl_audio_callback()中舶吗，接著再把剛剛賦值的ffp->is->audio_buf copy到stream中誓琼，stream從命名來(lái)看是一個(gè)流腹侣，流的另外一頭在哪里呢傲隶？

再返回到aout_thread_n()中：

SDL_Android_AudioTrack_write(env, atrack, buffer, copy_size);

這里的buffer就是剛剛的stream,該函數(shù)繼續(xù)調(diào)用:

 (*env)->SetByteArrayRegion(env, atrack->byte_buffer, 0, (int)size_in_byte, (jbyte*) data);
J4AC_AudioTrack__write(env, atrack->thiz, atrack->byte_buffer, 0, (int)size_in_byte);

這里先是把data拷貝到數(shù)組中跺株，為什么呢乒省？因?yàn)楹竺鏁?huì)把這個(gè)數(shù)組袖扛，也就是音頻幀傳遞給java蛆封，而SetByteArrayRegion()就是這里的一次轉(zhuǎn)換。
J4AC_AudioTrack__write()中繼續(xù)跟蹤會(huì)發(fā)現(xiàn):

jint J4AC_android_media_AudioTrack__write(JNIEnv *env, jobject thiz, jbyteArray audioData, jint offsetInBytes, jint sizeInBytes)
{
    return (*env)->CallIntMethod(env, thiz, class_J4AC_android_media_AudioTrack.method_write, audioData, offsetInBytes, sizeInBytes);
}

這就尷尬了垒迂，又調(diào)用到j(luò)ava里面去了，這里調(diào)用了java層的AudioTrack.java中的write()函數(shù)绣夺。

其實(shí)這里又用到了bilibili另外一個(gè)開(kāi)源項(xiàng)目:jni4android陶耍。這個(gè)項(xiàng)目可以直接在c里面生成一個(gè)java的裝飾類烈钞。這里用到的java就是AudioTrack.java毯欣，生成的文件就是AudioTrack.h和AudioTrack.c酗钞。后面基本就不用分析了吧砚作。在java里面基本都會(huì)用androidTrack吧葫录？網(wǎng)上教程也很多米同。

到了這里基本就結(jié)束了窍霞。音頻播放完成了但金，接下來(lái)會(huì)分析視頻播放流程冷溃，其實(shí)視頻播放流程和音頻潦差不多似枕，不過(guò)比音頻麻煩點(diǎn)凿歼。

** 如果大家還想了解ijkplayer的工作流程的話味赃，可以關(guān)注下android下的ijkplayer虐拓。**