ffmpeg源码分析一：概述

首先先看ffmpeg.c文件，有类似于如下的一些变量：

InputStream **input_streams = NULL;

int        nb_input_streams = 0;

InputFile   **input_files   = NULL;

int        nb_input_files   = 0;

OutputStream **output_streams = NULL;

int         nb_output_streams = 0;

OutputFile   **output_files   = NULL;

int         nb_output_files   = 0;

其中：

input_streams是输入流的数组，nb_input_streams是输入流的个数。

input_files是输入文件（也可能是设备）的数组，nb_input_files是输入文件的个数。

下面的输出相关的变量们就不用解释了。

可以看出，文件和流是分别保存的。于是，可以想象，结构InputStream中应有其所属的文件在InputFile中的序号（file_index）。输入流数组应是这样填充的：每当在输入文件中找到一个流时，就把它添加到input_streams中，所以一个输入文件对应的输入流在input_streams中是紧靠着的，于是InputFile结构中应有其第一个流在input_streams中的开始序号（ist_index）和被放在input_streams中的流的总个数（nb_streams）。

InputStream和InputFile结构体如下：

typedef struct InputStream {

    int file_index;

    AVStream *st;

    int discard;             /* true if stream data should be discarded */

    int decoding_needed;     /* true if the packets must be decoded in 'raw_fifo' */

    AVCodec *dec;

    AVFrame *decoded_frame;

    AVFrame *filter_frame; /* a ref of decoded_frame, to be sent to filters */

    int64_t       start;     /* time when read started */

    /* predicted dts of the next packet read for this stream or (when there are

     * several frames in a packet) of the next frame in current packet (in AV_TIME_BASE units) */

    int64_t       next_dts;

    int64_t       dts;       ///< dts of the last packet read for this stream (in AV_TIME_BASE units)

    int64_t       next_pts;  ///< synthetic pts for the next decode frame (in AV_TIME_BASE units)

    int64_t       pts;       ///< current pts of the decoded frame  (in AV_TIME_BASE units)

    int           wrap_correction_done;

    int64_t filter_in_rescale_delta_last;

    double ts_scale;

    int saw_first_ts;

    int showed_multi_packet_warning;

    AVDictionary *opts;

    AVRational framerate;               /* framerate forced with -r */

    int top_field_first;

    int guess_layout_max;

    int resample_height;

    int resample_width;

    int resample_pix_fmt;

    int      resample_sample_fmt;

    int      resample_sample_rate;

    int      resample_channels;

    uint64_t resample_channel_layout;

    int fix_sub_duration;

    struct { /* previous decoded subtitle and related variables */

        int got_output;

        int ret;

        AVSubtitle subtitle;

    } prev_sub;

    struct sub2video {

        int64_t last_pts;

        int64_t end_pts;

        AVFrame *frame;

        int w, h;

    } sub2video;

    int dr1;

    /* decoded data from this stream goes into all those filters

     * currently video and audio only */

    InputFilter **filters;

    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 {

    AVFormatContext *ctx;

    int eof_reached;      /* true if eof reached */

    int eagain;           /* true if last read attempt returned EAGAIN */

    int ist_index;        /* index of first stream in input_streams */

    int64_t input_ts_offset;

    int64_t ts_offset;

    int64_t last_ts;

    int64_t start_time;   /* user-specified start time in AV_TIME_BASE or AV_NOPTS_VALUE */

    int64_t recording_time;

    int nb_streams;       /* number of stream that ffmpeg is aware of; may be different

                             from ctx.nb_streams if new streams appear during av_read_frame() */

    int nb_streams_warn;  /* number of streams that the user was warned of */

    int rate_emu;

    int accurate_seek;

    int shared_encoding;

#if HAVE_PTHREADS

    pthread_t thread;           /* thread reading from this file */

    int non_blocking;           /* reading packets from the thread should not block */

    int finished;               /* the thread has exited */

    int joined;                 /* the thread has been joined */

    pthread_mutex_t fifo_lock;  /* lock for access to fifo */

    pthread_cond_t  fifo_cond;  /* the main thread will signal on this cond after reading from fifo */

    AVFifoBuffer *fifo;         /* demuxed packets are stored here; freed by the main thread */

#endif

} InputFile;

在输出流output_streams中，除了要保存其所在的输出文件在output_files中的序号（index），还应保存其对应的输入流在input_streams中的序号（ source_index），也应保存其在所属输出文件中的流序号（file_index）。而输出文件中呢，只需保存它的第一个流在output_streams中的序号（ost_index），但是为啥不保存输出文件中流的个数呢？我也不道，但我知道肯定不用保存也不影响实现功能（嘿嘿，相当于没说）．

OutputStream和OutputFile结构体如下：

typedef struct OutputStream {

    int file_index;          /* file index */

    int index;               /* stream index in the output file */

    int source_index;        /* InputStream index */

    AVStream *st;            /* stream in the output file */

    int encoding_needed;     /* true if encoding needed for this stream */

    int frame_number;

    int encoded_frame_number;

    /* input pts and corresponding output pts

       for A/V sync */

    struct InputStream *sync_ist; /* input stream to sync against */

    int64_t sync_opts;       /* output frame counter, could be changed to some true timestamp */ // FIXME look at frame_number

    /* pts of the first frame encoded for this stream, used for limiting

     * recording time */

    int64_t first_pts;

    /* dts of the last packet sent to the muxer */

    int64_t last_mux_dts;

    AVBitStreamFilterContext *bitstream_filters;

    AVCodec *enc;

    int64_t max_frames;

    AVFrame *filtered_frame;

    /* video only */

    AVRational frame_rate;

    int force_fps;

    int top_field_first;

    AVRational frame_aspect_ratio;

    /* forced key frames */

    int64_t *forced_kf_pts;

    int forced_kf_count;

    int forced_kf_index;

    char *forced_keyframes;

    AVExpr *forced_keyframes_pexpr;

    double forced_keyframes_expr_const_values[FKF_NB];

    /* audio only */

    int audio_channels_map[SWR_CH_MAX];  /* list of the channels id to pick from the source stream */

    int audio_channels_mapped;           /* number of channels in audio_channels_map */

    char *logfile_prefix;

    FILE *logfile;

    OutputFilter *filter;

    char *avfilter;

    char *filters;         ///< filtergraph associated to the -filter option

    char *filters_script;  ///< filtergraph script associated to the -filter_script option

    int64_t sws_flags;

    AVDictionary *opts;

    AVDictionary *swr_opts;

    AVDictionary *resample_opts;

    char *apad;

    OSTFinished finished;        /* no more packets should be written for this stream */

    int unavailable;                     /* true if the steram is unavailable (possibly temporarily) */

    int stream_copy;

    const char *attachment_filename;

    int copy_initial_nonkeyframes;

    int copy_prior_start;

    int keep_pix_fmt;

    AVCodecParserContext *parser;

} OutputStream;

typedef struct OutputFile {

    AVFormatContext *ctx;

    AVDictionary *opts;

    int ost_index;       /* index of the first stream in output_streams */

    int64_t recording_time;  ///< desired length of the resulting file in microseconds == AV_TIME_BASE units

    int64_t start_time;      ///< start time in microseconds == AV_TIME_BASE units

    uint64_t limit_filesize; /* filesize limit expressed in bytes */

    int shortest;

} OutputFile;

流和文件都准备好了，下面就是转换，那么转换过程是怎样的呢？还是我来猜一猜吧：

首先打开输入文件们，然后跟据输入流们准备并打开解码器们，然后跟据输出流们准备并打开编码器们，然后创建输出文件们，然后为所有输出文件们写好头部，然后就在循环中把输入流转换到输出流并写入输出文件中，转换完后跳出循环，然后写入文件尾，最后关闭所有的输出文件．

概述就先到这里吧，后面会对几个重要函数做详细分析。

原文链接：https://blog.csdn.net/austinblog/article/details/24800381