So, I'll describe how this stuff works. The main modules: 1. streamer.c: this is the input layer, this reads the file or the VCD or stdin. what it has to know: appropriate buffering by sector, seek, skip functions, reading by bytes, or blocks with any size. The stream_t structure describes the input stream, file/device. 2. demuxer.c: this does the demultiplexing of the input to audio and video channels, and their reading by buffered packages. The demuxer.c is basically a framework, which is the same for all the input formats, and there are parsers for each of them (mpeg-es, mpeg-ps, avi, avi-ni, asf), these are in the demux_*.c files. The structure is the demuxer_t. There is only one demuxer. 2.a. demux_packet_t, that is DP. Contains one chunk (avi) or packet (asf,mpg). They are stored in memory as in chained list, cause of their different size. 2.b. demuxer stream, that is DS. Struct: demux_stream_t Every channel (a/v) has one. This contains the packets for the stream (see 2.a). For now, there can be 3 for each demuxer : - audio (d_audio) - video (d_video) - DVD subtitle (d_dvdsub) 2.c. stream header. There are 2 types (for now): sh_audio_t and sh_video_t This contains every parameter essential for decoding, such as input/output buffers, chosen codec, fps, etc. There are each for every stream in the file. At least one for video, if sound is present then another, but if there are more, then there'll be one structure for each. These are filled according to the header (avi/asf), or demux_mpg.c does it (mpg) if it founds a new stream. If a new stream is found, the ====> Found audio/video stream: messages is displayed. The chosen stream header and its demuxer are connected together (ds->sh and sh->ds) to simplify the usage. So it's enough to pass the ds or the sh, depending on the function. For example: we have an asf file, 6 streams inside it, 1 audio, 5 video. During the reading of the header, 6 sh structs are created, 1 audio and 5 video. When it starts reading the packet, it chooses the stream for the first found audio & video packet, and sets the sh pointers of d_audio and d_video according to them. So later it reads only these streams. Of course the user can force choosing a specific stream with -vid and -aid switches. A good example for this is the DVD, where the english stream is not always the first, so every VOB has different language :) That's when we have to use for example the -aid 128 switch. Now, how this reading works? - demuxer.c/demux_read_data() is called, it gets how many bytes, and where (memory address), would we like to read, and from which DS. The codecs call this. - this checks if the given DS's buffer contains something, if so, it reads from there as much as needed. If there isn't enough, it calls ds_fill_buffer(), which: - checks if the given DS has buffered packages (DP's), if so, it moves the oldest to the buffer, and reads on. If the list is empty, it calls demux_fill_buffer() : - this calls the parser for the input format, which reads the file onward, and moves the found packages to their buffers. Well it we'd like an audio package, but only a bunch of video packages are available, then sooner or later the: DEMUXER: Too many (%d in %d bytes) audio packets in the buffer error shows up. So everything is ok 'till now, I want to move them to a separate lib. Now, go on: 3. mplayer.c - ooh, he's the boss :) The timing is solved odd, since it has/recommended to be done differently for each of the formats, and sometimes can be done in many ways. There are float variables called a_frame and v_frame, they store the just played A/V position in seconds. The structure of the playing loop : while(not EOF) { fill audio buffer (read & decode audio) + increase a_frame read & decode a single video frame + increase v_frame sleep (wait until a_frame>=v_frame) display the frame apply A-V PTS correction to a_frame check for keys -> pause,seek,... } When playing (a/v), it increases the variables by the duration of the played a/v. - with audio this is played bytes / sh_audio->o_bps Note: i_bps = number of compressed bytes for one second of audio o_bps = number of uncompressed bytes for one second of audio (this is = bps*samplerate*channels) - with video this is usually == 1.0/fps, but I have to note that fps doesn't really matters at video, for example asf doesn't have that, instead there is "duration" and it can change per frame. MPEG2 has "repeat_count" which delays the frame by 1-2.5 ... Maybe only AVI and MPEG1 has fixed fps. So everything works right until the audio and video are in perfect synchronity, since the audio goes, it gives the timing, and if the time of a frame passed, the next frame is displayed. But what if these two aren't synchronized in the input file? PTS correction kicks in. The input demuxers read the PTS (presentation timestamp) of the packages, and with it we can see if the streams are synchronized. Then MPlayer can correct the a_frame, within a given maximal bounder (see -mc option). The summary of the corrections can be found in c_total . Of course this is not everything, several things suck. For example the soundcards delay, which has to be corrected by MPlayer! The audio delay is the sum of all these: - bytes read since the last timestamp: t1 = d_audio->pts_bytes/sh_audio->i_bps - if Win32/ACM then the bytes stored in audio input buffer t2 = a_in_buffer_len/sh_audio->i_bps - uncompressed bytes in audio out buffer t3 = a_buffer_len/sh_audio->o_bps - not yet played bytes stored in the soundcard's (or DMA's) buffer t4 = get_audio_delay()/sh_audio->o_bps From this we can calculate what PTS we need for the just played audio, then after we compare this with the video's PTS, we have the difference! Life didn't get simpler with AVI. There's the "official" timing method, the BPS-based, so the header contains how many compressed audio bytes belong to one second of frames. Of course this doesn't always work... why it should :) So I emulate the MPEG's PTS/sector method on AVI, that is the AVI parser calculates a fake PTS for every read chunk, decided by the type of the frames. This is how my timing is done. And sometimes this works better. In AVI, usually there is a bigger piece of audio stored first, then comes the video. This needs to be calculated into the delay, this is called "Initial PTS delay". Of course there are 2 of them, one is stored in the header and not really used :) the other isn't stored anywhere, this can only be measured... 3.a. audio playback: Some words on audio playback: Not the playing is hard, but: 1. knowing when to write into the buffer, without blocking 2. knowing how much was played of what we wrote into The first is needed for audio decoding, and to keep the buffer full (so the audio will never skip). And the second is needed for correct timing, because some soundcards delay even 3-7 seconds, which can't be forgotten about. To solve this, the OSS gives several possibilities: - ioctl(SNDCTL_DSP_GETODELAY): tells how many unplayed bytes are in the soundcard's buffer -> perfect for timing, but not all drivers support it :( - ioctl(SNDCTL_DSP_GETOSPACE): tells how much can we write into the soundcard's buffer, without blocking. If the driver doesn't support GETODELAY, we can use this to know how much the delay is. - select(): should tell if we can write into the buffer without blocking. Unfortunately it doesn't say how much we could :(( Also, doesn't/badly works with some drivers. Only used if none of the above works. 4. Codecs. They are separate libs. For example libac3, libmpeg2, xa/*, alaw.c, opendivx/*, loader, mp3lib. mplayer.c calls them if a piece of audio or video needs to be played. (see the beginning of 3.) And they call the appropriate demuxer, to get the compressed data. (see 2.) We have to pass the appropriate stream header as parameter (sh_audio/ sh_video), this should contain all the needed info for decoding (the demuxer too: sh->ds). The codecs' seprating is underway, the audio is already done, the video is work-in-progress. The aim is that mplayer.c won't have to know which are the codecs and how to use 'em, instead it should call an init/decode audio/video function. 5. libvo: this displays the frame. The constants for different pixelformats are defined in img_format.h, their usage is mandatory. Each vo driver _has_ to implement these: query_format() - queries if a given pixelformat is supported. return value: flags: 0x1 - supported (by hardware or conversion) 0x2 - supported (by hardware, without conversion) 0x4 - sub/osd supported (has draw_alpha) IMPORTANT: it's mandatorial that every vo driver support the YV12 format, and one (or both) of BGR15 and BGR24, with conversion, if needed. If these aren't supported, not every codec will work! The mpeg codecs can output only YV12, and the older win32 DLLs only 15 and 24bpp. There is a fast MMX-using 15->16bpp converter, so it's not a significant speed-decrease! The BPP table, if the driver can't change bpp: current bpp has to accept these 15 15 16 15,16 24 24 24,32 24,32 If it can change bpp (for example DGA 2, fbdev, svgalib), then if possible we have to change to the desired bpp. If the hardware doesn't support, we have to change to the one closest to it, and do conversion! init() - this is called before displaying of the first frame - initializing buffers, etc. draw_slice(): this displays YV12 pictures (3 planes, one full sized that contains brightness (Y), and 2 quarter-sized which the colour-info (U,V). MPEG codecs (libmpeg2, opendivx) use this. This doesn't have to display the whole frame, only update small parts of it. draw_frame(): this is the older interface, this displays only complete frames, and can do only packed format (YUY2, RGB/BGR). Win32 codecs use this (DivX, Indeo, etc). draw_alpha(): this displays subtitles and OSD. It's a bit tricky to use it, since it's not a part of libvo API, but a callback-style stuff. The flip_page() has to call vo_draw_text(), so that it passes the size of the screen and the corresponding draw_alpha() implementation for the pixelformat (function pointer). The vo_draw_text() checks the characters to draw, and calls draw_alpha() for each. As a help, osd.c contains draw_alpha for each pixelformats, use this if possible! flip_page(): this is called after each frame, this diplays the buffer for real. This is 'swapbuffers' when double-buffering. 6. libao2: this control audio playing As in libvo (see 5.) also here are some drivers, based on the same API: static int control(int cmd, int arg); This is for reading/setting driver-specific and other special parameters. Not really used for now. static int init(int rate,int channels,int format,int flags); The init of driver, opens device, sets sample rate, channels, sample format parameters. Sample format: usually AFMT_S16_LE or AFMT_U8, for more definitions see dec_audio.c and linux/soundcards.h files! static void uninit(); Guess what. Ok I help: closes the device, not (yet) called when exit. static void reset(); Resets device. To be exact, it's for deleting buffers' contents, so after reset() the previously received stuff won't be output. (called if pause or seek) static int get_space(); Returns how many bytes can be written into the audio buffer without blocking (making caller process wait). If the buffer is (nearly) full, has to return 0! If it never gives 0, MPlayer won't work! static int play(void* data,int len,int flags); Plays a bit of audio, which is received throught the "data" memory area, with a size of "len". The "flags" isn't used yet. It has to copy the data, because they can be overwritten after the call is made. Doesn't really have to use all the bytes, it has to give back how many have been used (copied to buffer). static int get_delay(); Has to return how many bytes are in the audio buffer. Be exact, if possible, since the whole timing depends on this! In the worst case, return the size of the buffer. !!! Because the video is synchronized to the audio (card), it's very important !!! that the get_space and get_delay are working!