Recherche avancée

Sur d’autres sites (15334)

• avformat with custom IO - threadsafe ?

  30 mars 2020, par Daniel

  Consider the following pseudo-code snippet :

  



  init {
  avdevice_register_all
  avformat_network_init
}

thread {
  // open input
  av_find_input_format
  avformat_open_input
  avformat_find_stream_info

  // open output
  avformat_alloc_output_context2
  avio_alloc_context(outputbuffer,size,1,extradata,0,&staticWrite,0)
  avformat_new_stream
  avcodec_parameters_copy
  avformat_write_header

  loop {
    av_read_frame
    av_write_frame
  }
}

staticWrite {
  // here comes data
}


  



  When I have only one thread everything works fully good.
However when there are more threads (2,3), there are unexpected crashes (mainly segfaults).

  



  Now my goal with this question is not finding my segfaults, but I want to get a confirmation that the above snippet is good in terms of multi-threaded avformat, or if not, what else do I need to be able to use avformat's functions from multiple threads ?

  



  ffmpeg was compiled with this config :

  



  configuration: --enable-cross-compile --cross-prefix=arm-linux-gnueabihf- --arch=armhf --target-os=linux --prefix=. --enable-mmal --enable-omx --enable-omx-rpi --enable-gpl --enable-nonfree --enable-pthreads --extra-ldflags='-Llib -Wl,-rpath-link,lib' --extra-libs=-ldl --extra-cflags=-Iinclude


  


• h264_metadata : Localize code for display orientation

  17 juin 2019, par Andreas Rheinhardt

  h264_metadata : Localize code for display orientation
  The recent changes to h264_metadata (enabled by the recent changes to
  
  ff_cbs_write_packet) made it possible to add side_data to the output
  
  packet at any place, not only after the output packet has been written
  
  and the properties of the input packet copied. This means that one can
  
  now localize the code to add display orientation side-data to the packet
  
  to the place dealing with said display-orientation.
  Furthermore, the documentation of av_display_rotation_set states that
  
  the matrix will be fully overwritten by it, so there is no need to
  
  allocate it with av_mallocz.
  Signed-off-by : Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
  

  • [DH] libavcodec/h264_metadata_bsf.c

• Using ffprobe to get number of keyframes in raw AVI file *without* processing entire file ?

  26 juillet 2018, par aggieNick02

  This question and answer cover how to get the framecount and keyframe count from an AVI file, which is very useful. I’ve got a raw AVI file and want to count the number of keyframes (equivalent to non-dropped frames for raw AVI), but it takes a long time to process through a raw AVI file.

  There is some way to get this information without fully processing the file, as VirtualDub provides both framecount and key framecount in the file information, as well as total keyframe size, almost instantly for a 25-second raw 1920x1080 AVI. But ffprobe requires count_frames to populate nb_read_frames, which takes some good processing time.

  I can do some math with the file’s size and the frame’s width/height/format to get a fairly good estimate of the number of frames, but I’m worried the overhead of the container could be enough to throw the math off for very short clips. (For my 25 second clip, I get 1286.12 frames, when there are really 1286.)

  Any thoughts on if there is a way to get this information programatically with ffprobe or ffmpeg without processing the whole file ? Or with another API on windows ?