Recherche avancée

Sur d’autres sites (3641)

• avcodec/mips/mpegvideo_mmi : Version 2 of the optimizations for loongson mmi

  17 mai 2016, par Zhou Xiaoyong

  avcodec/mips/mpegvideo_mmi : Version 2 of the optimizations for loongson mmi
  1. no longer use the register names directly and optimized code format
  
  2. to be compatible with O32, specify type of address variable with mips_reg and handle the address variable with PTR_ operator
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavcodec/mips/mpegvideo_mmi.c

• avcodec_open segmentation fault

  24 juin 2014, par user3715403

  Good afternoon,

  I got a copy of the ffmpeg’s api-example.c (https://www.ffmpeg.org/doxygen/0.6/api-example_8c-source.html). I tested the function video_encode_example() and got the code working as expected.

  I then re-factored this function into different class methods and now I get a seg fault with the call to avcodev_open(). I then changed the code to call the original video_encode_example() from a class method and the calls to avcodec... were successful.

  It appears that when called from a function (eg video_encode_example()), the behaviour is as expected, but avcodev_open fails when called from a class method. I traced the code and in both cases the values for the avcodev_open() arguments are assigned (and similar) pointer values. cgdb reports that the seg fault occurs in avcodev_open2()

  Any suggestions ?
  The code was compiled and tested on a Ubuntu 12.04 64 bit machine) using the libav package available thru apt-get (libavcodec.so.53.35.0)

  Regards,
  Daniel

  Added comment following the first response :

  From the link above, I copied the function video_encode_example() and placed the call inside a class constructor. This part is working.

  VideoEncoder::VideoEncoder( const std::string& aFileName)
  
      : mCodec( NULL)
  
      , mCodecContext( NULL)
  
      , picture (NULL)
  
      , mFileName( aFileName)
  
      , mFileHandler( NULL)
  
  {
  
      // must be called before using avcodec lib
  
      ::avcodec_init();
  
  
  
      // register all the codecs
  
      ::avcodec_register_all();
  
  
  
      video_encode_example( aFileName.c_str());
  
      return;
  
      ...
  
  }

  The re-factored part includes splitting the avcodev call (from the original video_encode_example() function) to different VideoEncoder methods. The constructor now looks like :

  VideoEncoder::VideoEncoder( const std::string& aFileName)
  
      : mCodec( NULL)
  
      , mCodecContext( NULL)
  
      , picture (NULL)
  
      , mFileName( aFileName)
  
      , mFileHandler( NULL)
  
  {
  
      // must be called before using avcodec lib
  
      ::avcodec_init();
  
  
  
      // register all the codecs
  
      ::avcodec_register_all();
  
  
  
      // find the mpeg1 video encoder
  
      mCodec = ::avcodec_find_encoder( CODEC_ID_MPEG1VIDEO);
  
      if (!mCodec) {
  
          fprintf( stderr, "codec not found\n");
  
          exit( 1);
  
      }
  
  
  
      mCodecContext = ::avcodec_alloc_context3( mCodec);
  
      picture = ::avcodec_alloc_frame();
  
  
  
      // put sample parameters
  
      mCodecContext->bit_rate = 400000;
  
  
  
      // frames per second
  
      mCodecContext->time_base = (AVRational){1,25};
  
      mCodecContext->gop_size = 10;               // emit one intra frame every ten frames
  
      mCodecContext->max_b_frames = 1;
  
      mCodecContext->pix_fmt = PIX_FMT_YUV420P;
  
  
  
      // open it
  
      // Initializes the AVCodecContext to use the given AVCodec. Prior to using this function the context has to be allocated.
  
      if (::avcodec_open( mCodecContext, mCodec) < 0) {
  
          fprintf(stderr, "could not open codec\n");
  
          exit( 1);
  
      }
  
  
  
      mFileHandler = fopen( mFileName.c_str(), "wb");
  
      if (!mFileHandler) {
  
          fprintf( stderr, "could not open %s\n", mFileName.c_str());
  
          exit( 1);
  
      }
  
  
  
  }

  The call to avcodec_open causes the seg fault. In addition, the fault occurs whether using avcodev_.. or ::avcodev_...

• aarch64 : me_cmp : Don't do uaddlv once per iteration

  12 juillet 2022, par Martin Storsjö

  aarch64 : me_cmp : Don't do uaddlv once per iteration
  The max height is currently documented as 16 ; the max difference per
  
  pixel is 255, and a .8h element can easily contain 16*255, thus keep
  
  accumulating in two .8h vectors, and just do the final accumulationat the
  
  end. This should work for heights up to 256.
  This requires a minor register renumbering in ff_pix_abs16_xy2_neon.
  Before :       Cortex A53    A72    A73   Graviton 3
  
  pix_abs_0_0_neon :   97.7   47.0   37.5   22.7
  
  pix_abs_0_1_neon :  154.0   59.0   52.0   25.0
  
  pix_abs_0_3_neon :  179.7   96.7   87.5   41.2
  
  After :
  
  pix_abs_0_0_neon :   96.0   39.2   31.2   22.0
  
  pix_abs_0_1_neon :  150.7   59.7   46.2   23.7
  
  pix_abs_0_3_neon :  175.7   83.7   81.7   38.2
  Signed-off-by : Martin Storsjö <martin@martin.st>
  

  • [DH] libavcodec/aarch64/me_cmp_neon.S