Recherche avancée

Sur d’autres sites (5096)

• Revision 35436 : Petites pétouilles en passant par là (écriture aux dernières normes ...

  22 février 2010, par marcimat@… — Log

  Petites pétouilles en passant par là (écriture aux dernières normes ISO)…

• Revision 36317 : Il y a certains cas où l’on ne peut pas passer par revision_mot notamment ...

  16 mars 2010, par kent1@… — Log

  Il y a certains cas où l’on ne peut pas passer par revision_mot notamment si on s’insère dans un formulaire d’édition d’article dans un pipeline pre_edition ... On ne passe pas la gestion des conflits dans ce cas là

• ac3enc_fixed : convert to 32-bit sample format

  9 janvier 2021, par Lynne

  ac3enc_fixed : convert to 32-bit sample format
  The AC3 encoder used to be a separate library called "Aften", which
  
  got merged into libavcodec (literally, SVN commits and all).
  
  The merge preserved as much features from the library as possible.
  The code had two versions - a fixed point version and a floating
  
  point version. FFmpeg had floating point DSP code used by other
  
  codecs, the AC3 decoder including, so the floating-point DSP was
  
  simply replaced with FFmpeg's own functions.
  
  However, FFmpeg had no fixed-point audio code at that point. So
  
  the encoder brought along its own fixed-point DSP functions,
  
  including a fixed-point MDCT.
  The fixed-point MDCT itself is trivially just a float MDCT with a
  
  different type and each multiply being a fixed-point multiply.
  
  So over time, it got refactored, and the FFT used for all other codecs
  
  was templated.
  Due to design decisions at the time, the fixed-point version of the
  
  encoder operates at 16-bits of precision. Although convenient, this,
  
  even at the time, was inadequate and inefficient. The encoder is noisy,
  
  does not produce output comparable to the float encoder, and even
  
  rings at higher frequencies due to the badly approximated winow function.
  Enter MIPS (owned by Imagination Technologies at the time). They wanted
  
  quick fixed-point decoding on their FPUless cores. So they contributed
  
  patches to template the AC3 decoder so it had both a fixed-point
  
  and a floating-point version. They also did the same for the AAC decoder.
  
  They however, used 32-bit samples. Not 16-bits. And we did not have
  
  32-bit fixed-point DSP functions, including an MDCT. But instead of
  
  templating our MDCT to output 3 versions (float, 32-bit fixed and 16-bit fixed),
  
  they simply copy-pasted their own MDCT into ours, and completely
  
  ifdeffed our own MDCT code out if a 32-bit fixed point MDCT was selected.
  This is also the status quo nowadays - 2 separate MDCTs, one which
  
  produces floating point and 16-bit fixed point versions, and one
  
  sort-of integrated which produces 32-bit MDCT.
  MIPS weren't all that interested in encoding, so they left the encoder
  
  as-is, and they didn't care much about the ifdeffery, mess or quality - it's
  
  not their problem.
  So the MDCT/FFT code has always been a thorn in anyone looking to clean up
  
  code's eye.
  Backstory over. Internally AC3 operates on 25-bit fixed-point coefficients.
  
  So for the floating point version, the encoder simply runs the float MDCT,
  
  and converts the resulting coefficients to 25-bit fixed-point, as AC3 is inherently
  
  a fixed-point codec. For the fixed-point version, the input is 16-bit samples,
  
  so to maximize precision the frame samples are analyzed and the highest set
  
  bit is detected via ac3_max_msb_abs_int16(), and the coefficients are then
  
  scaled up via ac3_lshift_int16(), so the input for the FFT is always at least 14 bits,
  
  computed in normalize_samples(). After FFT, the coefficients are scaled up to 25 bits.
  This patch simply changes the encoder to accept 32-bit samples, reusing
  
  the already well-optimized 32-bit MDCT code, allowing us to clean up and drop
  
  a large part of a very messy code of ours, as well as prepare for the future lavu/tx
  
  conversion. The coefficients are simply scaled down to 25 bits during windowing,
  
  skipping 2 separate scalings, as the hacks to extend precision are simply no longer
  
  necessary. There's no point in running the MDCT always at 32 bits when you're
  
  going to drop 6 bits off anyway, the headroom is plenty, and the MDCT rounds
  
  properly.
  This also makes the encoder even slightly more accurate over the float version,
  
  as there's no coefficient conversion step necessary.
  SIZE SAVINGS :
  
  ARM32 :
  
  HARDCODED TABLES :
  
  BASE           - 10709590
  
  DROP  DSP      - 10702872 - diff :   -6.56KiB
  
  DROP  MDCT     - 10667932 - diff :  -34.12KiB - both :   -40.68KiB
  
  DROP  FFT      - 10336652 - diff : -323.52KiB - all :   -364.20KiB
  
  SOFTCODED TABLES :
  
  BASE           -  9685096
  
  DROP  DSP      -  9678378 - diff :   -6.56KiB
  
  DROP  MDCT     -  9643466 - diff :  -34.09KiB - both :   -40.65KiB
  
  DROP  FFT      -  9573918 - diff :  -67.92KiB - all :   -108.57KiB
  ARM64 :
  
  HARDCODED TABLES :
  
  BASE           - 14641112
  
  DROP  DSP      - 14633806 - diff :   -7.13KiB
  
  DROP  MDCT     - 14604812 - diff :  -28.31KiB - both :   -35.45KiB
  
  DROP  FFT      - 14286826 - diff : -310.53KiB - all :   -345.98KiB
  
  SOFTCODED TABLES :
  
  BASE           - 13636238
  
  DROP  DSP      - 13628932 - diff :   -7.13KiB
  
  DROP  MDCT     - 13599866 - diff :  -28.38KiB - both :   -35.52KiB
  
  DROP  FFT      - 13542080 - diff :  -56.43KiB - all :    -91.95KiB
  x86 :
  
  HARDCODED TABLES :
  
  BASE           - 12367336
  
  DROP  DSP      - 12354698 - diff :  -12.34KiB
  
  DROP  MDCT     - 12331024 - diff :  -23.12KiB - both :   -35.46KiB
  
  DROP  FFT      - 12029788 - diff : -294.18KiB - all :   -329.64KiB
  
  SOFTCODED TABLES :
  
  BASE           - 11358094
  
  DROP  DSP      - 11345456 - diff :  -12.34KiB
  
  DROP  MDCT     - 11321742 - diff :  -23.16KiB - both :   -35.50KiB
  
  DROP  FFT      - 11276946 - diff :  -43.75KiB - all :    -79.25KiB
  PERFORMANCE (10min random s32le) :
  
  ARM32 - before -  39.9x - 0m15.046s
  
  ARM32 - after  -  28.2x - 0m21.525s
  
                         Speed :  -30%
  ARM64 - before -  36.1x - 0m16.637s
  
  ARM64 - after  -  36.0x - 0m16.727s
  
                         Speed : -0.5%
  x86   - before - 184x -    0m3.277s
  
  x86   - after  - 190x -    0m3.187s
  
                         Speed :   +3%
  

  • [DH] doc/encoders.texi
  • [DH] libavcodec/Makefile
  • [DH] libavcodec/ac3enc.c
  • [DH] libavcodec/ac3enc.h
  • [DH] libavcodec/ac3enc_fixed.c
  • [DH] libavcodec/ac3enc_float.c
  • [DH] libavcodec/ac3enc_template.c
  • [DH] libavcodec/version.h
  • [DH] tests/fate/ac3.mak
  • [DH] tests/fate/ffmpeg.mak
  • [DH] tests/ref/fate/unknown_layout-ac3
  • [DH] tests/ref/lavf/rm