git.videolan.org Git - ffmpeg.git/rss log

Les articles publiés sur le site

• avcodec/aarch64/vvc : Optimize NEON version of vvc_dmvr

  3 mars, par Krzysztof Pyrkosz

  avcodec/aarch64/vvc: Optimize NEON version of vvc_dmvr
  
  This patch replaces blocks of instructions performing rounding and
  widening shifts with one-liners achieving the same result.
  
  Before and after on A78
  dmvr_8_12x20_neon:                                      86.2 ( 6.90x)
  dmvr_8_20x12_neon:                                      94.8 ( 5.93x)
  dmvr_8_20x20_neon:                                     141.5 ( 6.50x)
  dmvr_12_12x20_neon:                                    158.0 ( 3.76x)
  dmvr_12_20x12_neon:                                    151.2 ( 3.73x)
  dmvr_12_20x20_neon:                                    247.2 ( 3.71x)
  dmvr_hv_8_12x20_neon:                                  423.2 ( 3.75x)
  dmvr_hv_8_20x12_neon:                                  434.0 ( 3.69x)
  dmvr_hv_8_20x20_neon:                                  706.0 ( 3.69x)
  
  dmvr_8_12x20_neon:                                      77.2 ( 7.70x)
  dmvr_8_20x12_neon:                                      66.5 ( 8.49x)
  dmvr_8_20x20_neon:                                      92.2 ( 9.90x)
  dmvr_12_12x20_neon:                                     80.2 ( 7.38x)
  dmvr_12_20x12_neon:                                     58.2 ( 9.59x)
  dmvr_12_20x20_neon:                                     90.0 (10.15x)
  dmvr_hv_8_12x20_neon:                                  369.0 ( 4.34x)
  dmvr_hv_8_20x12_neon:                                  355.8 ( 4.49x)
  dmvr_hv_8_20x20_neon:                                  574.2 ( 4.51x)
  
  Signed-off-by: Martin Storsjö <martin@martin.st>
  

  • [DH] libavcodec/aarch64/vvc/inter.S

• swscale/aarch64 : dotprod implementation of rgba32_to_Y

  3 mars, par Krzysztof Pyrkosz

  swscale/aarch64: dotprod implementation of rgba32_to_Y
  
  The idea is to split the 16 bit coefficients into lower and upper half,
  invoke udot for the lower half, shift by 8, and follow by udot for the
  upper half.
  
  Benchmark on A78:
  bgra_to_y_128_c:                                       682.0 ( 1.00x)
  bgra_to_y_128_neon:                                    181.2 ( 3.76x)
  bgra_to_y_128_dotprod:                                 117.8 ( 5.79x)
  bgra_to_y_1080_c:                                     5742.5 ( 1.00x)
  bgra_to_y_1080_neon:                                  1472.5 ( 3.90x)
  bgra_to_y_1080_dotprod:                                906.5 ( 6.33x)
  bgra_to_y_1920_c:                                    10194.0 ( 1.00x)
  bgra_to_y_1920_neon:                                  2589.8 ( 3.94x)
  bgra_to_y_1920_dotprod:                               1573.8 ( 6.48x)
  
  Signed-off-by: Martin Storsjö <martin@martin.st>
  

  • [DH] libswscale/aarch64/input.S
  • [DH] libswscale/aarch64/swscale.c

• avcodec/mpeg12enc : Simplify writing bits

  3 mars, par Andreas Rheinhardt

  avcodec/mpeg12enc: Simplify writing bits
  
  Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
  

  • [DH] libavcodec/mpeg12enc.c

• avcodec/mpegvideo : Mark ff_mpv_common_defaults() as av_cold

  3 mars, par Andreas Rheinhardt

  avcodec/mpegvideo: Mark ff_mpv_common_defaults() as av_cold
  
  Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
  

  • [DH] libavcodec/mpegvideo.c

• avcodec/speedhqenc : Inline ff_speedhq_mb_y_order_to_mb()

  3 mars, par Andreas Rheinhardt

  avcodec/speedhqenc: Inline ff_speedhq_mb_y_order_to_mb()
  
  It is an extremely simple function that is only called once,
  so it should be inlined.
  
  Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
  

  • [DH] libavcodec/speedhqenc.c
  • [DH] libavcodec/speedhqenc.h