git.videolan.org Git - x264.git/summary

Les articles publiés sur le site

• x86 : Faster pixel_ssd_nv12

  17 septembre 2016, par Henrik Gramner

  x86: Faster pixel_ssd_nv12
  
  Also drop the MMX2 version to simplify things.
  

  • [DH] common/pixel.c
  • [DH] common/x86/pixel-a.asm
  • [DH] common/x86/pixel.h

• x86 : SSE zigzag_scan_4x4_field

  11 septembre 2016, par Henrik Gramner

  x86: SSE zigzag_scan_4x4_field
  
  Replaces the MMX2 version, one cycle faster.
  
  Also change the checkasm test to use the correct alignment macro.
  

  • [DH] common/dct.c
  • [DH] common/x86/dct-a.asm
  • [DH] common/x86/dct.h
  • [DH] tools/checkasm.c

• x86 : AVX2 mbtree_propagate_list

  7 septembre 2016, par Henrik Gramner

  x86: AVX2 mbtree_propagate_list
  
  SIMD part is around 25% faster than AVX on Haswell, around 7%
  faster when including the runtime of the scalar C wrapper.
  

  • [DH] common/x86/const-a.asm
  • [DH] common/x86/mc-a2.asm
  • [DH] common/x86/mc-c.c
  • [DH] common/x86/trellis-64.asm

• x86 : Move predict_16x16_dc_left calculations to asm

  7 septembre 2016, par Henrik Gramner

  x86: Move predict_16x16_dc_left calculations to asm
  
  1-2 cycles faster and avoids some code duplication to decrease code size.
  
  Also drop the MMX2 implementation in favor of SSE2 to simplify things.
  

  • [DH] common/pixel.c
  • [DH] common/x86/predict-a.asm
  • [DH] common/x86/predict-c.c
  • [DH] common/x86/predict.h

• aarch64 : implement x264_plane_copy_swap_neon

  26 août 2016, par Janne Grunau

  aarch64: implement x264_plane_copy_swap_neon
  
  plane_copy_swap_c: 27054
  plane_copy_swap_neon: 4152
  

  • [DH] common/aarch64/mc-a.S
  • [DH] common/aarch64/mc-c.c