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

Les articles publiés sur le site

• x86-64 : 64-bit variant of AVX2 hpel_filter

  10 mai 2013, par Jason Garrett-Glaser

  x86-64: 64-bit variant of AVX2 hpel_filter
  
  ~5% faster than 32-bit.
  

  • [DH] common/x86/mc-a2.asm
  • [DH] common/x86/mc-c.c
  • [DH] common/x86/x86util.asm

• OpenCL support improvement/refactoring

  6 mai 2013, par Anton Mitrofanov

  OpenCL support improvement/refactoring
  
  Autoload the OpenCL library so that it's not required to run an openCL-enabled
  build of x264.
  
  Update X264_BUILD, which should have been changed with the first patch.
  

  • [DH] Makefile
  • [DH] common/common.c
  • [DH] common/common.h
  • [DH] common/frame.c
  • [DH] common/opencl.c
  • [DH] common/opencl.h
  • [DH] configure
  • [DH] encoder/encoder.c
  • [DH] encoder/slicetype-cl.c
  • [DH] extras/cl.h
  • [DH] extras/cl_platform.h
  • [DH] x264.h

• x86 : AVX2 high bit-depth denoise_dct

  6 mai 2013, par Henrik Gramner

  x86: AVX2 high bit-depth denoise_dct
  
  28->15 cycles
  
  Also reorder instructions to use fewer registers, 3 cycles faster on Ivy Bridge with 64-bit Windows.
  

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

• x86 : AVX2 high bit-depth quant

  4 mai 2013, par Henrik Gramner

  x86: AVX2 high bit-depth quant
  
  quant_4x4: 13->6 cycles
  quant_4x4_dc: 14->8 cycles
  quant_8x8: 47->24 cycles
  quant_4x4x4: 48->25 cycles
  

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

• x86 : Don’t use explicitly aligned versions of SAD on AVX CPUs

  3 mai 2013, par Henrik Gramner

  x86: Don't use explicitly aligned versions of SAD on AVX CPUs
  
  On modern CPUs movdqu isn't slower than movdqa when used on aligned data and using the same code in both cases saves cache.
  
  This was already done for the high bit-depth AVX2 implementation but the aligned version still exists as dead code so remove that.
  

  • [DH] common/pixel.c
  • [DH] common/x86/pixel.h
  • [DH] common/x86/sad16-a.asm