1. 30 Mar, 2016 6 commits
  2. 29 Mar, 2016 7 commits
  3. 28 Mar, 2016 17 commits
  4. 26 Mar, 2016 6 commits
  5. 25 Mar, 2016 4 commits
    • Yi Luo's avatar
      8x8/16x16 HT types V_DCT to H_FLIPADST SSE2 optimization · 770bf715
      Yi Luo authored
      - Wrote function: fidtx8_sse2() and fidtx16_sse2().
      - Turned on vp10_fht8x8_sse2()/vp10_fht16x16_sse2() for new types.
      - Updated 8x8/16x16 unit tests for accuracy/speed.
      - Running 20K times with random numbers and getting through
        tx type from V_DCT to H_FLIPADST, SSE2 speed improvement:
        8x8: ~131%
        16x16: ~66%
      
      Change-Id: Ibbb707e932a08fec3b1f423a7dab280a1d696c9a
      770bf715
    • Alex Converse's avatar
      Fix failing test: VP10/SuperframeTest.TestSuperframeIndexIsOptional/0 · c5d118f1
      Alex Converse authored
      Failing since: 66f2f65e Merge MISC_FIXES
      
      Change-Id: I8135d6a9d74972c595f1b6294fb842e68f91e50e
      c5d118f1
    • Marco's avatar
      vp9: 1 pass vbr mode at speed 5: switch to use mv.search to NSTEP. · f94124cf
      Marco authored
      Change only affects 1 pass, vbr, speed = 5 (real-time mode).
      
      Some improvement for high motion content.
      AvgPSNR/SSIM metrics for ytlive set all up, on average ~2%,
      some clips (high motion ones) up 4/5%.
      
      Encoder speed down: on mynintendo_x1.1280_720.y4m: 47fps -> 44fps.
      
      Change-Id: I9e3eaa6392dcb6b5b44ee6f43004f97ba859bc11
      f94124cf
    • Yue Chen's avatar
      A major speed up for obmc experiment · e63792e5
      Yue Chen authored
      Skip checking obmc when regular inter predictor is not so good (the
      rd-cost for Y residual is greater than the total rd of the best mode
      so far.)
      
      Performance change compared to full rd search:
        +0.006% lowres, -0.056% midres
      Encoding time :
        1.14X baseline (was 1.42X)
      
      Change-Id: I11350f955a20e1a2331be458537a915e09fbedf3
      e63792e5