1. 27 Jun, 2017 1 commit
    • Wei-Ting Lin's avatar
      Fix token-encoding errors in NCOBMC · 1d46d903
      Wei-Ting Lin authored
      mi[0] is not set properly when encoding all mvs in a
      super-block first. After this patch NCOBMC can function
      properly.
      
      Change-Id: I149a50184c4823c0d3b82b6b21c7608e639668e6
      1d46d903
  2. 26 Jun, 2017 7 commits
    • Yaowu Xu's avatar
      quantize.c: convert to int before apply sign · d43d6777
      Yaowu Xu authored
      This change makes the conversions similar to those in av1_quantize.c,
      and fix ubsan warnings shown in nightly tests.
      
      Change-Id: I90851a80dcb9f052a32bf22199fd9ef8ff927725
      d43d6777
    • James Zern's avatar
      aom_dsp.cmake: add highbd_quantize_intrin_avx2.c · 284c8830
      James Zern authored
      added in:
      193422e7 Add avx2 highbd_quantize_b
      
      Change-Id: Ie4ba48042ffd36d69d2bf200bba12a1d924c8f9c
      284c8830
    • Yushin Cho's avatar
      daala-dist: high bit depth support · 8ab875d6
      Yushin Cho authored
      Change-Id: Idafef140d3425a9a9f66cb8864a804c4d2a89a70
      8ab875d6
    • Adrian Grange's avatar
      Remove redundant code line · 0d6ab64e
      Adrian Grange authored
      Change-Id: Iaf9e8192905ea24946d6e2831be5739eaa564fae
      0d6ab64e
    • Yushin Cho's avatar
      Fix daala-dist for var-tx · 0474912c
      Yushin Cho authored
      The var-tx has its own suite of tx size/type RD search functions,
      which recursively split the partition into square tx blocks.
      
      The Daala-dist requires access to 8x8 pixels (both decoded and predicted)
      since it measures the distortion for multiple of a 8x8 pixels.
      Thus, if tx block is smaller than 8x8, it waits until all of sub8x8 blocks
      are RD searched (with MSE) then replaces the MSE of 8x8 pixels with
      daala-dist's calculated distortion for 8x8 pixels.
      
      It is also applied to luma pixels only.
      
      Change-Id: Ic4891e89b4ef05cf880aa26781d2d06ccf3142de
      0474912c
    • Lester Lu's avatar
      New experiment: LGT · ad8290b8
      Lester Lu authored
      In previous ADSTs, DST-7 and DST-4 are used for length 4 and length
      8/16/32, respectively. In this LGT experiment we explore transforms
      between DST-4 and DST-7. When CONFIG_LGT flag is on, adst4 and adst8
      are replaced by lgt4 and lgt8, the intermediate transforms with
      pre-chosen parameters.
      
      The LGTs applied here are lgt4_160 and lgt8_170, where the numbers
      mean the self-loop weights times 100. The associated values for DST-7
      and DST-4 are 100 and 200.
      
      ovr_psnr:
      lowres: -0.140
      midres: -0.131
      hdres: -0.078
      
      These changes are not applied to the highbd scenario in the
      current version.
      
      Change-Id: I20600456da8766528b2b6b11aa28801e70af498e
      ad8290b8
    • Yushin Cho's avatar
      Fix daala_dist to handle visible pixels only · 75b01004
      Yushin Cho authored
      - If invisible pixels, av1_daala_dist() simply use source pixles for dst.
      - Added av1_daala_dist_diff() which inputs diff signal instead of dst.
      
      - Refactored daala_dist code so that av1_daala_dist() and _diff()
      is called inside av1's distortion calculation functions, pixel_sse() and
      sum_squares_visible().
      
      Change-Id: Id857db52fe19856d92c46a9e84ac2962c01ae045
      75b01004
  3. 24 Jun, 2017 9 commits
  4. 23 Jun, 2017 6 commits
    • Fergus Simpson's avatar
      scaling: Fix scaling numerator arguments for KFs · 7a3f4b35
      Fergus Simpson authored
      0835e7b80 left out the required changes to aomenc.c for the KF numerator
      arguments for resize and superres to work. This patch adds them.
      
      Change-Id: I350b01c8b187188de5313fffaa15c1ec9f052469
      7a3f4b35
    • Wei-Ting Lin's avatar
      Add NCOBMC_ADAPT_WEIGHT experiment flag · 9c6f8547
      Wei-Ting Lin authored
      NCOBMC_ADAPT_WEIGHT allow using different interpolation kernels
      to combine overlapped predictions generated using mvs in the
      neighboring blocks.
      This experiment will build on top of MOTION_VAR and it might conflict
      with WARPED_MOTION in the developing stage, so will only be effective
      when MOTION_VAR is on and WARPED_MOTION is off.
      
      Change-Id: I4f1b6e55b6146ed443955751c09bfa22ef2f33e8
      9c6f8547
    • David Barker's avatar
      Fix bug with chroma-sub8x8 + resizing · 11132ce3
      David Barker authored
      The special path in build_inter_predictors for chroma blocks
      corresponding to sub8x8 luma blocks would always fetch the
      scale factors from 'xd', which correspond (at least in the
      decoder) to the references of the block for which is_chroma_reference
      returns true.
      
      The correct behaviour is to fetch the scale factors from 'ref_buf',
      which corresponds to the references of the block currently being
      predicted.
      
      This patch fixes some encode/decode mismatches which were caused
      by the above behaviour when the various reference frames had
      different sizes.
      
      Change-Id: I48a0a167ea25d47d08018016cf8b77885b3b5d6b
      11132ce3
    • Debargha Mukherjee's avatar
      Do not find transformation for very few points · afe7c5fd
      Debargha Mukherjee authored
      Adds check to make sure that find transformation functions are
      never called for 0 points.
      
      Change-Id: I2d7cf40aace535b1d708d6189aea9c1e0f7c281b
      afe7c5fd
    • James Zern's avatar
      variance_test: move Subpel* from tuples to TestParams · d598f04d
      James Zern authored
      this normalizes these tests with the regular variance ones both in
      implementation and test list output
      
      Change-Id: Iaa549f2e2a054d716c24f5a64baf700747c55295
      d598f04d
    • Yaowu Xu's avatar
      Revert "Always use the 16-bit coding path" · cde0a31b
      Yaowu Xu authored
      BUG=aomedia:620
      
      This reverts commit 4da880bd.
      
      Change-Id: I40a2295083b3328ef743d3eab9c2aa514dc698eb
      cde0a31b
  5. 22 Jun, 2017 9 commits
    • Steinar Midtskogen's avatar
      Silence warnings in VS · 079acac1
      Steinar Midtskogen authored
      BUG=aomedia:615
      
      Change-Id: I827e857d310020705a5292ef8fe817bc042d8dd0
      079acac1
    • Yaowu Xu's avatar
      Prevent divide-by-zero · 9180b6e8
      Yaowu Xu authored
      Change-Id: Id22615d461bf16272d1b2e2c72ae7e00db8bcb5c
      9180b6e8
    • Yaowu Xu's avatar
      convert to int before apply sign · bdda9d4e
      Yaowu Xu authored
      avoids overflow of unsigned integer.
      
      Change-Id: Ic92974b508bb0cd6fc680203ffa6cff14d644ff7
      bdda9d4e
    • Jingning Han's avatar
      Fix compiler warning in joint_motion_search · cb637674
      Jingning Han authored
      Avoid compiler warning when global-motion is off.
      
      Change-Id: Ie6a0d3e4efc0e06b263e8c8c0c0dc153738c3804
      cb637674
    • Tom Finegan's avatar
      Add missing dependency to CMake test object libraries · d2a74d6b
      Tom Finegan authored
      The test libs omitted the dependency on aom.
      
      Change-Id: I47fe769252151cf0bf81f245452e22512963c9fa
      d2a74d6b
    • Zoe Liu's avatar
      Add entropy stats dump out for individual frame context type · a56f916e
      Zoe Liu authored
      Change-Id: Id0cd184e8b3cea085ecc3adbc7fea7bb765c7986
      a56f916e
    • Yi Luo's avatar
      Add avx2 highbd_quantize_b · 193422e7
      Yi Luo authored
      - First pass encoding time reduces ~10.9% on i7-6700
        at 100 frames, 1080p.
      - avx2 works for coeff number >= 8 cases; coeff number < 8
        case will be implemented by sse2.
      - Unit test is added type B/FP/DC.
      
      Change-Id: Ibe5b7807c64e6dfc2d59c470ed50a6e8ca94ef7c
      193422e7
    • Yushin Cho's avatar
      Fix daala-dist, rd tx search · 04eb9594
      Yushin Cho authored
      Previously, for block >=8x8, and tx < 8x8,
      we skipped setting the early-exit flag in block_rd_txfm() because
      distortion for sub8x8 tx block is from MSE but reference (best)
      is from daala-dist.
      However, not setting early-exit flag turned out to be the reason
      for a regression in MSE probe mode of daala-dist because
      it loses the chance to set rd_stats properly.
      
      On the other hand, there is still a small regression, say 0.05% psnr bd-rate,
      which seems to occur in the case that a tx block in a partition has chosen
      the skipped rd_cost since it is smaller than non-skip rd_cost and
      set the early-exit flag to 0 (so, not exit), but the daala-dist applied
      to the whole partition cannot access the same info but can choose from
      two kinds of rd_costs:
      1) all tx blocks are skipped (even if a tx block has non-zero coeff) and 0 bits
      2) sum of final distortion of all tx blocks (i.e. non-zero coeff decoded)
      and bits to encode coeffs.
      
      Change-Id: I2ec69972aa1f22d465293cb9e8d5e18ef2c6f7f3
      04eb9594
    • Yaowu Xu's avatar
      Add missing accumulation cross threads · a0cc9aa8
      Yaowu Xu authored
      BUG=aomedia:618
      
      Change-Id: Ie96ccc363462a28527c99a72e97b7acaf2ab0ff8
      a0cc9aa8
  6. 21 Jun, 2017 8 commits