1. 19 Apr, 2017 10 commits
    • James Zern's avatar
      use storage.googleapis for testdata download · f9687c5b
      James Zern authored
      replace downloads.webmproject.org with the canonical
      storage.googleapis.com/... form. this appears less likely to fail when
      dealing with multiple concurrent connections.
      original change in libvpx:
      22f7aca09 use storage.googleapis for testdata download
      Change-Id: Ie1bc329fecb336f089a662451b8ebf0c0d9dc2d6
    • James Zern's avatar
      Revert "Enable ext-refs as default" · 829dbf29
      James Zern authored
      This reverts commit acbb5dfe.
      This causes build warnings / failures with a combination of experiments.
      Change-Id: If8364845b7c1455957956c464da46805275c90c0
    • Fergus Simpson's avatar
      bitstream-debug: Merge queue_pop and queue_push · 2f905d26
      Fergus Simpson authored
      The bitstream_queue_pop and bitstream_queue_push functions were
      implemented twice each, one implementation for CONFIG_DAALA_EC being set
      and for it not being set. This could cause issues with debugging since
      there's two potential paths, meaning breakpoints and edits need to be
      inserted twice.
      The functions were already minimally different, so this patch merges the
      implementation pairs to simplify the code.
      Change-Id: I37cdd96470f0ae41b25dc28da50e2c7a1b9af01e
    • James Zern's avatar
      test-data: restrict large yuv to perf tests · 86251126
      James Zern authored
      niklas_1280_720_30.yuv (~619M) is only required with
      Change-Id: I143f7e83215b4b7214cf45b752c94a9ee05093d4
    • Zoe Liu's avatar
      Enable ext-refs as default · acbb5dfe
      Zoe Liu authored
      Change-Id: I53c844724794e13116152d0492bbcdd3ece12d17
    • David Barker's avatar
      Fix some rdopt bugs in ext-inter · cb03dc34
      David Barker authored
      * Refactor write_inter_mode and cost_mv_ref slightly
      * In handle_inter_mode: When discount_newmv_test returns 1,
        ext-inter is enabled, and we're costing up a compound mode,
        we would previously call cost_mv_ref with mode=NEARESTMV and
        comp_pred=1. But this is inconsistent when ext-inter is enabled!
        Fix this by changing the behaviour of cost_mv_ref to (effectively)
        derive comp_pred from the mode passed in, and switch to
        mode=NEAREST_NEARESTMV in the above case.
      * Fix incorrect motion vector costing for some compound modes
      * Fix a case where compound blocks were mis-costed. This primarily
        affects ext-inter *without* wedge or compound-segment enabled,
        but should still have some small impact when those experiments
        are enabled.
        The cause of this is that we only spend bits to send the
        interinter compound type if is_any_masked_compound_used(bsize)
        returns true. But the rdopt loop forgot to check this condition
        in one case.
      Change-Id: I8b54cb3552f9ec384fd5ea260e12685168fbccc1
    • Steinar Midtskogen's avatar
      Retune CDEF dering threshold adjustment · fade4637
      Steinar Midtskogen authored
      Change the adjustment range from [50% ... 300%] to [25% ... 100%].
      Adjustments above 100% were very rare, and capping the adjustment at
      100% adds SIMD optimisation opportunities.  And lowering the other end
      to 25% seems to help compresson slightly.
      Low latency, used-cpu=0:
         PSNR | PSNR Cb | PSNR Cr | PSNR HVS |    SSIM | MS SSIM | CIEDE 2000
      -0.0453 | -0.1118 | -0.1127 |  -0.0689 | -0.0429 | -0.0814 |    -0.0762
      High latency, used-cpu=0:
         PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
      -0.0303 |  0.0583 |  0.1740 |  -0.0440 | 0.0033 | -0.0042 |     0.0040
      Change-Id: Id999158330a53e8c3383cd0e53a91c7f59fe062a
    • Steinar Midtskogen's avatar
      Signal CDEF damping in the frame header · 0c966a50
      Steinar Midtskogen authored
      Change-Id: I57f232623e647f029b007de8ddb203c47ca3c11c
    • Michael Bebenita's avatar
      Fix CDEF SIMD code for cb4x4 · efacce93
      Michael Bebenita authored
      Change-Id: Iec83111328d5343250b9a35dc2896541393f3efa
    • Sean Purser-Haskell's avatar
      Fix array bounds overflow in warped test. · a57f77f2
      Sean Purser-Haskell authored
      Change-Id: Ie92cd0a4a97312f40fe8604ac671dc43188b5647
  2. 18 Apr, 2017 10 commits
    • Alex Converse's avatar
      intrabc: Add experiment · 576f0652
      Alex Converse authored
      Change-Id: If73aee341df8e9fe2d2a4e46f0bf4a0180cd4b5f
    • Ralph Giles's avatar
      Don't export kMinCompressedSize. · 19944291
      Ralph Giles authored
      kMinCompressedSize was defined as a global constant, so it
      was exported by default and possibly causing duplicate
      symbol issues for applications. In particular, libvpx also
      exports this symbol.
      Instead, move the definition into the only function which
      references is. As a local variable it is hidden from the
      linker and not part of the global symbol table.
      Change-Id: If8b3bc4c487197237adb04749cd3dd4792e39893
    • Fangwen Fu's avatar
      Set palette_throughput experiment default on. · 2d422fb1
      Fangwen Fu authored
      Change-Id: I6c0897f9d432121c425ae1cea34bd7f72d207c72
    • Ryan Lei's avatar
      unify the intra prediction angle options for different block sizes · c81ea6e0
      Ryan Lei authored
      this change unifies the intra prediction angles so that 8x8 block has
      the same number of angles and angle steps as 16x16 and 32x32 blocks.
      the benefit is that HW implementation can have the same logic to
      handle all block sizes. otherwise, prediction angles for 8x8 is
      different from other block sizes, which requies special logic.
      Change-Id: I99f8d82b4b96fa1442a657b56a93e326f1eec321
    • Angie Chiang's avatar
      Redo txfm/quant in av1_txk_search · 2ed03a37
      Angie Chiang authored
      Change-Id: I622d499187f3881b274ca6cf3745f51fa0103b18
    • Angie Chiang's avatar
      Add txk_sel exp · cd9b03fb
      Angie Chiang authored
      This will separate the transform kernel selection from lv_map
      experiment such that we can evaluate each feature's performance
      Note that txk_sel is build on top of lv_map
      Change-Id: I5bd1ea99be30000efcdc2bcd42de002b78b1c3c8
    • Timothy B. Terriberry's avatar
      ec_smallmul: Convert CDFs to iCDFs. · f6c807c5
      Timothy B. Terriberry authored
      Hoists the iCDF conversion outside of the daala code.
      We directly store 32768 - cdf[i] in each cdf, to avoid having to
      convert the whole array every time a symbol is coded.
      This works with ec_multisymbol, new_tokenset, and ec_adapt.
      Compared to Change-Id Idbbd3743e9189146cb519d5b984bdabd69e3f4c0,
      this improves decoder runtimes by 1.15% at QP=55 and 2.64% at
      The overall slowdown of ec_smallmul is now 0.12% at QP=55 and
      0.44% at QP=20.
      Encoder output should not change, and all streams should remain
      decodable without decoder changes.
      Change-Id: I06b8b75b667bb1bc4ddffc78f895e48a09f4c578
    • Yi Luo's avatar
      Deliver the eob threshold to inverse transform · f8e87b46
      Yi Luo authored
      Change-Id: Iaa8ab77eb4a982759939fb5fc475f699cb21a4e1
    • Yi Luo's avatar
      Move width branch out of height loop · a4d87992
      Yi Luo authored
      - AVX2 Copy and average functions are faster,
        Copy function: ~4%-57%
        Avg function:  ~17%-54%
      Change-Id: Ib1732cd90eb353379ef50ecbb1e207860969f1c3
    • zhipin deng's avatar
      enable mv_compress by default · 1c25c758
      zhipin deng authored
      Change-Id: Idcf587227a7e508344d00a4d456c747b4f8d109b
  3. 17 Apr, 2017 12 commits
    • Thomas Daede's avatar
      Remove rt deadline. · 8082614e
      Thomas Daede authored
      The "good" speed levels are universally better than the "rt" ones,
      running faster to achieve the same quality.
      rt mode also turned off alt refs and lag-in-frames, but these
      are still accessible separately (and the low latency test case
      explicitly sets them).
      Some features were used by the rt scale and not the good scale.
      Two additional "good" levels, 7 and 8, were added to accomidate
      these features and not reduce test coverage.
      Change-Id: I3a6a78ddb664511762c197bc41f3a9909665b565
    • Fergus Simpson's avatar
      bitstream-dbg: Add missing include to decodeframe · 350a9b76
      Fergus Simpson authored
      If daala_ec is disabled while bistream_debug is enabled, decodeframe.c
      fails to compile due to aom_util/debug_util.h not being included
      This patch just adds the missing include so that decodeframe.c will
      still build with bitstream_debug enabled and daala_ec disabled.
      Change-Id: I2e7eebdd17431a804be5b00239a9ae581e73f731
    • Sarah Parker's avatar
      Avoid exiting tx size search when tx-size is square · 36661926
      Sarah Parker authored
      This fixes a mismatch in ext-tx + rect-tx introduced
      by a refactor in 2d147c16.
      Change-Id: I5bc40b2dcd7bf53b20c1bea454aae49f2d5dfe67
    • Yushin Cho's avatar
      Skip adding zero siginal to prediction with DC only idct · 27acc478
      Yushin Cho authored
      If DC only idct gives zero, then we can skip the steps which
      add zero signal to predicted signal.
      DC only idct cases will occur more frequently at lower bit rates.
      Similar changes can be done for C version of high bit depth idct functions.
      Change-Id: I53af22904568f7043091710da70ca8299bf361c5
    • Michael Bebenita's avatar
      Fix inspection mi grid size. · b5bf51ec
      Michael Bebenita authored
      Change-Id: I152837d872ce5b234c62c7190a2350cea710d579
    • Michael Bebenita's avatar
      Compress analyzer data using RLE. · f1207b60
      Michael Bebenita authored
      This significantly reduces the size of the exported
      data and improves analyzer decoding time.
      Change-Id: I0553c76630009754038a86b634b545db24fb3dd6
    • Yaowu Xu's avatar
      Fix build for motion-var and ext-inter · a631722c
      Yaowu Xu authored
      Change-Id: If7a3dc0006543c30f55f4ed2b33f28b69395bd60
    • Yue Chen's avatar
      Changes for chroma u8x8 obmc · f0dfc03d
      Yue Chen authored
      (1) Add a macro DISABLE_CHROMA_U8X8_OBMC to enable(one-sided)/turn
      off obmc in under 8x8 chroma blocks.
      (2) When it is enabled, use the above neighbor in chroma 4x4 obmc
      Change-Id: I2373e6698fd56da7f99653d77be2b7284b3e74f2
    • Angie Chiang's avatar
      Turn on tx_type search when sb_type < BLOCK_8X8 · c2ebfbfc
      Angie Chiang authored
      Change-Id: I8d9686c36ebf7057c4432bb5a29ade78dc21d0a7
    • Jingning Han's avatar
      Fix enc/dec mismatch in global motion · 3ca0e672
      Jingning Han authored
      Resolve an enc/dec mismatch issue when global motion is turned on.
      Change-Id: Ifad02fab4412588527ec7e8b24ab9b8c6327e5ba
    • Jingning Han's avatar
      Singularity handling in Gaussian elimination · 041c67b9
      Jingning Han authored
      When the fwd Gaussian elimination process encounters diagonal
      element as zero value, the linear equation does not have unique
      solution. Return the linear solver state as unsolvable in such
      case. This resolves a floating point exception issue due to divided
      by zero in the loop restoration filter.
      Change-Id: I3c67525691a3003f9f470e8a0d5b4ee187cba963
    • Debargha Mukherjee's avatar
      Reduce prec of matrices/vectors for warp estimate · 8c41024b
      Debargha Mukherjee authored
      Reduces precision of matrices by 2 bits.
      No material change in performance.
      Change-Id: I549a27da1dcb381fb329c345ee280dbd86b45bac
  4. 16 Apr, 2017 2 commits
  5. 15 Apr, 2017 3 commits
  6. 14 Apr, 2017 3 commits