1. 20 Dec, 2017 8 commits
    • Timothy B. Terriberry's avatar
      Fix bustage caused by 8089315a with daala_tx. · 501acee3
      Timothy B. Terriberry authored
      The inverse transform API was changed to pass in an unpadded 32x32
      block of coefficients for transforms larger than 32x32, but the
      code path actually used for daala_tx was not modified to pad it out
      to the full size like the others were.
      
      Change-Id: Ibda5d20a9d839ba41f8a1a0308c414111219da92
      501acee3
    • Hui Su's avatar
      Palette: replace floating point numbers with integers · 5891f98f
      Hui Su authored
      Performance change is within noise range.
      
      Change-Id: I9ca2ea6dfdc629513836637283ae3f964a1dd497
      5891f98f
    • Cyril Concolato's avatar
      Moving tile_info upwards in the frame header · 3b5afc1e
      Cyril Concolato authored
      Change-Id: Idce7902ed4f0a283d8ae6643402a450f3e5caf9a
      3b5afc1e
    • Yue Chen's avatar
      Clean ups and bug-fixes for filter_intra · 994dba25
      Yue Chen authored
      Clean up: clean up unused entropy related arrays
      Bug-fixes: resolve compiling errors when it is enabled with
      entropy_stats and ext_skip
      
      Change-Id: I2893e1b5c09e847fb4677f6e62709464208b76f8
      994dba25
    • Debargha Mukherjee's avatar
      Correct/Cleanup min_tx_size computation · 3ebb0d00
      Debargha Mukherjee authored
      Change-Id: I17fd041e9e1cead6b2b6339f0efa55c1ee9938af
      3ebb0d00
    • Imdad Sardharwalla's avatar
      Fix inconsistent use of ref_mv_idx. · c23ad632
      Imdad Sardharwalla authored
      "+ 1" has been added in a few locations to match the indices for the
      context for the CDFs and the motion vectors.
      
      Suggested fix (of removing +1s entirely) can then be implemented
      without further affecting the overall behaviour of the encoder and
      decoder.
      
      BUG=aomedia:858
      
      Change-Id: I650a618d3fda1d27bd3550d25ad1ac52e4365061
      c23ad632
    • Steinar Midtskogen's avatar
      Enable CDEF for chroma for 4:2:2 and 4:4:0 · ab6c9c77
      Steinar Midtskogen authored
      This enables CDEF for chroma when the horizontal and vertical
      subsampling differs.  Since the direction search is only performed in
      luma, the following corrections to the direction index are done for
      chroma to correct the distortion resulting from the subsampling:
      
      4:2:2:   4:4:0:
      0 -> 7   0 -> 1
      1 -> 0   1 -> 2
      2 -> 2   2 -> 2
      3 -> 4   3 -> 2
      4 -> 5   4 -> 3
      5 -> 6   5 -> 4
      6 -> 6   6 -> 6
      7 -> 6   7 -> 0
      
      This improves the chroma PSNR BDR by about 2% for 4:2:2 content at
      cpu-used=4, low delay configuration.  4:2:0 and 4:4:4 content is
      unchanged.
      
      Change-Id: Iee92d3697bc5a6fc9b1f340a63243a334935b433
      ab6c9c77
    • Frederic Barbier's avatar
      Remove EXT_PARTITION_TYPES_AB experiment · 3f53da76
      Frederic Barbier authored
      This experiment has been abandonned for AV1.
      
      Change-Id: Ic706b3533ecf502ab316505ba0a9a8eeeb296612
      3f53da76
  2. 19 Dec, 2017 13 commits
    • Zoe Liu's avatar
      Exclude distance-weighted prediction for ext-skip · 5664419e
      Zoe Liu authored
      For ext-skip, the use of distance-weighted prediction has only shown
      a coding gain in the AMCY low delay scenario. It is hence not being
      considered and compound-average is used exlusively.
      
      Change-Id: I57439eda1246e179d6404bbec366c6c610af3a36
      5664419e
    • Urvang Joshi's avatar
      Tweak for recode_loop_test_global_motion() func. · 02aade85
      Urvang Joshi authored
      There were two assignments to 'recode' earlier, out of which the 2nd one
      was a no-op.
      
      Change-Id: Ia181c736177ad9886c2d0dc4a2002d5bc2fb7742
      02aade85
    • Lei's avatar
      add new experiment loopfiltering_across_tiles_ext · 7bb501d5
      Lei authored
      based on the latest discussion in the HW working group about how loop
      filter should be integrated with tiles, the following decisions have been
      made:
      1. two seperated flages should be added for
      loop_filter_across_tiles_enabled for horizontal tile boundary and
      vertical tile boundary.
      2. encoder and decoder should only check these two flags to determine
      whether loop filtering (including deblocking, CDEF and loop restoration)
      should cross tile boundaries (vertical and/or horizontal) or not
      regardless the horitontal depepdent tile flag.
      
      This change list implemented the support for two seperated
      loop_filter_across_tiles_enabled flags for vertical and horizontal tile
      boundaries. The new experiment is disabled as default before it is
      adopted.
      
      Change-Id: I814377947517f5419c08b004a3b71b950d01eadd
      7bb501d5
    • Yue Chen's avatar
      Remove filter_intra mode info for UV plane · b0571874
      Yue Chen authored
      Change-Id: I92f73b78dfd0e25a61897e8b6e9e28ff210376d5
      b0571874
    • Dake He's avatar
      [lv_map_multi] BR coding · e2d8f183
      Dake He authored
      1. Use separate contexts for transforms of size 32x32 and above.
      2. Initialize cdf directly
      
      Change-Id: I3373143788e81b2547d99e420e591892ef2641a2
      e2d8f183
    • Jingning Han's avatar
      Conditionally drop cutting tails in lv-map coeff optimization · 8be58fad
      Jingning Han authored
      When the optimized non-zero tail is longer that 2, the likelihood
      that forcing to drop all the tail coefficients would give better
      rate-distortion optimization is much lower. This commit makes the
      encoder to check such conditions and speed up the lv-map/multi
      encoding speed.
      
      Local test clips at high target bit-rate / quality show 20% overall
      encoding speed up. The coding performance change for lowres is
      0.014%.
      
      Change-Id: Ifcf1d18e453a83110ca8be794aee26f3e0237605
      8be58fad
    • Sebastien Alaiwan's avatar
      Remove ADAPT_SCAN experiment · 23c511e7
      Sebastien Alaiwan authored
      This experiment has been abandonned for AV1.
      
      Change-Id: Ib3bb07d62f2025ce50dc9bc1b3f9fc22488519a7
      23c511e7
    • Steinar Midtskogen's avatar
      Make CDEF only do luma if monochrome · 35e53713
      Steinar Midtskogen authored
      Change-Id: I499fe2be873944e73334f681a0e8138cfccbea29
      35e53713
    • Rupert Swarbrick's avatar
      Match braces in rdopt.c · d16d8ee2
      Rupert Swarbrick authored
      This makes motion by function easier in some editors (such as Emacs or
      vi), which get a bit confused by
      
        #if 1
        foo {
        #else
        foo {
        #endif
          bar;
        }
      
      Change-Id: Ie9d8c34d3b9fafdac634db5f3647f3cc1611ebee
      d16d8ee2
    • Rupert Swarbrick's avatar
      Balance curly braces in encoder.c · a84faf23
      Rupert Swarbrick authored
      This makes code motion a bit easier in editors. To do so, remove some
      copy-paste code in scale_references and rewrite a duplicated loop in
      set_high_precision_mv.
      
      This patch should have no effect on behaviour.
      
      Change-Id: Ia9e72f2c89a7f7f7a1d3cb77822ff8db8b5faaf1
      a84faf23
    • Yue Chen's avatar
      Clean-up abandoned options for filter_intra · 250dd963
      Yue Chen authored
      Code for the following options is removed.
      (1) Disable filter_intra in sub8x8 tx blocks
      (2) Filter intra implementaion based on 1x1, 2x2, and 4x4 unit
      
      Change-Id: Ie898f6e8155709beff030b0ffb876754f14fb173
      250dd963
    • Luc Trudeau's avatar
      [CFL] Cache DC_PRED during CfL-RDO · 467205ac
      Luc Trudeau authored
      By default, the DC_PRED is not cached (this includes
      decoding). During cfl_rd_pick_alpha(), DC_PRED caching
      is enabled, the DC_PRED is cached after the first time it
      is computed (for each plane) and then it is reused when
      testing all the other scaling parameters.
      
      Change-Id: Ie8ba0bb0427c4d9be8de5b44e6330e8a78b9c7d9
      467205ac
    • Luc Trudeau's avatar
      [CleanUp] Wrap intra_mode_to_tx_type_context · 9cea993b
      Luc Trudeau authored
      To avoid using a UV_PREDICTION_MODE with intra_mode_to_tx_type_context,
      a wrapper function is added. Using a UV_PREDICTION_MODE with
      intra_mode_to_tx_type_context is problematic with UV_CFL_PRED which
      needs to be converted to DC_PRED.
      
      Change-Id: I68320e34dd5e616d0f15ca2c1c951440243968cf
      9cea993b
  3. 18 Dec, 2017 14 commits
    • Urvang Joshi's avatar
      Re-enable TX64X64 by default. · efeafa85
      Urvang Joshi authored
      Incompatibilities with other experiments has been fixed.
      
      BUG=aomedia:1058
      
      Change-Id: Id8358404c8f452bccf05ce3c1d91a633b215f07d
      efeafa85
    • Cheng Chen's avatar
      Bug fix · a70cd7b5
      Cheng Chen authored
      Remove BLOCK_2X2, BLOCK_2X4, BLOCK_4X2, which has been removed
      from codebase.
      
      Change-Id: I1c4ac7d148b7e95b1b104df7b9664cbb31928ddc
      a70cd7b5
    • Cheng Chen's avatar
      LPF_SB: make it work with seg and ext_delta_q · 5dcef7d2
      Cheng Chen authored
      Clean up unused code. Now lpf_sb works as baseline loop filter. It
      is compatible with seg feature and ext_delta_q.
      
      Change-Id: If5deb10d9f35cc3037463e7e8420bcd6b2b0255e
      5dcef7d2
    • Cheng Chen's avatar
      JNT_COMP: add SIMD and interface for high bit-depth · bf3d4964
      Cheng Chen authored
      Add high bit-depth macro definitions:
      highbd_jnt_sad
      highbd_8(10/12)_jnt_sub_pixel_avg.
      
      Add SIMD functions:
      aom_highbd_jnt_comp_avg_pred_sse2
      aom_highbd_jnt_comp_avg_upsampled_pred_sse2
      
      This patch also solves the seg fault caused by low bit-depth and
      high bit-depth paths
      
      BUG=aomedia:967
      BUG=aomedia:944
      
      Change-Id: Iea69f114e81ca226a30d84a540ad846f1b94b8d6
      bf3d4964
    • Luc Trudeau's avatar
      [CleanUp] Remove block from av1_mode_context_analyzer · 15a18e3d
      Luc Trudeau authored
      av1_mode_context_analyzer is always called with block == -1.
      
      Change-Id: Ibc9e7e87ea95f3651f7505b82674a6fd795ae985
      15a18e3d
    • Hui Su's avatar
      Add call to aom_clear_system_state() in pallete RDO · 80668233
      Hui Su authored
      BUG=aomedia:1144
      
      Change-Id: Ia8fb6df1a72bd1ff86c7955b681ec0d99c7a99eb
      80668233
    • Yushin Cho's avatar
      [dist-8x8] Disable asserts with CFL turned on · d2f281ba
      Yushin Cho authored
      The dist-8x8 has asserts which check whether the sum of distortions of
      sub8x8 partitions exactly match to the distortion directly measured on 8x8 pixels.
      Currently, these asserts fail if CFL is turned on and the reason is unknown.
      Until it is fixed, disable those asserts.
      
      Change-Id: I27ea52f8f6f651360cd1adc740d3964cf4bf9869
      d2f281ba
    • Yushin Cho's avatar
      [PVQ] Delete leftover code · c5c17093
      Yushin Cho authored
      Deletng the code since it is in fact for the PVQ, not for dist-8x8.
      
      The relavant code removal for PVQ is commit:
      d0b77ac7
      "Remove PVQ code" on 2017-10-20, which has removed the PVQ code.
      
      Change-Id: I6b58d0acedb6bf278f35ac6c76a7a0a481cd3edb
      c5c17093
    • Debargha Mukherjee's avatar
      Make lv-map work with 64x16/16x64 transforms · 695abd72
      Debargha Mukherjee authored
      Some refactoring done to enable reuse of code.
      
      Change-Id: I28cea413a1bf7e6b8ce82ee50c5b5603e8185a2a
      695abd72
    • Angie Chiang's avatar
      Restore the best_eob in av1_search_txk_type · 4e16ea64
      Angie Chiang authored
      Change-Id: I679eea672e183a34cdbfeef6eb4f3bd5b549b85d
      4e16ea64
    • Angie Chiang's avatar
      Fix assertion of txk_type check · 1640bfd4
      Angie Chiang authored
      Change-Id: I6db5563ec4f4992a899090cf15c3f5f5125a0764
      1640bfd4
    • Zoe Liu's avatar
      Expand skip mode support for all comp frames · 4b847e19
      Zoe Liu authored
      All compound predicted frames now have the choice to either turn on
      or off the use of the skip mode. The current encoder design only
      turns on the use of skip mode when the temporal distances of the two
      reference frames to the current frame are only different by 1.
      
      This patch also includes a fix on the calculating of the temporal
      distance between the two reference frames to the current frame for
      skip mode.
      
      Performance wise, this patch does not have noticeable impact as the
      encoder stays with the same choice with the frame-level skip mode
      flag.
      
      Change-Id: I34f370940b3b25d2ab429b8721344133ae6288ad
      4b847e19
    • Cheng Chen's avatar
      LPF_SB: cleanup unused functions and code path · 76224b0e
      Cheng Chen authored
      Since we've decided to use guess filter levels for lpf_sb, I remove
      unused searching functions and coding paths to make codebase clean
      and easy to manage.
      
      Change-Id: I3497add59fc1eaf36acdd0acb85e3bcb6486d01d
      76224b0e
    • Cheng Chen's avatar
      Speed up by dropping some ref frames in compound search · c683bf9b
      Cheng Chen authored
      Record distortion for each single ref in rd. Rank according to their
      distortions. Then in compound search, drop the combination of ref
      frames of the largest and second largest distortions
      
      This patch shows neutral performance on google test using lowres
      with 20 frame.
      
      Local tests show ~5% speed up over baseline.
      
      Change-Id: I722fe66a0551f5f8a044c57c55caa74e46db7ee8
      c683bf9b
  4. 17 Dec, 2017 1 commit
  5. 16 Dec, 2017 4 commits