GitLab

Change-Id: I63fc3f1f010e77c6dc033f37e3e91ade17a55099

Adaptively reduce the best rate-distortion cost value in the
recursive transform block partition search. For bus CIF at 1000 kbps
this reduces the encoding time from 1864 seconds to 1756 seconds,
about 6% speed up.

Change-Id: I5433a1825c0f8b13fcc5ab7e19713a98969d53fc

Change-Id: Ie2c34490dc50cb242bcd701308e6b55243883b15

1) Integrate it with supertx
2) Integrate it with chroma_sub8x8

Change-Id: If4bb906d442d15bae3741192029ec851c48d3948

A separate prediction mode struct is added to allow
for uv-only modes (like CfL). Note: CfL will be
added as a separate mode in an upcoming commit.

Results on Subset1 (Compared to 4266a7ed with CfL enabled)
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0000 |  0.0000 |  0.0000 |   0.0000 | 0.0000 |  0.0000 |     0.0000

Change-Id: Ie80711c641c97f745daac899eadce6201ed97fcc

MRC_DCT uses a mask based on the prediction signal to modify the
residual before applying DCT_DCT. This adds all necessary functions
to perform this transform and makes the prediction signal available
to the 32x32 txfm functions so the mask can be created. I am still
experimenting with different types of mask generation functions and
so this patch contains a placeholder. This patch has no impact on
performance.

Change-Id: Ie3772f528e82103187a85c91cf00bb291dba328a

When convolve_round is turned on, both lbd/hbd use use 32-bit buf
Therefore, they use the same mask/blending functions

Change-Id: Icfc6db818c0a53216108e42161acac07303e6c1c

Change-Id: I94d66c65d78235e1025703caf79ccca43208d604

Around 0.9% improvement on screen_content set (encoding 30 frames).

Change-Id: Ic4c9333c9af5993bc41e513b9e766450b3a951eb

Change-Id: I416a7f99e292a6304bc24d93ab580650768d5e21

The soft coefficient optimization process would monotonically
increase the transform block distortion and decrease the
coefficient rate cost. Such observation provides a lower bound
on the rate-distortion cost for the given transform block. This
commit compares this lower bound against the best available
rate-distortion cost value and skips unnecessary optimization
process. It speeds up the baseline encoding process by 15%.

Change-Id: Ida8098a2820cef60d59ec1e72f0bbb1acbd98165

Use three metrics to identify the still gf group.
Performance:
lowres: pamphlet_cif -1.395; bowing_cif -0.989;
        others remain same. Overall -0.064
midres: snow_mnt_480p -0.827. others remain same.
        Overall -0.028

Change-Id: I22a6429c7ebdad2c36ec73c7a69cabc07e8208b7

The existing code has a case where we set a variable to equal
xd->ref_mv_stack[mbmi->ref_frame[0]][1 + mbmi->ref_mv_idx]
even for compound blocks. However, the range of allowable
values for mbmi->ref_mv_idx is determined by the ref_mv_count
for the *combined* ref frame, not for the first single ref frame.

This means that, if we have more ref-mv candidates for the combined
ref frame than for the first single ref frame, then we can sometimes
fetch uninitialized data.
In every case where this happens, we immediately overwrite
the destination with the correct mv, but it is still preferable
to avoid reading uninitialized data.

This patch moves the code block to avoid this bug. In addition,
the variable (nearmv[0]) is only used when the mode equals NEARMV,
so the condition on its assignment is changed to reflect that.

Change-Id: I3bd268dc80d8065d5189999232b8a0f826d40a95

This experiment replaces the 8-point Type-II DCT and 8-point Type-IV DST
 scaling vp9 transforms with the 8-point orthonormal Daala transforms.
These have reduced complexity and are perfect reconstruction at the cost
 of a slightly worse coding performance.
This is because the Daala transforms expect the input to be shifted by 4
 bits but the output scale of the vp9 transforms is only 3 bits.

subset-1:

monty-square-baseline-subset1 ->
  monty-square-dct8-subset1@2017-07-17T21:37:44.281Z

  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0019 | -0.0011 | -0.0585 |  -0.0111 | 0.0305 |  0.0317 |     0.0187

objective-1-fast:

monty-square-baseline-o1f ->
  monty-square-dct8-o1f@2017-07-17T21:37:15.735Z

  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0285 |  0.0129 | -0.5080 |   0.0529 | 0.0345 |  0.0441 |     0.0054

Change-Id: I2b775495398fb717204a295397c3c5e3ca938183

Because #if ... #else ... put the '{' on the same line, dangling { or } occurs,
which causes automatic syntax analyzer, such as 'Ctrl-Shifht-P' in Eclipse
or '%' of vi, fail to find matching { and }.

For some developers, this can make quick reading and/or understaning blocks of code
almost impossible.

Three function or blocks are repaird.
1. av1_rd_pick_inter_mode_sb() {...}

2. for (midx = 0; midx < MAX_MODES; ++midx) {...}
   in av1_rd_pick_inter_mode_sb()

3. handle_inter_mode() {...}

Change-Id: Ib5ac63b8c7f9870a491fac337ae3f58c57ce5e46

Make the codec account for the 64x64 processing unit constraint
when producing the mask for overlapped filter.

Change-Id: I3e596492ae522abe678369b0c9710441549e817e

For 128x128 level blocks, process the overlapped prediction in
the unit of 64x64. This allows hardware design to reuse the 64x64
processing unit in 128x128 level block coding.

Change-Id: I3967b8e3c1c697f96a50e59a0957fc69b67e6f8e

Change-Id: Id556e8d77c5871ddae338baa1abfb93b7aa207e9

This change does not alter the bitstream

Resuls on Subset1 (compared to 70a80a81 with cfl)
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0000 |  0.0000 |  0.0000 |   0.0000 | 0.0000 |  0.0000 |     0.0000

Change-Id: I7672eb4cde3c649ebba32610f7e56500e378c062

Change-Id: Ie1f5db74d3b22a46abd09e0207770935622074a5

The performance stay the same compared to turn on adpat_scan for
all txfms

The  performance of adapt_scan on top of ext_tx are
lowres: 0.517%
midres: 0.484%
hdres:  0.639%

Change-Id: Ic148dbe9c0af4554bbbe3f7012625f389f997289

Change-Id: Ibd0bda593717aab2894c36dd59925a0bcf654113

- Use 'tx_size' in function signatures.
- filter_intra_taps_3 and filter_intra_taps_4 updated to support
  TX_SIZES_ALL (thanks to yuec@)

With these changes, filter-intra works correctly with rect-intra-pred.
So, we remove the temporary workaround for this.

Change-Id: Ide0f593419c21a74c08c61859f8dad918ca169fa

Change-Id: I894ab72f529a598ae4534e0bd982e08863d75fcb

This workaround is temporary, until filter-intra can work with rectangular
blocks.

Tested OK:
make clean; ../../configure --disable-install-docs --enable-unit-tests
--enable-debug --enable-aom-highbitdepth --enable-experimental
--enable-adapt-scan --enable-dual-filter --enable-ext-inter
--enable-ext-intra --enable-ext-refs --enable-ext-tx
--enable-filter-intra --enable-loop-restoration --enable-rect-tx
--enable-compound-segment --enable-interintra --enable-wedge
make -j
./test_libaom

Change-Id: I4554d1f25de9448b22465e93a7616df0c206e298

Change-Id: I00f98e0e03a44aa894566b7a47bb0f32b56da766

Worst case SSE for a 12-bit 64x64 block requires 48 bits
(2*(12+log(64)+log(64))). As such, the dist variable must
be int64.

Results on Subset1 (compared to 19b5c8fa with CfL enabled)

  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0030 |  0.0001 |  0.0100 |   0.0026 | 0.0024 | -0.0008 |     0.0028

Change-Id: I1364c089c223b96daed942175a915fed0f6f1023

- Modified temporal filter to take extra buffer parameter
- Added bilinear interpolation, can be turned on by setting
BGSPRITE_INTERPOLATION define to 1.
- Added mean filter to bgsprite for blending modes
- Bilinear interpolation is turned OFF by default
- Mean filtering turned ON by default (was previously median filtering)
- Allow for multiple references for alt_ref in encodeframe

Change-Id: Icfe9054049339d353b4befff11e96c12ca73ed4e

For some reason xd->cur_buf->bit_depth appears to always be 0.
Use xd->bd instead, as this appears to have the correct value.

Along with commit 49bb8f8c, this fixes the HBD internal quality
loss introduced in commit 5d24b6f0.

BUG=aomedia:649

Change-Id: Ie97e9caf3b13f7493f873258f3863e1314466f1d

Change-Id: Icbb373913b9f7e86aa0d1a8d28df796530a1e001

Change-Id: I41fe97db4cee333e601356253f28186ef4036237

Change-Id: Iadfcc19db19295c493dc7c26718e68720cb782cd

Change-Id: Ib03758eb5ab32ee7d0f87d216bc5fad591578597

This integration only covers low bitdepth mode for now

The performance of Convolve_round on top of compound_segment
revives from 0.475% to 0.612% on lowres

Change-Id: I21606c79d0a22c0834966730358267c082d8071e

Tile groups are now an integral part of the codec.

Change-Id: I620a88ec7a44b057d5cce0bf6cf602822a3339a9

This experiment was provisionally adopted on 2017-06-27.

Change-Id: I5ebce1df7cec42804df553a26848ddfe8a449a59

Fix width/height for U, V planes.
Previously, width/height will be right shifted multiple times
in the loop.

Change-Id: I3581cb948c1918e39c30eb5c99ac958df5d3ed31

For chroma DC_PRED, cfl_predict_block and av1_predict_intra_block are
redundant,

Results on Subset1 (compared to  e1c27392 with CfL enabled)

  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0000 |  0.0000 |  0.0000 |   0.0000 | 0.0000 |  0.0000 |     0.0000

Change-Id: Id50f3682eaccb5370e29d66af8a9967a30d135e6

Change-Id: I1a43ed3c13ce1f8a0a6673f8d54fc82abcfeca5b

Change-Id: I8987dbc05dc79d33eeadebb427095fafa319e9fd