GitLab

- Add 16-bit saturation in fdct_round_shift().
- Add extreme value tests and round trip error tests.
- Fix inv 4x8 txfm calculation accuracy.
- Fix 4x8, 8x4, 8x16, 16x8, 16x32, 32x16 extreme value tests.
- BDRate: lowres: -0.034
          midres: -0.036
          hdres:  -0.013
BUG=webm:1340

Change-Id: I48365c1e50a03a7b1aa69b8856b732b483299fb5

Enable the ref-mv experiment by default.

Change-Id: Icb4f37aab0dc866430569f0ee82805de1498bee3

Merge the modifications in simp_pred_mv to ref_mv.

Change-Id: I5662124c1ac7133d76697522f06e42b277a4ad8d

This removes the dependency of context decoding from the candidate
list ranking in hardware implementation.

Change-Id: Iff34e7e6e78c9b32f435ded30629e43e73ac2f13

Replace the int_mv storage in the ref_mv_stack struct with unit8_t
integer to indicate the prediction difference level.

Change-Id: I0bfbcab350a46ecdd5bc47bd918cce9614365227

Includes:
Various table updates and fixes to support 64x64 transforms.
Entropymode updates to support tx_size expansion to 64x64.
tx_mode changes to support an ALLOW_64sXx64 transform mode.

Change-Id: Ib9098cfe27d0c015fe3be6ae13e7d09576771b9e

Bring rd_stats back to proper starting point for the second mode.
Bug introduced during code refactoring preparing for RD_DEBUG:
https://aomedia-review.googlesource.com/#/c/5689/

Change-Id: I9ad10144ae3cc395a6efa700d047741c7734470c

Introduced in I2c0bc4ba540886dd196e87a78b205407ad3e866b

Change-Id: I9ff49e7d37c5a1fea8dc8b5659a53697f76320df

Change-Id: Ibc9e904ba9e82d6ab638d88404847da7fc9fcce9

Fix a case where handle_inter_mode could return without restoring
the original values of xd->plane[i].dst.buf. This fixes an
assertion failure in aom_blend_a64_mask, as well as very slightly
improving the rd costing accuracy.

Change-Id: I7f720bcf676c6aa9858c1a4c8f6571a76a9b0772

This commit fixes several issues that prevent experiments to build
when enabled together. The experiments are:
ec-adapt
ext-inter
ext-partition

BUG=aomedia:80
BUG=aomedia:88

Change-Id: I5b790f16ad0ae2450f0be3aa26329b6a57a6a5c5

BUG=aomedia:80
BUG=aomedia:88

Change-Id: Ia8cd086ea1568235baf7c2fba2edd5a0e41af584

Adds 64x64 default scan orders
Adds 64x64 coefficient entropy models (just copy the 32x32 ones)
Entropy context updates for 64x64 transform
Various misc. changes to support 64x64 transforms

Change-Id: I2c0bc4ba540886dd196e87a78b205407ad3e866b

Only process reference motion vector stack for the used reference
frames at the decoder. This improves the decoding speed by 15%.

Change-Id: I4f7c11d4607f9bcb756b6c375723d017ec6ece22

This commit re-synchronize the reference frame type use cases in
the dynamic motion vector referencing system.

Change-Id: Ib25231c716db4176cd67cbdc889472a06b607194

BUG=aomedia:88

Change-Id: I216ae046bed89f1dc919fc9ebb5c191d491ba621

BUG=aomedia:88

Change-Id: Ie45052d8c91e96d1fda048bc4240cec396fc33d8

This is a manual cherry-pick from libvpx/master:
e61d82bd4fab408b32011568a52524b6a82bcf25

Change-Id: I29bbcea604507d016d4783c7ed1697712814478d

BUG=aomedia:80

Change-Id: I8b0c74c44d454630d8f32a6b0124f5684d0b1cdf

This commit revert the previous fix of the ubsan warning for unsigned
int overflow, and use a better fix by  moving the offs-- inside the
while loop to avoid "0-1" situation.

Change-Id: Id4a3e03859ebcdf264df0808412b30841028f87c

Previously, any uses of global motion inside supertx blocks were not
counted correctly. This caused encode/decode mismatches when every use
of global motion in a frame occurred inside supertx blocks.
This happened in, for example, AV1/ActiveMapTest.Test/0

This patch makes update_state_supertx count global motion usages in
the same way update_state does, and fixes the above problem.

Change-Id: Id500d5a24c565774fa3aa3b52cd3fdbeab75b486

To facilitate further experiments.
Fixes border, uses standard interpolation filters for
translation only global model, and some refactoring.

Change-Id: I38af6cb83723f92baed23e28d7fc9a0a9544757e

If a coding block is in the top row of a 64x64 (or 128x128) block,
its top-right corner should be available. No major coding performance
change is observed.

Change-Id: I0ee23db313cd53b34b7bc28230705f1cb1f8fe89

Change-Id: Iad227b07b6c3f5beebef971cabec787b01e6f51f

In https://aomedia-review.googlesource.com/#/c/5864/ , some
code to stop the decoder at a preselected point was left enabled.
This code should only be uncommented when debugging, so comment
it out by default.

Change-Id: Ie168e8a1588ba92971e3ff1a056f597a7dfca136

Also replaced "ToDo" with "TODO" in comments.

BUG=aomedia:86

Change-Id: I5cb044ce5bfdba5bde17d5546b57619ced8dbf77

Change-Id: I7b0a00627a9807c42345d3d28c3499c208c5e792

Uses a |difference| ^ 0.6 metric instead of clamped abs.
Improves lowres to -0.552%

Change-Id: Ic5159c4e20e5c4de8e7b1501bd68e014c9d1cf4b

To avoid unsigned int overflow

Change-Id: Ie070fea50b4b6c4a0e80c61371a2366dd79ed570

These transforms need to be updated to match c versions after they
are finalized for AV1.

Change-Id: Iddc3369b58ac3ce8d9c0c08724390c477ae89065

Adds a feature to recode if global motion is used for a reference but
has very few blocks in the frame actually using it.

lowres improves to -0.512% on average.

Change-Id: I61a36770e1b7103b9a27706909443c3f14ee4e42

Change-Id: I809eb52e8a632189c49b8ea0a2b5de760cc2a34c

av1_txfm.h: left shift of a negative number
av1/encoder/quantize.c: unsigned int overflow
aom_dsp/entenc.c: unsigned int overflow

Change-Id: I6143e68f7d6e2621f97900808c8ef7ee0ad0c814

av1/common/x86/av1_fwd_txfm1d_sse4.c
av1/encoder/rdopt.c

Change-Id: Ia386f78da64029b5f68b68a41485cc068dbf4098

Change-Id: I48c252cf1c8ad9fb6e94e3cadef6061688ab7ad4

The global motion detection only works on 8-bit buffers,
so any frames using 16-bit buffers are now down-converted
to 8 bits when necessary.

Change-Id: I4f88f4ccd449e73a2292cda70fe573dc49fcb8a0

There is a rare edge case in ransac(), causing the tests
AV1/DatarateTestLarge.ChangingDropFrameThresh/{4,5,6}
to fail when global motion is enabled. The sequence of events is:

* Since GLOBAL_MOTION_TYPES = 3, we try to generate a ROTZOOM model
* A model is generated at ransac.c:215, and its first 4 components
  are copied to best_params. The last four components of best_params
  are left as {0, 1, 0, 0}.
* We then finish the trial_count loop, and call find_transformation
  one final time to generate a refined model. But this fails, and
  best_params is not overwritten.
* get_wmtype decides that this is an AFFINE model, since
  wmmat[4] != -wmmat[3] and wmmat[5] != wmmat[2].
* We try to encode this global motion model. But the entropy coder
  cannot encode an AFFINE model, and crashes at daalaboolwriter.h:74
  (with OD_ASSERT(symb != -1))

The fix is to copy 8 entries to best_params regardless of the
model type we are generating, in line with the changes in
https://aomedia-review.googlesource.com/#/c/5589/

Change-Id: I6dbdfb997924f8ddf7ea3a1d557463264ea63cbe

Change-Id: I8f48db56e3f55b1d2fcc8a06aaa51e5c383e651f

In several places bits are overwritten in the bitstream. These
functions avoid zeroing bytes during writing so that this can
happen correctly when the number of bits is not 8*N.

Re-addresses the attempted fix in
133c57c3

which broke threaded encoding tests, which relied on re-using
byte buffers.

Change-Id: I682c5e3a7869eac7ad475584db8bf170d47a56c9

BUG=aomedia:79

Merge of nextgenv2 caused av1_optimize_b not to take account
of different inverse quantization when quant matrices are on.

Change-Id: I1b8da2e110ce201183be777663222e3d73c9f17b