GitLab

lowres: -0.401% BDRATE (tempete -5.4%)

Also includes some fixes and cleanups.

Change-Id: I82922a453fad59fad2f12829d2ef1e56b2d20949

Change-Id: I4d145daa33ed2f4b7f01e81ae2f0c17551982d80

And some minor format tweak

Change-Id: If0d33bf22f161827b43a89649679a4e95feb325f

This commit supports the recursive transform block partition to
start with both square and rectangular transform block sizes.

Change-Id: Idaf29d50cb1f4876b43e5ba82c2609708c6b1926

Default MTU size is 0, which implies a fixed number
of tile groups. MTU matching overrides fixed tile
group. MTU matching will succeed unless a single
tile is bigger than the MTU.

--mtu-size is in bytes, and includes headers but
not transport/wrapper overheads

Change-Id: I2b70bd41b175b54273b02d836f2a84011f617a7c

In this combination, the parameter 'supertx_enabled' was never
passed to read_inter_block_mode_info, so the address of the
function of the same name was unintentionally used instead.
This always evaluates to true, which caused lots of test
failures (eg, with test filter *EndToEnd*)

Change-Id: If7a31f517aeb170a3c7cf14ed65e1d8511781718

Change-Id: If7a117b78473d5b1f02b0219ddc314c7be7fbbbf

The switch to using daala_ec by default broke the compile
when used with ext-tx, as some code which should have been
inside #if !CONFIG_EXT_TX was not inside such a block.

Change-Id: I4cec9755bd4e6134a7d3036c4482f952a815cafc

In both cases, the data to be signalled (for clpf/dering) is
stored at the topmost superblock level, but previously this data
would not be signalled for 128x128 superblocks.

For dering, it was also assumed that all superblocks were of the
maximum possible size. When ext_partition was enabled and we
chose to use a 64x64 superblock, this resulted in an out-of-bounds
access to cm->mi_grid_visible, leading to a segfault.

Change-Id: I81b9c037ed8db8be16d7e4cf6041e1df7525cc97

Since non PVQ path in av1_encode_block_intra() was refactored
in nextgenv2 so that both inter and intra block txfm_quant,
i.e. av1_encode_block() and av1_encode_block_intra(), are done
similar way, such as both call av1_xform_quant().
This is also nice for PVQ, since it can reuse pvq enc helper call
in av1_xform_quant() for intra block mode.

Change-Id: I45aaf5b899552bf947d777cdb2d92e665d4f217e

This commit allows the rate-distortion optimization process to
skip repeated computation of the 32x32 transform block
rate-distortion cost in the recursive transform partition scheme.

Change-Id: Ifc2ac03d99ab221deae057e17ada7a81bef6dfcb

Reset the transform block size context to the actual block size
if the skip flag is turned on.

Change-Id: I821afac2dcd4c3a17389e9954b7b2575f87d07d7

For these files:
av1/common/common_data.h
av1/encoder/encodemb.c
av1/encoder/quantize.c
av1/encoder/rdopt.c

Change-Id: I379e861d55c2875d0366df24f986b29796a70087

A helper function av1_update_txb_coeff_cost() is created to do so.

Change-Id: I67be97d8dce36bafde11d2eb27c862945632f0cb

Change-Id: If6479b165e32caead29f7cedcfa64e637be2fb98

BUG=aomedia:79

Change-Id: I09c9547295e4ab540861d6d4def24718dcefc854

The definition of TX_SIZES has changed after nextgenv2 is merged,
which included non-square transforms. TX_SIZES is used as part of
condition doing recode in tx_size/type RDO when PVQ is enabled.

Change-Id: I9a66a72f0f4391eb375290aa0dabc638c68c6c68

Also includes some refactoring and cleanups.

Change-Id: I2c2528c434a1e9e9b898251fa69489d884463929

Update source file lists to stay in sync with the rest of aom.

BUG=https://bugs.chromium.org/p/aomedia/issues/detail?id=76

Change-Id: Ief04b0d96949719bc6ff9ff10686f5f34c4d742c

BUG=aomedia:80

Change-Id: Ie4bccf053d2c24dcb64519650bcbcef4baffcdae

Change-Id: I749c177001f073fea6182a4be3f54733fed52567

Change-Id: I88df7c226f9d14e5c3546e0582bea2c38d0efbd6

Remove the CONFIG_EXT_TX flag.

Change-Id: I7e35036156e116a2c00c982a8d57473258e002ea

Change-Id: Iaaf9244678cde74cf3e216f3b8c9d293f206f27e

Change-Id: I403596a9d96b13276b318c5c1c7921fdd741a163

Change-Id: I112224cbe02649793b06fe38f8b257ed53f050de

Overwriting was only guaranteed to work if a whole number
of bytes were being overwritten.

Change-Id: I5e72cb337ec6ff691e93288de9f751b583654a17

Compile error was caused by one of previous commits, 7c2b7f25.

Change-Id: I51885cb03190ab736e4bcbae13bce40857da83b6

Use the unified sub_tx_size table to fetch the transform block
size for next recursion.

Change-Id: I73af3192b529cc9f3fc45873a037542112046cb3

Make the UV transform block table support recursive transform block
partitioning. This doesn't change the coding statistics.

Change-Id: I90bcf7c52c597a77afe130f1f1ed5024b6888737

Change-Id: I58232696f29e2570aa5b5e4b020b4dca59be1f5c

Fix the merge issue due to pvq. This fixes significant key frame
performance loss.

Change-Id: I5fef4648cfc9e94f63b553ac15e051a6c4d4ca13

Unify the use case of largest transform block size fetch for skip
blocks.

Change-Id: I5dd60eb74b266fa7cd7944d8f823dbdae75e90af

The rans experiment is dead. The ans experiment with the ec_multisymbol
experiment also turned on takes its place.

Change-Id: Ie9f30ec7cf73aae6b2ea580a7b1f208485a8a7a7

Remove redundant #if statements there.

Change-Id: If6e6000c76899b0c3072e741a22bc3efa7d33739

Replace repeated is_inter_block() check with a constant is_inter
variable.

Change-Id: I1215f72d5fcbf167cf5252333e7d9fbe5c9ff808

This CL doesn't change any coding behavior.
The functions modified in this CL include
choose_smallest_tx_size()
choose_largest_tx_size()
choose_tx_size_type_from_rd()
txfm_rd_in_plane()
txfm_yrd()

This change make the code cleaner and will facilitate rd_debug
implementation

Change-Id: I19b1b2384dce8b46b28d4c9baf80ff92998d05b6

This is to facilitate debugging process in var_tx experiment

Change-Id: Ibd5ea7f6054c598b8e686abb4e8158ef28c67aab

BUG=https://bugs.chromium.org/p/aomedia/issues/detail?id=76

Change-Id: I9778eaa1dd67d36bae7953cbd8cbcd893f0aefac

Change-Id: Ic9679c0551f809807bd5c1196c302b072e7fa613