GitLab

This makes sure that all eps values are at least 4 since otherwise
the computation will not fit within 32 bits.

BUG=aomedia:893

Change-Id: I4815a865be8db792d0481172a2dfa0bc0a817f73

Change-Id: Ic4e04ef959e10b60dae6de8d6da50e7a281e654c

enable new_multisymbol by default since it has been enabled.

Change-Id: I6df662c6ae949ac9f834f788ff93b12a1a1eff28

Change-Id: Ia4b25afaed5dbd48c8e68fe42a479d7ddb645876

The patch is to enable loop filtering for intra block copy.

Apply loop filtering right after each superblock is encoded.
Use a constant filter level for now. In future updates, the filter
level should be optimized by LPF_SB and better guess value.

Change-Id: I0a5797b5ba5384c00044a8ec4a495dc75aa0f3fe

Add av1_allow_palette() to control whether palette mode should be enabled.

Change-Id: Iee24636451be42eb36093dc3453bc39c7e686276

Change-Id: I7a94cdef41e5e451247de939313feb58cd991e7f

Reduce the inter mode context model dependency to be within the
nearest neighbor search.

Change-Id: I0b9e7d360919a809b5400f233040bb474ab56768

This breaks the inter mode context dependency on the full spatial
neighbor search.

Change-Id: Ic27fb63531ce2c949756ea10752a7a8e7b7b1046

Make the newmv context model dependent only on the nearest spatial
neighbors reference motion vectors count and the newmv coded block
counts.

Change-Id: I8db9ca453c4fefddafba5f3476ee0d733f87dfc0

Function  SSE2 vs C  AVX2 vs C
4x4       ~4.5x
4x8       ~4.5x
8x4       ~11.7x
8x8       ~12.7x
8x16      ~14.0x
16x8                 ~21.7x
16x16                ~24.0x
16x32                ~28.7x
32x16                ~20.5x
32x32                ~24.4x

Change-Id: Iaca49727d8df17b7f793b774a8d51a401ef8a8d1

This was getting the wrong count of restoration tiles for chroma
planes with subsampling and a smaller restoration tilesize.

BUG=aomedia:886

Change-Id: I5c9c17ed4ad91111bcc6fa6205a9550b53f84a64

This is to reduce encoder meomory footprint, tests show no performance
difference between 17 and 25, as the largest alt-ref group currently
is 16.

Change-Id: Ia629c30f93c5db33a2d2fae4b757c8b0d589eb4a

Plain resize is now unaffected by this flag. That is, plain resizing is
performed for both width and height always.

Change-Id: I652204d68ca5266f0a1a413927828038aa5dfdf6

This allows it to be enabled/disabled during configure/cmake and also
allows us to specify dependencies.

If 0: superres scaling happens for both width and height.
If 1: superres scaling happens ONLY for width.

Change-Id: Id00c2a9c9886b638e06c6b92de453b85bee1d8fa

They're expected in the root of the config dir and not
in the examples sub dir.

Change-Id: I26e28e5a341f5bf8db4554269db198501172345e

In this experiment, sharp image discontinuity in the predicted
block is detected. Based on this discontinuity, we choose
particular LGTs as row and column transforms.

Bitstream syntax, entropy coding, and RD search for LGT are added.
One binary symbol is used to signal whether LGT is used. This
experiment can work independently with the lgt experiment.

lowres: -0.414% for key frames, -0.151% overall
midres: -0.413% for key frames, -0.161% overall

Change-Id: Iaa2f2c2839c34ca4134fa55e77870dc3f1fa879f

Change-Id: Idb1a4bf4dd655bde22862d76f6fa70457381a770

This is flag will allow us to calculate the context indexes of
any two consecutive non-zero binaries in parallel

Moreover, we can set MIN_SCAN_IDX_REDUCE_CONTEXT_DEPENDENCY to X,
which let first X coefficients be immune from the context
dependency reduction act

Change-Id: I75b71452996161ba06ec449021c7dea8e3899800

This aims at facilitate the experiment about reduce context
dependency

Change-Id: I3d026bda1118cf613001efa32deed62997d5e3bb

Change-Id: I7a73dbe72b618e795191cc31bc32e31ad99d8587

When early skipped in var-tx, distortion is set the same as sse.
If so, use pixel domain sse (i.e. skip error) since is more accureate
than sse from transform domain.

Change-Id: Id3cbc66ea6318108c031413646f3d06250e75e7e

The "txb_split_count" counter should be properly updated.

BUG=aomedia:864

Change-Id: I3fb34a818c3f474085c4a2980a2d3b68bd33fb12

Change-Id: I6d68f03e3f9b1e40b05503f6bb4055e2fd870893

Make the codec account for the 64x64 processing unit constraint
when generating secondary predictions and applying overlapped
filter.

This issue was addressed in commit 440d4254 and 501294ce, but
afterwards some features are not fully retained in an obmc
refactoring commit.

Change-Id: I6f16e6fccb966d45034d5b55447c9d9cb70e02cb

Function speedup on i7-6700:
D117   sse2   ssse3
4x4    ~1.8x
8x8           ~3.4x
16x16         ~5.5x
32x32         ~2.9x

D135   sse2   ssse3
4x4    ~1.9
8x8           ~3.3x
16x16         ~5.3x
32x32         ~3.6x

D153   sse2   ssse3
4x4    ~1.9x
8x8           ~2.8x
16x16         ~5.5x
32x32         ~3.6x

Change-Id: I43ab5fa8dcbcfa51acbde554abf3e5d7d336f391

Most of the fixes are related to replacing BLOCK_64X64 with
cm->sb_size.

Fixes the AV1/AqSegmentTest.TestNoMisMatchExtDeltaQ/* tests that
were breaking before with ex-partition.

Change-Id: I19d6045b422a93891b8cf4f8a929def97a595058

If you have a structure, foo_t, with an alignment request then Visual
Studio won't allow you to declare a function

  void use_foo(foo_t x);

The reasoning is that x might be passed on the stack, and their ABI
doesn't allow them to guarantee that x is aligned appropriately. More
strangely, this isn't allowed either:

 void use_some_foos(foo_t x[10]);

This is functionally equivalent to:

 void use_windows_foos(foo_t *x);

(except that you can't tell how long the array should be from the
function signature).

Since Visual Studio is supposed to allow the latter form, use that
instead.

Change-Id: Icd449fc1058606fa7e48a6f791091bbb42a73b2c

This was triggered by a visual studio compile warning:

  cdef_test.cc(128):
  warning C4804: '>>': unsafe use of type 'bool' in operation

However the code is rather hard to parse for humans too: when I first
looked, I thought this was something to do with C++ templating...

The new version is equivalent but defines max_pos in an outer
loop (and a smaller indent).

Change-Id: I0c5cabeee44d0839a7956a4ab1cf4ec5abfcc9ee

The rd_stats->sse is already updated by
"rd_stats->sse += tmp << 4;",
which is measured by pixel_diff_dist(), i.e. in pixel domain and
w/o quantization().

Change-Id: I4dc20a7e80af9dd846aa5de4298cb56e7f0d8f7e

Change-Id: Ie4382b8a1c0f87ce50e9afefd1cef8ca55435c61

Change-Id: Ibea4c2c732b16851ad16b475ea40f021d5b5d5b3

When a neighboring block uses global motion, use the mv
computed at the center of the current block as the candidate vector
rather than the mv computed at the center of the neighboring block.

0.15% improvement on cam_lowres

Change-Id: I79eff8bf27a7aa84ae4a6d56e4a10c41a4438fb9

Temporarily reverting this to allow investigation of a couple of BUGS

BUG=aomedia:881
BUG=aomedia:887 (merged into #881)

This reverts commit b1d3eda9.

Change-Id: I2605deb7b8fefa4236d78c8695025dc42316edd2

This patch fixes a bug in select_tx_type_yrd. The function works by
looping over possible transform types to find the best option (calling
select_tx_size_fix_type for each). Whenever there's a new best
candidate, the code copies information about the transform from the
mbmi structure into stack-allocated "best candidate" structures. At
the end, it copies the "best candidate" data back to mbmi.

Before the patch, if ref_best_rd was small, each call to
select_tx_size_fix_type might return INT64_MAX (because they don't
find anything better than ref_best_rd) and so we'd never actually copy
anything to the "best candidate" structures. Then, at the end of the
function, we'd merrily overwrite mbmi with whatever happened to be on
the stack, causing general mayhem when something tried to read the
data from mbmi later.

This patch exits early if no candidates were found. It also adds an
assertion saying that if no candidates were found, ref_best_rd must
have been less than INT64_MAX. This should hopefully catch any bugs
where the continue keywords in the loop stop us ever actually calling
select_tx_size_fix_type.

Change-Id: I54b998148281dd80f98d1570f736964593dc753f

For large blocks this is about 8x the speed of the C version. The code
needs SSE 4.1 for the PMULLD instruction that we use to do SIMD 32-bit
multiplies.

The patch uses av1_convolve_scale_test (written already to test the
low bit depth path) to make sure the optimised code matches the C
version.

Change-Id: I9304d6bb3d2cb31390de93ed08ff1a852e3ace86

For large blocks this is almost 8x the speed of the C version. The
code needs SSE 4.1 for the PMULLD instruction that we use to do SIMD
32-bit multiplies.

This patch also makes av1_convolve_scale_test actually test something,
making sure the optimised code matches the C version. The slightly
excessive generality in the test (all the templating) is because of a
following patch, which is for the high bit depth path and can then use
most of the same test code.

Change-Id: I6732bc6b2378ffaadae5aa6441100cf660f7ee11

Since 32x32 transform use DCT only, we can avoid update other
types of transform

Change-Id: I51dd8ec71975187d249d7e25130e994a48cac5c1

0.15% improvement on lowres set

Change-Id: If16a8e07797c64508f9e2d9b26ae874ac53c57a4

There's a bitstream conformance requirement that says that any block
must subsample to a valid block size with the current subsampling
mode. For example, this means that BLOCK_4X8 is illegal if there is
subsampling in only the horizontal direction (since there is no
BLOCK_2X8).

This patch checks the bitstream is conformant as it reads partition
information in decodeframe.c

BUG=aomedia:875

Change-Id: I18139aa76d6f965282402edbb0b68959478a46c3