The frame mi has an 1 row and 1 column offset from the beginning of
the allocated memory.
cm->mi = cm->mip + cm->mi_stride + 1;
This offset can be removed. This doesn't change the bitstream.

Also, fixed 1 bug in first_pass code to update xd->mi for each block.

Change-Id: I2327aaadc62d2b00baa5a3c5135281d7fc864ed7

(1) We use an initial variable shift, and then
(2) A multiply + 16 bit shift for bit-depth = 8, OR
    A multiply + 17 bit shift for bit-depth = 10 and 12

All the constants (shifts and multipliers) for different block sizes are
based on "Algorithm 1" in
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1467632

Note:
This also reverts test hashes etc to the version before
commit 37ebf187.

BUG=aomedia:1191

Change-Id: I2aaebc3e95958a3f145c0408aa5cbcc85e30c8dc

only googletest/include is needed for the gtest.h sub-includes.

Change-Id: Ib821a596103377c2f7a0df5f99ead209874c8ee6

Change-Id: I15beaeaf46d7c12795a7db369228c14b57a0b258

Fix compliler warnings.
Also fix incompatible input date type for aom_uleb_decode().

Change-Id: I61442ff4a2f8c21ffbc3c8cdd827ec2abdf9be25

Change-Id: Ia0f1f2e928f2cef740aabd2b39eaa410fcb1f084

Change-Id: I6cbf928fe2bfc634cecbfce8f0942e02d5a17db1

Under CONFIG_OBU, there are a large number of duplicate lines for the
handling of INTER_FRAME and S_FRAME in the packing of the
uncompressed frame header. Indeed, there are only two lines of
difference between the handling of these two different frame types. By
removing the two lines, it has been confirmed the code between the two
"if" branches are completely identical.

Further, by unifying the handling of these two frame types, the
incorrect handling on S_FRAME under CONFIG_AMVR (default off) is fixed.

Change-Id: Iad5bb9704a871e61bcdb108223c6c6e131e341f2

Change-Id: I912aeade82c2b17adf2cbfcba1c58d05b8eb7e8a

Some leftover from 69ae7b86.

Change-Id: I4c37538fe8e13761631bd9521e954af0730c4e0b

Change-Id: I4230cd0cda845fbe65ed75b6dc8139004a792758

Change-Id: Iba30805ba15c6c7cf16bd30e0eebfc4e9aeb2f17

Always use ext-intra style intra predictors. This is the same behavior
as when intra-edge(adopted) is on.

BUG=aomedia:1399

Change-Id: I404923844e4cc81162fa649fc8b7dc695af63545

BUG=aomedia:1363

Change-Id: Ie5609e102a5fcf0d8ff9fe592624722eac1309c5

The coding order becomes
for()
  read base
  read br
for()
  read sign
for()
  read golomb

BUG=aomedia:1369

Change-Id: I85fdcbb592f1b26a07807c6dd23cf7e2a6a64ca4

Change-Id: I9f6d50b21454b28b7ff76959ee8f3367ffd50d17

Move coeff decoding logics into read_coeffs_reverse()

Change-Id: Iea7fa09d387af1d797de3df1cf2878516e039203

Change-Id: I49e6425819b819cf604778688f44df886a44b162

This is to separate the logic of getting lower level coeff ctx
for eob and non-eob position

BUG=aomedia:1369

Change-Id: Id1d85d7fbb67e57c58479e00b633d794381132ea

This speed up av1_read_coeffs_txb by 10%
The over decoding time is reduced by 1.3%

BUG=aomedia:1369

Change-Id: Ic6e2f8dc4e04389678c7af9e742c31917c1c8df5

SGR currently calculates two filters with different radii. This patch
adds the experiment SKIP_SGR in which a filter is skipped if the radius
of the filter is set to 0.

SSE4.1 and SIMD code has also been updated.

Change-Id: I77e879e3636bfbacab2b6fc2ab426f35000c8e92

BUG=aomedia:1346

Change-Id: I4247f1387022cfd3b69116f6f259eebbf61db791

- Make them faster by setting cpu-used to 1.
- Drop unused cruft and redundant args to aomenc.
- Add some shell util functions for producing "fast" encode
  settings and frame count limits while ensuring lag-in-frames
  is sane for the given limit.

Change-Id: I61288f9f120201decb25fdfe9516b75d84cfb2ee

Change-Id: Ia788d299883f55cb57c4282e816752b6f0a197a2

This experiment has been abandonned for AV1.

Change-Id: I393e188420450b838fa53e8d7a4a00d41a4a2c6d

Currently set cm->frame_refs_short_signaling as 0 by default. It will
write the one extra bit to the frame header, and leave all the encoder
setup the same as the baseline. It is in the TODO list for the encoder
side implementation for cm->frame_refs_short_signaling == 1.

BUG=aomedia:1392

Change-Id: Iabcc7618c9a52b7bb975b69e2d134f49a969944b

There are 4 delta_lflevels (FRAME_LF_COUNT=4), corresponding to 2 for the luma
plane and 1 for each of the U and V planes. In monochrome mode we do not need
to use/encode/decode the last two values, and so in this case the for loops all
terminate at FRAME_LF_COUNT - 2 rather than FRAME_LF_COUNT.

BUG=aomedia:1410

Change-Id: Ieb60cce9e39287a385e7f74d9009aac1f0f23e54

Turn of CONFIG_RD_DEBUG when jnt_comp is on, to avoid stack size
overflow.

Make subpel processing for width <= 4 correct.

Change-Id: Ic1de96ff2eff4a80543e19531fa75511b0a2f427

This mirrors the behavior of the corresponding option for CDEF.

Change-Id: Ifbdacb1bedbde2fa740086593384567a2f44b342

This experiment has been adopted, we can simplify the code
by dropping the associated preprocessor conditionals.

Change-Id: Ic3438799335c6cd05f170302f49bd47e1f705c24

Set the maximum frame distance to be 32. If a reference motion
vector needs to span across two frames at distance more than 32,
drop the use of this reference motion vector.

BUG=aomedia:1394

This also clears ubsan warnings in unit tests.

Change-Id: I98b72b5fd2a885661542bbc5f8da1886d77f889f

Allow the right side rectangular block to access the top right
neighbor if available in vert_a partition.

BUG=aomedia:1343

Change-Id: I92369676ae1bea2d6dd5f09af612f4177560ae29

Change-Id: I278be8e79f2eb99f711c3d8f98c801f2f1f7bd38

Make test input aligned, then jnt_comp's subpel_avg_variance simd
implementations pass unit tests.

Change-Id: Ic85442f3252ba4e8707e35fbd6344fd662aafd74

To match the corresponding prediction angle.

Change-Id: I0441a0a25f165c1fdeb4eba915c6ca83b0514c4d

Add a flag disable_intra_edge_filter in the uncompressed header
of key/all-intra frames to disable corner, edge, and upsampling
filter with ext-intra modes.

This adds an experiment flag: intra_edge2

BUG=aom:1398

Change-Id: I799b541a92f8bc928c8c34e918de89313c22efb9

The do_post_rounding flag can now be removed since it is equivalent
to is_compound.

Change-Id: I3a18d853571b7363a87e65601c82d1f345b3dcd2

This patch simplifies the setting and use of the do_post_rounding
flag consistently across all convolve/warp operations. It is always
set now to 1 for compound modes, but the masked blending operation
can subsequently turn it off.
To accomplish this, for the warped case, the use of the conv_params->dst
buffer is made consistent with the other non-warp convolves,
in the sense that it is used only when is_compound is 1.

Change-Id: If3a37ffa65c3ca75f1df66ca427e6b5df86ac72f

BUG=aomedia:1363

Change-Id: I8ca90f1b0d7dfe92078c611f66dcb19e70a6c1e0

For FRAME_REFS_SIGNALING, when error resilience mode is on for inter
frames, no frame refs signaling options will be provided. Hence no
additional bit is added and the reference frame signaling stays the
same as baseline.

BUG=aomedia:1392

Change-Id: I8ff9688c278373acaae3ee4fb74e9a392132f54d