17183 Commits

Author SHA1 Message Date
James Zern
3735def667 vpx_fdctNxN_1_sse2: reduce store size
only output[0] needs to be set, store_output is more involved than a
movdqa in the high bitdepth case

Change-Id: I2cbd85d7cf74688bdf47eb767934fe42e02bff67
2016-04-04 16:02:06 -07:00
Marco
6877acb67e vp8-denoiser: Avoid doing the mcomp if we don't denoise.
Avoid doing the mcomp in denoiser if we don't denoise the
block (because of motion/SSE/skin threshold, etc).
This can reduce encoding time (with denoiser enabled) by ~1.5-2%.

Change-Id: Ia699b68dfd37b89cdf3a82b8aa40e8c8f98a3d4f
2016-04-04 13:25:38 -07:00
Angie Chiang
ff8c490b9a Branch dct to new implementation for bd12
Change-Id: I9281935653aacce22ac3100f79fb956c249e2bf3
2016-04-04 12:40:10 -07:00
Marco
5572ab1d02 vp9-noise estimate: Adhust threshold for Low level detection.
Change-Id: I1c78255c81cb4ac218964c80f8ecaf29f95b875b
2016-04-04 12:33:08 -07:00
Tom Finegan
79d5aac74f Fix crash when mkvparser returns a NULL BlockEntry.
BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1090

Change-Id: I64b7b19581f15d53048e3026c3f8946e344fbe26
2016-04-04 11:50:44 -07:00
Yi Luo
250935cab3 Optimized HBD 4x4 variance calculation
vpx_highbd_8/10/12_variance4x4_sse4_1 improves performance ~7%-11%.

Change-Id: Ida22bb2a2f7a58037cfd73e186d4f6267a960c02
2016-04-04 11:28:59 -07:00
Angie Chiang
f1060f5bc4 Change dct32x32's range
Bitdepth 10/12:
Fit coefficient range into 32 bits
Fit codfficient * const range into 32 bits

Bitdepth 8:
Fit coefficient range into 16 bits
Fit codfficient * constant range into 32 bits

Change-Id: I50b5a3132e8a9f5155c971ab0f6eb52876d2b5ca
2016-04-04 11:21:11 -07:00
Angie Chiang
39b3c025fa Fit dct's stage range into 32-bit when bitdepth is 12
Change-Id: I807e60c6dcacc50c087adcbdb1df022f8541efc5
2016-04-04 11:13:44 -07:00
Marco Paniconi
71cf9fac83 Merge "vp9-noise estimation: Increase threshold for Low-level." 2016-04-04 17:33:35 +00:00
Marco
9ff73fe092 vp9-noise estimation: Increase threshold for Low-level.
This make it more likely clean/low-noise content will
be set as LowLow, and hence no denoising will be done.
Also set early exit on denoising for small blocks.

Change-Id: I4a72bba3e6c5e2d523d304c39deacc9c39bf216c
2016-04-04 09:32:22 -07:00
Marco
a2307ff095 vp9-denoiser. Code cleanup and bugfix.
Some cleanup and bugfix: pass mi_row/mi_col (not mv_col/mv_row)
to build_inter_predictors. This only affects case where
the frame is resized, but since denoising is not done on resized
frames, the fix has not effect currently.

Change-Id: I36617a7f0b43b6f49976745f15d400977e6ffa46
2016-04-04 09:11:00 -07:00
Geza Lore
f0290cd127 Refactor get_partition to be universal.
Change-Id: I3a2fe4073bb94c5afc24d9274e6edcdb3aed934f
2016-04-04 15:22:25 +01:00
Geza Lore
e0dbfdeedc Minor refactoring of partition type processing.
Change-Id: Idcb1e94298d4b7d8832d285548ec2d2ced4b2988
2016-04-04 14:51:10 +01:00
James Zern
89b1c9d4be Merge "dct32x32_test: create fn pointers w/'&' ref" 2016-04-02 19:38:36 +00:00
James Zern
50e05d0651 dct32x32_test: create fn pointers w/'&' ref
this helps some toolchains (vs9) resolve the type of the parameter

Change-Id: I116030361fc956e10de82b64c81d5917cddbcd5a
2016-04-02 11:17:29 -07:00
James Zern
c21d437052 vpx_fdct32x32_1_msa: fix accumulator overflow
Change-Id: I33a5432eda3416382e1cea06b45082c0c65faa75
2016-04-02 11:04:38 -07:00
James Zern
c98f8e04e5 dct32x32_test: add PartialTrans32x32Test, Random
Change-Id: I644f35b0b0a540bf5b6ebb6dbc05b3c95736f42b
2016-04-01 19:45:21 -07:00
James Zern
a0359b8c90 Merge changes I8e9692fc,Ifc06110c
* changes:
  vpx_fdctNxN_1_c: remove unnecessary store
  vpx_fdct32x32_1_c: fix accumulator overflow
2016-04-01 23:12:51 +00:00
Marco Paniconi
2f0067c62b Merge "vp9-denoiser: Adjust usage of skin and consec_zeromv." 2016-04-01 21:26:08 +00:00
Angie Chiang
75ae90f7a9 Pass tx_type into get_tx_scale
Change-Id: I8a8df9fdefa492f66cf2cd29b0b081ad69b5d85e
2016-04-01 12:53:10 -07:00
James Zern
f4cae05cd4 vpx_fdctNxN_1_c: remove unnecessary store
only output[0] needs to be set, the other values will be ignored in this
case.

Change-Id: I8e9692fc0d6d85700ba46f70c2e899a956023910
2016-04-01 12:21:59 -07:00
Marco
7467ee0458 vp9-denoiser: Adjust usage of skin and consec_zeromv.
Change-Id: I250258fc70ec0fcf1b501708c5d4b84685cf882e
2016-04-01 11:50:15 -07:00
Marco
bc6c199785 vp8 denoiser: Some adjustments to usage of skin and motion.
Switch to use new skin model.

And fix condition for denoising skin block.
Previous condition did not denoise skin blocks if the selected
mode was non-zero motion in current frame. Modify condition to
also force no denoising if that mode was not selected as zero motion
now and for at least "x" past frames in a row (x = 2).

Change-Id: I00753e3fe45b9a308a7ef43c58f11868e3bfc6b0
2016-04-01 11:36:45 -07:00
Alex Converse
d649065ea3 Merge "Remove duplicate ans parameter in bitstream functions." into nextgenv2 2016-04-01 18:35:11 +00:00
Alex Converse
c961bcc594 Merge "ANS experiment: Use ANS everywhere." into nextgenv2 2016-04-01 18:34:50 +00:00
Alex Converse
0f68d80420 Remove duplicate ans parameter in bitstream functions.
Change-Id: Icd459209dae328f90c9a875259fe5d201b2a4e45
2016-04-01 11:33:06 -07:00
Alex Converse
fb9186d68d ANS experiment: Use ANS everywhere.
Use ANS for all entropy coded data in VP10 including the compressed header and
modes and motion vectors. ANS tokens continue to be used for DCT tokens.

Change-Id: Idf709a747150601e4d95d81ecfb3dc7253d349df
2016-04-01 11:32:31 -07:00
Debargha Mukherjee
2fba8189de Merge "Loopfilter fix" into nextgenv2 2016-04-01 17:48:09 +00:00
Angie Chiang
9f879b3c5f Merge "change vp10_fwd_txfm2d_#x#_sse2 to vp10_fwd_txfm2d_#x#_sse4_1" into nextgenv2 2016-04-01 17:25:23 +00:00
Angie Chiang
2c2b9bd455 Merge "Remove redundant code from vp10_fwd_txfm2d.c" into nextgenv2 2016-04-01 17:25:13 +00:00
Angie Chiang
1b755039c6 Merge "Simplify rounding in vp10_[fwd/inv]_txfm[1/2]d_#x#" into nextgenv2 2016-04-01 17:24:50 +00:00
Angie Chiang
0a9eedfbef Merge "Add vp10_fwd_txfm2d_sse2" into nextgenv2 2016-04-01 17:24:34 +00:00
Scott LaVarnway
a2a97b869f VP9: Refactor vp9_decode_block_tokens()
Change-Id: I30ab27808ec903f9490f36621fb16c197bd35d16
2016-04-01 04:57:39 -07:00
James Zern
74ed95a33e Merge "disable vp9_diamond_search_sad_avx" 2016-04-01 03:33:51 +00:00
James Zern
569c0dfc77 Merge "vpx_ports/mem_ops.h: clear some -Wconversion warnings" 2016-04-01 03:31:29 +00:00
Debargha Mukherjee
f7457f5e89 Loopfilter fix
Fixes mismatch introduced in
https://chromium-review.googlesource.com/#/c/336645

Change-Id: I15cded221c18dbf87b5029bc464e975d5c7c40e3
2016-03-31 19:57:42 -07:00
Yaowu Xu
a416d5bd2d Fix a build issue
Change-Id: Ifdb32c487632098496bf59fcc76c518f8f0426d2
2016-03-31 16:06:24 -07:00
Debargha Mukherjee
2a6389bb8b Merge "Fix interpolation values and decouple interintra" into nextgenv2 2016-03-31 21:47:10 +00:00
Debargha Mukherjee
4aac819a18 Fix some static analysis issues with resize
Change-Id: I2e8ada1ae95bd5577344d6f898e6ad4723e38fbd
2016-03-31 14:39:34 -07:00
Debargha Mukherjee
2be211e971 Fix interpolation values and decouple interintra
Decouples interintra modes and probability models from regular
intra modes, to enable creating/optimizing new interintra modes.
Also, fixes interpolation values for 128x128 interintra and obmc.

Change-Id: I5c2016db49b8f029164e5fe84c6274d4e02ff90e
2016-03-31 12:12:51 -07:00
Debargha Mukherjee
6d3fc82b7f Merge changes Id20526d0,Iee08d975 into nextgenv2
* changes:
  Refactor loopfilter level arrays to 2D.
  Rename MI_BLOCK_SIZE and MI_MASK macros.
2016-03-31 18:48:20 +00:00
Yi Luo
48aa76ac35 Fixed HBD variance unit test on incorrect bit mask operation
- Change logical AND (&&) to arithmetic AND (&).
- Disable failed unit tests for next-step fixing.
https://bugs.chromium.org/p/webm/issues/detail?id=1166

Change-Id: I75cf8d114fa995731ae1cf4c782b134781274a1f
2016-03-31 09:17:59 -07:00
Jingning Han
aae7e0f6a4 Merge "Refactor the sub8x8 block motion search control" into nextgenv2 2016-03-31 15:50:38 +00:00
Geza Lore
10232eda8e Refactor loopfilter level arrays to 2D.
Change-Id: Id20526d0b6d1371dc9f45cb8b5f24b6974da7bc4
2016-03-31 15:52:12 +01:00
Geza Lore
511da8cbe5 Rename MI_BLOCK_SIZE and MI_MASK macros.
Rename MI_BLOCK_SIZE.* -> MAX_MIB_SIZE.* (MIB is for MI Block).
Rename MI_MASK.* -> MAX_MIB_MASK.*

There are no functional changes.

This is in preparation for coding the superblock size at the frame
level, which will require some of these constants to become variables.
The new names better reflect future semantics, and hence make the code
clearer.

Change-Id: Iee08d97554cf4cc16a5dc166a3ffd1ab91529992
2016-03-31 09:57:41 +01:00
Alex Converse
615482af92 Merge "Use write_modes_b_wrapper throughout." into nextgenv2 2016-03-31 06:04:03 +00:00
Marco
8fab3e82ca vp9-1 pass vbr: Use average QP for active_worst_quality, for inter-frames.
And replace some operation with shift.

Change-Id: If1ea874ee05b7238d5f954e55bb3ceda804a091a
2016-03-30 17:43:47 -07:00
James Zern
057c1c4034 disable vp9_diamond_search_sad_avx
this results in different output than C, observed with 1080p input at
speed 2.

BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1168

Change-Id: Ie58cf20057f4531d1b1d19c7b7eae9e642587ce5
2016-03-30 17:31:28 -07:00
Hui Su
cce6688c31 Merge "Set block size upper bound for Palette mode" into nextgenv2 2016-03-31 00:23:11 +00:00
James Zern
44a9edf406 vpx_ports/mem_ops.h: clear some -Wconversion warnings
not strictly necessary, but allows projects using '-Wconversion
-Wno-sign-conversion' to reuse these headers.

Change-Id: Id1398d726c90173ccba9aea66798fcef6f20fa23
2016-03-30 17:02:10 -07:00