This patch adds bit account infrastructure to the bit reader API.
When configured with --enable-accounting, every bit reader API
function records the number of bits necessary to decoding a symbol.
Accounting symbol entries are collected in global accounting data
structure, that can be used to understand exactly where bits are
spent (http://aomanalyzer.org). The data structure is cleared and
reused each frame to reduce memory usage. When configured without
--enable-accounting, bit accounting does not incur any runtime
overhead.
All aom_read_xxx functions now have an additional string parameter
that specifies the symbol name. By default, the ACCT_STR macro is
used (which expands to __func__). For more precise accounting,
these should be replaced with more descriptive names.
Change-Id: Ia2e1343cb842c9391b12b77272587dfbe307a56d
To get ready for pulling AV1 to nextgenv2
Replace the experimental flag by MOTION_VAR. Rename major variables.
Change-Id: If6cf4f37b9319c46d8f90df551cc7295d66ca205
- av1_fht32x32 AVX2 function level time reduction ~89% compared to C.
- av1_fht32x32_avx2() on DCT_DCT improves 42.62% over aom_fdct32x32_avx2()
But function replacement must go with the corresponding inverse txfm.
- No obvious user level time reduction due to 32x32 TX_TYPE selection.
- Zero high 128b YMM to avoid AVX-SSE transition penalties
(fix 16x16 case).
- Added 32x32 AVX2 unit tests to verify bitexact.
- AVX2 optimization summary:
On CPU i7-6700, based on 16x16/32x32 fwd txfm optimization results:
C to AVX2: function level time reduction, ~86-89%.
SSE2 to AVX2: function level time reduction, ~51%.
Change-Id: Idd0cd8bf066a61c7117140ef15ab6c1f8eb4b036
Cherry-Picked the following commits:
0defd8f Changed "WebM" to "AOMedia" & "webm" to "aomedia"
54e6676 Replace "VPx" by "AVx"
5082a36 Change "Vpx" to "Avx"
7df44f1 Replace "Vp9" w/ "Av1"
967f722 Remove kVp9CodecId
828f30c Change "Vp8" to "AOM"
030b5ff AUTHORS regenerated
2524cae Add ref-mv experimental flag
016762b Change copyright notice to AOMedia form
81e5526 Replace vp9 w/ av1
9b94565 Add missing files
fa8ca9f Change "vp9" to "av1"
ec838b7 Convert "vp8" to "aom"
80edfa0 Change "VP9" to "AV1"
d1a11fb Change "vp8" to "aom"
7b58251 Point to WebM test data
dd1a5c8 Replace "VP8" with "AOM"
ff00fc0 Change "VPX" to "AOM"
01dee0b Change "vp10" to "av1" in source code
cebe6f0 Convert "vpx" to "aom"
17b0567 rename vp10*.mk to av1_*.mk
fe5f8a8 rename files vp10_* to av1_*
Change-Id: I6fc3d18eb11fc171e46140c836ad5339cf6c9419
Mannually cherry-picked:
1579133 Use OD_DIVU for small divisions in temporal_filter.
0312229 Replace divides by small values with multiplies.
9c48eec Removing divisions from od_dir_find8()
0950ed8 Merge "Port active map / cyclic refresh fixes to vp10."
efefdad Port active map / cyclic refresh fixes to vp10.
1eaf748 Port switch to 9-bit rate cost to aom.
0b1606e Only build deringing code when --enable-dering.
e2511e1 Deringing cleanup: don't hardcode the number of levels
8fe5c5d Rename dering_in to od_dering_in to sync with Daala
4eb1380 Makes second filters for 45-degree directions horizontal
7f4c3f5 Removes the superblock variance contribution to the threshold
3dc56f9 Simplifying arithmetic by using multiply+shift
cf2aaba Return 0 explicitly for OD_ILOG(0).
49ca22a Use the Daala implementation of OD_ILOG().
8518724 Fix compiler warning in od_dering.c.
485d6a6 Prevent multiple inclusion of odintrin.h.
51b7a99 Adds the Daala deringing filter as experimental
Note that a few of the changes were already in libvpx codebse.
Change-Id: I1c32ee7694e5ad22c98b06ff97737cd792cd88ae
spatial/temporal scalability are not supported in VP10 currently.
+ remove the unused vp10/encoder/skin_detection.[hc]
this also enables DatarateTestLarge for VP10 which passes with no
experiments enabled. these were removed previously when only the SVC
tests should have been:
134710a Disable tests not applicable to VP10
Change-Id: I9ee7a0dd5ad3d8cc1e8fd5f0a90260fa43da387c
- Add unit tests to verify the bit-exact result.
- In speed test, function speed (for each mode/tx_size)
improves about 23%~35%.
- On E5-2680, park_joy_1080p, 10 frames, --kf-max-dist=1,
encoding time improves about 1%~2%.
Change-Id: Id89f313d44eea562c02e775a6253dc4df7e046a9
- HBD encoder speed improvement (SSE4.1):
Enable CONFIG_VP9_HIGHBITDEPTH, on Xeon E5-2680,
50 frames, park_joy_1080p, 12-bit,
Encoding time reduces from 4846481 to 4177471 (ms)
- Add unit test to verify bit-exact and EOB calculation
Change-Id: I08e8ef3549ddad5ab36d86e78557df3b288537ea
- Made source buffers pointers to const.
- Renamed vpx_blend_mask6b to vpx_blend_a64_mask. This is more
indicative that the function does alpha blending. The 6, or 6b
suffix was misleading, as the max mask value (64) does not fit into
6 bits.
- Added VPX_BLEND_* macros to use when needing to blend scalars.
- Use VPX_BLEND_A256 in combine_interintra to be more explicit about
the operation being done.
- Added versions of vpx_blend_a64_* which take 1D horizontal/vertical
masks directly and apply them to all rows/columns
(vpx_blend_a64_hmask and vpx_blend_a64_vmask). The SSE4.1 optimzied
horizontal version now falls back on the 2D version. This can be
improved upon if it show up high enough in a profile.
- All vpx_blend_a64_* functions now support block sizes down to 1x1
(ie: a single pixel). This is for usage convenience. The SSE4.1
optimized versions fall back on the C implementation if
w <= 2 or h <= 2. This can again be improved if it becomes hot code.
Change-Id: I13ab3835146ffafe3e1d74d8e9cf64a5abe4144d
test/assertion_helpers.h
test/randomise.{cc,h}
test/snapshot.h
Modfiy blend_mask6_test.cc not to rely on these.
Change-Id: I88b8933fe0a729a606797e5cd421795a544c612d
This reinstates the tests from commit
efda2831e5f758b4f350679b5c55c0b9282449b0 with the appropriate
fixes for 32 bit x86 builds.
Change-Id: Ib331906c5b448ca964895ee9cbfd4266f67d1089
- Apply signal direction/4-pixel vertical/8-pixel vertical
parallelism.
- Add unit test to verify the bit exact result.
- Overall encoding time improves ~24% on Xeon E5-2680 CPU.
Change-Id: I104dcbfd43451476fee1f94cd16ca5f965878e59
Since combining VPX_DL_REALTIME with VPX_RC_FIRST_PASS is basically
nonsense, ignore the user's pass setting when this happens and
behave as if the requested encode is a single pass encode.
BUG=webm:1233
Change-Id: I5ee4c4e5838c4ca6d24988890aae490b10826db2
this file is shared between vp9 & vp10; this makes it available in the
presence of --disable-vp9
BUG=webm:1235
Change-Id: Iaf060c3c09afd2c7df69995b0c01589f78d4945e
This reverts commit efda2831e5f758b4f350679b5c55c0b9282449b0.
This commit causes segmentation fault at SSE2/SumSquares2DTest.RandomValues/0
Change-Id: I171937e4daf6f15323e8206418773deb03bd8c53
We can optimize wedge partition selection by pre-computing the
residuals of the 2 underlying predictors, and then blend these
to compute the sse of the compound predictor, without actually
having to compute and subtract the compound predictor.
Similarly we can pre-compute a proxy array which we can use to
cheaply check which mask sign would have lower sse.
Details are in wedge_utils.c.
Mathematically these are equivalence transformations, but due to the
finite precision the encoder output will be perturbed, though on
average this should make 0% difference.
ext-inter gains about ~4.5% speedup.
Change-Id: Ib2657c3209ae161b4090b58b4b6c392641bf2792
This is to replace vp10/common/reconinter.c:build_masked_compound.
Functionality is equivalent, but the interface is slightly more
generic.
Total encoder speedup with ext-inter: ~7.5%
Change-Id: Iee18b83ae324ffc9c7f7dc16d4b2b06adb4d4305
