]>
granicus.if.org Git - libvpx/log
Geza Lore [Fri, 17 Jun 2016 14:32:25 +0000 (15:32 +0100)]
Fix warnings from gtest under GCC 5 or newer.
Change-Id: I9661f2fe9d315dccae69caa70d929b5d9d93b7db
Jingning Han [Fri, 17 Jun 2016 00:25:50 +0000 (00:25 +0000)]
Merge "Rework table access operations in vp10_optimize_b function" into nextgenv2
Yue Chen [Thu, 16 Jun 2016 22:03:44 +0000 (22:03 +0000)]
Merge "Make supertx skip bits observe segment level coding." into nextgenv2
Jingning Han [Thu, 16 Jun 2016 16:50:19 +0000 (09:50 -0700)]
Rework table access operations in vp10_optimize_b function
Localize table access. This provides another 10% speed-up to
the unit.
Change-Id: Ib902121f412f78e2bd501b9799c8c64462f803b5
Debargha Mukherjee [Thu, 16 Jun 2016 18:25:04 +0000 (18:25 +0000)]
Merge "Change supertx syntax order." into nextgenv2
Zoe Liu [Thu, 16 Jun 2016 16:41:30 +0000 (09:41 -0700)]
Disable the unit test of ArfFreq for BIDIR_PRED
The test in arf_freq assumes any no-show frame as ALTREF_FRAME and
then calculate the minimum run between two consecutive ALTREF_FRAME's
based on this assumption. As BWDREF_FRAME is also a no-show frame and
the minimum run between two consecutive BWDREF_FRAME's may vary
between 1 and any arbitrary positive number as long as it does not
exceed the golden frame group interval, this test does not apply to
the experiment of BIDIR_PRED.
Change-Id: I70efb2c691fdc18601dbb8a7735ac2f27817e75a
Geza Lore [Thu, 16 Jun 2016 16:26:23 +0000 (17:26 +0100)]
Make supertx skip bits observe segment level coding.
Change-Id: Id918d502c8f89e236bcb51949d7ad34efa017321
Geza Lore [Thu, 16 Jun 2016 16:13:55 +0000 (17:13 +0100)]
Change supertx syntax order.
Move the supertx skip bit and transform type past the recursive
prediction blocks. This is in preparation for using the segment level
skip feature for supertx blocks.
Change-Id: I8319414b0734144a9264e8a4a60940b6716b12a8
Zoe Liu [Thu, 16 Jun 2016 16:15:07 +0000 (16:15 +0000)]
Merge "Fix the superframe unit test for BIDIR_PRED" into nextgenv2
Geza Lore [Thu, 16 Jun 2016 12:16:07 +0000 (13:16 +0100)]
Plug leak of variance tree.
The speed features can change per frame, so remove condition on
releasing the variance tree.
Change-Id: I651c87a1504266d737e6d98f14fd3ed30d84e01d
Debargha Mukherjee [Thu, 16 Jun 2016 12:15:47 +0000 (12:15 +0000)]
Merge "Use correct size load in vpx_avg_4x4_sse2." into nextgenv2
Debargha Mukherjee [Thu, 16 Jun 2016 11:46:35 +0000 (11:46 +0000)]
Merge "Split some slower tests based on cpu-used" into nextgenv2
Geza Lore [Thu, 16 Jun 2016 10:08:14 +0000 (11:08 +0100)]
Use correct size load in vpx_avg_4x4_sse2.
The old version used 64 bit loads, and then ignored the top half
of the result. This can cause asan failures if we read past the end
of a buffer. Switched to using 32 bit loads instead.
Change-Id: I57da127a26f869fb4b4f700b55408f6dc2fbbc1a
Debargha Mukherjee [Tue, 14 Jun 2016 21:50:30 +0000 (14:50 -0700)]
Split some slower tests based on cpu-used
Change-Id: Idf84475fe06666d5c73c9d86dfc5c23bef170086
James Zern [Thu, 16 Jun 2016 03:33:04 +0000 (20:33 -0700)]
cosmetics,test.mk: fix a typo
Change-Id: Ib74a494e1cf50a356f51e8185e19ca66fcb896a2
James Zern [Thu, 16 Jun 2016 01:30:22 +0000 (18:30 -0700)]
rename vp9_end_to_end_test.cc -> end_to_end_test.cc
this is shared between vp9/10
BUG=webm:1235
Change-Id: I2f44b15268a33453a1c1e0c691d4fc1fc12d0263
James Zern [Thu, 16 Jun 2016 01:28:30 +0000 (18:28 -0700)]
vp9_end_to_end_test: enable in vp10-only builds
this file is shared between vp9 & vp10; this makes it available in the
presence of --disable-vp9
BUG=webm:1235
Change-Id: Iaf060c3c09afd2c7df69995b0c01589f78d4945e
Zoe Liu [Wed, 15 Jun 2016 22:39:51 +0000 (15:39 -0700)]
Fix the superframe unit test for BIDIR_PRED
Change-Id: I2ef8e479893403581711abc020509c6863c2035d
Sarah Parker [Thu, 9 Jun 2016 19:36:45 +0000 (12:36 -0700)]
Add EndToEndTestLarge for VP10 non-highbitdepth
The current test case is only run for vp9 and vp10 when HBD
is enabled. This was mistakenly removed in:
d53f9a3 Enable VP10 HBD PSNR checking unit test
Change-Id: I88b8168ad1efd805d759238a037653a2901bf50d
Jingning Han [Wed, 15 Jun 2016 19:16:30 +0000 (19:16 +0000)]
Merge "Refactor trellis optimization process" into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:51:38 +0000 (16:51 +0000)]
Merge "Pick up correct dequant for segment with new-quant." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:50:55 +0000 (16:50 +0000)]
Merge "Disable loop restoration when LPF_PICK_MINIMAL_LPF." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:50:15 +0000 (16:50 +0000)]
Merge "Remove magic number from traversal (CYCLIC_REFRESH_AQ)." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:49:33 +0000 (16:49 +0000)]
Merge "Select segment based loopfilter strength for supertx blocks." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:48:13 +0000 (16:48 +0000)]
Merge "Remove now superfluous argument from predict_b_extend." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:47:58 +0000 (16:47 +0000)]
Merge "Rework supertx segment handling and adaptive quantization." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:47:28 +0000 (16:47 +0000)]
Merge "Minor refactor of decode_block for supertx." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:45:57 +0000 (16:45 +0000)]
Merge "Re-initialise quantiser after changing segment." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:45:21 +0000 (16:45 +0000)]
Merge "Refactor variance aq." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:44:37 +0000 (16:44 +0000)]
Merge "Pass segment id explicitly to quantizer init." into nextgenv2
Debargha Mukherjee [Wed, 15 Jun 2016 16:43:51 +0000 (16:43 +0000)]
Merge "Fix estimate_wedge_sign with high bit-depth." into nextgenv2
Jingning Han [Wed, 15 Jun 2016 16:07:10 +0000 (16:07 +0000)]
Merge "Rework transform quantization pipeline" into nextgenv2
Jingning Han [Wed, 15 Jun 2016 16:07:03 +0000 (16:07 +0000)]
Merge "Refactor the trellis optimization process" into nextgenv2
Jingning Han [Fri, 10 Jun 2016 03:44:00 +0000 (20:44 -0700)]
Refactor trellis optimization process
This commit refactors the trellis coefficient optimization process.
It saves multiplications used to generate the final dequantized
coefficients. It also removes two memset operations on quantized
and dequantized coefficient sets.
The trellis coefficient optimization is on average running over
10% faster.
Change-Id: If3aa26d2a706c3012bf2b7ac059bf1825250e81f
Geza Lore [Wed, 15 Jun 2016 09:49:37 +0000 (10:49 +0100)]
Fix estimate_wedge_sign with high bit-depth.
This fixes some crashes in
VP10/EndToEndTestLarge.EndtoEndPSNRTest/ with high bit-depth and
ext-inter.
Change-Id: I10f0f08e1be4bd5c388616074d4aa3f91a2fda7a
Hui Su [Wed, 15 Jun 2016 02:46:06 +0000 (02:46 +0000)]
Merge "Speed up ext-intra inter frame encoding" into nextgenv2
Hui Su [Wed, 15 Jun 2016 02:45:59 +0000 (02:45 +0000)]
Merge "ext-intra: refactor rd loop in interframe" into nextgenv2
Jingning Han [Thu, 9 Jun 2016 18:20:26 +0000 (11:20 -0700)]
Rework transform quantization pipeline
This commit reworks the transform and quantization unit. It enables
the use of adaptive quantization for intra modes. This further
improves the compression performance:
lowres 0.36%
midres 0.79%
hdres 0.73%
The key frame coding performance is improved:
lowres 1.7%
midres 1.9%
hdres 3.3%
The overall coding gains are:
lowres 1.1%
midres 1.8%
hdres 2.3%
Change-Id: Iaec1a3a4c1d5eac883ab526ed076d957060479dd
Hui Su [Tue, 14 Jun 2016 23:11:25 +0000 (23:11 +0000)]
Merge "Fix rate cost calculation for ext-intra" into nextgenv2
Hui Su [Tue, 14 Jun 2016 22:37:19 +0000 (22:37 +0000)]
Merge "Handle intra modes when tx type speed feature is enabled" into nextgenv2
Jingning Han [Tue, 14 Jun 2016 21:54:08 +0000 (21:54 +0000)]
Merge "Fix enc/dec mismatch in non-420 settings" into nextgenv2
hui su [Mon, 13 Jun 2016 17:52:19 +0000 (10:52 -0700)]
Handle intra modes when tx type speed feature is enabled
Change-Id: I9dc156214f3b3ded33ab30d558124b3151548161
hui su [Thu, 2 Jun 2016 18:53:00 +0000 (11:53 -0700)]
Speed up ext-intra inter frame encoding
Skip filter intra mode search when regular intra modes have large
rd cost.
Encoding speed improvement: 8%.
Compression performance drop: 0.02% / 0.09% / 0.03% on
lowres / midres / hdres
Change-Id: I94d3e48781bff6ae6895a54f271dd65c959bb976
hui su [Tue, 31 May 2016 21:40:04 +0000 (14:40 -0700)]
ext-intra: refactor rd loop in interframe
Move filter intra modes search to the end, after regular
mode search.
On average no performance changes.
Change-Id: I9293c8fdf706ebf831fbd61c6bb81959790f4848
hui su [Fri, 10 Jun 2016 17:37:24 +0000 (10:37 -0700)]
Fix rate cost calculation for ext-intra
It was broken by commit
8ee640f979 .
Change-Id: I26b9eba810c74849b0805e64da2d269ab0685cb9
Jingning Han [Tue, 14 Jun 2016 20:45:02 +0000 (20:45 +0000)]
Merge "Make tx_type speed feature default" into nextgenv2
Geza Lore [Tue, 14 Jun 2016 15:47:06 +0000 (16:47 +0100)]
Pick up correct dequant for segment with new-quant.
Change-Id: Id73500305a7cc581c11461c9ddb1b22dd8f5d8f4
Geza Lore [Tue, 14 Jun 2016 12:41:20 +0000 (13:41 +0100)]
Disable loop restoration when LPF_PICK_MINIMAL_LPF.
The speed feature sf->lpf_picl == LPF_PICK_MINIMAL_LPF is used
to disable loop filtering. This did not work with the loop-restoration
experiment, but now it is respected.
Note that this speed feature is only used in real-time cpu-used >= 8
settings.
Change-Id: I193723c9ac5f802ec31d8c8b4d37650796e065fd
Geza Lore [Tue, 14 Jun 2016 12:12:45 +0000 (13:12 +0100)]
Remove magic number from traversal (CYCLIC_REFRESH_AQ).
mi->stride now depends on the maximum superblock size, and hence
the constant 8 padding is no longer appropriate. Traverse the array
using mi->stride instead.
Change-Id: I8e84b9fe1728f6663f8c10765fe32206375f1e71
Geza Lore [Fri, 10 Jun 2016 08:32:21 +0000 (09:32 +0100)]
Select segment based loopfilter strength for supertx blocks.
Segment based loopfilter strength for supertx coded blocks is now
selected based on the minimum of all segment IDs within a supertx
coded block (same as the quantiser settings).
Change-Id: Ib056bd0d05f6a1d3b512a76deb4e2ad4db0f7dc4
Geza Lore [Thu, 9 Jun 2016 15:56:04 +0000 (16:56 +0100)]
Remove now superfluous argument from predict_b_extend.
Change-Id: I7a76756842af9ce806c6e0e1f98f294af748e8bd
Geza Lore [Thu, 9 Jun 2016 16:06:02 +0000 (17:06 +0100)]
Rework supertx segment handling and adaptive quantization.
Segment level quantizer settings for supertx coded blocks are now
selected based on the minimum of all segment IDs within a supertx
coded block.
This also fixes the 3 adaptive quantization modes with supertx.
Change-Id: Ib5db099539d4f82f240e1d745d6e5264f8b34cde
Geza Lore [Thu, 9 Jun 2016 16:02:10 +0000 (17:02 +0100)]
Minor refactor of decode_block for supertx.
Exit early from function when supertx is used, rather than putting
the bulk of the function body in a single conditional.
Change-Id: I41f388a45bd46e4a6ee1c51f26782ed9bddff4e5
Geza Lore [Thu, 9 Jun 2016 14:29:29 +0000 (15:29 +0100)]
Re-initialise quantiser after changing segment.
When using VARIANCE_AQ, we can change the segment assignment after
initialising the quantiser in set_offsets, so re-initialise it when
we do so.
Change-Id: I1f168553aaf0ade419f0d4bf05820cd591b87659
Geza Lore [Thu, 9 Jun 2016 14:12:27 +0000 (15:12 +0100)]
Refactor variance aq.
Explicitly signal when the segment map is being refreshed when
using VARIANE_AQ. This simplifies decisions about when the segment id
needs to be set from the previous segment map vs based on the current
variance.
Change-Id: Ieb12c950e9cfbc3f53f4d184880071dea805563c
Geza Lore [Thu, 9 Jun 2016 12:31:37 +0000 (13:31 +0100)]
Pass segment id explicitly to quantizer init.
This is purely refactoring in preparation of fixing supertx segment
handling
Change-Id: I74bcae34241fdf2b592e1cd45b67af77b9e16c9a
Debargha Mukherjee [Tue, 14 Jun 2016 14:44:55 +0000 (14:44 +0000)]
Merge "A crash fix for supertx / ext-inter combination." into nextgenv2
Jingning Han [Mon, 13 Jun 2016 17:30:52 +0000 (10:30 -0700)]
Fix enc/dec mismatch in non-420 settings
This commit makes the dual filter experiment work with non-420
settings. It fixes unit test failure in EndToEndTestLarge.
Change-Id: I04f7afdee78f91389d9ff72947efa152098af930
James Zern [Tue, 14 Jun 2016 00:04:56 +0000 (00:04 +0000)]
Merge "Revert "Add 1D version of vpx_sum_squares_i16"" into nextgenv2
Debargha Mukherjee [Mon, 13 Jun 2016 20:14:54 +0000 (13:14 -0700)]
A crash fix for supertx / ext-inter combination.
Change-Id: I9860376c98aa3b25f5bf86ed13d4a7631fa6b153
Jingning Han [Wed, 8 Jun 2016 22:43:53 +0000 (15:43 -0700)]
Refactor the trellis optimization process
Speed up the trellis optimization unit by 10%.
Change-Id: If055f6c0589a405c008d2900bb8fbc11b1246f66
Jingning Han [Mon, 13 Jun 2016 19:08:14 +0000 (12:08 -0700)]
Make tx_type speed feature default
Revisit the compression performance and complexity trade-off after
making the SIMD version of trellis optimizations. Before that,
reduce the transform-quantization function calls temporarily. This
would cause about 0.3% performance drop for lowres set.
Change-Id: I16917a6bd5c44ec6cd8cd0b59f3c336c4fd96dd2
James Zern [Mon, 13 Jun 2016 19:08:04 +0000 (19:08 +0000)]
Merge "active_map_refresh_test: fix missing file w/vp10-only" into nextgenv2
Jingning Han [Mon, 13 Jun 2016 17:36:19 +0000 (17:36 +0000)]
Merge "Trellis based adaptive quantization" into nextgenv2
James Zern [Sat, 11 Jun 2016 16:49:02 +0000 (09:49 -0700)]
active_map_refresh_test: fix missing file w/vp10-only
Change-Id: I6413b7622a3c8524ec0409e087cf7c92f79e4f2d
Debargha Mukherjee [Fri, 10 Jun 2016 23:18:38 +0000 (23:18 +0000)]
Merge "Some refactoring to support warped motion mode" into nextgenv2
Zoe Liu [Fri, 10 Jun 2016 22:58:30 +0000 (15:58 -0700)]
Fix one typo in the comment
Change-Id: Ie98fd60426b18980ec85572f3cfc9ce0b97a5361
Alex Converse [Fri, 10 Jun 2016 21:52:56 +0000 (21:52 +0000)]
Merge "Turn on ActiveMapTest speeds [0,5) with all experiments." into nextgenv2
Jingning Han [Wed, 8 Jun 2016 00:54:20 +0000 (17:54 -0700)]
Trellis based adaptive quantization
This commit combines uniform quantizer with trellis based coefficient
level optimization. It improves the codebase compression performance:
lowres 0.8%
midres 1.0%
hdres 1.6%
Note that the current trellis optimization unit is using C code. This
will make the cost of the overall quantization process slower. A number
of optimizations will come up next.
Change-Id: Id441dd238e4844409d0f08f82604be777f3f5282
Debargha Mukherjee [Thu, 9 Jun 2016 07:57:09 +0000 (00:57 -0700)]
Some refactoring to support warped motion mode
Change-Id: I15d54a3ae48b2b33082668116792c6595bdb3ddb
Debargha Mukherjee [Fri, 10 Jun 2016 18:47:17 +0000 (18:47 +0000)]
Merge "Adds higher precision for homography model 3rd row" into nextgenv2
Sarah Parker [Fri, 10 Jun 2016 17:26:33 +0000 (17:26 +0000)]
Merge "Move new quant experiment from nextgen" into nextgenv2
Jingning Han [Fri, 10 Jun 2016 16:04:38 +0000 (16:04 +0000)]
Merge "Add MIN_TX_SIZE definition" into nextgenv2
Sarah Parker [Tue, 10 May 2016 22:32:42 +0000 (15:32 -0700)]
Move new quant experiment from nextgen
This experiment implements non-uniform quantization where
the width of the bins increases gradually to more closely
match a laplacian distribution of the coeficcients.
Performance Gain:
derflr: 0.15%
hevcmr: 0.675%
Change-Id: I25234244e3bcd94b87c1f77cf682190b61c8ef94
James Zern [Fri, 10 Jun 2016 06:38:31 +0000 (23:38 -0700)]
Revert "Add 1D version of vpx_sum_squares_i16"
This reverts commit
f19700fe52850d051e505ec1b085f25060f7d054 .
This crashes in SSE2/SumSquares2DTest.RandomValues/0 under x86 due to
alignment issues
Change-Id: I135d83ba6a7894c09d7c7a139b7eaf876416b40c
James Zern [Fri, 10 Jun 2016 03:49:28 +0000 (03:49 +0000)]
Merge "Revert "Optimize wedge partition selection."" into nextgenv2
Angie Chiang [Fri, 10 Jun 2016 02:15:06 +0000 (19:15 -0700)]
Revert "Optimize wedge partition selection."
This reverts commit
efda2831e5f758b4f350679b5c55c0b9282449b0 .
This commit causes segmentation fault at SSE2/SumSquares2DTest.RandomValues/0
Change-Id: I171937e4daf6f15323e8206418773deb03bd8c53
Sarah Parker [Thu, 9 Jun 2016 23:29:53 +0000 (16:29 -0700)]
Fix vp9_end_to_end_test for vp10 HBD
This test is failing when no experiments are turned on. PSNR is
31.96 when the threshold is 32.
broken since:
0d6980d Remove swap buffer speed feature
Change-Id: I3c29815b40d5282c37f52f4345b56992f8558b2e
Debargha Mukherjee [Thu, 9 Jun 2016 22:36:07 +0000 (22:36 +0000)]
Merge "Add warped motion config flag" into nextgenv2
Debargha Mukherjee [Thu, 9 Jun 2016 09:22:48 +0000 (02:22 -0700)]
Add warped motion config flag
Change-Id: I4b5e1251dc215073384e168a6f845ae059d6c4f2
Aamir Anis [Thu, 9 Jun 2016 20:57:09 +0000 (20:57 +0000)]
Merge "Updated loop restoration" into nextgenv2
Alex Converse [Thu, 9 Jun 2016 20:50:05 +0000 (13:50 -0700)]
Turn on ActiveMapTest speeds [0,5) with all experiments.
Change-Id: I7da9e6a85648aa69e5e20d825b717d51e3c6809c
Alex Converse [Thu, 10 Mar 2016 23:11:31 +0000 (15:11 -0800)]
Port active map / cyclic refresh fixes to VP10.
Bring commits
575e81f and
3d6b8a6 to VP10. These changes predate
the creation of the active map cyclic refresh test.
BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1224
Change-Id: I3559b6933ffa5649926a4b214e45ed0fae523a25
Debargha Mukherjee [Thu, 9 Jun 2016 11:05:26 +0000 (04:05 -0700)]
Adds higher precision for homography model 3rd row
Also adds a function to integerize a double model.
Change-Id: Ie09b3e165492cf66ab81fe25d4bc2422a5e6defd
Jingning Han [Wed, 8 Jun 2016 21:46:54 +0000 (21:46 +0000)]
Merge "Take out skip_recode speed feature" into nextgenv2
Jingning Han [Wed, 8 Jun 2016 19:45:54 +0000 (19:45 +0000)]
Merge "Remove swap buffer speed feature" into nextgenv2
Jingning Han [Wed, 8 Jun 2016 17:22:52 +0000 (10:22 -0700)]
Add MIN_TX_SIZE definition
Change-Id: I399d601d40827ac383a6687cbeaec59e9a9c63e4
Jingning Han [Wed, 8 Jun 2016 17:04:34 +0000 (10:04 -0700)]
Take out skip_recode speed feature
The assumption doesn't hold true in the current codebase. Remove
this speed feature to simplify the codebase.
Change-Id: I9b69f484c9b7cd612b825047cc5b2fce63ee0af7
Jingning Han [Wed, 8 Jun 2016 16:49:03 +0000 (09:49 -0700)]
Remove swap buffer speed feature
The inter prediction residual can undergo different transform types
during the rate-distortion optimization search. The assumption used
in this speed feature no longer holds true. This commit removes the
related code to clean up the codebase and clear out unit test
failure in higher speed setting.
Change-Id: I7f7cd4df2345ed3e607c9fae75b38cd2dbde0cac
Jingning Han [Tue, 7 Jun 2016 23:44:01 +0000 (23:44 +0000)]
Merge "Add tx type speed feature to recursive transform block partitioning" into nextgenv2
Jingning Han [Tue, 7 Jun 2016 23:42:43 +0000 (23:42 +0000)]
Merge "Rework the tx type speed feature" into nextgenv2
Angie Chiang [Tue, 7 Jun 2016 22:02:28 +0000 (22:02 +0000)]
Merge "Move #if out of TEST_P in vp10_fwd/inv_txfm2d_test.cc" into nextgenv2
Zoe Liu [Tue, 7 Jun 2016 21:34:55 +0000 (21:34 +0000)]
Merge "Fix a RD performance bug in bipredictive frames" into nextgenv2
Alex Converse [Tue, 7 Jun 2016 20:28:25 +0000 (20:28 +0000)]
Merge "Turn ActiveMapTest back on." into nextgenv2
Debargha Mukherjee [Tue, 7 Jun 2016 20:23:44 +0000 (20:23 +0000)]
Merge "Pick up top left mbmi for supertx decode." into nextgenv2
Debargha Mukherjee [Tue, 7 Jun 2016 20:23:20 +0000 (20:23 +0000)]
Merge "Zero segment counter before accumulating." into nextgenv2
Alex Converse [Tue, 7 Jun 2016 16:49:10 +0000 (09:49 -0700)]
Turn ActiveMapTest back on.
If it's creating problems with some experiments, disable it under the
actual conditions where it doesn't work and file a bug.
Change-Id: Iab9f4bfe42ea926d49d371918da25f9a8938a20f
Jingning Han [Tue, 7 Jun 2016 16:40:17 +0000 (09:40 -0700)]
Add tx type speed feature to recursive transform block partitioning
Change-Id: I45440a72b4287d98cbe21b72defc67138a8eb953
Geza Lore [Tue, 7 Jun 2016 18:30:01 +0000 (19:30 +0100)]
Pick up top left mbmi for supertx decode.
This ensures using the correct segment_id downstream in
reconstruct_inter_block.
Change-Id: Ia8b6ec60de51fa2e26c326d3c537abb18aea75ae
Jingning Han [Mon, 6 Jun 2016 21:58:50 +0000 (14:58 -0700)]
Rework the tx type speed feature
This commit re-works the transform type speed feature. It moves
the transform type selection outside of the coding mode loop. This
avoids repeated motion search if the best prediction mode is
chosen as NEWMV. It improves the speed performance for clips that
contain more motion activities.
For mobile_cif at 1000 kbps, this makes the baseline encoding 7%
faster and makes the encoding with dynamic motion vector referencing
scheme enabled 10% faster.
Change-Id: I93e2714b3e461303372c4b66a4134ee212faffd1