1015 Commits

Author SHA1 Message Date
Jingning Han
2a39296d57 Merge "Fix encoder failure in segmentation mode" into nextgenv2 2016-05-02 15:39:12 +00:00
Jingning Han
e729d28c08 Fix encoder failure in segmentation mode
This commit fixes an encoder segment fault in the codebase, when
the segmentation feature is turned on. The issue was introduced in

5cce322 Porting ext_partition experiment from nextgen

Change-Id: Ifb4c06c5a6976114a8bd061d40d0338a136abaaf
2016-04-29 17:59:26 -07:00
Yi Luo
299c5fc202 HBD hybrid transform 8x8 SSE4.1 optimization
- Tx_type: DCT_DCT, DCT_ADST, ADST_DCT, ADST_ADST.
- Update bit-exact unit test against current C version.
- HBD encoder speed improves ~3.8%.

Change-Id: Ie13925ba11214eef2b5326814940638507bf68ec
2016-04-29 17:04:52 -07:00
Debargha Mukherjee
88fe7871be Refactor wedge generation
Change-Id: I2ec4f562e28a4673477e20186f9d6167b24b76b8
2016-04-28 17:51:21 -07:00
Debargha Mukherjee
cf3ee22597 Merge "Make the backward updates work with bitshifts" into nextgenv2 2016-04-28 20:25:53 +00:00
Debargha Mukherjee
e4bf50b9b9 Make the backward updates work with bitshifts
Removes integer divides from backward updates for VP10.
Currently this is put in as part of the entropy experiment.
Coding efficiency change is in the noise level.

Change-Id: I5b3c0ab6169ee6d82d0ca1778e264fd4577cdd32
2016-04-28 11:51:18 -07:00
Debargha Mukherjee
7ff7943455 Brings back near-near compound mode into ext-inter
lowres: improves by 0.1%

Change-Id: I245019916bf47c6e24bc8c3953b86715ab0193c9
2016-04-28 11:34:13 -07:00
Geza Lore
bf93b38561 Fix some mismatches when using ext-interp.
With ext interp, write_switchable_interp_filter calls
vp10_is_interp_needed, which needs access to the reference frame
buffers to check if they are scaled, the ref frame buffer pointer
at this point used to be uninitialized in the encoder resulting in
bitstream syntax mismatch when the encoder/decoder did not read/write
the interp filter element consistently.

Change-Id: Ie0be2a19cbfcb5639a751aa857458e91c23b8fe3
2016-04-28 18:09:54 +01:00
Alex Converse
13c3757067 Merge "buf_ans: Misc cleanup." into nextgenv2 2016-04-27 23:02:39 +00:00
Alex Converse
918a2fd726 Use signed variables in the lookahead.
Wrap around behavior is enforced manually and we use the values in
arithmetic involving negative integers.

Change-Id: I199706b6f3af91f4fb6fe2ef302fbbc6d0cf5785
2016-04-27 16:01:20 -07:00
Alex Converse
d3fe3b2abb Avoid an unsigned overflow in invert_quant
Change-Id: I16a570b2af66b6580d1cd6f8345a25f079009bf4
2016-04-27 15:07:22 -07:00
Hui Su
338c9e704a Merge "ext-intra: completely remove floating point operations" into nextgenv2 2016-04-27 22:00:22 +00:00
Alex Converse
fc838a04be buf_ans: Misc cleanup.
Change-Id: I18a3ef2ee6cdda57abcd27683b30b4e3136182c0
2016-04-27 14:10:15 -07:00
Debargha Mukherjee
bd76fc0492 Merge "Turn skip recode off temporarily for ref-mv" into nextgenv2 2016-04-27 20:43:32 +00:00
hui su
6e39af3697 ext-intra: completely remove floating point operations
No performance changes

Change-Id: Ia489041253423ddf8ebc7e2d41fbfb9e138109f0
2016-04-27 12:08:38 -07:00
Jingning Han
22a68fb047 Merge "Fix compound mv costing for ref-mv." into nextgenv2 2016-04-27 17:15:16 +00:00
Debargha Mukherjee
a671241a6e Turn skip recode off temporarily for ref-mv
To fix tests in VP10/AltRefForcedKeyTestLarge.Frame1IsKey/*

Change-Id: I2f4f9fea515c9935d57006b709a9dd524f174b25
2016-04-27 09:55:30 -07:00
Debargha Mukherjee
bc982cc994 Merge "Initialize dummy variables." into nextgenv2 2016-04-27 16:36:24 +00:00
Geza Lore
264d5c446e Fix compound mv costing for ref-mv.
I believe this is necessary for computing the correct rate,
when not doing joint_motion_search.

Change-Id: I7634d6d7a5e6f0a6998edb4d577dd047d80df3c8
2016-04-27 13:37:29 +01:00
Geza Lore
d29ec48504 Initialize dummy variables.
Valgrind flags these up as needed by handle_inter_mode.
Initializing fixes some assertion failures in the unit tests with
only ref-mv enabled.

Change-Id: I4d56c356692745dbecd9f790cdbb8dbfbaf72d55
2016-04-27 13:35:12 +01:00
Geza Lore
4e177393f0 Fix ext-tile without ext-partition.
Default case (when ext-partition was not configured) was incorrect
in encoder tile size initialization.

BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1197

Change-Id: Ibe57cb1dc16b9fa300573816fc16d2d2f6849fc6
2016-04-27 11:14:48 +01:00
Yue Chen
02e941d371 Merge "Remove double counting for mv costs" into nextgenv2 2016-04-26 21:40:13 +00:00
Jingning Han
2f2448aec9 Merge "Rework motion vector precision limit" into nextgenv2 2016-04-26 21:31:10 +00:00
Yue Chen
34177e673d Remove double counting for mv costs
The bug is introduced by commit 1a0352d, in which mv costs are
counted twice in joint_motion_search() in ext_inter experiment.

Change-Id: Ibace453df999d3c2e781d73f1f0912038fee2d4e
2016-04-26 13:01:52 -07:00
Hui Su
3f7a709676 Merge "ext-intra: get rid of some floating operations." into nextgenv2 2016-04-26 18:53:33 +00:00
Jingning Han
8678ab4c55 Rework motion vector precision limit
This commit enables 1/8 luma component motion vector precision
for all motion vector cases. It improves the compression performance
of lowres by 0.13% and hdres by 0.49%.

Change-Id: Iccfc85e8ee1c0154dfbd18f060344f1e3db5dc18
2016-04-26 10:14:26 -07:00
Hui Su
1e93a3e64e Merge "Keep track of zcoeff_blk in tx size/type search" into nextgenv2 2016-04-26 16:41:50 +00:00
hui su
ad50c226e6 ext-intra: get rid of some floating operations.
No performance changes.

Change-Id: Idd4043090fec09e57520bc970ed2e39e6f7e1a5e
2016-04-25 14:44:42 -07:00
Debargha Mukherjee
022da62579 Merge "Clear X87 register state before using double." into nextgenv2 2016-04-25 21:42:23 +00:00
Yi Luo
333ff883e1 Merge "HBD hybrid transform 4x4 SSE4.1 optimization" into nextgenv2 2016-04-25 19:43:35 +00:00
Geza Lore
23c4116ebb Clear X87 register state before using double.
MMX and X87 floating point instructions cannot be mixed freely on
the 32 bit x86 architecture.

This fixes a lot of unit tests in the 32bit build with
--enable-ext-intra.

BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1196

Change-Id: I0e1c3565f4b9cb4fc2d716e94d9c40e68b36fac8
2016-04-25 10:30:20 -07:00
Alex Converse
d4fe243cdf Merge "Raise the probability resolution for rANS tokens to 10-bits per symbol" into nextgenv2 2016-04-25 17:11:16 +00:00
Yi Luo
a4593f17ca HBD hybrid transform 4x4 SSE4.1 optimization
- Optimization on tx_type: DCT_DCT, DCT_ADST, ADST_DCT, ADST_ADST.
- Overall encoder speed improves ~4.5%-6%.
- Update bit-exact unit test against current C version.

Change-Id: If751c030612245b1c2470200c9570cf40d655504
2016-04-25 09:53:09 -07:00
Jingning Han
221c09aa99 Merge "Refactor sub-pixel motion search" into nextgenv2 2016-04-25 16:12:51 +00:00
James Zern
0aa6435c45 vp10/rdopt: quiet unused variable warning
when CONFIG_REF_MV and CONFIG_EXT_INTER are enabled

Change-Id: I17fa2b5fe0e1878333099cc5fa2b1ee36636b4d3
2016-04-23 16:59:45 +00:00
Jingning Han
cdf989adb7 Silence compiler above-boundary warnings
Change-Id: I6d806f92e8d38d5b0b01bc8e0fd97bd8839c84df
2016-04-22 15:49:34 -07:00
Alex Converse
1f57aa38cd Raise the probability resolution for rANS tokens to 10-bits per symbol
Change-Id: I397b5a9371c85d1df401d261143c985623e9def6
2016-04-22 15:48:11 -07:00
Jingning Han
77d451ecca Refactor sub-pixel motion search
Unify the rate cost used in the motion estimation process.

Change-Id: I8e52ca9f29eee3469553433302b62fb02a038919
2016-04-22 22:27:09 +00:00
Jingning Han
0dccc85c98 Replace left shift with multiplications
This avoids the potential risk in left shift of negative numbers.

Change-Id: I7aecb499ee6ce7342b172adc4741de5c6c107a24
2016-04-22 20:56:02 +00:00
Jingning Han
3f6ec144e5 Fix an enc/dec mismatch issue in ext-inter experiment
This commit fixes an encoding decision process issue that could
trigger enc/dec mismatch in the ext-inter experiment.

Change-Id: I6f10d1fd2fd1aa04e51df04c39a65cf72ac66c42
2016-04-22 20:49:29 +00:00
Yi Luo
cf7f00691f Change hybrid transform function argument from TXFM_2D_CFG* to int
Unit test shows manually developed SSE4.1 code would performs ~30%
  better if TXFM_2D_CFG configuration is set in lower level. This
  change only updates function signature. There is no performance
  impact.

Change-Id: I62692bd50a21ffc8a944bbd6c155c0a2020ad77b
2016-04-21 18:37:21 -07:00
Alex Converse
8f2fa04181 Unbreak the non-var_tx build.
Change-Id: I76cc3d88122de42f035fbf6508bdf3fd7c995012
2016-04-21 13:27:19 -07:00
Debargha Mukherjee
53968c3917 Merge "Fix uninitialized blk_skip for VAR TX." into nextgenv2 2016-04-21 19:56:17 +00:00
Alex Converse
67058089b4 Merge "Move ZERO_TOKEN into the ANS coef tokenset." into nextgenv2 2016-04-21 18:08:04 +00:00
Alex Converse
fcea1485bb Merge "Store ANS token CDFs in the FRAME_CONTEXT rather than in a global table." into nextgenv2 2016-04-19 23:29:14 +00:00
Alex Converse
3829cd2f2f Move ZERO_TOKEN into the ANS coef tokenset.
Change-Id: I87943e027437543ab31fa3ae1aa8b2de3a063ae5
2016-04-19 15:29:47 -07:00
Jingning Han
1a0352d18e Merge "Handle zero motion vector residual" into nextgenv2 2016-04-19 21:20:08 +00:00
hui su
7dffb43267 Keep track of zcoeff_blk in tx size/type search
Prevent potential problems when per transform block
zero forcing is re-enabled (a To-Do).

Change-Id: I03b0ab2a86d88058441f2ca18994cfd2e6329898
2016-04-19 11:45:11 -07:00
hui su
ad59b08f76 Adjust optimize_b RD parameters
Coding gain:
lowres  0.44%
midres  0.24%
hdres   0.32%

Change-Id: Ie558203b2b2bf5c16cd49b114df3d696c4f35049
2016-04-19 09:54:08 -07:00
hui su
e43c21112d Enable optimize_b for intra blocks
Coding gain:
lowres  0.05%
midres  0.10%
hdres   0.18%

Change-Id: I508b150c02588f911a8ddddfe73c770f0819fe10
2016-04-19 09:50:45 -07:00