platform/upstream/libvpx.git
11 years agosegmentation: make local symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:39:44 +0000 (21:39 -0700)]
segmentation: make local symbols static.

Change-Id: Ia30924dc095663b04794902522308d822c1a5ad0

11 years agoencodeframe: make local symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:38:07 +0000 (21:38 -0700)]
encodeframe: make local symbols static.

Change-Id: I77bc38b53428ff3e6456b5bc3516418892a58c1e

11 years agoentropymode: put print_mode_contexts under #ifdef MODE_STATS.
Ronald S. Bultje [Tue, 30 Oct 2012 04:35:35 +0000 (21:35 -0700)]
entropymode: put print_mode_contexts under #ifdef MODE_STATS.

This is the condition under which it is called in onyx_if.c. Also remove
the unused function print_mv_ref_cts().

Change-Id: I51ea3720d46f86d136e2215e01cf9d6c7dfc41ea

11 years agossim: remove unused function dssim().
Ronald S. Bultje [Tue, 30 Oct 2012 04:30:09 +0000 (21:30 -0700)]
ssim: remove unused function dssim().

Change-Id: I5898f64da77e43a3860a77f3bc49d64b725c278c

11 years agodetokenize: make local symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:28:08 +0000 (21:28 -0700)]
detokenize: make local symbols static.

Change-Id: Ie80d13ce095faa8bca21bcdca3ca8249d9abc398

11 years agofirstpass: make local symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:26:53 +0000 (21:26 -0700)]
firstpass: make local symbols static.

Change-Id: Iee0f707abbfe427b10058f6cf3be21b89d6a6e65

11 years agobitstream: give exported symbols a vp9_ prefix; make local symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:25:22 +0000 (21:25 -0700)]
bitstream: give exported symbols a vp9_ prefix; make local symbols static.

Change-Id: I3aa5b515c1eae19ae40ba9808d133590c95f7d13

11 years agoalloccommon: give exported symbols a vp9_ prefix.
Ronald S. Bultje [Tue, 30 Oct 2012 04:24:12 +0000 (21:24 -0700)]
alloccommon: give exported symbols a vp9_ prefix.

Also include the correct header in ratectrl.c so it picks up these
function prototypes.

Change-Id: Ic51b13119f1625f0691917f2713b23d0249dece7

11 years agoMake implicit_segmentation-related code an experiment.
Ronald S. Bultje [Tue, 30 Oct 2012 04:15:42 +0000 (21:15 -0700)]
Make implicit_segmentation-related code an experiment.

This way, the code is not compiled in by default, thus decreasing
overall binary size.

Change-Id: I85cac8f5a22a51a7d99c820ef6d6ed179d4106a0

11 years agodecodframe: make local symbol static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:08:59 +0000 (21:08 -0700)]
decodframe: make local symbol static.

Change-Id: Id55f3cd3c2f5857dbfd78703452740dfd63cdf41

11 years agotokenize: make local symbols static; give exported symbols a vp9_ prefix.
Ronald S. Bultje [Tue, 30 Oct 2012 04:07:40 +0000 (21:07 -0700)]
tokenize: make local symbols static; give exported symbols a vp9_ prefix.

Change-Id: I23aa5547b1e74f31327258dda0e330aae7ca1d35

11 years agoencodemb: make locally used functions static.
Ronald S. Bultje [Tue, 30 Oct 2012 04:02:36 +0000 (21:02 -0700)]
encodemb: make locally used functions static.

Change-Id: I8341a19cb4f2234841a40fcbf560315f079b0108

11 years agoonyx_if: remove unused function, and make locally used functions static.
Ronald S. Bultje [Tue, 30 Oct 2012 01:20:32 +0000 (18:20 -0700)]
onyx_if: remove unused function, and make locally used functions static.

Change-Id: Ica08c2c9cda9623389dde3f276dfed77c36e3a11

11 years agoreconintra: make locally used symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 01:15:04 +0000 (18:15 -0700)]
reconintra: make locally used symbols static.

Also remove them from the header file.

Change-Id: I38182f70268fd47985fb1d8d87cdeb207f67e475

11 years agordopt: make locally used symbols static.
Ronald S. Bultje [Tue, 30 Oct 2012 01:04:33 +0000 (18:04 -0700)]
rdopt: make locally used symbols static.

Change-Id: Id33c448084ca5c1c0525a05c3c4f35d6b0805fbe

11 years agoratectrl: make locally used symbol static.
Ronald S. Bultje [Tue, 30 Oct 2012 01:02:04 +0000 (18:02 -0700)]
ratectrl: make locally used symbol static.

Change-Id: I4f317f413fbbcc59f84ff77899ea208fb0dccb2a

11 years agombgraph: remove unused function, and make local function static.
Ronald S. Bultje [Tue, 30 Oct 2012 00:59:45 +0000 (17:59 -0700)]
mbgraph: remove unused function, and make local function static.

Change-Id: I295c7bcf13b630cb9395585ede7311d85d785f25

11 years agoFix superblock experiment.
Ronald S. Bultje [Tue, 30 Oct 2012 00:58:18 +0000 (17:58 -0700)]
Fix superblock experiment.

Also merge some duplicate code related to the superblock experiment
in the RD loop.

Change-Id: Ic93f1d4d1ed81220fd7ecf6e65da2821a215b2de

11 years agoMerge "Name space clean up." into experimental
Ronald S. Bultje [Mon, 29 Oct 2012 21:41:47 +0000 (14:41 -0700)]
Merge "Name space clean up." into experimental

11 years agoMerge "A clean up of the down_copy function" into experimental
Ronald S. Bultje [Mon, 29 Oct 2012 21:41:39 +0000 (14:41 -0700)]
Merge "A clean up of the down_copy function" into experimental

11 years agoremove fdct invoke macros
Jim Bankoski [Mon, 29 Oct 2012 18:25:56 +0000 (11:25 -0700)]
remove fdct invoke macros

Remove the fdct invoke macro calls

Change-Id: Ica2431c655819fa012133ee7abc75a16761e5fd6

11 years agoFixes invoke macro commit...
Jim Bankoski [Mon, 29 Oct 2012 15:58:05 +0000 (08:58 -0700)]
Fixes invoke macro commit...

Change-Id: Ic186269b51ffb3c344635bfa9825bed25be2c5ae

11 years agoName space clean up.
Paul Wilkins [Mon, 29 Oct 2012 13:44:18 +0000 (13:44 +0000)]
Name space clean up.

Preparation for project restructuring.
Added vp9_ prefix on some function names that have global scope.
Added static declaration on some that dont.

Change-Id: If072f78b4300e8c17cfeed82c5d17b59946dcc5e

11 years agoInconsistent reference to MacroblockD data structure.
Paul Wilkins [Mon, 29 Oct 2012 10:36:11 +0000 (10:36 +0000)]
Inconsistent reference to MacroblockD data structure.

Cleaned up some inconsistent references using both xd-> and
x->e_mbd. to access the same data structure in the same function.

Change-Id: Ieb496fa22bf1feec6aa7bc70b941ea4f16e0f8b5

11 years agoMerge "invoke macro removal encodemb" into experimental
Paul Wilkins [Mon, 29 Oct 2012 12:42:58 +0000 (05:42 -0700)]
Merge "invoke macro removal encodemb" into experimental

11 years agoinvoke macro removal encodemb
Jim Bankoski [Sun, 28 Oct 2012 17:38:23 +0000 (10:38 -0700)]
invoke macro removal encodemb

Change-Id: I321280abcf48f3dc16e194d29bde2bd3baec6006

11 years agoA clean up of the down_copy function
Deb Mukherjee [Fri, 26 Oct 2012 22:54:37 +0000 (15:54 -0700)]
A clean up of the down_copy function

Change-Id: I0c689fb44e9e91539c450d26672d7a1d92e92900

11 years agoMerge "Faster 8t filtering" into experimental
Scott LaVarnway [Fri, 26 Oct 2012 21:53:50 +0000 (14:53 -0700)]
Merge "Faster 8t filtering" into experimental

11 years agoMerge "Extend edges correctly during actual frame encoding also." into experimental
Ronald S. Bultje [Fri, 26 Oct 2012 18:05:51 +0000 (11:05 -0700)]
Merge "Extend edges correctly during actual frame encoding also." into experimental

11 years agoMerge "Explicit MV reference experiment." into experimental
Paul Wilkins [Fri, 26 Oct 2012 16:42:37 +0000 (09:42 -0700)]
Merge "Explicit MV reference experiment." into experimental

11 years agoExplicit MV reference experiment.
Paul Wilkins [Thu, 25 Oct 2012 12:58:21 +0000 (13:58 +0100)]
Explicit MV reference experiment.

Coding and costing of mv reference signal.

Issues in updating MV ref with COMPANDED_MVREF_THRESH
to be resolved. Ideally the MV precision should be defined based
on absolute MV magnitude not as now the MV ref magnitude.

Update to mv counts moved into bitstream.c because otherwise
if the motion reference is changed at the last minute the encoder
and decoder get out of step in terms of the counts used to update
entropy probs.

Code working on a few test clips but no results yet re benefit vs
signaling cost and no tuning of red loop to test lower cost alternatives
based on the available reference values.

Patch 3. Added check to make sure we don't pick a reference
that would give rise to an uncodeable / out of range residual.

Patch 6-7: Attempt to rebase. OK to submit but best to leave flag off for now.

Patch 9. Remove print no longer needed.

Change-Id: I1938c2ffe41afe6d3cf6ccc0cb2c5d404809a712

11 years agoExtend edges correctly during actual frame encoding also.
Ronald S. Bultje [Fri, 26 Oct 2012 01:09:02 +0000 (18:09 -0700)]
Extend edges correctly during actual frame encoding also.

Should fix B_PRED-related encoder/decoder mismatches.

Change-Id: I16f808dffd19094e02e8562ba58cc1016155ce93

11 years agoFaster 8t filtering
Scott LaVarnway [Fri, 26 Oct 2012 00:24:50 +0000 (17:24 -0700)]
Faster 8t filtering

Quickly modified the ssse3 sixtap filters to support eight taps.  For the test
clip used, a 23+% boost in decoder performance was seen.  We can
revisit later and improve further.

Change-Id: I5f59860459e80d6fa23e6cc0fd91296a969f5240

11 years agoMerge "Always extend macroblock edges in the decoder." into experimental
Debargha Mukherjee [Thu, 25 Oct 2012 20:52:29 +0000 (13:52 -0700)]
Merge "Always extend macroblock edges in the decoder." into experimental

11 years agoAlways extend macroblock edges in the decoder.
Ronald S. Bultje [Thu, 25 Oct 2012 19:46:26 +0000 (12:46 -0700)]
Always extend macroblock edges in the decoder.

This fixes edge extension if SB [0,1] doesn't use B_PRED but [1,1] does.

Change-Id: I48354a9dba0de16472938824f5e6db87ce61979f

11 years agoAdded sse2 instrinsic version of vp8_sad3x16
Scott LaVarnway [Mon, 22 Oct 2012 23:38:58 +0000 (16:38 -0700)]
Added sse2 instrinsic version of vp8_sad3x16

1.6% boost in decoder performance for the clip used.

Change-Id: I91f3c4573fd3d10afbf18930f279af7ae2223e3a

11 years agoMerge "Added sse2 instrinsic version of vp8_sad16x3" into experimental
Scott LaVarnway [Thu, 25 Oct 2012 19:19:26 +0000 (12:19 -0700)]
Merge "Added sse2 instrinsic version of vp8_sad16x3" into experimental

11 years agoAdded sse2 instrinsic version of vp8_sad16x3
Scott LaVarnway [Mon, 22 Oct 2012 23:16:04 +0000 (16:16 -0700)]
Added sse2 instrinsic version of vp8_sad16x3

3.7% boost in decoder performance for the clip used.

Change-Id: I74f28486a9352b472b36e21b5eaf30eff35e9199

11 years agoFix 4x4 intra prediction on the edge of SB rows.
Ronald S. Bultje [Thu, 25 Oct 2012 19:05:37 +0000 (12:05 -0700)]
Fix 4x4 intra prediction on the edge of SB rows.

Change-Id: I87d571008c73f0a8514e0a864405aadb82fd1bc0

11 years agoFix yet another typo in splitmv/tx_select code.
Ronald S. Bultje [Thu, 25 Oct 2012 16:17:47 +0000 (09:17 -0700)]
Fix yet another typo in splitmv/tx_select code.

Change-Id: I6a28cc87af275fc267b3cc8d90d642dcc870c249

11 years agoGeneralize coefficient tokenizing.
Ronald S. Bultje [Thu, 25 Oct 2012 16:14:21 +0000 (09:14 -0700)]
Generalize coefficient tokenizing.

Change-Id: Ie33f7d0f90d0e5862be975d676a700d06f79eec6

11 years agoGeneralize EOB token stuffing.
Ronald S. Bultje [Wed, 24 Oct 2012 23:59:00 +0000 (16:59 -0700)]
Generalize EOB token stuffing.

Change-Id: Ibb51f3f8c3575d4621890220b20887a13a46a04e

11 years agoMerge "Fix typo in splitmv/tx_select code." into experimental
Ronald S. Bultje [Wed, 24 Oct 2012 22:21:05 +0000 (15:21 -0700)]
Merge "Fix typo in splitmv/tx_select code." into experimental

11 years agoFix typo in splitmv/tx_select code.
Ronald S. Bultje [Wed, 24 Oct 2012 20:03:51 +0000 (13:03 -0700)]
Fix typo in splitmv/tx_select code.

Change-Id: I2823043634b0fa617c2715d63ff18595b7209287

11 years agoFix bad merge of coef_probs update
John Koleszar [Wed, 24 Oct 2012 19:59:22 +0000 (12:59 -0700)]
Fix bad merge of coef_probs update

Stray CONFIG_HYBRIDTRANSFORM caused decode mismatch.

Change-Id: I1d77bad9b366f7a6ceb3cb45269ed5badb5c4c53

11 years agocoef_probs: remove duplicate read/update code
John Koleszar [Fri, 19 Oct 2012 22:35:36 +0000 (15:35 -0700)]
coef_probs: remove duplicate read/update code

Refactor per-transform copy & paste into a common function
update_coef_probs_common() and read_coef_probs_common(). The dry-run and
bit-writing loops in the encoder are still obvious candidates to be made
common, but they start to diverge a bit in the next commit, so are left
as-is for now.

Change-Id: I896bd3f4a073a6296eab7e92463fef79d8c6c08c

11 years agoMerge "Fix "_FORTIFY_SOURCE" redefined warning" into experimental
John Koleszar [Wed, 24 Oct 2012 16:36:40 +0000 (09:36 -0700)]
Merge "Fix "_FORTIFY_SOURCE" redefined warning" into experimental

11 years agoFix "_FORTIFY_SOURCE" redefined warning
Yunqing Wang [Wed, 24 Oct 2012 16:14:36 +0000 (09:14 -0700)]
Fix "_FORTIFY_SOURCE" redefined warning

On Ubuntu 12.04, we got the following warning message:
<command-line>:0:0: warning: "_FORTIFY_SOURCE" redefined
 [enabled by default]
<built-in>:0:0: note: this is the location of the previous definition

This was already fixed in VP8 configure file. Did the same change in
experimental branch to stop this warning.

Change-Id: Id162e5fd8841585ae806df6560b2f7536ea307c0

11 years agoMerge "Fixed the MSVC compiling error with correct cast" into experimental
Yaowu Xu [Tue, 23 Oct 2012 21:31:38 +0000 (14:31 -0700)]
Merge "Fixed the MSVC compiling error with correct cast" into experimental

11 years agoFixed the MSVC compiling error with correct cast
Scott LaVarnway [Tue, 23 Oct 2012 20:35:13 +0000 (13:35 -0700)]
Fixed the MSVC compiling error with correct cast

Change-Id: Ia904f4ec72500d29f1361ce305d8f3231e592f47

11 years agoMerge "Change eob[] array type in splitmv RD loop to a regular integer." into experim...
pascal massimino [Tue, 23 Oct 2012 21:00:01 +0000 (14:00 -0700)]
Merge "Change eob[] array type in splitmv RD loop to a regular integer." into experimental

11 years agoRevert "make the instrinsic code build with MS compilers"
Yaowu Xu [Tue, 23 Oct 2012 20:00:23 +0000 (13:00 -0700)]
Revert "make the instrinsic code build with MS compilers"

This reverts commit b0e3ca126189123ddec27ebba6aa62290e64adb6.

Change-Id: I9c5aa463461b187160ad01fbc1795ae4f5263b2c

11 years agoMerge "changed to avoid code confusion" into experimental
Yaowu Xu [Tue, 23 Oct 2012 18:54:25 +0000 (11:54 -0700)]
Merge "changed to avoid code confusion" into experimental

11 years agoMerge "make the instrinsic code build with MS compilers" into experimental
Yaowu Xu [Tue, 23 Oct 2012 18:54:07 +0000 (11:54 -0700)]
Merge "make the instrinsic code build with MS compilers" into experimental

11 years agomake the instrinsic code build with MS compilers
Scott LaVarnway [Tue, 23 Oct 2012 16:39:22 +0000 (09:39 -0700)]
make the instrinsic code build with MS compilers

to enable build under windows/MSVC

Change-Id: Ida41cc5b3c8d0fec9512c2f5c5feb64e07b44805

11 years agoChange eob[] array type in splitmv RD loop to a regular integer.
Ronald S. Bultje [Tue, 23 Oct 2012 16:23:23 +0000 (09:23 -0700)]
Change eob[] array type in splitmv RD loop to a regular integer.

Change-Id: I240d6b50069fd3f35cc4fed2f4507796f0ef25e9

11 years agoMerging in the Switchable interp experiment
Deb Mukherjee [Mon, 22 Oct 2012 21:43:01 +0000 (14:43 -0700)]
Merging in the Switchable interp experiment

There is a macro DEFAULT_INTERP_FILTER defined in encoder/onyx_if.c that
is set as EIGHTTAP for now - so SWITCHABLE is not really used. Ideally,
this should be SWITCHABLE but that would make the encoder quite a bit slower.
We will change the default filter to SWITCHABLE once we find a faster way to
search for switchable filters.

Change-Id: Iee91832cdc07e6e14108d9b543130fdd12fc9874

11 years agoMerge "Change entropy context assignment style from *(x + y) = z to x[y] = z." into...
Ronald S. Bultje [Tue, 23 Oct 2012 05:04:28 +0000 (22:04 -0700)]
Merge "Change entropy context assignment style from *(x + y) = z to x[y] = z." into experimental

11 years agoMerging in the TX_SELECT experiment
Deb Mukherjee [Mon, 22 Oct 2012 23:52:28 +0000 (16:52 -0700)]
Merging in the TX_SELECT experiment

Change-Id: I0d306b0062be59bc785b65882691c956c8b5e7c8

11 years agoMerge of the newmventropy experiment
Deb Mukherjee [Mon, 22 Oct 2012 19:19:00 +0000 (12:19 -0700)]
Merge of the newmventropy experiment

Removes the old mv encoding code.

Change-Id: I1026d48cc5ac15ccb776f98463e929254c3dc7da

11 years agoChange entropy context assignment style from *(x + y) = z to x[y] = z.
Ronald S. Bultje [Mon, 22 Oct 2012 21:06:52 +0000 (14:06 -0700)]
Change entropy context assignment style from *(x + y) = z to x[y] = z.

Change-Id: Ibc4425bb6e01606751021e57c93106c365a2da21

11 years agochanged to avoid code confusion
Yaowu Xu [Mon, 22 Oct 2012 20:58:26 +0000 (13:58 -0700)]
changed to avoid code confusion

The previous double if and else code snippets were rather confusing.

Change-Id: Id1b6152fa0e471beb9b20407aa406e109c1471e3

11 years agoMerge "Added rtcd support vp8_sad16x3 and vp8_sad3x16" into experimental
Scott LaVarnway [Mon, 22 Oct 2012 20:57:03 +0000 (13:57 -0700)]
Merge "Added rtcd support vp8_sad16x3 and vp8_sad3x16" into experimental

11 years agoAdded rtcd support vp8_sad16x3 and vp8_sad3x16
Scott LaVarnway [Mon, 22 Oct 2012 20:45:42 +0000 (13:45 -0700)]
Added rtcd support vp8_sad16x3 and vp8_sad3x16

Change-Id: I5bca7b7a4b230082d36ac6fb84db84137ad177d7

11 years agoMerge changes I02e7f64a,Ide954b00,Idc8b5977 into experimental
Ronald S. Bultje [Mon, 22 Oct 2012 19:54:39 +0000 (12:54 -0700)]
Merge changes I02e7f64a,Ide954b00,Idc8b5977 into experimental

* changes:
  Fix another typo in 4x4-transform-for-i8x8-intra-pred coeff contexts.
  8x8 transform support in splitmv.
  Use SPLITMV_PARTITIONING instead of a plain integer type.

11 years agoMerge "Merging the hybrid transform experiments" into experimental
Deb Mukherjee [Mon, 22 Oct 2012 19:18:46 +0000 (12:18 -0700)]
Merge "Merging the hybrid transform experiments" into experimental

11 years agoMerging the hybrid transform experiments
Deb Mukherjee [Mon, 22 Oct 2012 18:55:29 +0000 (11:55 -0700)]
Merging the hybrid transform experiments

Change-Id: I99f1982b30a630a9a070a8326d83b34a33cba14c

11 years agoFix another typo in 4x4-transform-for-i8x8-intra-pred coeff contexts.
Ronald S. Bultje [Mon, 22 Oct 2012 18:55:06 +0000 (11:55 -0700)]
Fix another typo in 4x4-transform-for-i8x8-intra-pred coeff contexts.

Change-Id: I02e7f64a7c99f6c222b9ef21341b0e801d646810

11 years ago8x8 transform support in splitmv.
Ronald S. Bultje [Mon, 22 Oct 2012 18:49:00 +0000 (11:49 -0700)]
8x8 transform support in splitmv.

For splitmv, where partitioning is 8x16, 16x8 or 8x8, this patch
uses the 8x8 transform (instead of the 4x4) if txfm_mode is
ALLOW_8X8 or ALLOW_16X16. For TX_MODE_SELECT, splitmv can indicate
which of the 2 transform sizes (4x4 or 8x8) it wants to use.

Gains (with hybridtx4x4/8x8/16x16 and tx_select experiments
enabled) on derf: +0.9%, HD: +0.4%, STD/HD: +0.8% (SSIM or overall
PSNR, both metrics show similar improvements).

Change-Id: Ide954b000b415548ed92a7ac78e24f36e60fcf06

11 years agoMerge "Re-enable specialized sse2 block loop filter functions" into experimental
Scott LaVarnway [Mon, 22 Oct 2012 18:39:08 +0000 (11:39 -0700)]
Merge "Re-enable specialized sse2 block loop filter functions" into experimental

11 years agoMerge "sse2 intrinsic version of vp8_mbloop_filter_vertical_edge()" into experimental
Scott LaVarnway [Mon, 22 Oct 2012 18:38:37 +0000 (11:38 -0700)]
Merge "sse2 intrinsic version of vp8_mbloop_filter_vertical_edge()" into experimental

11 years agoMerge "sse2 intrinsic version of vp8_mbloop_filter_horizontal_edge()" into experimental
Scott LaVarnway [Mon, 22 Oct 2012 18:38:21 +0000 (11:38 -0700)]
Merge "sse2 intrinsic version of vp8_mbloop_filter_horizontal_edge()" into experimental

11 years agoUse SPLITMV_PARTITIONING instead of a plain integer type.
Ronald S. Bultje [Mon, 22 Oct 2012 18:25:48 +0000 (11:25 -0700)]
Use SPLITMV_PARTITIONING instead of a plain integer type.

This can be used to distinguish between 16x8, 8x16, 8x8 and 4x4
partitioning modes when choosing splitmv as a MB mode.

Change-Id: Idc8b59772e1a80ccc4ad44d63c5c2ec3fc061a3c

11 years agoMerge "Invert SB probability." into experimental
Ronald S. Bultje [Mon, 22 Oct 2012 18:18:01 +0000 (11:18 -0700)]
Merge "Invert SB probability." into experimental

11 years agoMerge "Remove variance vtable from rtcd" into experimental
Jim Bankoski [Mon, 22 Oct 2012 18:12:07 +0000 (11:12 -0700)]
Merge "Remove variance vtable from rtcd" into experimental

11 years agoInvert SB probability.
Ronald S. Bultje [Mon, 22 Oct 2012 18:05:56 +0000 (11:05 -0700)]
Invert SB probability.

It currently counts the probability that the branch is true, but it
should count the probability that the branch is false.

Change-Id: I963825da2e7a7ed3a613eb23ffd085e427dc36e5

11 years agoMerge "Fix use of wrong context in i8x8 coefficient cost calculation." into experimental
Ronald S. Bultje [Mon, 22 Oct 2012 18:03:25 +0000 (11:03 -0700)]
Merge "Fix use of wrong context in i8x8 coefficient cost calculation." into experimental

11 years agoRemove variance vtable from rtcd
Jim Bankoski [Mon, 22 Oct 2012 03:47:57 +0000 (20:47 -0700)]
Remove variance vtable from rtcd

Change-Id: Idd2722a538423b451e1e3495f89a7141480493d6

11 years agoMerge "Allow B_VL_PRED & B_LD_PRED modes with Superblocks" into experimental
Deb Mukherjee [Sun, 21 Oct 2012 22:35:24 +0000 (15:35 -0700)]
Merge "Allow B_VL_PRED & B_LD_PRED modes with Superblocks" into experimental

11 years agoAllow B_VL_PRED & B_LD_PRED modes with Superblocks
Deb Mukherjee [Fri, 19 Oct 2012 22:12:12 +0000 (15:12 -0700)]
Allow B_VL_PRED & B_LD_PRED modes with Superblocks

Allows B_VL_PRED & B_LD_PRED modes to be used for all blocks
within a MB in B_PRED mode. These modes were temporarily
disabled with super-block coding.

Change-Id: I973b9bdb82c3da5f12d7cc963162a28805f25303

11 years agoFix use of wrong context in i8x8 coefficient cost calculation.
Ronald S. Bultje [Sat, 20 Oct 2012 05:05:50 +0000 (22:05 -0700)]
Fix use of wrong context in i8x8 coefficient cost calculation.

Change-Id: Ica58d0c7d8addb5ddd5d6e1fdea8f0f96e110862

11 years agoRe-enable specialized sse2 block loop filter functions
Scott LaVarnway [Fri, 19 Oct 2012 23:02:50 +0000 (16:02 -0700)]
Re-enable specialized sse2 block loop filter functions

vp8_loop_filter_bh and vp8_loop_filter_bv

Change-Id: I0a77d38c701a6cce96cb8371c7f3ab7e2e7f878c

11 years agosse2 intrinsic version of vp8_mbloop_filter_vertical_edge()
Scott LaVarnway [Fri, 19 Oct 2012 22:52:12 +0000 (15:52 -0700)]
sse2 intrinsic version of vp8_mbloop_filter_vertical_edge()

First sse2 version of vp8_mbloop_filter_vertical_edge().  For now,
intrinsics are being used until the bitstream is finalized.  This function
will be revisited later for further performance improvements.

For the test clip used, a 34+% decoder performance improvement
was seen.  This will vary depending on material.

Change-Id: I455b438bc8d8af76cf7533ac42eda5f689b21f7c

11 years agocalculate probs consistently
John Koleszar [Thu, 18 Oct 2012 23:27:30 +0000 (16:27 -0700)]
calculate probs consistently

There were several different methods for calculating bitstream
probabilities in use. Consolodate these into a pair of functions,
get_prob() and get_binary_prob().

Change-Id: I5534f517f74027fee16d89c9baefaafea8156b2f

11 years agolint-hunks: better support for working tree
John Koleszar [Thu, 18 Oct 2012 21:34:53 +0000 (14:34 -0700)]
lint-hunks: better support for working tree

When run with no arguments, report warnings in the diff between the
working tree and HEAD. With arguments, report warnings in the diff
between the named commit and its parents.

Change-Id: Ie10dcdecb303edf8af51bad645cc11206a1fc26b

11 years agoRemove bc, bc2 from pbi,cpi,xd
John Koleszar [Wed, 17 Oct 2012 23:47:38 +0000 (16:47 -0700)]
Remove bc, bc2 from pbi,cpi,xd

Pass the bool coder to be used explicitly. This avoids cases where two
different bool coders can be addressed from the same function. Also be
more consistent with bool coder variable naming, start to standardize
on 'bc'.

Change-Id: I1c95e2fdbe24ebe8c0f84924daa1728e3b054a31

11 years agoSome cleanups and fixes.
Deb Mukherjee [Mon, 15 Oct 2012 23:41:41 +0000 (16:41 -0700)]
Some cleanups and fixes.

Separates the logic on transform type selection previously spread out
over a number of files into a separate function. Currently the tx_type
field in b_mode_info is not used, but still left in there to eventually
use for signaling the transform type in the bitstream.

Also, now for tx_type = DCT_DCT, the regular integer DCT is used, as
opposed to the floating point DCT used in conjuction with hybrid
transform.

Results change somewhat due to the transform change, but are within
reasonable limits. The hd/std-hd sets are slightly up, while derf/yt
are slightly down.

Change-Id: I5776840c2239ca2da31ca6cfd7fd1148dc5f9e0f

11 years agoMerge "Fix encode breakout skip handling." into experimental
Ronald S. Bultje [Thu, 18 Oct 2012 22:21:30 +0000 (15:21 -0700)]
Merge "Fix encode breakout skip handling." into experimental

11 years agoMerge "Detect skippable blocks in splitmv RD loop case." into experimental
Ronald S. Bultje [Thu, 18 Oct 2012 22:21:14 +0000 (15:21 -0700)]
Merge "Detect skippable blocks in splitmv RD loop case." into experimental

11 years agosse2 intrinsic version of vp8_mbloop_filter_horizontal_edge()
Scott LaVarnway [Thu, 18 Oct 2012 21:29:26 +0000 (14:29 -0700)]
sse2 intrinsic version of vp8_mbloop_filter_horizontal_edge()

First sse2 version of vp8_mbloop_filter_horizontal_edge().  For now,
intrinsics are being used until the bitstream is finalized.  This function
will be revisited later for further performance improvements.
For the test clip used, a 31+% decoder performance improvement
was seen.  This will vary depending on material.

Change-Id: I03ed3a7182478bdd1f094644ff3e0442625600e7

11 years agoMerge "Use correct context for optimize_b() in vp8_optimize_mby_8x8()." into experimental
Ronald S. Bultje [Thu, 18 Oct 2012 19:39:10 +0000 (12:39 -0700)]
Merge "Use correct context for optimize_b() in vp8_optimize_mby_8x8()." into experimental

11 years agoA fix for crashes when #macroblock rows is odd.
Deb Mukherjee [Thu, 18 Oct 2012 19:00:23 +0000 (12:00 -0700)]
A fix for crashes when #macroblock rows is odd.

Change-Id: If9fc2df4051ccfb8d804dae9d1be08d12d0f81ce

11 years agolint-hunks: exit status for only affected lines
John Koleszar [Thu, 18 Oct 2012 04:43:18 +0000 (21:43 -0700)]
lint-hunks: exit status for only affected lines

Prior to this patch, if there were any lint errors, this script would
exit with an error, even if those errors were not in the hunks being
tested by this script. This change makes it so that if any lint lines
are printed, an error is returned.

Change-Id: I69c8bef4367ccf25d287508f29e587b1f4426143

11 years agoFix encode breakout skip handling.
Ronald S. Bultje [Thu, 18 Oct 2012 01:00:56 +0000 (18:00 -0700)]
Fix encode breakout skip handling.

Change-Id: Ib42a5dbded27abf2a01fd656a47a1e96b1fe2912

11 years agoMerge "lint-hunks: support operating on arbirary revs" into experimental
John Koleszar [Wed, 17 Oct 2012 23:44:46 +0000 (16:44 -0700)]
Merge "lint-hunks: support operating on arbirary revs" into experimental

11 years agoDetect skippable blocks in splitmv RD loop case.
Ronald S. Bultje [Wed, 17 Oct 2012 21:32:17 +0000 (14:32 -0700)]
Detect skippable blocks in splitmv RD loop case.

Change-Id: Id8a7265dc721e5cdcaa144e0041beeb32a98f51c

11 years agoMerge "cosmetic: fix '= & ' construct" into experimental
pascal massimino [Wed, 17 Oct 2012 22:17:21 +0000 (15:17 -0700)]
Merge "cosmetic: fix '= & ' construct" into experimental

11 years agoMerge "rtcd: merge rtcd.c from master" into experimental
John Koleszar [Wed, 17 Oct 2012 22:16:20 +0000 (15:16 -0700)]
Merge "rtcd: merge rtcd.c from master" into experimental

11 years agoMerge "consolidate update_mb_segmentation_map data" into experimental
John Koleszar [Wed, 17 Oct 2012 22:15:59 +0000 (15:15 -0700)]
Merge "consolidate update_mb_segmentation_map data" into experimental

11 years agoMerge changes Ic711a378,I43a46c36,If19ee74a into experimental
John Koleszar [Wed, 17 Oct 2012 22:15:10 +0000 (15:15 -0700)]
Merge changes Ic711a378,I43a46c36,If19ee74a into experimental

* changes:
  Move remaining per-frame data into partition 0
  Interleave modes/residual per macroblock
  Force interleaved decoding