platform/upstream/libvpx.git
12 years agoCleanups for hybrid transform
Deb Mukherjee [Wed, 12 Sep 2012 02:36:28 +0000 (19:36 -0700)]
Cleanups for hybrid transform

Some cleanups on the transform size and type selection logic.

Change-Id: If2e9675459482242cf83b4f7de7634505e3f6dac

12 years agoMerge "Revert "Hybrid transform cleanups"" into experimental
Yaowu Xu [Tue, 11 Sep 2012 22:00:51 +0000 (15:00 -0700)]
Merge "Revert "Hybrid transform cleanups"" into experimental

12 years agoRevert "Hybrid transform cleanups"
Debargha Mukherjee [Tue, 11 Sep 2012 21:52:20 +0000 (14:52 -0700)]
Revert "Hybrid transform cleanups"

This reverts commit c5af82b7ed6d89a4a12003b8d1c6b4befcadce92

There was a mismatch bug introduced by this patch. Hence reverting until it is fixed.

12 years agoMerge "Hybrid transform cleanups" into experimental
Deb Mukherjee [Mon, 10 Sep 2012 22:39:37 +0000 (15:39 -0700)]
Merge "Hybrid transform cleanups" into experimental

12 years agoHybrid transform cleanups
Deb Mukherjee [Mon, 10 Sep 2012 06:10:30 +0000 (23:10 -0700)]
Hybrid transform cleanups

Some cleanups that will make it easier to maintain the code
and incorporate upcoming changes on  entropy coding for the
hybrid transforms.

Change-Id: I44bdba368f7b8bf203161d7a6d3b1fc2c9e21a8f

12 years agoPartial merge of NEWBESTREFMV and NEW_MVREF.
Paul Wilkins [Fri, 7 Sep 2012 13:21:22 +0000 (14:21 +0100)]
Partial merge of NEWBESTREFMV and NEW_MVREF.

This commit merges those parts of the CONFIG_NEW_MVREF
that specifically relate to choosing a better set of candidate
MV references into the NEWBESTREFMV experiment.

CONFIG_NEW_MVREF will then be used for changes relating
to the explicit coding of a cost optimized MV reference in the
bitstream as part of MV coding.

Change-Id: Ied982c0ad72093eab29e38b8cd74d5c3d7458b10

12 years agoMV reference changes
Paul Wilkins [Fri, 7 Sep 2012 11:46:41 +0000 (12:46 +0100)]
MV reference changes

Extend experiment to use both vectors from MBs
coded using compound prediction as candidates.

In final sort only consider best 4 candidates
for now but make sure 0,0 is always one of them.

Other minor changes to new MV reference code.
Pass in Mv list to vp8_find_best_ref_mvs().

Change-Id: Ib96220c33c6b80bd1d5e0fbe8b68121be7997095

12 years agoNew motion vector entropy coding
Deb Mukherjee [Thu, 26 Jul 2012 20:42:07 +0000 (13:42 -0700)]
New motion vector entropy coding

Adds a new experiment with redesigned/refactored motion vector entropy
coding. The patch also takes a first step towards separating the
integer and fractional pel components of a MV. However the fractional
pel encoding still depends on the integer pel part and so they are
not fully independent. Further experiments are in progress to see
how much they can be decoupled without affecting performance.
All components including entropy coding/decoding, costing for MV
search, forward updates and backward updates to probability tables,
have been implemented.

Results so far:
derf: +0.19%
std-hd: +0.28%
yt: +0.80%
hd: +1.15%

Patch: Simplifies the fractional pel models:
derf: +0.284%
std-hd: +0.289%
yt: +0.849%
hd: +1.254%

Patch: Some changes in the models, rebased.
derf: +0.330%
std-hd: +0.306%
yt: +0.816%
hd: +1.225%

Change-Id: I646b3c48f3587f4cc909639b78c3798da6402678

12 years agohybrid transform of 16x16 dimension
Jingning Han [Wed, 29 Aug 2012 18:25:38 +0000 (11:25 -0700)]
hybrid transform of 16x16 dimension

Enable ADST/DCT of dimension 16x16 for I16X16 modes. This change provides
benefits mostly for hd sequences.

Set up the framework for selectable transform dimension.

Also allowing quantization parameter threshold to control the use
of hybrid transform (This is currently disabled by setting threshold
always above the quantization parameter. Adaptive thresholding can
be built upon this, which will further improve the coding performance.)

The coding performance gains (with respect to the codec that has all
other configuration settings turned on) are

derf:   0.013
yt:     0.086
hd:     0.198
std-hd: 0.501

Change-Id: Ibb4263a61fc74e0b3c345f54d73e8c73552bf926

12 years agoMerge "Add proper skip support to intra frames." into experimental
Ronald S. Bultje [Thu, 30 Aug 2012 16:11:41 +0000 (09:11 -0700)]
Merge "Add proper skip support to intra frames." into experimental

12 years agoAdd proper skip support to intra frames.
Ronald S. Bultje [Wed, 29 Aug 2012 17:43:20 +0000 (10:43 -0700)]
Add proper skip support to intra frames.

Change-Id: I7524883fb29f42303fb46a5bc6772fbcf8781d1d

12 years agoRemove some disabled code from pick_inter_mode_sb().
Ronald S. Bultje [Tue, 28 Aug 2012 16:51:52 +0000 (09:51 -0700)]
Remove some disabled code from pick_inter_mode_sb().

Change-Id: I323c429b453eec8c3b76c37f7e2a52da9dfe8d77

12 years agoDisable 16x16 temp buffer for superblocks.
Ronald S. Bultje [Tue, 28 Aug 2012 01:04:53 +0000 (18:04 -0700)]
Disable 16x16 temp buffer for superblocks.

Change-Id: I3d585bae446c16fcf7a39d72c1bcd6468a7bb1a1

12 years agoMerge "Fix file permissions" into experimental
John Koleszar [Fri, 24 Aug 2012 22:21:49 +0000 (15:21 -0700)]
Merge "Fix file permissions" into experimental

12 years agoMerge "change to properly account for coef scaling in unit tests" into experimental
Ronald S. Bultje [Fri, 24 Aug 2012 19:10:04 +0000 (12:10 -0700)]
Merge "change to properly account for coef scaling in unit tests" into experimental

12 years agochange to properly account for coef scaling in unit tests
Yaowu Xu [Fri, 24 Aug 2012 16:59:49 +0000 (09:59 -0700)]
change to properly account for coef scaling in unit tests

The transform functions in experimental branch absorbed a scaling
factor of 4 to allow quantization steps closer to unit quantizer.
This commit added scaling code in between forward and inverse
transform to properly account for the scaling factor.

Change-Id: I9a573ddc1ffa74973b34800a5da1a56dbabe0949

12 years agoNew Motion Reference Search
Paul Wilkins [Fri, 24 Aug 2012 14:44:01 +0000 (15:44 +0100)]
New Motion Reference Search

Alternative strategy for finding a list of candidate motion
vectors to use as reference values in mv coding and as
nearest and near.

Sort by sad in vp8_find_best_ref_mvs() rather than just
pick the best. Allow 0,0 as a best ref option but not a
nearest or near unless there are no alternatives.

Encode/Decode verified on at least some clips.

Some commented out experimental and stats code still in place.

Gain over existing code averages about 1% on derf (alll metrics)
with improvement on all clips. Other test results pending.

The entropy coding of the mode (nearest/near etc) still
depends upon and requires the old "findnear" code so
this needs looking at and may provide room for further gains.

Change-Id: I871d7cba1d1c379c4bad9bcccce1fb19c46b8247

12 years agoRemove duplicate code in C intra prediction.
Ronald S. Bultje [Wed, 22 Aug 2012 23:05:21 +0000 (16:05 -0700)]
Remove duplicate code in C intra prediction.

Change-Id: If4eb0fd60cc5bc2f32254e82e8d6cfa4120cb558

12 years agoFix file permissions
John Koleszar [Wed, 22 Aug 2012 17:16:05 +0000 (10:16 -0700)]
Fix file permissions

Change-Id: I8d1a2c664cbeb939cf23bcd414b23e19da760a14

12 years agoMerge "all_builds.py: move to tools/" into experimental
John Koleszar [Wed, 22 Aug 2012 17:03:06 +0000 (10:03 -0700)]
Merge "all_builds.py: move to tools/" into experimental

12 years agoMerge "all_builds.py: support for sharding builds" into experimental
John Koleszar [Wed, 22 Aug 2012 17:02:45 +0000 (10:02 -0700)]
Merge "all_builds.py: support for sharding builds" into experimental

12 years agoMerge "remove rotation experiment" into experimental
John Koleszar [Wed, 22 Aug 2012 17:01:39 +0000 (10:01 -0700)]
Merge "remove rotation experiment" into experimental

12 years agoMerge "SSE2 version of vectorized 8-tap filtering." into experimental
Christian Duvivier [Wed, 22 Aug 2012 00:17:42 +0000 (17:17 -0700)]
Merge "SSE2 version of vectorized 8-tap filtering." into experimental

12 years agoMerge "Fix mismatch caused by a rebase failure." into experimental
Ronald S. Bultje [Wed, 22 Aug 2012 00:12:41 +0000 (17:12 -0700)]
Merge "Fix mismatch caused by a rebase failure." into experimental

12 years agoFix mismatch caused by a rebase failure.
Ronald S. Bultje [Tue, 21 Aug 2012 23:45:51 +0000 (16:45 -0700)]
Fix mismatch caused by a rebase failure.

Change-Id: Ifff660de60fb12d745092eafc17f6a49f6e1fb53

12 years agoSSE2 version of vectorized 8-tap filtering.
Christian Duvivier [Tue, 21 Aug 2012 06:49:43 +0000 (23:49 -0700)]
SSE2 version of vectorized 8-tap filtering.

About 20% overall encoder speedup (vs. about 30% for sse4 version).

Change-Id: Ibf608a6a1bc94b14ec47e8046d3206b275b5a8bd

12 years agoFix compiler warnings
Daniel Kang [Tue, 21 Aug 2012 22:08:54 +0000 (15:08 -0700)]
Fix compiler warnings

Change-Id: I887cdcf90b87736f2a3805420aa41eed93324bee

12 years agoFix warning.
Christian Duvivier [Tue, 21 Aug 2012 21:58:31 +0000 (14:58 -0700)]
Fix warning.

Change-Id: I99311811820169d62fda3febac4bafacfc7a538b

12 years agoall_builds.py: move to tools/
John Koleszar [Tue, 21 Aug 2012 18:18:58 +0000 (11:18 -0700)]
all_builds.py: move to tools/

Change-Id: I64c470fb5a4f32a862cfb5424fb95baac47fcc24

12 years agoall_builds.py: support for sharding builds
John Koleszar [Tue, 21 Aug 2012 18:18:38 +0000 (11:18 -0700)]
all_builds.py: support for sharding builds

Allow sharding of the builds with the --shard= and --shards= options.

Change-Id: I9d5552ad2edc0b1210e96f0e94ce7dfd645c45f8

12 years agoMerge changes I96cd08c9,If8eeb8ff into experimental
John Koleszar [Tue, 21 Aug 2012 17:10:43 +0000 (10:10 -0700)]
Merge changes I96cd08c9,If8eeb8ff into experimental

* changes:
  hybridtransform8x8: fix compilation
  comp_intra_pred: fix compilation

12 years agoremove rotation experiment
John Koleszar [Tue, 21 Aug 2012 17:09:56 +0000 (10:09 -0700)]
remove rotation experiment

This is being reimplemented more generically in terms of affine
transforms.

Change-Id: I9300bfde5f8b93c708c64f59427087720f8ed782

12 years agohybridtransform8x8: fix compilation
John Koleszar [Tue, 21 Aug 2012 16:31:31 +0000 (09:31 -0700)]
hybridtransform8x8: fix compilation

corrected ifdefs to include CONFIG_HYBRIDTRANSFORM8X8 for missing
definitions

Change-Id: I96cd08c98d74b5ba10866f3fa30252073ce92fa5

12 years agoFix some compiler warnings.
Ronald S. Bultje [Tue, 21 Aug 2012 00:45:36 +0000 (17:45 -0700)]
Fix some compiler warnings.

Change-Id: Iccb56e3b966d1519ba49dfe4e1396479c6b54a6f

12 years agocomp_intra_pred: fix compilation
John Koleszar [Tue, 21 Aug 2012 00:38:44 +0000 (17:38 -0700)]
comp_intra_pred: fix compilation

Added missing parameters to calls to:
  vp8_build_intra_predictors_internal
  vp8_build_intra_predictors_mbuv_internal

Change-Id: If8eeb8ff23eff4572397b404fe61be5d0c950bbe

12 years agoMerge "Fix typo." into experimental
John Koleszar [Tue, 21 Aug 2012 00:09:30 +0000 (17:09 -0700)]
Merge "Fix typo." into experimental

12 years agoFix typo.
Ronald S. Bultje [Tue, 21 Aug 2012 00:07:24 +0000 (17:07 -0700)]
Fix typo.

Change-Id: I95217ce52675ad00637fc8f60d4d0dd054efd512

12 years agoMerge "Superblock coding." into experimental
Ronald S. Bultje [Tue, 21 Aug 2012 00:04:23 +0000 (17:04 -0700)]
Merge "Superblock coding." into experimental

12 years agoSuperblock coding.
Ronald S. Bultje [Mon, 20 Aug 2012 21:43:34 +0000 (14:43 -0700)]
Superblock coding.

This commit adds a pick_sb_mode() function which selects the best 32x32
superblock coding mode. Then it selects the best per-MB modes, compares
the two and encodes that in the bitstream.

The bitstream coding is rather simplistic right now. At the SB level,
we code a bit to indicate whether this block uses SB-coding (32x32
prediction) or MB-coding (anything else), and then we follow with the
actual modes. This could and should be modified in the future, but is
omitted from this commit because it will likely involve reorganizing
much more code rather than just adding SB coding, so it's better to let
that be judged on its own merits.

Gains on derf: about even, YT/HD: +0.75%, STD/HD: +1.5%.

Change-Id: Iae313a7cbd8f75b3c66d04a68b991cb096eaaba6

12 years agoFix warnings.
Christian Duvivier [Fri, 17 Aug 2012 22:37:51 +0000 (15:37 -0700)]
Fix warnings.

Change-Id: I01dd96da18fe22450bb372b10ba90eab52e29bdf

12 years agoMerge "A few more optimizations, about 1% overall speedup." into experimental
Christian Duvivier [Fri, 17 Aug 2012 00:03:08 +0000 (17:03 -0700)]
Merge "A few more optimizations, about 1% overall speedup." into experimental

12 years agoA few more optimizations, about 1% overall speedup.
Christian Duvivier [Thu, 16 Aug 2012 05:41:32 +0000 (22:41 -0700)]
A few more optimizations, about 1% overall speedup.

Unroll horizontal pass, no more intermediate buffer, faster special transpose.

Change-Id: I05df75be4e5f01420066cdf3c61a2edf35bedb64

12 years agoFixed decoder mismatch issue
Jingning Han [Thu, 16 Aug 2012 20:25:49 +0000 (13:25 -0700)]
Fixed decoder mismatch issue

Resolved the decoder mismatch issue due to quantization parameter
threshold for hybrid transform coding. The macroblock dequantizer
initialization is moved to be performed before coefficient
detokenization, since the (de)tokenization is now dependent on the
macroblock level quantization parameter.

Change-Id: I443da4992ebb70ae4114750b2f1363c0c628580e

12 years agoMerge "Set a default mv entropy if no tokens are coded." into experimental
Ronald S. Bultje [Thu, 16 Aug 2012 18:33:10 +0000 (11:33 -0700)]
Merge "Set a default mv entropy if no tokens are coded." into experimental

12 years agoSet a default mv entropy if no tokens are coded.
Ronald S. Bultje [Thu, 16 Aug 2012 18:09:31 +0000 (11:09 -0700)]
Set a default mv entropy if no tokens are coded.

This doesn't affect the result, since there are no MVs coded using this
entropy. It does, however, silence valgrind warnings about uninitialized
variables.

Change-Id: I6e21ba92df6ce5381bf58b8c349ef4373294a0b6

12 years agoFix valgrind error which causes non-deterministic results.
Daniel Kang [Thu, 16 Aug 2012 18:04:34 +0000 (11:04 -0700)]
Fix valgrind error which causes non-deterministic results.

Error was introduced in bf474e170fefbed60cf20fa3d2e9e229776469f0

Change-Id: I55bf114a1938c3a8869d8ff49cbe076eb9e83ea0

12 years agoFix stack overrun in calculating segmentation map coding method.
Ronald S. Bultje [Thu, 16 Aug 2012 05:20:24 +0000 (22:20 -0700)]
Fix stack overrun in calculating segmentation map coding method.

This caused spurious crashes on encodes, frustratingly often only on
high-definition material.

Change-Id: Ia8c1dfc420d598ca92108bd9cc2df7450e8d10a2

12 years agoMerge "First partial snapshot of vectorized 8-tap filtering." into experimental
Christian Duvivier [Thu, 16 Aug 2012 01:01:18 +0000 (18:01 -0700)]
Merge "First partial snapshot of vectorized 8-tap filtering." into experimental

12 years agoFirst partial snapshot of vectorized 8-tap filtering.
Christian Duvivier [Wed, 15 Aug 2012 01:19:09 +0000 (18:19 -0700)]
First partial snapshot of vectorized 8-tap filtering.

About 3.5x faster, 30% overall encoder speedup. Rest of optimizations
will come soon (see TODO section in filter_sse4.c).

Change-Id: If18108048bfd5345fc942e8574e4c7f58e0e86e0

12 years agoFix build with TX_16X16
Daniel Kang [Wed, 15 Aug 2012 16:43:36 +0000 (09:43 -0700)]
Fix build with TX_16X16

Change-Id: I1ccb2546e29803d0e9d69282409d549a28201967

12 years agoCode clean up.
Paul Wilkins [Wed, 15 Aug 2012 10:00:53 +0000 (11:00 +0100)]
Code clean up.

Further cases of inconsistent naming convention.

Change-Id: Id3411ecec6f01a4c889268a00f0c9fd5a92ea143

12 years agoFix warnings.
Christian Duvivier [Tue, 14 Aug 2012 21:12:00 +0000 (14:12 -0700)]
Fix warnings.

Change-Id: I4b911e4173da30c164bde7ea50bc80a70fbbb745

12 years agoFix build.
Christian Duvivier [Tue, 14 Aug 2012 20:38:11 +0000 (13:38 -0700)]
Fix build.

Change-Id: I27729a8631cffd54dfbae3ec0f07df7c883f7510

12 years agoShut up warnings added by -Wundef
Daniel Kang [Mon, 13 Aug 2012 23:50:03 +0000 (16:50 -0700)]
Shut up warnings added by -Wundef

Change-Id: I9c7ef4a75c37aa0e10df75e165e3066614c955ef

12 years agoMerge changes I0b51674f,I1ea6ebf9,I89076d93 into experimental
John Koleszar [Tue, 14 Aug 2012 16:26:16 +0000 (09:26 -0700)]
Merge changes I0b51674f,I1ea6ebf9,I89076d93 into experimental

* changes:
  lint_hunks: show style violations in the index
  intersect_diffs: split out diff classes
  ftfy: update to match current astyle rule

12 years agoMerge "Some build fixes" into experimental
Deb Mukherjee [Tue, 14 Aug 2012 16:21:40 +0000 (09:21 -0700)]
Merge "Some build fixes" into experimental

12 years agoMerge "Changed to use reference mv as nearest mv" into experimental
Yaowu Xu [Tue, 14 Aug 2012 16:12:31 +0000 (09:12 -0700)]
Merge "Changed to use reference mv as nearest mv" into experimental

12 years agoChanged to use reference mv as nearest mv
Yaowu Xu [Mon, 6 Aug 2012 17:51:20 +0000 (10:51 -0700)]
Changed to use reference mv as nearest mv

The reference motion vector selected by surrounding pixels that has
the best matching score is used as nearest motion vector.

The change has shown consistent gain on all test sets, compression
gains range from .2% to .6%. The variation is largely dependent on
various other experiments on or off.

Change-Id: I5552e1c2f6fc57c3e8818a5ee41ffda89af05e75

12 years agoSome build fixes
Deb Mukherjee [Tue, 14 Aug 2012 14:28:43 +0000 (07:28 -0700)]
Some build fixes

Fixes two typos that breaks the build introduced in recent check-ins.

Change-Id: I3f15f3c088c48a63fede8941c9a4ba05dd0403c2

12 years agoCode clean up.
Paul Wilkins [Tue, 14 Aug 2012 10:32:29 +0000 (11:32 +0100)]
Code clean up.

References to MACROBLOCKD that use "x" changed to "xd"
to comply with convention elsewhere that x = MACROBLOCK
and xd = MACROBLOCKD.

Simplify some repeat references using local variables.

Change-Id: I0ba2e79536add08140a6c8b19698fcf5077246bc

12 years agoFix typo, which adds skip testing for 16x16
Daniel Kang [Mon, 13 Aug 2012 21:18:09 +0000 (14:18 -0700)]
Fix typo, which adds skip testing for 16x16

Also add warnings for undefined macros in the C pre-processor

Change-Id: I1ec30e57c5a49fb72151a4cf140d7eeb0fb1d779

12 years agoShut up some compiler warnings on gcc
Daniel Kang [Mon, 13 Aug 2012 23:20:20 +0000 (16:20 -0700)]
Shut up some compiler warnings on gcc

Change-Id: I8548aebd3c39797b253eef472bfe1c40f2f7ec59

12 years agoFix warning/bug introduced in 4d5b81a80faa8333ca3a7aed53684d2167fbebd2.
Christian Duvivier [Mon, 13 Aug 2012 22:51:04 +0000 (15:51 -0700)]
Fix warning/bug introduced in 4d5b81a80faa8333ca3a7aed53684d2167fbebd2.

Change-Id: I082d2517022cbf2f472c143fe7772719d8295498

12 years agoCosmetics to rdopt.c
Daniel Kang [Mon, 13 Aug 2012 16:55:08 +0000 (09:55 -0700)]
Cosmetics to rdopt.c

Change-Id: Icc276609e409e4e234d240e5c12e137600e178ed

12 years agoCombine optimize_b and optimize_b_8x8
Daniel Kang [Fri, 10 Aug 2012 22:34:31 +0000 (15:34 -0700)]
Combine optimize_b and optimize_b_8x8

The 16x16 has not been done yet for speed reasons.

Change-Id: I066e369a065674d58986e937344c9303fed19450

12 years agoextend the range of encoding time
Yaowu Xu [Sat, 11 Aug 2012 00:05:46 +0000 (17:05 -0700)]
extend the range of encoding time

The commit also corrects unit of encoding from microsecond to
milisecond.

Change-Id: I030108e90f89f33dfad2d092dc8991db81979715

12 years agofix build problem for newbestrefmv experiment
Yaowu Xu [Sat, 11 Aug 2012 00:15:23 +0000 (17:15 -0700)]
fix build problem for newbestrefmv experiment

Change-Id: I7f531d55f076254c84402fed34af93fe3685fbdf

12 years agoFix some tabbing in rdopt.c
Daniel Kang [Fri, 10 Aug 2012 17:00:18 +0000 (10:00 -0700)]
Fix some tabbing in rdopt.c

Change-Id: Ie7e7727512cc1333bd16e7688cd1d8764371dde3

12 years agolint_hunks: show style violations in the index
John Koleszar [Fri, 13 Jul 2012 22:14:17 +0000 (15:14 -0700)]
lint_hunks: show style violations in the index

Shows the subset of style violations that intersect with lines modified
in the current index.

Change-Id: I0b51674fd43df118e8bbf55b684a1ca0bd0e1d40

12 years agoRefactor to remove some arguments from vp8_rd_pick_inter_mode
Daniel Kang [Fri, 10 Aug 2012 01:25:29 +0000 (18:25 -0700)]
Refactor to remove some arguments from vp8_rd_pick_inter_mode

Change-Id: I8e72279cc68f34d269705f06cdaf8f3d06eed635

12 years agoRefactor RD to take same codepath for single and comp pred
Daniel Kang [Thu, 9 Aug 2012 23:07:41 +0000 (16:07 -0700)]
Refactor RD to take same codepath for single and comp pred

Change-Id: Id38baf1b89648ef534e28be72f583137871f920c

12 years agoCode cleanup.
Paul Wilkins [Fri, 10 Aug 2012 13:12:43 +0000 (14:12 +0100)]
Code cleanup.

Add local variable in several places to reference the MB mode
info structure. Currently this is usually accessed in the code as
x->e_mbd.mode_info_context->mbmi.* or in some places
xd->mode_info_context->mbmi.*

Resolved some uses of x-> for the MACROBLOCKD structure.

Rebased without dependency on motion reference experiment.

Change-Id: If6718276ee4f2ef131825d1524dfdb02a3793aed

12 years agointersect_diffs: split out diff classes
John Koleszar [Fri, 13 Jul 2012 20:01:40 +0000 (13:01 -0700)]
intersect_diffs: split out diff classes

Seperate parsing functions so they can be reused for other diff
utilities

Change-Id: I1ea6ebf90ded128eec116cc1f326a28bdda2fb77

12 years agoftfy: update to match current astyle rule
John Koleszar [Wed, 8 Aug 2012 16:34:44 +0000 (09:34 -0700)]
ftfy: update to match current astyle rule

Change-Id: I89076d93968a5817d6d0330ec7fd6a6c73fe397f

12 years agoRefactor vp8_build_inter16x16_predictor functions
Daniel Kang [Thu, 9 Aug 2012 00:12:12 +0000 (17:12 -0700)]
Refactor vp8_build_inter16x16_predictor functions

Change-Id: I4fceefe76b14bd38a358dc4356708f7981bab339

12 years agoMore refactoring of diamond search.
Daniel Kang [Wed, 8 Aug 2012 18:52:22 +0000 (11:52 -0700)]
More refactoring of diamond search.

This should make merging inter code paths easier.

Change-Id: I5cb81f25d56fa4790b4e9cfa4bc32b9062c2cfdf

12 years agoRefactor SPLITMV in RD to a special case.
Daniel Kang [Wed, 8 Aug 2012 18:49:56 +0000 (11:49 -0700)]
Refactor SPLITMV in RD to a special case.

SPLITMV shares very little code with the other inter modes.

Change-Id: I0beee95de1fbedeea1f71650951ffc7db3a118eb

12 years agoBegin refactoring to help merge parts of rd inter.
Daniel Kang [Wed, 8 Aug 2012 18:13:53 +0000 (11:13 -0700)]
Begin refactoring to help merge parts of rd inter.

Also refactor some RD code, remove some comments, and cosmetic changes.

Change-Id: I43806ae573ab5a5036602ac4425d750bbfa1e812

12 years agoMerge "Merging in the sixteenth subpel uv experiment" into experimental
Deb Mukherjee [Thu, 9 Aug 2012 16:36:18 +0000 (09:36 -0700)]
Merge "Merging in the sixteenth subpel uv experiment" into experimental

12 years agoMerge "Import tools/ from master branch" into experimental
John Koleszar [Thu, 9 Aug 2012 16:12:17 +0000 (09:12 -0700)]
Merge "Import tools/ from master branch" into experimental

12 years agoNewline got lost in submission of "Partial import of "New RTCD implementation" from...
Christian Duvivier [Thu, 9 Aug 2012 00:20:23 +0000 (17:20 -0700)]
Newline got lost in submission of "Partial import of "New RTCD implementation" from master branch."

Change-Id: I3841eca46e29a108a79fe47af0cfa97f0bc80a2a

12 years agoMerging in the sixteenth subpel uv experiment
Deb Mukherjee [Wed, 8 Aug 2012 23:49:15 +0000 (16:49 -0700)]
Merging in the sixteenth subpel uv experiment

Merges this experiment in to make it easier to run tests on
filter precision, vectorized implementation etc.

Also removes an experimental filter.

Change-Id: I1e8706bb6d4fc469815123939e9c6e0b5ae945cd

12 years agoMerge "Partial import of "New RTCD implementation" from master branch." into experimental
Christian Duvivier [Wed, 8 Aug 2012 23:49:03 +0000 (16:49 -0700)]
Merge "Partial import of "New RTCD implementation" from master branch." into experimental

12 years agoPartial import of "New RTCD implementation" from master branch.
Christian Duvivier [Wed, 8 Aug 2012 22:47:36 +0000 (15:47 -0700)]
Partial import of "New RTCD implementation" from master branch.

Latest version of all scripts/makefile but rtcd_defs.sh is empty, all
existing functions are still selected using the old/current way.

Change-Id: Ib92946a48a31d6c8d1d7359eca524bc1d3e66174

12 years agoMerge "a tiny fix for MSVC build" into experimental
Debargha Mukherjee [Wed, 8 Aug 2012 23:12:23 +0000 (16:12 -0700)]
Merge "a tiny fix for MSVC build" into experimental

12 years agoMerge "Refactor setting up MVs and buffers in rdopt.c" into experimental
Daniel Kang [Wed, 8 Aug 2012 21:58:24 +0000 (14:58 -0700)]
Merge "Refactor setting up MVs and buffers in rdopt.c" into experimental

12 years agoa tiny fix for MSVC build
Yaowu Xu [Wed, 8 Aug 2012 21:41:37 +0000 (14:41 -0700)]
a tiny fix for MSVC build

Change-Id: Ib23b0711a4690e1edff3131ce48527b6805877fb

12 years agoMerge "Cleanup and commenting of the sub-pixel interpolation functions. Mostly as...
Christian Duvivier [Wed, 8 Aug 2012 19:45:46 +0000 (12:45 -0700)]
Merge "Cleanup and commenting of the sub-pixel interpolation functions. Mostly as a preparation for the upcoming vectorized implementation." into experimental

12 years agoCleanup and commenting of the sub-pixel interpolation functions. Mostly
Christian Duvivier [Wed, 8 Aug 2012 03:57:36 +0000 (20:57 -0700)]
Cleanup and commenting of the sub-pixel interpolation functions. Mostly
as a preparation for the upcoming vectorized implementation.

Change-Id: I0d1629943cee68fcb0ea8b6dd8aef265d050cc7a

12 years agoImport tools/ from master branch
John Koleszar [Wed, 8 Aug 2012 16:28:01 +0000 (09:28 -0700)]
Import tools/ from master branch

Change-Id: I476f43ea513e7c140f4e4f14df73dd3131cb2692

12 years agoRefactor setting up MVs and buffers in rdopt.c
Daniel Kang [Wed, 8 Aug 2012 01:55:28 +0000 (18:55 -0700)]
Refactor setting up MVs and buffers in rdopt.c

Change-Id: I94f08f2cfb37d11d71050dca3200f1d391269a6a

12 years agoSome refactoring of mcomp functions.
Daniel Kang [Tue, 7 Aug 2012 23:44:26 +0000 (16:44 -0700)]
Some refactoring of mcomp functions.

Change-Id: Ic7a7cb1199b085e98ede0e634619b3077c348d57

12 years agoChange the RD cost functions to take into account the seg_eob.
Daniel Kang [Tue, 7 Aug 2012 01:29:59 +0000 (18:29 -0700)]
Change the RD cost functions to take into account the seg_eob.

Change-Id: I2f96e5b205f70fdb559f01637c4b5730246c38a8

12 years agoclamp_mv_min_max -> vp8_clamp_mv_min_max
Daniel Kang [Tue, 7 Aug 2012 22:29:16 +0000 (15:29 -0700)]
clamp_mv_min_max -> vp8_clamp_mv_min_max

It's now used in more places

Change-Id: I63f8e1d827404e0b4f203bdb1df361e565d0779d

12 years agoMore refactoring and cosmetics to rdopt.c
Daniel Kang [Tue, 7 Aug 2012 21:39:25 +0000 (14:39 -0700)]
More refactoring and cosmetics to rdopt.c

Change-Id: I7803d719ead238189e75c52701c3bb6c1552f4b5

12 years agoReorganiziation of the RD loop to make more sense.
Daniel Kang [Tue, 7 Aug 2012 21:33:15 +0000 (14:33 -0700)]
Reorganiziation of the RD loop to make more sense.

Change-Id: Ia1e3b0ff9b70747b3725391de59f230edab767d7

12 years agoMinor cosmetic changes to rdopt.c.
Daniel Kang [Mon, 6 Aug 2012 23:21:23 +0000 (16:21 -0700)]
Minor cosmetic changes to rdopt.c.

Change-Id: Ib7b972486ccbbcb82fa4c05a5f6d7f8aaec32155

12 years agoCombine vp8_block_error_c and vp8_submb_error_c.
Daniel Kang [Mon, 6 Aug 2012 22:03:04 +0000 (15:03 -0700)]
Combine vp8_block_error_c and vp8_submb_error_c.

Change-Id: I58792bac6c7e5a524a6dcdb5241ef18bcc0c3a45

12 years agoA bit of temporary relief on encoder slowness
Yaowu Xu [Tue, 7 Aug 2012 20:55:49 +0000 (13:55 -0700)]
A bit of temporary relief on encoder slowness

The commit replaces run-time initialization of cosine constants with
static constant values, which provides ~30% relief on slow speed. The
real solution, however will be to implement integer versions of those
functions that current use float/double.

Change-Id: Ie3ff1793509653d78dd1aeaf88cc6737da1bc55f

12 years agoa new way of determining reference motion vector
Yaowu Xu [Mon, 6 Aug 2012 17:51:20 +0000 (10:51 -0700)]
a new way of determining reference motion vector

Using surrounding reconstructed pixels from left and above to select
best matching mv to use as reference motion vector for mv encoding.

Test results:
       AVGPSNR  GLBPSNR VPXSSIM
Derf:  1.107%   1.062%  0.992%
Std-hd:1.209%   1.176%  1.029%

Change-Id: I8f10e09ee6538c05df2fb9f069abcaf1edb3fca6