platform/upstream/libvpx.git
13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Fri, 4 Feb 2011 05:05:11 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I242ca4854cb21f3d63efb979bd6ecc9f06f67f33

13 years agoMake vp8_adjust_mb_lf_value return the updated value rather than
Gaute Strokkenes [Thu, 3 Feb 2011 19:03:49 +0000 (19:03 +0000)]
Make vp8_adjust_mb_lf_value return the updated value rather than
manipulating it in situ via a pointer.

Change-Id: If4a87a4eccd84f39577c0e91e171245f4954c5cf

13 years agoMerge "Avoid using an anonymous union."
John Koleszar [Thu, 3 Feb 2011 17:08:50 +0000 (09:08 -0800)]
Merge "Avoid using an anonymous union."

13 years agoMerge "Zero out block mv when an intra mode is selected"
Scott LaVarnway [Thu, 3 Feb 2011 15:16:52 +0000 (07:16 -0800)]
Merge "Zero out block mv when an intra mode is selected"

13 years agoMerge "Improved encoder threading"
Yunqing Wang [Thu, 3 Feb 2011 13:44:57 +0000 (05:44 -0800)]
Merge "Improved encoder threading"

13 years agoAvoid using an anonymous union.
Gaute Strokkenes [Wed, 2 Feb 2011 15:22:51 +0000 (15:22 +0000)]
Avoid using an anonymous union.

Change-Id: I5744269a35e2d696ecf40c1665efd572bfc9b6cb

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Wed, 2 Feb 2011 05:05:09 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I585615400697b77c50dd05480616f868f2637aa7

13 years agoZero out block mv when an intra mode is selected
Scott LaVarnway [Tue, 1 Feb 2011 17:55:51 +0000 (12:55 -0500)]
Zero out block mv when an intra mode is selected

instead of each time mode is tested.

Change-Id: Ief0f5586dafde54cc14d348dcecdacb182e7c1d5

13 years agoRemoved unnecessary B_MODE_INFO memset.
Scott LaVarnway [Tue, 1 Feb 2011 16:35:08 +0000 (11:35 -0500)]
Removed unnecessary B_MODE_INFO memset.

Change-Id: I2bcef6a8e47f88542861fd1356631ca934e2a0e7

13 years agoMoved rd calculation into vp8_pick_intra4x4mby_modes
Scott LaVarnway [Tue, 1 Feb 2011 16:26:04 +0000 (11:26 -0500)]
Moved rd calculation into vp8_pick_intra4x4mby_modes

Then removed unnecessary code.

Change-Id: I142658815d843c9396b07881dbdd8d387c43c90e

13 years agoRemoved intra_modes from vp8cx_encode_intra_macro_block
Scott LaVarnway [Tue, 1 Feb 2011 15:05:35 +0000 (10:05 -0500)]
Removed intra_modes from vp8cx_encode_intra_macro_block

Restructured function in order to eliminate the prediction
modes save/restore.  Code cleanup also.

Change-Id: I816e3b910de64d0f0f0ddc2398805c63263191e8

13 years agoImproved encoder threading
Attila Nagy [Wed, 26 Jan 2011 08:29:46 +0000 (10:29 +0200)]
Improved encoder threading

Reduce the number of sync points by letting each thread
continue imediatly with a new MB row.
Better multicore scaling, improves performance by 5-20% on ARM multicore.

Change-Id: Ic97e4d1c4886a842c85dd3539a93cb217188ed1b

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Tue, 1 Feb 2011 05:05:09 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: Id1d4bbe257cd126bb5f44347b896ddb659724f0b

13 years agoRemoved prediction_error accumulation
Scott LaVarnway [Tue, 1 Feb 2011 00:53:02 +0000 (19:53 -0500)]
Removed prediction_error accumulation

from vp8cx_encode_intra_macro_block.  prediction_error is used when
deciding if a frame should be a keyframe.  After reviewing this with
Yaowu, it was pointed out that vp8cx_encode_intra_macro_block
is only called for keyframes, so the accumulation is unnecessary.

Change-Id: Id79dc81b80d4f5d124f3a0dba1b923887e2e1ec8

13 years agoRemoved last_auto_filter_prediction_error
Scott LaVarnway [Tue, 1 Feb 2011 00:41:09 +0000 (19:41 -0500)]
Removed last_auto_filter_prediction_error

last_auto_filter_prediction_error is not really used.

Change-Id: Ic6e56c4076bbd250ef783ee1be46964c85f62864

13 years agoPossible bug in vp8cx_encode_intra_macro_block
Scott LaVarnway [Mon, 31 Jan 2011 22:43:18 +0000 (17:43 -0500)]
Possible bug in vp8cx_encode_intra_macro_block

vp8_pick_intra4x4mby_modes uses the passed in distortion
for an early breakout.  The best distortion was never saved
and the distortion for TM_PRED was always used.

Change-Id: Idbaf73027408a4bba26601713725191a5d7b325e

13 years agoMerge "Performance improvement of first pass"
Scott LaVarnway [Mon, 31 Jan 2011 21:02:23 +0000 (13:02 -0800)]
Merge "Performance improvement of first pass"

13 years agoMerge "change the threshold of DC check for encode breakout"
Yaowu Xu [Mon, 31 Jan 2011 19:00:46 +0000 (11:00 -0800)]
Merge "change the threshold of DC check for encode breakout"

13 years agoMerge "validate min_q against max_q"
John Koleszar [Mon, 31 Jan 2011 15:33:55 +0000 (07:33 -0800)]
Merge "validate min_q against max_q"

13 years agovalidate min_q against max_q
John Koleszar [Fri, 28 Jan 2011 16:56:18 +0000 (11:56 -0500)]
validate min_q against max_q

min_q is required to be <= max_q.

Change-Id: I28eccf96df3b52a94913762b54c4fbe0d021ce5e

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Mon, 31 Jan 2011 14:10:35 +0000 (09:10 -0500)]
Merge remote branch 'origin/master' into experimental

Conflicts:
configure

Change-Id: I18c2292256d2387ff09da209aa9cf6891e1864a0

13 years agoMerge "Changed condition for using RD in Intra Mode"
Adrian Grange [Mon, 31 Jan 2011 10:18:40 +0000 (02:18 -0800)]
Merge "Changed condition for using RD in Intra Mode"

13 years agochange the threshold of DC check for encode breakout
Yaowu Xu [Thu, 20 Jan 2011 00:21:01 +0000 (16:21 -0800)]
change the threshold of DC check for encode breakout

Previously, the DC check is to make sure there is no code-able
DC shift for quantizer Q0, which has been verified rather
conservative. This commit changes the criteria to have two
components, DC and AC, to address the conservativeness. First,
it checks if all AC energy is enough to contribute a single
non-zero quantized AC coefficient. Second, for DC, the decision
to skip further considers two possible scenarios: 1. There is
no code-able 2nd order DC coefficient at all; 2 The residue is
relatively flat, but the uniform DC change is very small, i.e.
less than 1/2 gray level per pixel.

Comparing to previous criteria, the new criteria is about 10%
to 15% faster in encoding time with a very small quality loss.
(threshold ~1000 and quality range 33db-45db)

It should be noted that this commit enables "automatic" static
threshold for encodebreakout if a non-zero small value is passed
in to encoder.

Change-Id: I0f77719a1ac2c2dfddbd950d84920df374515ce3

13 years agoMerge "Adds "armvX-none-rvct" targets"
Johann [Fri, 28 Jan 2011 17:03:58 +0000 (09:03 -0800)]
Merge "Adds "armvX-none-rvct" targets"

13 years agoChanged condition for using RD in Intra Mode
Adrian Grange [Fri, 28 Jan 2011 14:47:36 +0000 (14:47 +0000)]
Changed condition for using RD in Intra Mode

The condition for using RD when selecting the intra coding mode
for a MB is that the RD flag is set AND we're not in real-time
mode.

Previously the code used RD if either the RD flag was set OR
we were not using real-time mode.

Change-Id: Ic711151298468a3f99babad39ba8375f66d55a08

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Fri, 28 Jan 2011 13:33:52 +0000 (08:33 -0500)]
Merge remote branch 'origin/master' into experimental

Conflicts:
vp8/encoder/rdopt.c

Change-Id: Ic17907df70fff45c9e766b5d0cbab0c5f1a1095f

13 years agoInconsistent distortion metric in vp8_rd_pick_intra_mbuv_mode
Paul Wilkins [Fri, 28 Jan 2011 13:13:30 +0000 (13:13 +0000)]
Inconsistent distortion metric in vp8_rd_pick_intra_mbuv_mode

This function was using a variance metric compared to and SSE metric in
other places (eg. vp8_rd_inter_uv)

Change-Id: I9109fcc5a13bca9db1d7ead500fe14999ab233eb

13 years agoAdds "armvX-none-rvct" targets
Tero Rintaluoma [Mon, 24 Jan 2011 09:21:40 +0000 (11:21 +0200)]
Adds "armvX-none-rvct" targets

Adds following targets to configure script to support RVCT compilation
without operating system support (for Profiler or bare metal images).
 - armv5te-none-rvct
 - armv6-none-rvct
 - armv7-none-rvct

To strip OS specific parts from the code "os_support"-config was added
to script and CONFIG_OS_SUPPORT flag is used in the code to exclude OS
specific parts such as OS specific includes and function calls for
timers and threads etc. This was done to enable RVCT compilation for
profiling purposes or running the image on bare metal target with
Lauterbach.

Removed separate AREA directives for READONLY data in armv6 and neon
assembly files to fix the RVCT compilation. Otherwise
"ldr <reg>, =label" syntax would have been needed to prevent linker
errors. This syntax is not supported by older gnu assemblers.

Change-Id: I14f4c68529e8c27397502fbc3010a54e505ddb43

13 years agowarning: pointer targets differ in signedness
Johann [Thu, 27 Jan 2011 16:50:29 +0000 (11:50 -0500)]
warning: pointer targets differ in signedness

vp8/encoder/rdopt.c:728: warning: pointer targets in passing argument 3
of 'macro_block_yrd' differ in signedness
vp8/encoder/rdopt.c:541: note: expected 'int *' but argument is of type
'unsigned int *'

distortion is signed when calling macro_block_yrd is both other cases,
as well as for RDCOST

Change-Id: I5e22358b7da76a116f498793253aac8099cb3461

13 years agoclean up implicit declaration warnings for neon
Johann [Thu, 27 Jan 2011 16:31:59 +0000 (11:31 -0500)]
clean up implicit declaration warnings for neon

Change-Id: I6ca2d89f355839c4c770773c09fc69dcea7c1406
warning: implicit declaration of function
  'vp8_variance_halfpixvar16x16_[h|v|hv]_neon'
  'vp8_sub_pixel_variance16x16_neon_func'

13 years agoMerge "Removed unused members from VP8_COMP"
Scott LaVarnway [Thu, 27 Jan 2011 16:12:22 +0000 (08:12 -0800)]
Merge "Removed unused members from VP8_COMP"

13 years agoMerge "Remove copies of same functions"
Yunqing Wang [Thu, 27 Jan 2011 16:11:26 +0000 (08:11 -0800)]
Merge "Remove copies of same functions"

13 years agoMerge "Refine motion vector prediction for NEWMV mode"
Yunqing Wang [Thu, 27 Jan 2011 16:10:53 +0000 (08:10 -0800)]
Merge "Refine motion vector prediction for NEWMV mode"

13 years agoPerformance improvement of first pass
Scott LaVarnway [Wed, 26 Jan 2011 21:42:56 +0000 (16:42 -0500)]
Performance improvement of first pass

Improved the performance of the first pass only
(~6% on 720p test clip) by making use of LUT instead of the
float calculations.  Might try a SIMD version later.
Also started to make use of int_mv instead of
MV.

Change-Id: If2a217c7d6b59cd2c25c5553e0ca7e0502403af8

13 years agoRemove copies of same functions
Yunqing Wang [Wed, 26 Jan 2011 17:03:13 +0000 (12:03 -0500)]
Remove copies of same functions

Reduce the code size.

Change-Id: I2e1998557a3c8776e262c442fd758c25e17aff7a

13 years agoRemoved unused members from VP8_COMP
Scott LaVarnway [Mon, 24 Jan 2011 19:19:14 +0000 (14:19 -0500)]
Removed unused members from VP8_COMP

Change-Id: I8f3f2642b02975fbdb14982984a29821f80d30d3

13 years agoRationalize vp8_rd_pick_intra16x16mby_mode()
Paul Wilkins [Wed, 26 Jan 2011 18:46:34 +0000 (18:46 +0000)]
Rationalize vp8_rd_pick_intra16x16mby_mode()

Use the function macro_block_yrd() to calculate error and distortion
in keeping with what is done for inter frames.

The old code was using a variance metric for once case and an
SSE function for measuring distortion in the other case.

The function vp8_encode_intra16x16mbyrd() is no longer used.

Change-Id: Ic228cb00a78ff637f4365b43f58fbe5a9273d36f

13 years agoMerge "Correction to buffer update for non-viewable frames."
Paul Wilkins [Wed, 26 Jan 2011 17:33:48 +0000 (09:33 -0800)]
Merge "Correction to buffer update for non-viewable frames."

13 years agoMerge "cap the best quantizer for 2nd order DC"
Yaowu Xu [Wed, 26 Jan 2011 17:27:11 +0000 (09:27 -0800)]
Merge "cap the best quantizer for 2nd order DC"

13 years agoMerge "Adds vpx_vp8_enc_asm_offsets.c.o to OBJS-yes list"
John Koleszar [Wed, 26 Jan 2011 15:29:19 +0000 (07:29 -0800)]
Merge "Adds vpx_vp8_enc_asm_offsets.c.o to OBJS-yes list"

13 years agoAdds vpx_vp8_enc_asm_offsets.c.o to OBJS-yes list
Attila Nagy [Wed, 26 Jan 2011 15:06:09 +0000 (17:06 +0200)]
Adds vpx_vp8_enc_asm_offsets.c.o to OBJS-yes list

Change-Id: Ibd6e3bc82471839904b1086b499efc55f7c5cbaf

13 years agoCorrection to buffer update for non-viewable frames.
Paul Wilkins [Wed, 26 Jan 2011 12:52:31 +0000 (12:52 +0000)]
Correction to buffer update for non-viewable frames.

The code previously tested cpi->common.refresh_alt_ref_frame
but there are situations where this flag may be set for viewable frames.

The correct test should be !cm->show_frame.

Change-Id: Ia1a600622992a4a68fe1d38ac23bf6b34b133688

13 years agoMerge "Fix for incorrect variable declaration."
Paul Wilkins [Wed, 26 Jan 2011 09:53:53 +0000 (01:53 -0800)]
Merge "Fix for incorrect variable declaration."

13 years agocap the best quantizer for 2nd order DC
Yaowu Xu [Wed, 26 Jan 2011 06:24:22 +0000 (22:24 -0800)]
cap the best quantizer for 2nd order DC

This commit also removes artificial RDMULT cap for low quantizers.
The intention is to address some abnormal behavior of mode selections
at the low quantizer end, where many macroblocks were coded with
SPLITMV with all partitions using same motion vector including (0,0).
This change improves the compression quality substantially for high
quality encodings in both PSNR and SSIM terms. Overall effect on
mid/low rate range is also positive for all metrics, but smaller
in magnitude.

Change-Id: I864b29c4bd9ff610d2545fa94a19cc7e80c02667

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Wed, 26 Jan 2011 05:05:11 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I35581d7fd525fe972e750899e35de93867595e83

13 years agoFix for incorrect variable declaration.
Fritz Koenig [Tue, 25 Jan 2011 23:41:41 +0000 (15:41 -0800)]
Fix for incorrect variable declaration.

Commit 336aa0b7da8a35ba57400ce92fc016fc7fb35233 incorrectly
declared current_pos as and int, when it should have been
a FIRSTPASS_STATS pointer.

Change-Id: I0a51c7a86ebba8546c95dd5d9d1c1143d4613e40

13 years agoMerge "update sse2 regular quantizer"
Johann [Tue, 25 Jan 2011 21:40:28 +0000 (13:40 -0800)]
Merge "update sse2 regular quantizer"

13 years agoMerge "move new neon subpixel function"
Johann [Tue, 25 Jan 2011 21:09:05 +0000 (13:09 -0800)]
Merge "move new neon subpixel function"

13 years agoRefine motion vector prediction for NEWMV mode
Yunqing Wang [Tue, 25 Jan 2011 20:54:34 +0000 (15:54 -0500)]
Refine motion vector prediction for NEWMV mode

Adjust checking points in motion vector prediction to better cover
possible movements, and get a better prediction. Tests on test
clips showed a 0.1% improvement in SSIM, and no change in PSNR
and performance.

Change-Id: Ifdab05d35e10faea1445c61bb73debf888c9d2f8

13 years agoMerge "Fix issue 262, vp8cx_pack_tokens_into_partitions_armv5"
Johann [Tue, 25 Jan 2011 20:49:52 +0000 (12:49 -0800)]
Merge "Fix issue 262, vp8cx_pack_tokens_into_partitions_armv5"

13 years agomove new neon subpixel function
Johann [Tue, 25 Jan 2011 20:11:39 +0000 (15:11 -0500)]
move new neon subpixel function

previously wasn't guarded with ifdef ARMV7, causing a link error with
ARMV6

Change-Id: I0526858be0b5f49b2bf11e9090180b2a6c48926d

13 years agoMerge "Modify calling of NEON code in sub-pixel search"
Yunqing Wang [Tue, 25 Jan 2011 17:54:23 +0000 (09:54 -0800)]
Merge "Modify calling of NEON code in sub-pixel search"

13 years agoFix issue 262, vp8cx_pack_tokens_into_partitions_armv5
Attila Nagy [Mon, 17 Jan 2011 11:00:08 +0000 (13:00 +0200)]
Fix issue 262, vp8cx_pack_tokens_into_partitions_armv5

http://code.google.com/p/webm/issues/detail?id=262
Function was asuming that partitions have equal amount of mb_rows,
which is not always true.

Change-Id: I59ed40117fd408392a85c633beeb5340ed2f4b25

13 years agoMerge "Incorrect bit allocation in forced KF groups."
Paul Wilkins [Tue, 25 Jan 2011 13:32:26 +0000 (05:32 -0800)]
Merge "Incorrect bit allocation in forced KF groups."

13 years agoIncorrect bit allocation in forced KF groups.
Paul Wilkins [Tue, 25 Jan 2011 12:29:06 +0000 (12:29 +0000)]
Incorrect bit allocation in forced KF groups.

The old 2 pass code estimated error distribution when coding a
forced (by interval) key frame. The result of this was that in some
cases, when allocating bits at the GF group level within a KF
group there was either a glut of bits or starvation of bits at the end
of the KF group.

Added code to rescan and get the correct data once the position of
a forced key frame has been determined.

Change-Id: I0c811675ef3f9e4109d14bd049d7641682ffcf11

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Tue, 25 Jan 2011 05:05:10 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I3dd6219db4126c1ad526c7877ad1ce49c496155f

13 years agoconfigure.sh fix for visual studio
James Berry [Mon, 24 Jan 2011 21:48:21 +0000 (16:48 -0500)]
configure.sh fix for visual studio

-For targets with external build systems like visual
studio CC is not set so check_add_cflags will fail.
Only call this function if extra_cflags is set.

Change-Id: I3531bad69e9b6a59c5be1b0e8b6053ccccbc332c

13 years agoAdded vp8_update_zbin_extra
Scott LaVarnway [Wed, 29 Dec 2010 19:30:57 +0000 (14:30 -0500)]
Added vp8_update_zbin_extra

vp8cx_mb_init_quantizer was being called for every mode checked
in vp8_rd_pick_inter_mode.  zbin_extra is the only value that
really needs to be recalculated.  This calculation is disabled
when using the fast quantizer for mode selection.
This gave a small performance boost (~.5% to 1%).
Note: This needs to be verified with segmentation_enabled.

Change-Id: I62716a870b3c82b4a998bdf95130ff0b02106f1e

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Sat, 22 Jan 2011 05:05:13 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I2c3326f7e4d9e901f098e499973586e973e1b8fb

13 years agoMerge "Modify sub-pixel filters to eliminate unnecessary calculations"
Yunqing Wang [Fri, 21 Jan 2011 19:07:17 +0000 (11:07 -0800)]
Merge "Modify sub-pixel filters to eliminate unnecessary calculations"

13 years agoModify sub-pixel filters to eliminate unnecessary calculations
Yunqing Wang [Thu, 20 Jan 2011 18:01:30 +0000 (13:01 -0500)]
Modify sub-pixel filters to eliminate unnecessary calculations

In sub-pixel calculation, xoffset and yoffset mostly take some
specific values. Modified sub-pixel filter functions according to
these possible values to improve performance.

Change-Id: I83083570af8b00ff65093467914fbb97a4e9ea21

13 years agoModified static scene check.
Paul Wilkins [Fri, 21 Jan 2011 17:52:00 +0000 (17:52 +0000)]
Modified static scene check.

Added code to scan ahead a few frames when we see what
we think is a static scene in the two pass GF loop to see if the
conditions persist.

Moved calculation of decay rate out into a fuunction.

Change-Id: I6e9c67e01ec9f555144deafc8ae67ef25bffb449

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Fri, 21 Jan 2011 15:11:12 +0000 (10:11 -0500)]
Merge remote branch 'origin/master' into experimental

Conflicts:
configure

Change-Id: I87cdc3faac79e683038fa4a45c3ac542b8e931a5

13 years agoFurther work to reduce pulsing.
Paul Wilkins [Thu, 20 Jan 2011 18:01:20 +0000 (18:01 +0000)]
Further work to reduce pulsing.

These changes are specifically targeted at fade transitions to
static scenes. Here we want to place a GF/ARF immediately
after the fade and prevent an ARF just  before the fade.

Also some code lines and comment lines shortened to 80 chars
while I was there.

Change-Id: Iefdc09a4fa7b265048fc017246b73e138693950f

13 years agoUpdate configure scripts
Attila Nagy [Thu, 20 Jan 2011 12:46:57 +0000 (14:46 +0200)]
Update configure scripts

Add --extra-cflags as config parameter for user defined extra CFLAGS.
Add -g to asflags when debug enabled for arm targets.

Change-Id: Ibdde7cfdda6736c1c1db45e6466bd08504a51f15

13 years agoFixed use of motion percentage in KF/GF group calc
Adrian Grange [Thu, 20 Jan 2011 13:13:33 +0000 (13:13 +0000)]
Fixed use of motion percentage in KF/GF group calc

In both vp8_find_next_key_frame and define_gf_group,
motion_pct was initialised at the top of the loop before
next_frame stats had been read in.

This fix sets motion_pct after next_frame stats have
been read.

Change-Id: I8c0bebf372ef8aa97b97fd35b42973d1d831ee73

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Thu, 20 Jan 2011 05:05:23 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: Ia0840303fe1dc8c12f3389e7a1fe20b6d3c6b9c5

13 years agoexperiment extending the quantizer range
Yaowu Xu [Wed, 1 Dec 2010 23:50:14 +0000 (15:50 -0800)]
experiment extending the quantizer range

Prior to this change, VP8 min quantizer is 4, which caps the
highest quality around 51DB. This experimental change extends
the min quantizer to 1, removes the cap and allows the highest
quality to be around ~73DB, consistent with the fdct/idct round trip
error. To test this change, at configure time use options:

--enable-experimental --enable-extend_qrange

The following is a brief log of changes in each of the patch sets

patch set 1:
In this commit, the quantization/dequantization constants are kept
unchanged, instead scaling factor 4 is rolled into fdct/idct.
Fixed Q0 encoding tests on mobile:
  Before:    9560.567kbps Overall PSNR:50.255DB VPXSSIM:98.288
  Now:   18035.774kbps Overall PSNR:73.022DB VPXSSIM:99.991

patch set 2:
regenerated dc/ac quantizer lookup tables based on the scaling
factor rolled in the fdct/idct. Also slightly extended the range
towards the high quantizer end.

patch set 3:
slightly tweaked the quantizer tables and generated bits_per_mb
table based on Paul's suggestions.

patch set 4:
fix a typo in idct, re-calculated tables relating active max Q
to active min Q

patch set 5:
added rdmult lookup table based on Q

patch set 6:
fix rdmult scale: dct coefficient has scaled up by 4

patch set 7:
make transform coefficients to be within 16bits

patch set 8:
normalize 2nd order quantizers

patch set 9:
fix mis-spellings

patch set 10:
change the configure script and macros to allow experimental code
to be enabled at configure time with --enable-extend_qrange

patch set 11:
rebase for merge

Change-Id: Ib50641ddd44aba2a52ed890222c309faa31cc59c

13 years agoMerge "First pass loop bug."
Paul Wilkins [Wed, 19 Jan 2011 16:33:34 +0000 (08:33 -0800)]
Merge "First pass loop bug."

13 years agoFirst pass loop bug.
Paul Wilkins [Wed, 19 Jan 2011 15:50:22 +0000 (15:50 +0000)]
First pass loop bug.

Incorrect value loop_decay_rate used in GF loop.

The intent was to test the  cumulative value decay_accumulator.

Change-Id: I62928c63eb09f4f6936a45ebd1c23784d1c9681b

13 years agoMerge "Implement error tracking in the decoder"
John Koleszar [Wed, 19 Jan 2011 13:51:00 +0000 (05:51 -0800)]
Merge "Implement error tracking in the decoder"

13 years agoImplement error tracking in the decoder
Henrik Lundin [Thu, 16 Dec 2010 15:46:31 +0000 (16:46 +0100)]
Implement error tracking in the decoder

A new vpx_codec_control called VP8D_GET_FRAME_CORRUPTED. The output
from the function is non-zero if the last decoded frame contains
corruption due to packet losses.

The decoder is also modified to accept encoded frames of zero length.
A zero length frame indicates to the decoder that one or more frames
have been completely lost. This will mark the last decoded reference
buffer as corrupted. The data pointer can be NULL if the length is
zero.

Change-Id: Ic5902c785a281c6e05329deea958554b7a6c75ce

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Wed, 19 Jan 2011 05:05:16 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I7125984ec28a7094195b640d6659590c6eead533

13 years agoMerge "fix last frame buffer copy logic regression"
John Koleszar [Tue, 18 Jan 2011 20:54:57 +0000 (12:54 -0800)]
Merge "fix last frame buffer copy logic regression"

13 years agoModify calling of NEON code in sub-pixel search
Yunqing Wang [Tue, 18 Jan 2011 19:19:52 +0000 (14:19 -0500)]
Modify calling of NEON code in sub-pixel search

In vp8_find_best_sub_pixel_step_iteratively(), many times xoffset
and yoffset are specific values - (4,0) (0,4) and (4,4). Modified
code to call simplified NEON version at these specific offsets to
help with the performance.

Change-Id: Iaf896a0f7aae4697bd36a49e182525dd1ef1ab4d

13 years agovp8e -removed undefined max call
Jim Bankoski [Tue, 18 Jan 2011 16:14:17 +0000 (11:14 -0500)]
vp8e -removed undefined max call

Change-Id: I42a86b0488f44115f09551fc5ad6d711fd470f0d

13 years agoMerge "Further CQ, Key frame and ARF changes"
Paul Wilkins [Tue, 18 Jan 2011 16:04:46 +0000 (08:04 -0800)]
Merge "Further CQ, Key frame and ARF changes"

13 years agoFurther CQ, Key frame and ARF changes
Paul Wilkins [Tue, 18 Jan 2011 15:19:05 +0000 (15:19 +0000)]
Further CQ, Key frame and ARF changes

This code fixes a bug in the calculation of
the minimum Q for alt ref frames.

It also allows an extended gf/arf interval for sections
of clips that completely static (or nearly so).

Change-Id: I1a21aaa16d4f0578e5f99b13bebd78d59403c73b

13 years agoFix encoder real-time only configuration.
Attila Nagy [Mon, 10 Jan 2011 09:14:10 +0000 (11:14 +0200)]
Fix encoder real-time only configuration.

Remove allocation/deallocation of stats storage.
Remove full search functions in machine specific encoder inits.
Remove last pass validation in  validate_config.

Change-Id: I7f29be69273981a4fef6e80ecdb6217c68cbad4e

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Tue, 18 Jan 2011 05:05:16 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I3e51945a61446e1ee70caaf97ff7cf0c740bf1da

13 years agoFix CQ range and experimental KF sizing changes.
Paul Wilkins [Mon, 17 Jan 2011 17:23:11 +0000 (17:23 +0000)]
Fix CQ range and experimental KF sizing changes.

The CQ level was not using the q_trans[] array to convert
to a 0-127 range as per min and maxq

Experimental change to try and match the reconstruction
error for forced key frames approximately to that of the
previous frame by means of the recode loop. Though this
may cause extra recodes and the recode behavior has not
been optimized, it can only happen on forced key frames.

Change-Id: I1f7e42d526f1b1cb556dd461eff1a692bd1b5b2f

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Sat, 15 Jan 2011 05:05:09 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: Id03b5485239fdb2a5e2351a3393eeed847a6e641

13 years agoupdate sse2 regular quantizer
Johann [Tue, 11 Jan 2011 14:41:57 +0000 (09:41 -0500)]
update sse2 regular quantizer

about ~5% gain on 32bit. disabled for 64bit

unset executable bit on ssse3 version (cosmetic)

Change-Id: I1a5860839eb294ce4261f819caea2dcfa78e57ca

13 years agoMerge "KF/GF Pulsing"
Paul Wilkins [Fri, 14 Jan 2011 17:20:37 +0000 (09:20 -0800)]
Merge "KF/GF Pulsing"

13 years agoMerge "Testing of modes with Alt Ref frame"
Paul Wilkins [Fri, 14 Jan 2011 15:26:37 +0000 (07:26 -0800)]
Merge "Testing of modes with Alt Ref frame"

13 years agoMerge "Experimental change to help with ARNR problem."
Paul Wilkins [Fri, 14 Jan 2011 15:26:01 +0000 (07:26 -0800)]
Merge "Experimental change to help with ARNR problem."

13 years agoTesting of modes with Alt Ref frame
Paul Wilkins [Fri, 14 Jan 2011 14:52:15 +0000 (14:52 +0000)]
Testing of modes with Alt Ref frame

Previously when a frame was being overlaid on a previously coded
alt ref frame we only checked the alt ref 0,0 mode. Where there is
a possibility that the alt ref buffer is a filtered frame we should allow
the other prediction modes as normal or at the least allow use of
the last frame buffer.

Change-Id: I4d6227223d125c96b4f3066ec6ec9484fee7768c

13 years agoARNR filter pointer update bug fix
Adrian Grange [Fri, 14 Jan 2011 15:04:39 +0000 (15:04 +0000)]
ARNR filter pointer update bug fix

In cases where the frame width is not a multiple of 16 the
ARNR filter would go wrong.

In vp8_temporal_filter_iterate_c when updating pointers
at the end of a row of MBs,  the image size was
incorrectly used rather than using Num_MBs_In_Row
times 16 (Y) or 8 (U,V).

This worked when width is multiple of 16 but failed
otherwise.

Change-Id: I008919062715bd3d17c7aa2562ab58d1cb37053a

13 years agoExperimental change to help with ARNR problem.
Paul Wilkins [Fri, 14 Jan 2011 12:07:12 +0000 (12:07 +0000)]
Experimental change to help with ARNR problem.

Allow use of other reference frames for the ARF overlay frame
when ARNR filtering is enabled

Change-Id: Icd6a9fb38977a88fbe7cc9b9c18198eb454c0273

13 years agoKF/GF Pulsing
Paul Wilkins [Fri, 14 Jan 2011 11:34:53 +0000 (11:34 +0000)]
KF/GF Pulsing

This change is designed to try and reduce pulsing effects when moving
with a complex transition like a fade, into an easy or static section in
an otherwise difficult clip in CQ mode.

The active CQ level is relaxed down to the user entered level for frames that
are generating less than the passed in minimum bandwidth.

Change-Id: Id6d8b551daad4f489c087bd742bc95418a95f3f0

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Fri, 14 Jan 2011 05:05:08 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I229048ab797a4087437c265e6bb96db5c32e7b3d

13 years agoMerge "Moved ref frame calculations"
Scott LaVarnway [Thu, 13 Jan 2011 14:59:28 +0000 (06:59 -0800)]
Merge "Moved ref frame calculations"

13 years agoOne pass rate control correction.
Paul Wilkins [Thu, 13 Jan 2011 12:45:00 +0000 (12:45 +0000)]
One pass rate control correction.

Fixed discrepancy cpi->ni_frames vs cm->current_video_frame > 150.

Make one pass path explicit.

There is still scope for some odd behaviour around the transition
point at cpi->ni_frames > 150.

Change-Id: Icdee130fe6e2a832206d30e45bf65963edd7a74d

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Thu, 13 Jan 2011 05:05:10 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: I0859ddc498ff2d8f75fc4d2bc37127eb6188e231

13 years agoLimit key frame quantizer for forced key frames.
Paul Wilkins [Wed, 12 Jan 2011 17:08:42 +0000 (17:08 +0000)]
Limit key frame quantizer for forced key frames.

Where a key frame occurs because of a minimum interval
selected by the user, then these forced key frames ideally need
to be more closely matched in quality to the surrounding frame.

Change-Id: Ia55b1f047e77dc7fbd78379c45869554f25b3df7

13 years agoMerge remote branch 'origin/master' into experimental
John Koleszar [Wed, 12 Jan 2011 05:05:14 +0000 (00:05 -0500)]
Merge remote branch 'origin/master' into experimental

Change-Id: Id5da32e6d58a58e04a4dff9ca1df23ebb6c436b8

13 years agoMoved ref frame calculations
Scott LaVarnway [Tue, 11 Jan 2011 20:00:00 +0000 (15:00 -0500)]
Moved ref frame calculations

Moved ref frame calculations to outside of the
mode_index loop.

Change-Id: I06103fc7e8af88b54b84443acf6691d29b1272ac

13 years agoMerge "Add no_skip_block4x4_search flag in SPLITMV mode"
Yunqing Wang [Tue, 11 Jan 2011 16:34:24 +0000 (08:34 -0800)]
Merge "Add no_skip_block4x4_search flag in SPLITMV mode"

13 years agoMerge "use unaligned load"
Johann [Tue, 11 Jan 2011 16:25:22 +0000 (08:25 -0800)]
Merge "use unaligned load"

13 years agouse unaligned load
Johann [Tue, 11 Jan 2011 16:22:29 +0000 (11:22 -0500)]
use unaligned load

source buffer is not guaranteed to be aligned for odd size buffers

Change-Id: Id0b1fd40ba3bd6c994bcfada788feccd2b53c5a9