profile/ivi/libvpx.git
13 years agoMerge "gen_msvs_proj: write boolean for Debug attribute"
John Koleszar [Thu, 16 Jun 2011 14:19:46 +0000 (07:19 -0700)]
Merge "gen_msvs_proj: write boolean for Debug attribute"

13 years agogen_msvs_proj: write boolean for Debug attribute
James Zern [Wed, 15 Jun 2011 21:42:06 +0000 (14:42 -0700)]
gen_msvs_proj: write boolean for Debug attribute

Replace =1 with =true for yasm tool element. This aids in upgrading
e.g., vs9 project files to vs10.
build/x86-msvs/yasm.xml generated during conversion will require the
Separator attribute to be removed for the build to complete
successfully.

Change-Id: If75c4f9a925529740048882003e9d766c5ac4f0c

13 years agoMerge "Remove redundant check for KEY_FRAME in multithreaded decoder"
John Koleszar [Wed, 15 Jun 2011 17:18:06 +0000 (10:18 -0700)]
Merge "Remove redundant check for KEY_FRAME in multithreaded decoder"

13 years agoMerge "fix --disable-runtime-cpu-detect on x86"
John Koleszar [Wed, 15 Jun 2011 14:09:09 +0000 (07:09 -0700)]
Merge "fix --disable-runtime-cpu-detect on x86"

13 years agoGuard vpx_config.h against multiple inclusions
Attila Nagy [Wed, 15 Jun 2011 10:27:39 +0000 (13:27 +0300)]
Guard vpx_config.h against multiple inclusions

Change-Id: Iabe2be73af2b92c53687755b31b77448fba385d2

13 years agoRemove redundant check for KEY_FRAME in multithreaded decoder
Attila Nagy [Mon, 13 Jun 2011 12:13:10 +0000 (15:13 +0300)]
Remove redundant check for KEY_FRAME in multithreaded decoder

For Intra blocks is enough to check ref_frame == INTRA_FRAME.

Change-Id: I3e2d3064c7642658a9e14011a4627de58878e366

13 years agoMerge "Populate bmi for B_PRED only"
Scott LaVarnway [Tue, 14 Jun 2011 19:04:50 +0000 (12:04 -0700)]
Merge "Populate bmi for B_PRED only"

13 years agofix --disable-runtime-cpu-detect on x86
Johann [Tue, 14 Jun 2011 15:31:50 +0000 (11:31 -0400)]
fix --disable-runtime-cpu-detect on x86

Change-Id: Ib8e429152c9a8b6032be22b5faac802aa8224caa

13 years agoMerge "Fix RT only build"
Paul Wilkins [Tue, 14 Jun 2011 13:01:44 +0000 (06:01 -0700)]
Merge "Fix RT only build"

13 years agoFix RT only build
Tero Rintaluoma [Tue, 14 Jun 2011 10:39:06 +0000 (13:39 +0300)]
Fix RT only build

Moved encode_intra function from firstpass.c to encodeintra.c to
prevent linking problem in real-time only build. Also changed name
of the function to vp8_encode_intra because it is not a static.

Change-Id: Ibf3c6c1de3152567347e5fbef47d1d39564620a5

13 years agoUpdate -linux-rvct targets
Tero Rintaluoma [Tue, 14 Jun 2011 08:29:35 +0000 (11:29 +0300)]
Update -linux-rvct targets

- Updated -linux-rvct targets to support RVDS 4.0 and later.
- Changed optimization flag to -Otime because -O3 ruined performance
  for RVCT linux targets.
- Added support for --enable-small for RVCT
- RVCT created library should be able to link with GCC
- Supports building shared linux libraries

Change-Id: Ic62589950d86c3420fd4d908b8efb870806d1233

13 years agofix corrupt frame leak
James Zern [Tue, 14 Jun 2011 00:29:49 +0000 (17:29 -0700)]
fix corrupt frame leak

If setup_token_decoder reported an internal error the memory allocated
there would not be freed in the resulting call to _remove_decompressor.

Change-Id: Ib459de222d76b1910d6f449cdcd01663447dbdf6

13 years agoPopulate bmi for B_PRED only
Scott LaVarnway [Mon, 13 Jun 2011 21:14:11 +0000 (17:14 -0400)]
Populate bmi for B_PRED only

Small decode performance gain (~1%) on keyframes.  No
noticeable gains on encode.  Also changed pick_intra4x4mby_modes()
to read the above and left block modes for keyframes only.

Change-Id: I1f4885252f5b3e9caf04d4e01e643960f910aba5

13 years agoCalc ref_frame_cost once per frame
Scott LaVarnway [Mon, 13 Jun 2011 13:50:54 +0000 (09:50 -0400)]
Calc ref_frame_cost once per frame

instead of every macro block.

Change-Id: I2604e94c6b89e3a8457777e21c8c38406d55b165

13 years agoFix make clean for asm offset files
Tero Rintaluoma [Fri, 10 Jun 2011 11:05:53 +0000 (14:05 +0300)]
Fix make clean for asm offset files

Automatically created assembly offset files added to CLEAN-OBJS list
for proper cleanup. This will fix following build error:
1) Build for the workstation
./conigure
make
make clean
2) Build for ARM platform
./configure --target=armv7-linux-gcc
make ==> this will fail because it uses old asm_*_offset.asm files

Change-Id: Id5275c470390ca81b8db086a15ad75af39b80703

13 years agoMerge "bug fix mode_info_context not initialized for error-resilient"
John Koleszar [Thu, 9 Jun 2011 20:39:47 +0000 (13:39 -0700)]
Merge "bug fix mode_info_context not initialized for error-resilient"

13 years agoremove one set of 16x16 variance funcations
Yaowu Xu [Wed, 8 Jun 2011 20:45:29 +0000 (13:45 -0700)]
remove one set of 16x16 variance funcations

call to this set of functions are replaced by var16x16.

Change-Id: I5ff1effc6c1358ea06cda1517b88ec28ef551b0d

13 years agobug fix mode_info_context not initialized for error-resilient
James Berry [Thu, 9 Jun 2011 16:46:31 +0000 (12:46 -0400)]
bug fix mode_info_context not initialized for error-resilient

uninitialized xd->mode_info_context would crash
vpxenc for --error-resilient=1.

Change-Id: I31849e40281e3d65ab63257cfec5e93398997f0b

13 years agoUpdate keyframe activity in non-RD mode
John Koleszar [Wed, 8 Jun 2011 16:09:02 +0000 (12:09 -0400)]
Update keyframe activity in non-RD mode

Activity update is no longer dependent on being in RD mode, so update
it unconditionally.

Change-Id: Ib617a6fc210dfc045455e3e4467d7ee5e3d1fa0e

13 years agouse GCC inline magic
Johann [Wed, 8 Jun 2011 18:43:34 +0000 (14:43 -0400)]
use GCC inline magic

Better fix for #326. ICC happens to support the inline magic

Change-Id: Ic367eea608c88d89475cb7b05d73500d2a1bc42b

13 years agoMerge "Revert "Use shared object files for ELF""
Johann [Wed, 8 Jun 2011 18:47:18 +0000 (11:47 -0700)]
Merge "Revert "Use shared object files for ELF""

13 years agoRevert "Use shared object files for ELF"
Johann [Wed, 8 Jun 2011 18:36:04 +0000 (11:36 -0700)]
Revert "Use shared object files for ELF"

Broke RVCT. New magic coming for ICC. Stay tuned!

This reverts commit c73eb2ffff674e4e9fa7d5209d531aeec6ac5edd

13 years agoMerge "vp8_pick_inter_mode: remove best_bmodes"
John Koleszar [Wed, 8 Jun 2011 17:59:30 +0000 (10:59 -0700)]
Merge "vp8_pick_inter_mode: remove best_bmodes"

13 years agoMerge "vp8_pick_intra_mode: correct returned rate"
John Koleszar [Wed, 8 Jun 2011 17:58:36 +0000 (10:58 -0700)]
Merge "vp8_pick_intra_mode: correct returned rate"

13 years agoMerge "Move RD intra block mode selection to rdopt.c"
John Koleszar [Wed, 8 Jun 2011 17:51:50 +0000 (10:51 -0700)]
Merge "Move RD intra block mode selection to rdopt.c"

13 years agovp8_pick_inter_mode: remove best_bmodes
John Koleszar [Wed, 8 Jun 2011 17:50:50 +0000 (13:50 -0400)]
vp8_pick_inter_mode: remove best_bmodes

Since BPRED will be tested at most once, and SPLITMV is not enabled,
there's nothing to clobber the subblock modes, so there's no need to
save and restore them.

Change-Id: I7c3615b69190c10bd068a44df5488d6e8b85a364

13 years agoMerge "Removed unused function parameters"
Scott LaVarnway [Wed, 8 Jun 2011 17:20:28 +0000 (10:20 -0700)]
Merge "Removed unused function parameters"

13 years agovp8_pick_intra_mode: correct returned rate
John Koleszar [Wed, 8 Jun 2011 17:08:12 +0000 (13:08 -0400)]
vp8_pick_intra_mode: correct returned rate

The returned rate was always the 4x4 rate, instead of the rate
matching the selected mode.

Change-Id: I51da31f80884f5e37f3bcc77d1047d31e612ded4

13 years agoRemoved unused function parameters
Scott LaVarnway [Wed, 8 Jun 2011 17:01:09 +0000 (13:01 -0400)]
Removed unused function parameters

Change-Id: Ib641c624faec28ad9eb99e2b5de51ae74bbcb2a2

13 years agoAdjust errorperbit according to RDMULT in activity masking
Yaowu Xu [Tue, 7 Jun 2011 20:59:46 +0000 (13:59 -0700)]
Adjust errorperbit according to RDMULT in activity masking

In activity masking, RDO constant RDMULT is adjusted on a per MB basis
adaptive to activity with the MB. errorperbit, which is defined as
RDMULT/RDDIV, is a constant used in motion estimation. Previously, in
activity masking, errorperbit is not changed even when RDMULT is changed.
This commit changed to adjust errorperbit according to the change in
RDMULT.

Test in cif set showed a very small but consistent gain by all quality
metrics (average, overall psnr and ssim) when activity masking is on.

Change-Id: I07ded3e852919ab76757691939fe435328273823

13 years agoMerge "Further activity masking changes:"
Yaowu Xu [Wed, 8 Jun 2011 16:30:31 +0000 (09:30 -0700)]
Merge "Further activity masking changes:"

13 years agoMove RD intra block mode selection to rdopt.c
John Koleszar [Wed, 8 Jun 2011 16:05:05 +0000 (12:05 -0400)]
Move RD intra block mode selection to rdopt.c

This change is analogous to I0b67dae1f8a74902378da7bdf565e39ab832dda7,
which made the move for the non-RD path.

Change-Id: If63fc1b0cd1eb7f932e710f83ff24d91454f8ed1

13 years agoMove intra block mode selection to pickinter.c
John Koleszar [Wed, 8 Jun 2011 15:24:52 +0000 (11:24 -0400)]
Move intra block mode selection to pickinter.c

This commit moves the intra block mode selection from encodeframe.c
to pickinter.c (in the non-RD case). This allowed pick_intra_mbuv_mode
and pick_intra4x4mby_modes to be made static, and is a step towards
refactoring intra mode selection in the main pickinter loop. Gave a
small perf increase (~0.5%).

Change-Id: I0b67dae1f8a74902378da7bdf565e39ab832dda7

13 years agoFurther activity masking changes:
Paul Wilkins [Wed, 8 Jun 2011 15:00:59 +0000 (16:00 +0100)]
Further activity masking changes:

Some further re-structuring of activity masking code.
Still has various experimental switches.
Supports a metric based on intra encode.
Experimental comparison against a fixed activity target  rather
than a frame average, for altering rd and zbin.

Overall the SSIM performance is similar  to TT's original
code but there is a much smaller PSNR hit of circa
0.5% instead of 3.2%

Change-Id: I0fd53b2dfb60620b3f74d7415e0b81c1ac58c39a

13 years agoMerge "remove redundant functions"
Yaowu Xu [Tue, 7 Jun 2011 23:36:37 +0000 (16:36 -0700)]
Merge "remove redundant functions"

13 years agoMerge "adjust sad per bit constants"
Yaowu Xu [Tue, 7 Jun 2011 19:37:04 +0000 (12:37 -0700)]
Merge "adjust sad per bit constants"

13 years agoadjust sad per bit constants
Yaowu Xu [Fri, 20 May 2011 06:12:40 +0000 (23:12 -0700)]
adjust sad per bit constants

While investigating the effect of DC values on SAD and SSE in motion
estimation, a side finding indicates the two table of constants need
be adjusted. The adjustment was done by multiplying old constants by
90% with rounding. Also absorb the 1/2 scaling constant into the two
tables. Refer to change Ifa285c3e for background of the 1/2 factor.

Cif set test showed a very small gain on all metric.

Change-Id: I04333527a823371175dd46cb04a817e5b9a8b752

13 years agoMerge "Reduce overshoot in 1 pass rate control"
John Koleszar [Tue, 7 Jun 2011 19:30:37 +0000 (12:30 -0700)]
Merge "Reduce overshoot in 1 pass rate control"

13 years agoMerge "Wrapped asserts in critical code with CONFIG_DEBUG"
Scott LaVarnway [Tue, 7 Jun 2011 13:53:51 +0000 (06:53 -0700)]
Merge "Wrapped asserts in critical code with CONFIG_DEBUG"

13 years agoMerge "Removed unused function vp8_treed_read_num"
Scott LaVarnway [Tue, 7 Jun 2011 13:51:24 +0000 (06:51 -0700)]
Merge "Removed unused function vp8_treed_read_num"

13 years agoMerge "Use shared object files for ELF"
John Koleszar [Tue, 7 Jun 2011 13:44:32 +0000 (06:44 -0700)]
Merge "Use shared object files for ELF"

13 years agoWrapped asserts in critical code with CONFIG_DEBUG
Scott LaVarnway [Tue, 7 Jun 2011 13:34:47 +0000 (09:34 -0400)]
Wrapped asserts in critical code with CONFIG_DEBUG

Change-Id: I5b0aaca06f2e0f40588cb24fb0642b6865da8970

13 years agoRemoved unused function vp8_treed_read_num
Scott LaVarnway [Tue, 7 Jun 2011 13:32:51 +0000 (09:32 -0400)]
Removed unused function vp8_treed_read_num

Change-Id: Id66e70540ee7345876f099139887c1843093907f

13 years agoremove redundant functions
Yaowu Xu [Mon, 6 Jun 2011 23:42:58 +0000 (16:42 -0700)]
remove redundant functions

The encoder defined about 4 set of similar functions to calculate sum,
variance or sse or a combination of them. This commit removed one set
of these functions, get8x8var and get16x16var, where calls to the later
function are replaced with var16x16 by using the fact on a 16x16 MB:
    variance == sse - sum*sum/256

Change-Id: I803eabd1fb3ab177780a40338cbd596dffaed267

13 years agoRemove hex search's variance calculation while in real-time mode
Yunqing Wang [Thu, 2 Jun 2011 18:47:06 +0000 (14:47 -0400)]
Remove hex search's variance calculation while in real-time mode

In real-time mode motion search, there is no need to calculate
variance. This change improved encoding speed by 1% ~ 2%(speed=-5).

Change-Id: I65b874901eb599ac38fe8cf9cad898c14138d431

13 years agoMerge "neon fast quantize block pair"
Johann [Mon, 6 Jun 2011 20:42:58 +0000 (13:42 -0700)]
Merge "neon fast quantize block pair"

13 years agoMerge "adds preload for armv6 encoder asm"
Johann [Mon, 6 Jun 2011 20:32:13 +0000 (13:32 -0700)]
Merge "adds preload for armv6 encoder asm"

13 years agoUse shared object files for ELF
Johann [Tue, 31 May 2011 15:59:04 +0000 (11:59 -0400)]
Use shared object files for ELF

Fixes #326

Change-Id: I5f2a4257430ef62f674190acefd43a0474821288

13 years agoMerge "Removed unnecessary bmi motion vector stores."
Scott LaVarnway [Mon, 6 Jun 2011 14:57:39 +0000 (07:57 -0700)]
Merge "Removed unnecessary bmi motion vector stores."

13 years agoMerge "Don't allow very short GF groups even when the GF is predicted from an ARF."
John Koleszar [Mon, 6 Jun 2011 14:02:29 +0000 (07:02 -0700)]
Merge "Don't allow very short GF groups even when the GF is predicted from an ARF."

13 years agoReduce overshoot in 1 pass rate control
John Koleszar [Tue, 3 May 2011 15:17:05 +0000 (11:17 -0400)]
Reduce overshoot in 1 pass rate control

This patch attempts to reduce the peak bitrate hit by the encoder
when using small buffer windows.

Tested on the CIF set over 200-500kbps using these settings:

  --buf-sz=500 --buf-initial-sz=250 --buf-optimal-sz=250 \
  --undershoot-pct=100

Two pass encodes were tested at best quality. One pass encodes were
tested only at realtime speed 4:

  --rt --cpu-used=-4

The peak datarate (over the specified 500ms window) was measured
for each encode, and averaged together to get metric for
"average peak," computed as SUM(peak)/SUM(target). This patch
reduces the average peak datarate as follows:

  One pass:
    baseline:   1.29715
    this patch: 1.23664

  Two pass:
    baseline:   1.32702
    this patch: 1.37824

This change had a positive effect on our quality metrics as well:

  One pass CBR:
                    Min  / Mean / Max (pct)
    Average PSNR    -0.42 / 2.86 / 27.32
    Overall PSNR    -0.90 / 2.00 / 17.27
    SSIM            -0.05 / 3.95 / 37.46

  Two pass CBR:
                    Min  / Mean / Max (pct)
    Average PSNR    -4.47 / 4.35 / 35.99
    Overall PSNR    -3.40 / 4.18 / 36.46
    SSIM            -4.56 / 6.98 / 53.67

  One pass VBR:
                    Min  / Mean / Max (pct)
    Average PSNR    -5.21 /  0.01 / 3.30
    Overall PSNR    -8.10 / -0.38 / 1.21
    SSIM            -7.38 / -0.11 / 3.17
    (note: most values here were close to the mean, there were a few
     outliers on files that were very sensitive to golden frame size)

  Two pass VBR:
                    Min  / Mean / Max (pct)
    Average PSNR    0.00 / 0.00 / 0.00
    Overall PSNR    0.00 / 0.00 / 0.00
    SSIM            0.00 / 0.00 / 0.00

Neither one pass or two pass CBR mode adheres particularly strictly
to the short term buffer constraints, and two pass is less
consistent, even in the baseline commit. This should be addressed
in a later commit. This likely will hurt the quality numbers, as it
will have to reduce the burstiness of golden frames.

Aside: My work on this commit makes it clear that we need to make
rate control modes "pluggable", where you can easily write a new
one or work on one in isolation.

Change-Id: I1ea9a48f2beedd59891f1288aabf7064956b4716

13 years agoRemoved unnecessary bmi motion vector stores.
Scott LaVarnway [Fri, 3 Jun 2011 17:05:16 +0000 (13:05 -0400)]
Removed unnecessary bmi motion vector stores.

left_block_mv and above_block_mv will return the MB
motion vector for non SPLITMV macro blocks.

Change-Id: I58dbd7833b4fdcd44b6b72e98ec732c93c2ce4f4

13 years agoMerge "Removed B_MODE_INFO"
Scott LaVarnway [Fri, 3 Jun 2011 15:32:30 +0000 (08:32 -0700)]
Merge "Removed B_MODE_INFO"

13 years agoAdjust bounds checking for hex search in real-time mode
Yunqing Wang [Thu, 2 Jun 2011 21:33:17 +0000 (17:33 -0400)]
Adjust bounds checking for hex search in real-time mode

Currently, hex search couldn't guarantee the motion vector(MV)
found is within the limit of maximum MV. Therefore, very large
motion vectors resulted from big motion in the video could cause
encoding artifacts. This change adjusted hex search bounds
checking to make sure the resulted motion vector won't go out
of the range. James Berry, thank you for finding the bug.

Change-Id: If2c55edd9019e72444ad9b4b8688969eef610c55

13 years agoRemoved B_MODE_INFO
Scott LaVarnway [Thu, 2 Jun 2011 17:46:41 +0000 (13:46 -0400)]
Removed B_MODE_INFO

Declared the bmi in BLOCKD as a union instead of B_MODE_INFO.
Then removed B_MODE_INFO completely.

Change-Id: Ieb7469899e265892c66f7aeac87b7f2bf38e7a67

13 years agoDon't allow very short GF groups even when the GF is predicted from an ARF.
Ronald S. Bultje [Thu, 2 Jun 2011 16:14:51 +0000 (09:14 -0700)]
Don't allow very short GF groups even when the GF is predicted from an ARF.

This is basically a slightly modified version of the previous patch,
and it has a moderately positive effect (SSIM/PSNR both +0.08% avg
on derf-set). Most clips show no change, except waterfall/coastguard,
each ~ +0.8% SSIM/PSNR. You can see similar effects in other clips
by shortening their length to terminate at a very short last group
of frames.

Change-Id: I7a70de99ca1f9fe6a8b6ca7a6e30e8a4b64383e4

13 years agoMerge "further clean up of errorperbit and sadperbit"
Yaowu Xu [Thu, 2 Jun 2011 15:58:03 +0000 (08:58 -0700)]
Merge "further clean up of errorperbit and sadperbit"

13 years agofurther clean up of errorperbit and sadperbit
Yaowu Xu [Wed, 1 Jun 2011 18:41:05 +0000 (11:41 -0700)]
further clean up of errorperbit and sadperbit

this commit makes the usage errorperbit and sadperbit consistent for
encoding modes and passes. Removed all different magic weight factors
associated with errorperbit. Now 1/2 is used for both sadperbit16 and
sadperbit4, the /2 operation is merged into initializations of the 2
variables.

Tests on cif set show .23%, 0.18% and 0.19% gain by avg psnr, overall
psnr and ssim respectively.

Change-Id: Ifa285c3e065ce0a5a77addfc9f95aabf54ee270d

13 years agoMerge "Bugfix in vp8dx_set_reference"
John Koleszar [Wed, 1 Jun 2011 20:57:23 +0000 (13:57 -0700)]
Merge "Bugfix in vp8dx_set_reference"

13 years agoBugfix in vp8dx_set_reference
Henrik Lundin [Wed, 1 Jun 2011 19:41:12 +0000 (21:41 +0200)]
Bugfix in vp8dx_set_reference

The fb_idx_ref_cnt book-keeping was in error. Added an assert to
prevent future errors in the reference count vector. Also fixed a
pointer syntax error.

Change-Id: I563081090c78702d82199e407df4ecc93da6f349

13 years agoMerge "Fix code under #if CONFIG_INTERNAL_STATS."
John Koleszar [Wed, 1 Jun 2011 18:14:17 +0000 (11:14 -0700)]
Merge "Fix code under #if CONFIG_INTERNAL_STATS."

13 years agoFix code under #if CONFIG_INTERNAL_STATS.
Ronald S. Bultje [Wed, 1 Jun 2011 18:10:13 +0000 (11:10 -0700)]
Fix code under #if CONFIG_INTERNAL_STATS.

Change-Id: Iccbd78d91c3071b16fb3b2911523a22092652ecd

13 years agoremove some magic weights associated with sad_per_bit
Yaowu Xu [Thu, 1 Jul 2010 01:58:54 +0000 (18:58 -0700)]
remove some magic weights associated with sad_per_bit

sad_per_bit has been used for a number of motion vector search routines
with different magic weights: 1, 1/2 and 1/4. This commit remove these
magic numbers and use 1/2 for all motion search routines, also reformat
a number of source code lines to within 80 column limit.

Test on cif set shows overall effect is neutral on all metrics. <=0.01%

Change-Id: I8a382821fa4cffc9c0acf8e8431435a03df74885

13 years agoneon fast quantize block pair
Tero Rintaluoma [Mon, 9 May 2011 07:09:41 +0000 (10:09 +0300)]
neon fast quantize block pair

vp8_fast_quantize_b_pair_neon function added to quantize
two adjacent blocks at the same time to improve performance.
 - Additional 3-6% speedup compared to neon optimized fast
   quantizer (Tanya VGA@30fps, 1Mbps stream, cpu-used=-5..-16)

Change-Id: I3fcbf141e5d05e9118c38ca37310458afbabaa4e

13 years agoMerge "vp8_pick_inter_mode code cleanup"
Scott LaVarnway [Tue, 31 May 2011 19:31:46 +0000 (12:31 -0700)]
Merge "vp8_pick_inter_mode code cleanup"

13 years agovp8_pick_inter_mode code cleanup
Scott LaVarnway [Tue, 31 May 2011 18:24:42 +0000 (14:24 -0400)]
vp8_pick_inter_mode code cleanup

Small code cleanups before attempting to reduce the size
of bmi found in BLOCKD.

Change-Id: Ie9c14adb53afd847716a75bcce067d0e6c04f225

13 years agoInitialize first_time_stamp_ever
John Koleszar [Tue, 31 May 2011 16:37:45 +0000 (12:37 -0400)]
Initialize first_time_stamp_ever

Misplaced #endif caused first_time_stamp_ever to only be initialized if
CONFIG_INTERNAL_STATS was set.

Change-Id: I2296a4ab00f7dfb767583edcc5d59b94f48c0621

13 years agoadds preload for armv6 encoder asm
Tero Rintaluoma [Mon, 30 May 2011 08:10:03 +0000 (11:10 +0300)]
adds preload for armv6 encoder asm

Added preload instructions to armv6 encoder optimizations.
About 5% average speed-up on Tegra2 for VGA@30fps sequence.

Change-Id: I41d74737720fb71ce7a316f07555357822f3347e

13 years agoMerge "bug fix check frame buffer index before copy"
John Koleszar [Fri, 27 May 2011 19:35:06 +0000 (12:35 -0700)]
Merge "bug fix check frame buffer index before copy"

13 years agobug fix check frame buffer index before copy
James Berry [Fri, 27 May 2011 18:57:25 +0000 (14:57 -0400)]
bug fix check frame buffer index before copy

in onyx_if.c update_reference_frames() make
sure that frame buffer indexes are not equal
before preforming a buffer copy.  If two frames
share the same buffer the flags will already be
set correctly.

Change-Id: Ida9b5516d08e3435c90f131d2dc19d842cfb536e

13 years agoMerge "Use hex search for realtime mode speed>4"
Yunqing Wang [Fri, 27 May 2011 18:12:50 +0000 (11:12 -0700)]
Merge "Use hex search for realtime mode speed>4"

13 years agoUse hex search for realtime mode speed>4
Yunqing Wang [Fri, 27 May 2011 17:31:15 +0000 (13:31 -0400)]
Use hex search for realtime mode speed>4

Test showed using hex search in realtime mode largely speed up
encoding process, and still achieves similar quality like the
diamond search we have. Therefore, removed the diamond search
option.

Change-Id: I975767d0ec0539f9f6ed7fdfc09506e39761b66c

13 years agoMerge "Broken EC after MODE_INFO size reduction"
Scott LaVarnway [Fri, 27 May 2011 14:52:04 +0000 (07:52 -0700)]
Merge "Broken EC after MODE_INFO size reduction"

13 years agoMerge "Remove unused code"
Yunqing Wang [Fri, 27 May 2011 14:25:25 +0000 (07:25 -0700)]
Merge "Remove unused code"

13 years agoRemove unused code
Yunqing Wang [Fri, 27 May 2011 14:20:49 +0000 (10:20 -0400)]
Remove unused code

Hex search is not called in rdopt.c

Change-Id: I67347f03e13684147a7c77fb9e9147e440bb5e8e

13 years agoBroken EC after MODE_INFO size reduction
Scott LaVarnway [Thu, 26 May 2011 19:13:00 +0000 (15:13 -0400)]
Broken EC after MODE_INFO size reduction

This patch fixes the compiler errors and the seg fault
when running decode_with_partial_drops.

Change-Id: I7c75369e2fef81d53b790d5dabc327218216838b

13 years agoMerge "Do not copy data between encoder reference buffers."
John Koleszar [Thu, 26 May 2011 16:58:26 +0000 (09:58 -0700)]
Merge "Do not copy data between encoder reference buffers."

13 years agoMerge "fix the mix use of errorperbit and sadperbit"
Yaowu Xu [Thu, 26 May 2011 16:39:41 +0000 (09:39 -0700)]
Merge "fix the mix use of errorperbit and sadperbit"

13 years agoMerge "Use int_mv instead of MV in vp8_mv_cont"
Scott LaVarnway [Thu, 26 May 2011 14:01:38 +0000 (07:01 -0700)]
Merge "Use int_mv instead of MV in vp8_mv_cont"

13 years agofix the mix use of errorperbit and sadperbit
Yaowu Xu [Wed, 25 May 2011 23:07:37 +0000 (16:07 -0700)]
fix the mix use of errorperbit and sadperbit

error_per_bit and sad_per_bit were designed as estimates of a bit worth
of sum squared error and sum absolute difference respectively. Under
this assumption, error_per_bit should be used in combination with 2nd
order errors (variance or sum squared error) while sad_per_bit should
be used in combination with 1st order SADs in motion estimation. There
were a few places where sad_per_bit has been misused with variances,
this commit changes to use error_per_bit for those places, also changes
parameter names to properly indicate which constant is being used.

On cif set, the change has a universal gain by all metrics: 0.13% by
average/overall psnr and 0.1% by ssim.

Change-Id: I4850fdcc3fd6886b30f784bd843f13dd401215fb

13 years agoMerge " Use var8x8 instead of get8x8var in VP8_UVSSE"
Yunqing Wang [Wed, 25 May 2011 18:35:42 +0000 (11:35 -0700)]
Merge "  Use var8x8 instead of get8x8var in VP8_UVSSE"

13 years agoMerge "Return sse value in vp8_variance SSE2 functions"
Yunqing Wang [Wed, 25 May 2011 18:31:07 +0000 (11:31 -0700)]
Merge "Return sse value in vp8_variance SSE2 functions"

13 years agoremove code not in use
Yaowu Xu [Wed, 25 May 2011 16:35:54 +0000 (09:35 -0700)]
remove code not in use

Change-Id: I6e5e86235d341cce3b02abda26dbeb71940ed955

13 years agoReturn sse value in vp8_variance SSE2 functions
Yunqing Wang [Wed, 25 May 2011 13:26:29 +0000 (09:26 -0400)]
Return sse value in vp8_variance SSE2 functions

Minor modification.

Change-Id: I09511d38fd1451d5c4106a48acdb3f766ce59cb7

13 years ago Use var8x8 instead of get8x8var in VP8_UVSSE
Attila Nagy [Wed, 25 May 2011 09:54:34 +0000 (12:54 +0300)]
  Use var8x8 instead of get8x8var in VP8_UVSSE

  'sum' returned by get8x8var is not used and var8x8 has optimizations
  for more platforms.

Change-Id: I4a907fb1a05f285669fb0b95dc71d42182c980f6

13 years agoFix a bug happening while encoding at profile=3
Yunqing Wang [Tue, 24 May 2011 19:59:37 +0000 (15:59 -0400)]
Fix a bug happening while encoding at profile=3

While profile=3, there is no sub-pixel search. Distortion and SSE
have to calculated using get_inter_mbpred_error().

Change-Id: Ifb36e17eef7750af93efa7d0e2870142ef540184

13 years agoUse int_mv instead of MV in vp8_mv_cont
Scott LaVarnway [Tue, 24 May 2011 20:01:12 +0000 (16:01 -0400)]
Use int_mv instead of MV in vp8_mv_cont

Less operations.

Change-Id: Ibb9cd5ae66b8c7c681c9a654d551c8729c31c3ae

13 years agoRemoved unused variable warnings
Scott LaVarnway [Tue, 24 May 2011 19:17:03 +0000 (15:17 -0400)]
Removed unused variable warnings

Change-Id: I6e5e921f03dc15a72da89a457848d519647677a3

13 years agoMerge "MODE_INFO size reduction"
Scott LaVarnway [Tue, 24 May 2011 19:08:24 +0000 (12:08 -0700)]
Merge "MODE_INFO size reduction"

13 years agoMODE_INFO size reduction
Scott LaVarnway [Tue, 24 May 2011 17:24:52 +0000 (13:24 -0400)]
MODE_INFO size reduction

Declared the bmi in MODE_INFO as a union instead of B_MODE_INFO.
This reduced the memory footprint by 518,400 bytes for 1080
resolutions.  The decoder performance improved by ~4% for the
clip used and the encoder showed very small improvements. (0.5%)
This reduction was first mentioned to me by John K. and in a
later discussion by Yaowu.
This is WIP.

Change-Id: I8e175fdbc46d28c35277302a04bee4540efc8d29

13 years agoMerge "Fixing bug in VP8_SET_REFERENCE decoder control command"
John Koleszar [Tue, 24 May 2011 12:57:44 +0000 (05:57 -0700)]
Merge "Fixing bug in VP8_SET_REFERENCE decoder control command"

13 years agoMerge "Rewrite hex search function"
Yunqing Wang [Tue, 24 May 2011 12:26:16 +0000 (05:26 -0700)]
Merge "Rewrite hex search function"

13 years agoFixing bug in VP8_SET_REFERENCE decoder control command
Henrik Lundin [Mon, 23 May 2011 11:47:33 +0000 (13:47 +0200)]
Fixing bug in VP8_SET_REFERENCE decoder control command

In vp8dx_set_reference, the new reference image is written to an
unused reference frame buffer.

Change-Id: I9e4f2cef5a011094bb7ce7b2719cbfe096a773e8

13 years agoMerge "use get8x8var directly for non-subpixel motion case in VP8_UVSSE"
Yaowu Xu [Mon, 23 May 2011 21:49:56 +0000 (14:49 -0700)]
Merge "use get8x8var directly for non-subpixel motion case in VP8_UVSSE"

13 years agoRewrite hex search function
Yunqing Wang [Fri, 20 May 2011 19:26:32 +0000 (15:26 -0400)]
Rewrite hex search function

Reduced some bound checks in hex search function.

Change-Id: Ie5f73a6c227590341c960a74dc508cff80f8aa06

13 years agouse get8x8var directly for non-subpixel motion case in VP8_UVSSE
Yaowu Xu [Sun, 22 May 2011 04:51:21 +0000 (21:51 -0700)]
use get8x8var directly for non-subpixel motion case in VP8_UVSSE

VP8_UVSSE mistakenly used subpixvar8x8 to calculate SSE for non-subpixl
motion cases.

Change-Id: I4a5398bb9ef39c211039f6af4540546d4972e6a9

13 years agoMerge "bug fix active_worst_quality set below active_best_quality"
John Koleszar [Fri, 20 May 2011 18:23:10 +0000 (11:23 -0700)]
Merge "bug fix active_worst_quality set below active_best_quality"

13 years agoMerge "cleanup: collect twopass variables"
John Koleszar [Fri, 20 May 2011 18:20:44 +0000 (11:20 -0700)]
Merge "cleanup: collect twopass variables"

13 years agoMerge "Fixed iwalsh_neon build problems with RVDS4.1"
Johann [Fri, 20 May 2011 14:51:11 +0000 (07:51 -0700)]
Merge "Fixed iwalsh_neon build problems with RVDS4.1"

13 years agoMerge "revise two function definitions with less parameters"
Yaowu Xu [Fri, 20 May 2011 14:45:42 +0000 (07:45 -0700)]
Merge "revise two function definitions with less parameters"