Commit Graph

189 Commits

Author SHA1 Message Date
Jim Bankoski
85a541a421 Reapply 'Amend and improve VP8 multithreading implementation'
Reapply this patch:
ff0107f Amend and improve VP8 multithreading implementation

Amended the patch to add a unit test, and fix an asan error.

BUG=webm:851

Change-Id: I6572c03256169c64e80248bf5a5e99f59a2fc93c
2016-12-13 02:11:34 +00:00
Jim Bankoski
3e21d703ce Drop empty frames.
Change-Id: I2d45a6eb3aaca97eb61e8e7ef9e5114221091244
2016-10-14 06:28:14 -07:00
Marco
df66f8e830 vp8: Change default gf behavior for 1 pass cbr.
In 1 pass CBR, with error_resilience off, allow for
special logic to change the default gf behaviour.
In this CL: boost is turned off and the gf period
is set to a multiple of cyclic refresh period.

Change only affect 1 pass CBR mode, i.e, when the flag
gf_update_onepass_cbr is set.

Including the previous change (3ec8e11: to allow cyclic refresh
for error_resilience off), comparing metrics on RTC set for
error_resilience off vs on: avgPSNR/SSIM up by ~6%.

Change-Id: Id5b3fb62a4f04de5a805bd1b418f2b349574e0bc
2016-10-07 11:13:06 -07:00
James Zern
8281da74b9 vp8: convert some uses of unsigned long to size_t
similar to changes that were done in vp9 for encoded frame size
reporting. has the side-effect of quieting a -Wshorten-64-to-32 warning.

Change-Id: I89f74cb617fc29334ee351dc8dfaa3b8cfd4e5af
2016-09-19 18:35:59 -07:00
Alexander Potapenko
948a1f51d0 vp8: Remove TSAN warning around end of encode.
Tsan warns when run in one pass and there is a recode
loop.

Change-Id: Ice2ecb2270f09ebd49efbd49c0e4f77d32e23c0f
2016-09-08 14:36:32 +02:00
Johann Koenig
4d1540f8ce Merge changes from topic 'Wundef'
* changes:
  Enable -Wundef by default
  Define VP8_TEMPORAL_ALT_REF to !CONFIG_REALTIME_ONLY
  Remove CONFIG_DEBUG guards from assert()
  Remove unused function vpx_de_mblock
  Fix -Wundef warning for OUTPUT_FPF
  Fix -Wundef warning for __SANITIZE_ADDRESS__
2016-09-02 01:39:18 +00:00
Johann
1139f0dbc2 Define VP8_TEMPORAL_ALT_REF to !CONFIG_REALTIME_ONLY
Previously VP8_TEMPORAL_ALT_REF was only defined for non-realtime-only
builds. However, its value was checked with #if, not #ifdef.

Fixes -Wundef warnings.

BUG=webm:1069

Change-Id: If78d8731298f3f0d3662ffa25f973e7adaf67152
2016-08-31 23:01:57 -07:00
Marco
b6a5f6f740 vp8: Move loopfilter synchronization to end of encode_frame call.
Allow loopfilter to continue until encode_frame is completed.

Change-Id: I7bbccc3d409e263aab6a6ff24588d8b2a964a96e
2016-08-25 12:37:30 -07:00
Yunqing Wang
87c6c5224d Revert "Amend and improve VP8 multithreading implementation"
Reverted the patch because of possible performance issue.

Change-Id: I49944f827ccd38ed194c9f8d9cb9036fa9bf79e1
2016-07-21 12:28:25 -07:00
clang-format
81a6739533 vp8: apply clang-format
Change-Id: I7605b6678014a5426ceb45c27b54885e0c4e06ed
2016-07-15 19:28:44 -07:00
Yaowu Xu
003a9d20ad Port metric computation changes from nextgenv2
Change-Id: I4aceffcdf7af59ffeb51984f0345c3a4c7e76a9f
2016-06-24 13:52:50 -07:00
Johann
f4b5330c50 VP8_COMP: make frames_since_golden signed
This value is signed in vp9/10

Cleans warning in Android build:
comparison of integers of different signs: 'unsigned int' and 'int'
if (cpi->frames_since_golden == (cpi->current_gf_interval >> 1))
    ~~~~~~~~~~~~~~~~~~~~~~~~ ^   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Change-Id: Ie137724982f3a46c8c1820548c1960d62a4e96f2
2016-06-16 22:02:30 +00:00
Yunqing Wang
b198bcd528 Simplify the loopfilter synchronization logic in VP8 encoder
This patch was to fix a reported Hangouts deadlock/freezing issue
in VP8 encoder(issue 27232610). The original encoder loopfilter
synchronization happened in the following frame, which was prone
to causing problems in some complex use cases. This patch simplified
the synchronization logic.

More testing needs to be done.

Change-Id: I38fd3f35d11f98fae1e44546aa5e4c6d6e19c4be
2016-03-22 14:13:18 -07:00
Yunqing Wang
ff0107f60d Amend and improve VP8 multithreading implementation
There are flaws in current implementation of VP8 multithreading encoder
and decoder as reported in the following issue:
https://code.google.com/p/chromium/issues/detail?id=158922

Although the data race warnings are harmless, and wouldn't cause real
problems while encoding and decoding videos, it is better to fix the
warnings so that VP8 code could pass the TSan test.

To synchronize the thread-shared data access and maintain the speed
(i.e. decoding speed), use multiple mutexes based on mb_rows to reduce
the number of synchronizations needed, make the reads and writes of
the shared data protected, and reduce the number of mb_col writes by
nsync times.

The decoder speed tests showed < 3% speed loss while using 2 ~ 4
threads.

Change-Id: Ie296defffcd86a693188b668270d811964227882
2016-01-08 11:59:49 -08:00
Ronald S. Bultje
54d48955f6 vp8: change build_intra_predictors_mby_s to use vpx_dsp.
Change-Id: I2000820e0c04de2c975d370a0cf7145330289bb2
2015-09-30 18:45:40 -04:00
Johann
79a00d71bd Rename vp8 quantize.c
Move it to vp8_quantize.c and make sure to
use the full path for all vp8 includes of quantize.h

Change-Id: I284651ff681707385f4924ea7db1541905c1624a
2015-08-18 11:44:32 -07:00
Johann
6a82f0d7fb Move sub pixel variance to vpx_dsp
Change-Id: I66bf6720c396c89aa2d1fd26d5d52bf5d5e3dff1
2015-07-07 15:51:04 -07:00
Marco
976f7f42c1 VP8: For high overshoot, force drop frame and max-out QP.
This allows rate control to react to content of current frame being encoded.
Enabling this feature via the setting: screen_content_mode = 2.

Change-Id: Ib2c6670551d96f4907495d5b7b76bb8c49e673db
2015-05-19 08:19:21 -07:00
James Zern
60b9f685cb vp8/rdopt.h+onyx_int.h: add some missing prototypes
silences missing prototype warnings

Change-Id: Icd477e37b502205d0a60e7389e51b1ba17d8888e
2015-05-14 22:41:26 -07:00
Alex Converse
581731a95f vp8enc: Prevent out of bounds memory access.
Prevent out of bounds access when attempting to increase frame size

Change-Id: I710c40c692802a72963c9680c2125da17f9060a9
2015-01-27 16:44:31 -08:00
Marco
af898b56bb Various updates to vp8.
Change-Id: Icc7a816491897107764e4c936288e9000e6319b8
2014-12-03 16:01:28 -08:00
Marco
21a0dade42 vp8: Modify to use closest reference in zero_mv bias.
Modify zero_mv bias condition to include check that "closest" reference is last_frame.
This is needed for temporal layers, where the last_frame is not always the closest reference.
Also, constain zeromv_count to be for last_frame reference.

Change-Id: I7af54a809ebf01ef43b9933c9d4095b6cb189390
2014-08-14 15:55:20 -07:00
Marco Paniconi
91801176fe vp8: Add an aggressive denoising mode.
Change-Id: Ie4686e1b15af6bcc8d59d585bbeb996f38224522
2014-07-29 11:16:45 -07:00
Joey Parrish
18c08607e0 Add VPXD_SET_DECRYPTOR support to the VP9 decoder.
Change-Id: I88f86c8ff9af34e0b6531028b691921b54c2fc48
2014-04-23 16:11:54 -07:00
James Zern
513fae3ee6 vp8/encoder: add extern "C" to headers
Change-Id: I252f5f8a5d5ada65da08699774a7bb1eb2bd5b2e
2014-01-23 16:21:24 -08:00
James Zern
bb28520891 vp8/encoder: normalize include guards
Change-Id: I82834550503a43ff7ec8422342dc65136453b287
2013-12-16 19:41:01 -08:00
James Zern
5f30a0c687 VP[89]_COMMON: remove golden/altref frame counts
these are only used in the encoder.
frames_since_golden / frames_till_alt_ref_frame -> VP[89]_COMP

Change-Id: Ie14a6f46987bced685ddb449b85dc261caba6dfe
2013-07-18 14:09:21 -07:00
James Zern
9581eb6e8a use consistent framerate naming
s/frame_rate/framerate/g

Change-Id: I6fc3e088e419c5f46e3a9390dd8a2cad2677a2fc
2013-07-16 14:12:47 -07:00
Marco Paniconi
fb48bbf5c6 Fixes for run-time change in temporal layers.
Use a separate counter for resetting the pattern upon
a change in temporal layers, and set/initialize the
layer context parameters for the new temporal layer state.

Also moved the setting of layer configuation in init_config()
to a separate function.

Change-Id: Ic7fc023a1e5490020509e107004645098f4c00f0
2013-05-21 09:04:14 -07:00
Scott LaVarnway
699196dc3d Removed mode_chosen_counts
This eliminates a tsan warning.

Change-Id: I204d109b12adb4e347762e63d12859244f28d1a9
2013-03-20 14:41:24 -07:00
Johann
403145032d Merge "Use intrinsics for sse2 regular quantize" 2013-03-01 17:20:26 -08:00
Johann
eca59cad0b Use intrinsics for sse2 regular quantize
Remove dependency of this function on asm_offsets. ssse3/sse4 next.

Change quant_shift calculation so it be done using SIMD. Pre-calculate
as much as possible to simplify EOB selection.

Take advantage of qcoeff being zero'd by tying the if statements
together.

Speed parity with previous implementation with gcc x86_64 linux

Change-Id: Ife97556a1eca3a74b09def1a3d04084974dff1fb
2013-02-28 18:06:15 -08:00
James Zern
4e00060d29 vp8: make gf_group_bits 64-bit
avoids signed integer overflow; matches kf_group_bits

Change-Id: I193145cdc4fa53e70fba0a1731a03eb1a574931d
2013-02-22 12:45:28 -08:00
Scott LaVarnway
a0ad16e203 Moved error_bins to macroblock struct
Change-Id: Ic9956ddf1c2ddffcf7be7fdfc23ad9a2426fc47a
WIP: Fixing unsafe threading in VP8 encoder.
2012-12-10 17:32:58 -08:00
Scott LaVarnway
74efda4bd6 Moved zbin_mode_boost to macroblock struct
Fixing unsafe threading in VP8 encoder.

Change-Id: Ibf4c89a2043654834747811bc11eb283de0bb830
2012-12-10 12:42:24 -08:00
Scott LaVarnway
3a19eebe4d Moved zbin_over_quant to macroblock struct
Change-Id: I76fe20ade099573997404b8733cf7f79e82fb21e
WIP: Fixing unsafe threading in VP8 encoder.
2012-12-10 10:51:42 -08:00
Scott LaVarnway
f2b36a4de7 Removed check_gf_quality()
and various unused members in VP8_COMP along with other
code cleanups.

Change-Id: I56c6c0a77a51f5ac5cbd6071017bcbfd2623b7df
2012-12-05 08:56:42 -08:00
Scott LaVarnway
bfca084fcd Moving mbs_tested_so_far, mode_test_hit_counts to macroblock struct
Change-Id: Ifa78c0a953fab3e5dd7af0446924846c7022cd09
2012-12-04 16:52:47 -08:00
Scott LaVarnway
9961ad479a Merge "Moving rd_thresh_mult, rd_threshes to macroblock struct" 2012-12-03 12:05:48 -08:00
Scott LaVarnway
69d074841d Moving count_mb_ref_frame_usage to macroblock struct
Change-Id: I44e4e3869f231ae270cca98c9565f23c512e3ddf
2012-11-06 16:58:28 -08:00
Scott LaVarnway
80fe4723d2 Merge "Removed mb_norm_activity_map" 2012-11-06 16:28:54 -08:00
Scott LaVarnway
fe91e47bc7 Moving rd_thresh_mult, rd_threshes to macroblock struct
Change-Id: I650a593162280ab40e71e527ec6518303e2d5723
2012-11-06 16:27:00 -08:00
Scott LaVarnway
78bf8284cf Removed mb_norm_activity_map
Not used.

Change-Id: I07ba929e3ff82b4b0c86f731769cdf5434c7a816
2012-11-06 09:47:45 -08:00
Scott LaVarnway
ee28bb87b4 Moving _error counts to macroblock struct
Change-Id: I28ac1519d1594801fef9a623cb64598d3d751eb0
2012-11-06 09:21:54 -08:00
Scott LaVarnway
01824d1848 Moving MVcount to macroblock struct
Change-Id: Ie22841d096f3c86694b95bd06fc3a8ce1f032a10
2012-11-06 08:51:11 -08:00
Scott LaVarnway
95390b2b20 Moving ymode_count, uv_mode_count to macroblock struct
Change-Id: Ib73c7b2bee4cb2eb2528fa6b381fffe9503079a0
2012-11-05 12:25:18 -08:00
Scott LaVarnway
03c0af8747 Moved skip_true_count to macroblock struct
Change-Id: Ie9a26be7c9baa54a0e43a63ed6c77f2746477a9c
2012-11-05 11:02:35 -08:00
Scott LaVarnway
7ee44eef13 Moving coef_counts to macroblock struct
Change-Id: I289564a5a27f0d03ddc6f19c7838542ff22719be
2012-11-05 11:00:49 -08:00
Yunqing Wang
24c346dffc Add biasing to ZEROMV for videos with static background
For videos with big static background(such as video conferencing
clips), the mode decision was biased to ZEROMV in order to
obtain a stable background. The percentage of ZEROMV on last
frame was used to predict if there is static area in current frame,
and checking already-encoded neighboring macroblocks' motion
vectors to make sure the local area has low motion.

Change-Id: I05b3241d3a56a0bda88b6681e5646c1c8baf2e57
2012-08-21 14:56:02 -07:00
Yaowu Xu
d71ba03822 silent compiling warnings for VC9 build
Change-Id: Iaa947e640f27e6f6eaf7d845f243536bca2df513
2012-08-20 11:45:01 -07:00