Commit Graph

12199 Commits

Author SHA1 Message Date
Alex Converse
d8362a3887 Merge "Fix LAST SKIP when considering GOLDEN" 2014-11-07 16:21:37 -08:00
Deb Mukherjee
cc57c5e4af Iadst transforms to use internal low precision
Change-Id: I266777d40c300bc53b45b205144520b85b0d6e58
(cherry picked from commit a1b726117f)
2014-11-07 14:19:45 -08:00
Deb Mukherjee
27dce0f324 Test name changes to use SSE/SSE2 exactly
Change-Id: I3b5a478d198868c2796366f0ac59d0e2036308b8
2014-11-07 13:44:19 -08:00
Alex Converse
ce9ba97a9d Fix LAST SKIP when considering GOLDEN
Change-Id: I39d9f13fa34984ee9dad0c4f303ef672635f420e
2014-11-07 13:44:17 -08:00
Paul Wilkins
08d86bc904 Merge "Add intra complexity and brightness weight to first pass." 2014-11-07 09:22:12 -08:00
Yaowu Xu
98492c1091 Merge "Change the use of a reserved color space entry" 2014-11-07 06:24:59 -08:00
Paul Wilkins
31b6d7c1eb AQ1 - remove first pass weights.
Removed redundant weighting function tied for AQ1 from first
pass code.

Improvment in baseline AQ1 results:-
Derf  opsnr +0.142% SSIm +0.258%
YT  opsnr +0.173% SSIm +0.3%

Change-Id: I16ef91caf2d7f302cd5940cc5e2626d48ebcb212
2014-11-07 14:11:29 +00:00
Yaowu Xu
af3519a385 Change the use of a reserved color space entry
This commit rename a reserved color space entry to BT_2020, it intends
to provide support for VP9 bitstream to pass along the color space
type defined in BT.2020(Rec.2020)

please note this entry does not have any effect on encoding/decoding
behavior, but allow applications to the pass the information along
from encoding end to decoding end.

Change-Id: I4678520e89141ea5e8900f7bd1c0e95b710b7091
2014-11-06 19:14:21 -08:00
Jingning Han
754b05a4de Refactor nonrd_use_partition coding process
This commit integrates the non-RD mode decision process and the
encoding process into a single recursion scheme.

Change-Id: I6a7e72a0b84d567554801ebbe01ec75d54c1f77d
2014-11-06 17:00:48 -08:00
Johann
6eec73a747 Remove asm offset dependencies
The obj_int_extract code is no longer worth maintaining. It creates
significant issues when adapting for different build systems and no
longer offers as significant of a performance benefit due to
improvements in intrinsics.

Source files will remain until the various third-party builds are updated.

The neon fast quantizer has been moved to intrinsics. The armv6 version
has been removed because so few remaining targets require it.

Compilers and processors have improved significantly since the
pack_tokens code was written. The assembly is no longer faster than the
C code.

pack_tokens were the only optimizations for the armv5te targets so the targets
will be removed after the test infrastructure has been updated.

BUG=710

Change-Id: Ic785b167cd9f95eeff31c7c76b7b736c07fb30eb
2014-11-06 16:00:01 -08:00
Yunqing Wang
bf44117d5f Merge "Modify the frame context memory deallocation" 2014-11-06 13:08:57 -08:00
Jingning Han
417e754f56 Merge "Remove unused is_background function" 2014-11-06 12:03:15 -08:00
Jingning Han
e97f404e52 Merge "Rework cut-off decisions in cyclic refresh aq mode" 2014-11-06 12:03:07 -08:00
Yunqing Wang
1228433430 Modify the frame context memory deallocation
This patch was to fix the vpxdec fuzzing3 test failure. When an
error occurs, setjmp() is invoked, which calls the decoder
removing routine. In multiple thread situation, other threads
could try to access the frame context memory that is already
deallocated, thus causing a segfault.

An invalid unit test was added for this issue.

Change-Id: Ida7442154f3d89759483f0f4fe0324041fffb952
2014-11-06 11:34:19 -08:00
Paul Wilkins
5e935126a6 Add intra complexity and brightness weight to first pass.
The aim of this patch is to apply a positive weighting to
frames that have a significant number of blocks that are
of low spatial complexity and are dark. The rationale behind
this is that artifacts tend to be more visible in such frames.

In this patch the weight is only applied in regard to the distribution
of bits between frames. Hence if all the frames share similar
characteristics (as is the case for most of our short test clips) there
will be little or no net effect.

However, the effect can be seen on some longer form test content.

For example Tears of steel baseline test:
2323.09 Kbit/s opsnr 39.915 ssim 74.729
With this patch:-
2213.34 Kbit/s opsnr 39.963 ssim 74.808
(Sligtly better metrics and about 5% smaller)

The weighting may well need some further tuning along side changes
to the aq modes.

Change-Id: Ieced379bca03938166ab87b2b97f55d94948904c
2014-11-06 10:45:00 +00:00
Jingning Han
10da059b52 Remove unused is_background function
Change-Id: Ia540eac5f066ae95280c2f898370eddf0110c279
2014-11-05 21:19:23 -08:00
Jingning Han
caaf63b2c4 Rework cut-off decisions in cyclic refresh aq mode
This commit removes the cyclic aq mode dependency on
in_static_area and reworks the corresponding cut-off thresholds.
It improves the compression performance of speed -5 by 1.47% in
PSNR and 2.07% in SSIM, and the compression performance of speed
-6 by 3.10% in PSNR and 5.25% in SSIM. Speed wise, about 1% faster
in both settings at high bit-rates.

Change-Id: I1ffc775afdc047964448d9dff5751491ba4ff4a9
2014-11-05 21:17:09 -08:00
hkuang
e8860693ea Merge "Totally remove prev_mi in VP9 decoder." 2014-11-05 17:48:47 -08:00
hkuang
4cc7c5a17f Totally remove prev_mi in VP9 decoder.
This will save the memory and improve the decode speed due to
removing unnecessary memset of big prev_mi array for
all the key frames.

Decoding a all key frames 1080p video shows speed improve around 2%.

Change-Id: I6284a445c1291056e3c15135c3c20d502f791c10
2014-11-05 16:14:30 -08:00
Yaowu Xu
7e491de1c4 Merge "Fix visual studio 2013 compiler warnings" 2014-11-05 15:34:13 -08:00
Johann
92440e82fd Merge "vp8 quantize is now in intrinsics, not asm" 2014-11-05 14:05:26 -08:00
Yaowu Xu
2c4fee17bc Fix visual studio 2013 compiler warnings
For configured with --enable-vp9-highbitdepth

Change-Id: I2b181519d7192f8d7a241ad5760c3578255f24e6
2014-11-05 13:47:28 -08:00
Hui Su
2c95a3f374 Merge "Simplify interface of write_selected_tx_size and read_tx_size" 2014-11-05 13:33:09 -08:00
Johann
7d07f512cd Fix SSSE3 partial_idct_test detection
The test filter is not a prefix matcher. It requires test type to
contain no more than the optimization type. In this example, SSSE3_64
fails to match and the test is not skipped even when SSSE3 is not
available.

Change-Id: Ia74229a167c88da4e6da169012a7a77d438c3f75
2014-11-05 12:58:08 -08:00
Johann
d1b64deef4 vp8 quantize is now in intrinsics, not asm
Change-Id: Ie106c5335c9ba5aac81e23150e7026fb6ea6196e
2014-11-05 12:28:07 -08:00
Jingning Han
a7889cac9a Merge "Skip ref frame mode search conditioned on predicted mv residuals" 2014-11-05 12:04:10 -08:00
James Zern
df1b0cafd6 quantize_test: delete QuantizePair related tests
functions were removed in:
2134eb2 Remove pair quantization

fixes ARM build

Change-Id: I634de50c2752408381d70afb0f5088b61052853c
2014-11-05 13:11:34 +01:00
Johann
c4b3625393 Merge "Remove pair quantization" 2014-11-04 16:29:19 -08:00
Johann
412eaaf090 Merge "vp8 quantization -> intrinsics" 2014-11-04 16:28:46 -08:00
Hui Su
709c634b84 Simplify interface of write_selected_tx_size and read_tx_size
Change-Id: Ia2b2a895deefaaf7b34bf26df86add56dbab082c
2014-11-04 16:11:50 -08:00
Minghai Shang
9f9e30d7bf Merge "[spatial svc] Make spatial svc working for one pass rate control" 2014-11-04 15:57:16 -08:00
hkuang
825e270d01 Merge "Fix the memory leak due to missing free frame_mvs." 2014-11-04 14:43:47 -08:00
Johann
0593ce390d Merge "Avoid divide-by-zero in vp8 initialization" 2014-11-04 14:28:45 -08:00
hkuang
23da920a8e Fix the memory leak due to missing free frame_mvs.
Change-Id: I2ceee7341d906259002c0ea31ea009ae32c04bfd
2014-11-04 13:28:31 -08:00
Johann
b4957234af Avoid divide-by-zero in vp8 initialization
Check that the numerator is not zero. If it is, guess 30fps.

Fixes a clang IOC error in the quantize test. It's very unlikely for
this to occur in the wild because the setup in the quantize test is very
nonstandard.

Change-Id: Icdab7b81d4e168d3423e14db20787f960052e0c3
2014-11-04 12:52:10 -08:00
Minghai Shang
86c36a504d [spatial svc] Make spatial svc working for one pass rate control
Change-Id: Ibd9114485c3d747f9d148f64f706bf873ea473ac
2014-11-04 11:46:48 -08:00
Johann
dae280d537 Align structures in quantize test
MACROBLOCKD structures require alignment.

This fixes an issue caught by clang IOC.

Change-Id: Ibb5bcc122f531b4302a87e1144e4feaf46c1de64
2014-11-04 10:49:59 -08:00
Johann
9c6ce43df6 Add some indirection to the quantize test
Visual Studio and XCode require the address of the functions be passed.

Change-Id: Id39cab8e50061fdc6ac6018371a3a158c713e14b
2014-11-04 10:48:32 -08:00
Jingning Han
1e753387c8 Merge "Refactor sub-pixel motion search unit" 2014-11-04 09:11:15 -08:00
Jingning Han
1434f7695b Skip ref frame mode search conditioned on predicted mv residuals
This commit makes the RTC coding mode to conditionally skip the
reference frame mode search, when the predicted motion vector of
the current reference frame gives more than two times sum of
absolute difference compared to that of other reference frames.

It reduces the runtim by 1% - 4% for speed -5 and -6. The average
compression performance is improved by about 0.1% in both settings.

It is of particular benefit to light change scenarios. The
compression performance of test clip mmmovingvga.y4m is improved by
6.39% and 15.69% at high bit rates for speed -5 and -6, respectively.

Speed -5
vidyo1 16555 b/f, 40.818 dB, 12422 ms ->
       16552 b/f, 40.804 dB, 12100 ms

nik    33211 b/f, 39.138 dB, 11341 ms ->
       33228 b/f, 39.139 dB, 11023 ms

mmmoving 33263 b/f, 40.935 dB, 13508 ms ->
         33256 b/f, 41.068 dB, 12861 ms

Speed -6
vidyo1 16541 b/f, 40.227 dB, 8437 ms ->
       16540 b/f, 40.220 dB, 8216 ms

nik    33272 b/f, 38.399 dB, 7610 ms ->
       33267 b/f, 38.414 dB, 7490 ms

mmmoving 33255 b/f, 40.555 dB, 7523 ms ->
         33257 b/f, 40.975 dB, 7493 ms

Change-Id: Id2aef76ef74a3cba5e9a82a83b792144948c6a91
2014-11-04 09:10:19 -08:00
Yunqing Wang
6d90a9d289 Merge "WORKAROUND FIX FOR GCC4.9.1" 2014-11-03 16:56:38 -08:00
Johann
6bab322d49 Merge "vp8 quantize test" 2014-11-03 15:45:21 -08:00
Marco
343acaa8f2 Merge "Allow disable of refresh golden for more than 1 layer encoding." 2014-11-03 14:38:05 -08:00
Johann
53b68de565 vp8 quantize test
Test for Regular, Fast and Pair quantization

Change-Id: I0a26c164afe632db869099402189826c0d43f9a2
2014-11-03 14:15:34 -08:00
Jingning Han
e083f6bd08 Refactor sub-pixel motion search unit
This commit unfolds the legacy macro definitions used in the
sub-pixel motion search and refactors the operational flow for
later optimizations.

Change-Id: I3e3f770cad961d03d1a6eb0b2a0186cc77eaf2b8
2014-11-03 09:02:57 -08:00
Jingning Han
0ca5908ff6 Merge "Fix the THR_MODES array used in vp9_pick_inter_mode" 2014-11-03 08:46:42 -08:00
Yaowu Xu
2fe893c94f Merge "Fix speed 7 and speed 12 for rt" 2014-11-03 08:02:58 -08:00
Marco
d6b688375f Allow disable of refresh golden for more than 1 layer encoding.
The current logic was allowing for disabling golden refresh only
for two pass svc encoding. This change disables it as long as
more than 1 layer encoding is used (for example temporal layers under 1pass CBR).

Change-Id: I4dc5204a7ad365c821ec7963e93b59da82e1826b
2014-11-02 22:24:00 -08:00
Jingning Han
7e119e2946 Fix the THR_MODES array used in vp9_pick_inter_mode
Fix the alignment of entries fo intra prediction modes.

Change-Id: Ie32ad87cf90694efd591a4b1cc29c916c4cd56f7
2014-11-02 12:25:57 -08:00
levytamar82
86175a5788 WORKAROUND FIX FOR GCC4.9.1
In the function mb_lpf_horizontal_edge_w_avx2_16 the usage of the intrinsic
_mm256_cvtepu8_epi16 cause a compiler bug in gcc 4.9.1.
until it will be fixed I created a workaround that create the up convert by
using broadcast128+shuffle.
The bug was reported here:
https://code.google.com/p/webm/issues/detail?id=867

Change-Id: I73452e6806f42e0fadcde96b804ea3afa7eeb351
2014-11-01 11:27:28 -07:00