Commit Graph

18232 Commits

Author SHA1 Message Date
James Zern
10cb17aec0 Merge "runtime error fix: bitdepth_conversion_avx2.h" 2017-11-10 00:15:03 +00:00
Jerome Jiang
6246d8aa76 vp9: Fix mem rel for non-ref for external buffer.
Release frame buffers for non-ref when the decoder is destroyed.

Enable the non ref test.

BUG=b/68819248

Change-Id: Id87ef3b0a62318f9812e927cd957c05c859047fa
2017-11-09 15:47:21 -08:00
Jerome Jiang
0665b09661 Merge "vp9: SVC feature to use partition from lower resolution." 2017-11-09 23:28:44 +00:00
Jerome Jiang
fdb054a05d vp9: SVC feature to use partition from lower resolution.
For SVC with 3 spatial layers:
Add feature to copy/upscale partition from middle spatial layer
to the upper/highest resolution, when superblock sad is not high.

Enabled for speed >= 7 and only for non-reference frames.

Speedup ~3-4%, small loss in avgPNSR/SSIM of ~1%.

Change-Id: I7f0a2716c0fde28bade0f86159d11b7e31d6ab8d
2017-11-09 14:16:50 -08:00
Scott LaVarnway
2387024f41 runtime error fix: bitdepth_conversion_avx2.h
Change-Id: I7364a157de39eb7137b599808474b8d46d19d376
2017-11-09 12:26:43 -08:00
Johann Koenig
bdb8b3ad86 Merge "fail early on oversize frames" 2017-11-09 19:50:04 +00:00
Scott LaVarnway
62ab5e99c1 vpx: [x86] add vp9_block_error_fp_avx2()
SSE2 asm vs AVX2 intrinsics speed gains:
blocksize   16: ~1.00
blocksize   64: ~1.17
blocksize  256: ~1.67
blocksize 1024: ~1.81

Change-Id: I2a86db239cf57e3ff617890ccb2d236aba83ad5e
2017-11-09 05:02:31 -08:00
paulwilkins
d6e29868ac Fix to frames considered in arf boost calculation.
For a chosen interval "i" the existing arf boost calculation examined frames
+/- (i-1) frames from the current location in the second pass.

This change checks to make sure that the forward search does not extend
beyond the next key frame in the event that the distance to the next key
frame is < (i - 1).

Small metrics gains on all our  test sets but these are localized to a few clips
(e.g. midres set psnr-hvs sintel -2.59% but overall average was only -0.185%)

Change-Id: I26fc9ce582b6d58fa1113a238395e12ad3123cf6
2017-11-09 10:46:10 +00:00
Jerome Jiang
adbb4c4d32 Merge "vp9: Add nonref frame buffer test." 2017-11-09 04:41:10 +00:00
Jerome Jiang
a68bbcff29 vp9: Add nonref frame buffer test.
The new test will run a SVC bitstream which has non ref frames.
It checks the number of buffer acquired and released to make sure all
external frame buffers are released.

Add a new test bitstream:
vp90-2-22-svc_1280x720_1.webm
which has 400 frames in total, and 1 spatial layer and 2 temporal layers.
There is one non ref frame every other frame.

Disabled for now. Will be enabled with the fix.

BUG=b/68819248

Change-Id: I0515336fd9809a9e1fceba90e4dce53dabaf53a5
2017-11-08 18:41:33 -08:00
Johann Koenig
cf8039c25f Merge "Support building AVX-512 and implement sadx4 for AVX-512" 2017-11-08 16:28:40 +00:00
paulwilkins
93e83fd7cf CVBR command line option.
Added command line control of Corpus VBR.

The new corpus vbr mode is a variant of standard
VBR (end-usage=0) where the complexity distribution
mid point is passed in rather than calculated for a specific
clip or chunk.

The new variant is enabled by setting a new command line
parameter --corpus-complexity to a zero value. Omitting
this parameter or setting it to 0 will cause the codec to use
standard vbr mode.

The correct value for a given corpus needs to be derived
experimentally using a training set such that the average
rate for the corpus is close to the target value.

For example our using our low res test set with upper and lower
vbr limits of 50%-150% and a corpus complexity value of 650
gives a similar average data rate across the set to using standard
vbr. However, with the corpus mode easier clips will be allocated
fewer bits and harder clips more bits rather than having the same
rate target for all.

Change-Id: I03f0fc8c6fb0ee32dc03720fea6a3f1949118589
2017-11-08 10:41:04 +00:00
Marco
6fbc354c97 Nonrd_pickmode: avoid computing UV cost when early_term is set.
For nonrd_pickmode: if early_term is set there should be
no need to include UV in rdcost (when color_sensitivity is set).

Neutral change on RTC and RTC_derf metrics, for speed >= 5.
No change for ytlive metrics.

Very small speed gain (~0.5%) on some clips with strong color content.

Change-Id: Ifc00928ecd935fc71e94935ceef0ae7481249f07
2017-11-06 10:22:14 -08:00
Kyle Siefring
b383a17fa4 Support building AVX-512 and implement sadx4 for AVX-512
The added AVX-512 support requires the subset of AVX-512 added in Skylake-X.

Change-Id: I39666b00d10bf96d06c709823663eb09b89265b7
2017-11-03 13:37:23 -04:00
Marco
eb7d431cb5 Compound prediction mode for nonrd pickmode.
Allow for compound prediction mode in nonrd_pickmode for ZEROMV.
For real-time encoding, 1 pass with non-zero lag-in-frames.

Added speed feature to control the feature.
Enabled for speed >=6 for now, under VBR mode.

avgPSNR/SSIM metrics positive on ytlive set, for speed 6:
some clips up by ~3-5%, some clips neutral gain, average gain
across clips is ~1%.

Small/negligible decrease in speed.

Change-Id: I7a60c7596e69b9a928410c5ee2f9141eecd8613d
2017-11-03 10:13:05 -07:00
Johann
5fe82459ec fail early on oversize frames
Even though frame_size is calculated in uint64_t, it winds up in an int
size value.

This was exposed with the msan test because the memset is called with
(int)frame_size, leading to a segfault.

Change-Id: I7fd930360dca274adb8f3e43e5e6785204808861
2017-11-03 09:49:13 -07:00
Jerome Jiang
3ba9a2c8b2 Merge "vp9: Move allocation of vt2 after early exits." 2017-11-01 16:58:01 +00:00
Jerome Jiang
34805d6d0d vp9: Move allocation of vt2 after early exits.
Remove the memory deallocation on the early exits.

Change-Id: I00b4a814ae6705105ecab89644d055ca3311d9f4
2017-10-31 17:04:04 -07:00
Jerome Jiang
0c84b9b703 Merge "vp9: Reduce stack usage of choose_partitioning." 2017-10-31 21:42:18 +00:00
Jerome Jiang
18b470f486 vp9: Reduce stack usage of choose_partitioning.
Move vt2 to heap.
Reduce the stack usage from ~87K to ~44K.

BUG=b/68362457

Change-Id: I8f5f93712934d59a8cc4564378172d409a736a2e
2017-10-31 13:10:27 -07:00
Jerome Jiang
c77822615e Merge "vp9: Reduce stack usage of choose_partioning." 2017-10-30 23:39:41 +00:00
Jerome Jiang
cc47231187 vp9: Reduce stack usage of choose_partioning.
Change type of sum_square_error from int64_t to uint32_t.
Change type of sum_error from int64_t to int32_t.

This reduces the stack usage from ~131K to ~87K.

BUG=b/68362457

Change-Id: I147d7c7b226bceb4f0817bb86848e1fa9d9ac149
2017-10-30 13:53:20 -07:00
James Zern
acb9460929 vp8: correct if/else '{' placement
swap '{' and c-style comments removing a few redundant ones along the
way; covers most leftovers from the clang-tidy run against an
x86_64-linux config.

Change-Id: I67a45596f80a12389faca49c5be440875092a7df
2017-10-27 12:27:10 -07:00
Scott LaVarnway
3bf02ad74a vpx: hadamard: use ptrdiff_t instead of int for stride
Eliminates the following instruction for the x86 (64 bit)
intrinsic code:

movslq %esi,%rax

Change-Id: I8f5ebd40726f998708a668b0f52ea7a0576befae
2017-10-26 11:41:48 -07:00
Kyle Siefring
037e596f04 Merge "Optimize convolve8 SSSE3 and AVX2 intrinsics" 2017-10-24 19:22:36 +00:00
Kyle Siefring
ae35425ae6 Optimize convolve8 SSSE3 and AVX2 intrinsics
Changed the intrinsics to perform summation similiar to the way the assembly does.

The new code diverges from the assembly by preferring unsaturated additions.

Results for haswell

SSSE3
Horiz/Vert  Size  Speedup
Horiz       x4    ~32%
Horiz       x8    ~6%
Vert        x8    ~4%

AVX2
Horiz/Vert  Size  Speedup
Horiz       x16   ~16%
Vert        x16   ~14%

BUG=webm:1471

Change-Id: I7ad98ea688c904b1ba324adf8eb977873c8b8668
2017-10-24 10:39:48 -04:00
Scott LaVarnway
e0aa6b24aa Merge "vpx: [x86] vpx_hadamard_16x16_avx2() highbitdepth fix" 2017-10-23 22:02:59 +00:00
Marco
0738d90169 vp9-svc: Allow for adapt_rd_thresh with row-mt.
Set adaptive_row_thresh_mt = 1 at speed >= 7,
for svc when multi-threading is used with row-mt.
This allow the adaptive_rd_thresh feature to be used
in the nonrd-pickmode.

~1-2% speedup for SVC encoding with small quality
loss (< 0.6%) on RTC set.

Change-Id: Iab9878dff117bccdaef3e4d0645165db9808cdfc
2017-10-23 11:47:18 -07:00
Scott LaVarnway
512bf4e029 vpx: [x86] vpx_hadamard_16x16_avx2() highbitdepth fix
Use an intermediate buffer before storing to coeffs when
highbitdepth is enabled.

Change-Id: I101981a1995f1108ad107c55c37d6e09eadb404b
2017-10-23 08:49:32 -07:00
Scott LaVarnway
4906cea027 vpx: [x86] vpx_hadamard_16x16_avx2() improvements
~10% performance gain.  Fixed the cosmetics noted in the
previous commit.

Change-Id: Iddf475f34d0d0a3e356b2143682aeabac459ed13
2017-10-20 08:55:06 -07:00
Scott LaVarnway
b58259ab55 Merge "vpx: [x86] add vpx_hadamard_16x16_avx2()" 2017-10-19 23:32:10 +00:00
Paul Wilkins
199971d606 Merge "Corpus VBR tweak for undershoot." 2017-10-19 10:07:45 +00:00
Paul Wilkins
0c493cbe2b Merge "Increase precision of some debug stats output for corpus VBR." 2017-10-19 10:07:30 +00:00
Paul Wilkins
d8c34a2552 Merge "Prevent double application of min rate in two pass." 2017-10-19 10:06:33 +00:00
Scott LaVarnway
55c126a5d7 vpx: [x86] add vpx_hadamard_16x16_avx2()
This version is ~1.91x faster than the sse2 version.  When
highbitdepth is enabled, it is ~1.74x.

Change-Id: I2b0e92ede9f55c6259ca07bf1f8c8a5d0d0955bd
2017-10-18 18:00:00 -07:00
Jerome Jiang
401e6d48bf Merge "Add datarate test for vp8 ROI." 2017-10-18 19:39:26 +00:00
Jerome Jiang
bd6d82e881 Add datarate test for vp8 ROI.
BUG=webm:1470

Change-Id: Icbc848837e64eacc49491dcc26b4c5802af2ee13
2017-10-18 11:19:59 -07:00
Jerome Jiang
ec2fced451 Merge "vp8: Enable use of ROI map." 2017-10-18 18:16:44 +00:00
Kyle Siefring
b3a36f7946 Merge "Refactor x86/vpx_subpixel_8t_intrin_avx2.c" 2017-10-18 16:19:52 +00:00
Shiyou Yin
df15220a89 Merge "vp8: [loongson] optimize idct with mmi" 2017-10-18 00:55:36 +00:00
Jerome Jiang
dbb8926b86 vp8: Enable use of ROI map.
Disable cyclic refresh if ROI is used and add flag to properly handle
the static_thresh deltas.
Remove the ROI test for cyclic refresh (it's allowed but disabled if ROI
is used).
Add an example in vpx_temporal_svc_encoder.c. Turned off by default.

BUG=webm:1470

Change-Id: Ief9ba1d7f967bc00511b412b491c3f70943bfbda
2017-10-17 15:23:03 -07:00
Linfeng Zhang
9336e01621 Merge changes I17fff122,Ic149e3cb
* changes:
  Add 4 to 3 scaling SSSE3 optimization
  Test extreme inputs in frame scale functions
2017-10-17 16:03:29 +00:00
Linfeng Zhang
0d2e95193b Merge "Generalize CheckScalingFiltering in ConvolveTest" 2017-10-17 16:03:07 +00:00
Kyle Siefring
55805e2786 Refactor x86/vpx_subpixel_8t_intrin_avx2.c
Change-Id: I6539111dfb35a43028e9755785b2e9ea31854305
2017-10-17 11:57:40 -04:00
Shiyou Yin
577d4fa792 vp8: [loongson] optimize idct with mmi
1. vp8_dequant_idct_add_y_block_mmi
2. vp8_dequant_idct_add_uv_block_mmi

Change-Id: I9987147be2685ac79d4b045d1d56f6709ee1223c
2017-10-17 03:27:31 +00:00
Linfeng Zhang
580d32240f Add 4 to 3 scaling SSSE3 optimization
Note this change will trigger the different C version on SSSE3 and
generate different scaled output.

Its speed is 2x compared with the version calling vpx_scaled_2d_ssse3().

Change-Id: I17fff122cd0a5ac8aa451d84daa606582da8e194
2017-10-16 15:42:42 -07:00
Marco
a9248457b1 Adjust threshold in gf_boost for 1 pass vbr
Small inncrease the sad_thresh1, avoids some false
detection of possible scene changes within lag.

Small improvement in few clips on ytlive, otherwise neutral change.

Change-Id: Ia79b53bb657bbce65a7aac7d20666b6373d5af8b
2017-10-13 15:33:51 -07:00
Paul Wilkins
12df840777 Merge "Further Corpus VBR change." 2017-10-13 15:59:58 +00:00
Paul Wilkins
eaa593d293 Merge "Corpus Wide VBR test implementation." 2017-10-13 15:59:45 +00:00
paulwilkins
8842ee0b0d Corpus VBR tweak for undershoot.
In cases of strong undershoot adjust Q range down faster.

Change-Id: I84982beceb3c9b6dc50e52e4a6e891c7dd395d03
2017-10-13 10:27:15 +01:00