Johann [Wed, 15 Mar 2017 17:40:58 +0000 (10:40 -0700)]
vp9 temporal filter: sse4 implementation
Approximates division using multiply and shift.
Speeds up both sizes (8x8 and 16x16) by 30 times.
Fix the call sites to use the RTCD function.
Delete sse2 and mips implementation. They were based on a previous
implementation of the filter. It was changed in Dec 2015: ece4fd5d2247c9512b31a93dd593de567beaf928
Marco [Tue, 25 Apr 2017 23:11:19 +0000 (16:11 -0700)]
vp9: SVC: Adjust some speed settings for temporal layers.
Make some speed setting changes for temporal enhancement layers,
and remove the switch in subpel_force_stop for the aggressive_base_mv
in non-rd pickmode.
Gain some 2-3% speed with little/negligible quality loss.
vp9: speed >= 8: Skip uv variance in model_rd_sb_y_large
For speed >= 8 and color_sensitivity not set, skip the transform
skipping test in UV planes.
Add a new condition to check noise level to skip chroma check
for speed >= 8 if y_sad is high.
1~2% speedup on ARM for speed 8.
Borg tests show neutral results in both rtc and rtc_derf.
Marco [Fri, 14 Apr 2017 18:32:19 +0000 (11:32 -0700)]
vp9; Reduce artifact in non-rd pickmode for lighting changes.
Add a low-variance high-sumdiff to the superblock content state
and use it to limit the mv and bias some decisions in non-rd pickmode.
Only affects speed >= 6.
Reduces artifact for lighting changes.
Small/no difference in metrics on RTC set.
Yunqing Wang [Mon, 24 Apr 2017 19:06:49 +0000 (12:06 -0700)]
Make the row based multi-threaded encoder deterministic
This patch followed allow_exhaustive_searches feature modification and
continued to modify the encoder to achieve the determinism in the row
based multi-threaded encoding. While row-mt = 1 and using multiple
threads, the adaptive feature in encoder was disabled, which gave
BDRate gain(at speed 1, -0.6% ~ -0.7%; at speed 2, -0.46% ~ -0.59%),
but some encoder speed losses(7% ~ 10% at speed 1 and 3% ~ 6% at
speed 2). These speed losses were acceptable considering the speed
gains obtained from row-mt.
Yunqing Wang [Thu, 20 Apr 2017 00:00:08 +0000 (17:00 -0700)]
Make allow_exhaustive_searches feature no longer adaptive
A previous patch turned on allow_exhaustive_searches feature only for
FC_GRAPHICS_ANIMATION content. This patch further modified the feature
by removing the exhaustive search limit, and made it no longer adaptive.
As a result, the 2 counts that recorded the number of motion searches
were removed, which helped achieve the determinism in the row based
multi-threading encoding. Tests showed that this patch didn't cause
the encoder much slower.
Used exhaustive_searches_thresh for this speed feature, and removed
allow_exhaustive_searches. Also, refactored the speed feature code
to follow the general speed feature setting style.
In non-rd pickmode (speed >= 5), avoid duplication of computations in
model_rd_for_sb_y when the speed feature use_simple_block_yrd is
enabled (or for high bitdepth build under certain conditions).
QVGA, VGA and HD have 1.23%, 2.68% and 1.7% speedup on ARM for speed 8,
respectively.
Slightly faster, the other dc predictors cannot be faster since
the computation speedup is overwhelmed by the time spent reading
dst to write just the 8x8 part.
Yunqing Wang [Wed, 19 Apr 2017 23:32:59 +0000 (16:32 -0700)]
Only allow allow_exhaustive_searches for FC_GRAPHICS_ANIMATION content
The allow_exhaustive_searches feature improves the encoding quality
of FC_GRAPHICS_ANIMATION content a lot. For non-FC_GRAPHICS_ANIMATION
content, the quality test result is almost neutral. This patch makes
this feature to be used only for FC_GRAPHICS_ANIMATION content.
The motivation of doing that is to make this feature no longer adaptive,
which will be implemented in the following patch.
Replace by CAST_TO_BYTEPTR/SHORTPTR.
The rule is: if a short ptr is casted to a byte ptr, any offset
operation on the byte ptr must be doubled. We do this by casting to
short ptr first, adding offset, then casting back to byte ptr.
Marco [Tue, 18 Apr 2017 16:43:32 +0000 (09:43 -0700)]
vp9: Disable some SVC tests for now.
Disable the 1 pass CBR SVC tests with temporal_layers > 1.
Issue with the commit 863f860, which will cause encoder/decoder
mismatch due to skipping encoder loopfilter for non-reference frames.
Marco [Mon, 17 Apr 2017 16:19:03 +0000 (09:19 -0700)]
vp9: Datarate test: modify frame flags for 3 TL.
Modify the frame flags to update the ARF on top layer,
for the tests:
VP9/DatarateTestVP9Large.BasicRateTargeting3TemporalLayers
VP9/DatarateTestVP9Large.BasicRateTargeting3TemporalLayersFrameDropping
This is needed to fix the encode/decoder mismatches caused by 863f860,
and removed in the revert e9b7f98.
Marco [Tue, 11 Apr 2017 23:17:18 +0000 (16:17 -0700)]
vp9: Avoid encoder loopfilter for non-reference frames.
Useful for SVC, where the top layer enhancement frames may
not update any reference buffers, as is the case for the
patterns in the 1 pass CBR SVC when #temporal_layers > 1.
~3% encoder speedup for SVC patterns with temporal layers
in 1 pass CBR mode.
Updated the SVC datarate tests for the mismatch frames.
Set the frame-dropper off in some tests with #temporal_layers > 1
so we can correctly set #mismatch frames. Adjusted rate target
threshold for tests where frame-dropper was turned off.
Yunqing Wang [Mon, 10 Apr 2017 17:57:41 +0000 (10:57 -0700)]
Fix an integer overflow in vp9_mcomp.c
The MV unit test revealed an integer overflow issue in vp9_mcomp.c.
This was caused if the MV was very large. In mv_err_cost(), when
mv->row = 8184, mv->col = 8184 and ref_mv is 0, mv_cost = 34363
and error_per_bit = 132412, causing the overflow.
Marco [Mon, 10 Apr 2017 21:56:46 +0000 (14:56 -0700)]
vp9: 1 pass CBR: avoid nonrd_pick_partition on segment.
For speed 5, 1 pass CBR: Don't use the nonrd_pick_partition
on the segment, rather use choose_partitioning followed by
nonrd_select_partition (as is done on base segment).
Little/no quality loss on RTC and RTC_derf (< 0.3%),
speedup of at least 5%.
Marco [Thu, 6 Apr 2017 17:25:55 +0000 (10:25 -0700)]
vp9; Move the denoising condition for speed 5.
Move the condition for effectively disabling the denoising
for speed 5 into the vp9_denoiser_denoise().
This is cleaner, and also moving the condition into vp9_denoiser_denoise
will keep the denoiser buffer updated with the current source.
This allows for more consistent behavior if speed is changed midstream.
projects / libvpx / log