Mans Rullgard [Tue, 30 Jul 2013 17:11:06 +0000 (18:11 +0100)]
vp9: neon: optimise loads in horiz convolve functions
Loading to single lanes in multiple registers is expensive since
it requires a read and write of each register which saturates
the register file access. Loading to single registers followed
by a separate transpose reduces this pressure.
Deb Mukherjee [Thu, 1 Aug 2013 19:56:12 +0000 (12:56 -0700)]
Adds a source variance computation function
Adds a function to compute source variance for various
sb_types to be used for pruning mode and partition searches.
[The existing activity measure function is currently specialized
for only 16x16 MBs and needs to be updated].
Jingning Han [Thu, 1 Aug 2013 19:45:16 +0000 (12:45 -0700)]
Remove unused vp9_short_idct10_32x32_add
The inverse 32x32 transform detects all zero entries and skips the
computations accordingly per 8 rows in the first 1-D operation. The
function vp9_short_idct10_32x32_add performs differently and is not
used anywhere, hence removed.
Jingning Han [Wed, 31 Jul 2013 23:50:34 +0000 (16:50 -0700)]
Optimize 32x32 2D inverse DCT for speed-up
This commit exploits the sparsity of quantized coefficient matrix.
It detects each 32x8 array and skip the corresponding inverse
transformation if all entries are zero.
For ped1080p at 8000 kbps, this on average reduces the runtime of
32x32 inverse 2D-DCT SSE2 function from 6256 cycles -> 5200
cycles. It makes the overall encoding process about 2% faster at
speed 0. The speed-up is more pronounceable for the decoding process.
Jingning Han [Thu, 1 Aug 2013 00:02:06 +0000 (17:02 -0700)]
Remove unnecessary arguments in rd_pick_ref_frame
This commit removes redundant arguments passing in the function of
rd_pick_reference_frame. This resolves the clang warnings about
potential use of uninitialized values.
Jingning Han [Tue, 30 Jul 2013 22:47:12 +0000 (15:47 -0700)]
Make the use of ref_frame index consistent
Refactor the frame buffer referencing in choose_partition and make
it consistent with other places. This means to prevent potential
issues when we extend reference frame buffer.
Jingning Han [Mon, 29 Jul 2013 21:54:31 +0000 (14:54 -0700)]
Skip redundant tokenization in rd loop
This commit makes the encoder skip the redundant tokenization process
in the rate-distortion optimization search loop, while updating the
entropy contexts accordingly. It makes the speed 0 encoding process
about 0.5% faster at no performance change.
Ronald S. Bultje [Wed, 24 Jul 2013 22:13:58 +0000 (15:13 -0700)]
Inverse dimension order in token_cost array.
This allows us to increment the position at the band-level only as
we go from one band to the next; more importantly, that allows us to
use an add instead of multiply instruction, and omit the instruction
altogether if the band doesn't change from one coef to the next, thus
being slightly faster (probably more noticeable on systems where a
multiply is expensive, like arm).
Jingning Han [Sat, 27 Jul 2013 00:01:51 +0000 (17:01 -0700)]
Shortcut 8x8/16x16 inverse 2D-DCT
This commit brought back the shortcut implementation of 8x8/16x16
inverse 2D-DCT. When the eob <= 10, it skips the inverse transform
operations on row 4:7/4:15 in the first round. For bus_cif at 1000
kbps, this provides about 2% speed-up at speed 0.
Jingning Han [Fri, 26 Jul 2013 21:11:37 +0000 (14:11 -0700)]
Special handle on DC only inverse 8x8 2D-DCT
This commit enables a special handle for the 8x8 inverse 2D-DCT,
where only DC coefficient is quantized to be non-zero. For bus_cif
at 2000 kbps, it provides about 1% speed-up at speed 0.
Paul Wilkins [Wed, 24 Jul 2013 13:07:37 +0000 (14:07 +0100)]
Auto min and max partition size experiment.
Speed feature experiment to set an upper and lower
partition size limit based on what has been seen
in spatial neighbors.
This seems to gives quite reasonable speed gains in local
(10-15%) and when used with speed 0 the losses are small
(0.25% derf, 0.35% stdhd). However, for now I am only
enabling it on speed 1 as there may be clashes with the existing
temporal partition selection in speed 2.
Using a tighter min / max around the range derived from the
neighbors increases speed further but at the cost of a
bigger quality loss. However, I think this spatial method could
be combined with data from either the last frame or a variance
method (or both) to refine the range of minimum and maximum
partition size. I.e. consider the min and max from spatial and
temporal neighbors and the variance recommendation.
Yunqing Wang [Fri, 26 Jul 2013 02:17:46 +0000 (19:17 -0700)]
Modify static threshold calculation
Used 3 * standard_deviation in internal threshold calculation
instead of fit curve. This actually approached the algorithm
better.
For comparison, similar tests were done:
The overall psnr loss is less than before.
1. derf set:
when static-thresh = 1, psnr loss is 0.329%;
when static-thresh = 500, psnr loss is 0.970%;
2. stdhd set:
when static-thresh = 1, psnr loss is 0.922%;
when static-thresh = 500, psnr loss is 1.307%;
Similar speedup is achieved. For example,
clip bitrate static-thresh psnr time
akiyo(cif) 500 0 48.952 5.077s(50f)
akiyo 500 500 48.866 4.169s(50f)
Yunqing Wang [Thu, 11 Jul 2013 18:15:00 +0000 (11:15 -0700)]
Add encoding option --static-thresh
This option exists in VP8, and it was rewritten in VP9 to support
skipping on different partition levels. After prediction is done,
we can check if the residuals in the partition block will be all
quantized to 0. If this is true, the skip flag is set, and only
prediction data are needed in reconstruction. Based on DCT's energy
conservation property, the skipping check can be estimated in
spatial domain.
The prediction error is calculated and compared to a threshold.
The threshold is determined by the dequant values, and also
adjusted by partition sizes. To be precise, the DC and AC parts
for Y, U, and V planes are checked to decide skipping or not.
Test showed that
1. derf set:
when static-thresh = 1, psnr loss is 0.666%;
when static-thresh = 500, psnr loss is 1.162%;
2. stdhd set:
when static-thresh = 1, psnr loss is 1.249%;
when static-thresh = 500, psnr loss is 1.668%;
For different clips, encoding speedup range is between several
percentage and 20+% when static-thresh <= 500. For example,
clip bitrate static-thresh psnr time
akiyo(cif) 500 0 48.923 5.635s(50f)
akiyo 500 500 48.863 4.402s(50f)
Removing unused constants, macros, and function declarations. Using
ROUND_POWER_OF_TWO macro, vp9_zero, vp9_copy where possible. Moving
#include from *.h to *.c. Merging for loops for motion vectors.
Jingning Han [Tue, 23 Jul 2013 22:53:09 +0000 (15:53 -0700)]
Make coeff_optimize initialized per-plane
This commit makes the initialization of trellis coeff optimization
a per-plane operation, thereby eliminating the redundant steps in
encode_sby and encode_sbuv. It makes the encoder at speed 0 slightly
faster.
The feature that uses small partition results as a measure to skip
mode evaluation at larger partition requires the flags to be reset.
The reset was missing in the code path that calls rd_use_partition().
Martin Storsjo [Wed, 24 Jul 2013 13:06:49 +0000 (16:06 +0300)]
msvs: Generate proper configurations for mixed platforms
Prior to 73c4e284, the generated .sln files didn't contain any
information about the different configurations when using .vcxproj
project files. The MSVS IDE was able to fill this in just fine when
loaded though.
When building for ARM, the obj_int_extract project still is built
for x86, in order for the build process to be able to use
obj_int_extract.exe.
Now that configuration info is generated, it breaks current ARM
setups, since the configurations generated by gen_msvs_sln.sh only
included configurations from the last parsed project file (as
mentioned in the comment).
In these setups, the MSVS IDE generated a third meta-platform, called
"Mixed Platforms". This meta-platform points to either ARM or
Win32 as platform in each of the individual projects.
When the MSVS IDE generated this automatically, it also included
the original ARM and Win32 platforms as separate choices, but these
can be omitted since they don't make sense.
projects / libvpx / log