Merge sb32x32 and sb64x64 functions; allow for rectangular sizes. Code
gives identical encoder results before and after. There are a few
macros for rectangular block sizes under the sbsegment experiment; this
experiment is not yet functional and should not yet be used.
Change-Id: I71f93b5d2a1596e99a6f01f29c3f0a456694d728
Restructure the code to avoid the majority of per-block-size
switches, code duplication, etc. All block types (mb/sb32/sb64)
can be handled by the same code.
Change-Id: I4022718d66e31a15a7074e43f3b98cd0a5124ea7
Clamp only the motion vectors inferred from neighboring reference
macroblocks. The motion vectors obtained through motion search in
NEWMV mode are constrained during the search process, which allows
a relatively larger referencing region than the inferred mvs.
Hence further clamping the best mv provided by the motion search may
affect the efficacy of NEWMV mode.
Synchronized the decoding process. The decoded mvs in NEWMV modes
should be guaranteed to fit in the effective range. Put a mv range
clamping function there for security purpose.
This improves the coding performance of high motion sequences, e.g.,
derf set:
foreman 0.233%
husky 0.175%
icd 0.135%
mother_daughter 0.337%
pamphlet 0.561%
stdhd set:
blue_sky 0.408%
city 0.455%
also saw sunflower goes down by -0.469%.
Change-Id: I3fcbba669e56dab779857a8126a91b926e899cb5
Start grouping data per-plane, as part of refactoring to support
additional planes, and chroma planes with other-than 4:2:0
subsampling.
Change-Id: Idb76a0e23ab239180c818025bae1f36f1608bb23
Took vp9_setup_scale_factors_for_frame() out from
vp9_setup_interp_filters(), so that it is only called once per
frame instead of per macroblock. Decoder tests showed a 1.5%
performance gain.
Change-Id: I770cb09eb2140ab85132f82aed388ac0bdd3a0aa
General code cleanup in loopfilter code. Modification of setup_frame_size,
so now VP9_COMMON is modified in one place after all width/height checks
passed.
Change-Id: Iedf32df43a912d7aae788ed276ac6c429973f6fe
Rename the file and clean up includes. In the future we would like to
pattern match the files which need additional compiler flags.
Change-Id: I2c76256467f392a78dd4ccc71e6e0a580e158e56
The patch adds the flexibility to use standard EOB based coding
on smaller block sizes and nzc based coding on larger blocksizes.
The tx-sizes that use nzc based coding and those that use EOB based
coding are controlled by a function get_nzc_used().
By default, this function uses nzc based coding for 16x16 and 32x32
transform blocks, which seem to bridge the performance gap
substantially.
All sets are now lower by 0.5% to 0.7%, as opposed to ~1.8% before.
Change-Id: I06abed3df57b52d241ea1f51b0d571c71e38fd0b
Almost all arguments for vp9_build_inter32x32_predictors_sb and
vp9_build_inter64x64_predictors_sb can be deduced from the first macroblock
argument.
Change-Id: I5d477a607586d05698d5b3b9b9bc03891dd3fe83
Extracting setup_frame_size and update_frame_context functions. Introducing
vp9_read_prob function as shortcut for (vp9_prob)vp9_read_literal(r, 8).
Change-Id: Ia5c68fd725b2d1b9c5eb20f69cacb62361b5a3dd
Wrote sse2 version of vp9_short_idct_32x32 function. Compared
to c version, the sse2 version is 5X faster.
Change-Id: I071ab7378358346ab4d9c6e2980f713c3c209864
This gains about 0.2% on derf, 0.1% on hd and 0.4% on stdhd. I can put
this under an experimental flag if wanted, just trying to get my patch
queue in shape.
Change-Id: Ibe1a30fe0e0b07bec4802e0f3ff0ba22e505f576
Adds an experiment to use a weighted prediction of two INTER
predictors, where the weight is one of (1/4, 3/4), (3/8, 5/8),
(1/2, 1/2), (5/8, 3/8) or (3/4, 1/4), and is chosen implicitly
based on consistency of the predictors to the already
reconstructed pixels to the top and left of the current macroblock
or superblock.
Currently the weighting is not applied to SPLITMV modes, which
default to the usual (1/2, 1/2) weighting. However the code is in
place controlled by a macro. The same weighting is used for Y and
UV components, where the weight is derived from analyzing the Y
component only.
Results (over compound inter-intra experiment)
derf: +0.18%
yt: +0.34%
hd: +0.49%
stdhd: +0.23%
The experiment suggests bigger benefit for explicitly signaled weights.
Change-Id: I5438539ff4485c5752874cd1eb078ff14bf5235a
These are mostly just for experimental purposes. I saw small gains (in
the 0.1% range) when playing with this on derf.
Change-Id: Ib21eed477bbb46bddcd73b21c5c708a5b46abedc
Now that the first AC coefficient in both directions use the same DC
as their context, there no longer is a purpose in letting both have
their own band. Merging these two bands allows us to split bands for
some of the very high-frequency AC bands.
In addition, I'm redoing the banding for the 1D-ADST col/row scans. I
don't think the old banding made any sense at all (it merged the last
coefficient of the first row/col in the same band as the first two of
the second row/col), which was clearly an oversight from the band being
applied in scan-order (rather than in their actual position). Now,
coefficients at the same position will be in the same band, regardless
what scan order is used. I think this makes most sense for the purpose
of banding, which is basically "predict energy for this coefficient
depending on the energy of context coefficients" (i.e. pt).
After full re-training, together with previous patch, derf gains about
1.2-1.3%, and hd/stdhd gain about 0.9-1.0%.
Change-Id: I7a0cc12ba724e88b278034113cb4adaaebf87e0c
Pearson correlation for above or left is significantly higher than for
previous-in-scan-order (absolute values depend on position in scan, but
in general, we gain about 0.1-0.2 by using either above or left; using
both basically just makes this even better). For eob branch skipping,
we continue to use the previous token in scan order.
This helps about 0.9% on derf after re-training on a limited data set.
Full re-training and results on larger-resolution clips are pending.
Note that this commit breaks trellis, so we can probably get further
gains out of it by fixing trellis at some later point.
Change-Id: Iead68e296fc3a105cca746b5e3da9555d6010cfe
Reoptimizes the 8-tap smooth filter.
Results:
derf: +0.101%
yt: +0.157%
hd: +0.791%
stdhd: +0.264%
The next step will be to reoptimize the other two filters.
Change-Id: I3d256a510ad9c7c30c33fae4a70fb43dfc708ed0
Lower case variable names, removing redundant variables, declaration and
initialization on the same line.
Change-Id: Ie0c6c95b14103990eb6a9d7784f8259c662e1251
Adds a per-frame, strength adjustable, in loop deringing filter. Uses
the existing vp9_post_proc_down_and_across 5 tap thresholded blur
code, with a brute force search for the threshold.
Results almost strictly positive on the YT HD set, either having no
effect or helping PSNR in the range of 1-3% (overall average 0.8%).
Results more mixed for the CIF set, (-0.5 min, 1.4 max, 0.1 avg).
This has an almost strictly negative impact to SSIM, so examining a
different filter or a more balanced search heuristic is in order.
Other test set results pending.
Change-Id: I5ca6ee8fe292dfa3f2eab7f65332423fa1710b58
Replaces the default tables for single coefficient magnitudes with
those obtained from an appropriate distribution. The EOB node
is left unchanged. The model is represeted as a 256-size codebook
where the index corresponds to the probability of the Zero or the
One node. Two variations are implemented corresponding to whether
the Zero node or the One-node is used as the peg. The main advantage
is that the default prob tables will become considerably smaller and
manageable. Besides there is substantially less risk of over-fitting
for a training set.
Various distributions are tried and the one that gives the best
results is the family of Generalized Gaussian distributions with
shape parameter 0.75. The results are within about 0.2% of fully
trained tables for the Zero peg variant, and within 0.1% of the
One peg variant.
The forward updates are optionally (controlled by a macro)
model-based, i.e. restricted to only convey probabilities from the
codebook. Backward updates can also be optionally (controlled by
another macro) model-based, but is turned off by default. Currently
model-based forward updates work about the same as unconstrained
updates, but there is a drop in performance with backward-updates
being model based.
The model based approach also allows the probabilities for the key
frames to be adjusted from the defaults based on the base_qindex of
the frame. Currently the adjustment function is a placeholder that
adjusts the prob of EOB and Zero node from the nominal one at higher
quality (lower qindex) or lower quality (higher qindex) ends of the
range. The rest of the probabilities are then derived based on the
model from the adjusted prob of zero.
Change-Id: Iae050f3cbcc6d8b3f204e8dc395ae47b3b2192c9
Lower case variable names, code simplification by using already defined
clamp and read_le16 functions.
Change-Id: I8fd544365bd8d1daed86d7b2ae0843e4ef80df08
Wrote sse2 version of vp9_short_idct10_16x16 function. Compared
to c version, the sse2 version is 2.3X faster.
Change-Id: I314c4f09369648721798321eeed6f58e38857f26
Wrote sse2 version of vp9_short_idct16x16 function. Compared to c
version, the sse2 version is over 2.5X faster.
Change-Id: I38536e2b846427a2cc5c5423aaf305fd0e605d61
Renaming Width to width, Height to height and Version to version in
several structs and function signatures.
Change-Id: I084c3f7e747cb2ce3345aff27a3dff9b13a87543
Wrote sse2 functions of vp9_short_idct8x8 and vp9_short_idct10_8x8.
Compared to c version, the sse2 version is 2X faster. The decoder
test didn't show noticeable gain since 8x8 idct doesn't take much
of decoding time (less than 1% in my test).
Change-Id: I56313e18cd481700b3b52c4eda5ca204ca6365f3
If the intended display size is different than the size the frame is
coded at, then send that size explicitly in the bitstream. Adds a new
bit to the frame header to indicate whether the extra size fields
are present.
Change-Id: I525c66f22d207efaf1e5f903c6a2a91b80245854
This fix resolves some of the mismatch issues being seen
recently. While this is the right thing to do when tiling
is used for this experiment, it is not the underlying cause
of the the mismatches.
Something else is causing writing outside of the allowable
frame area in the encoder leading to this mismatch.
Change-Id: If52c6f67555aa18ab8762865384e323b47237277
Updates the YV12_BUFFER_CONFIG structure to be crop-aware. The
exiting width/height parameters are left unchanged, storing the
width and height algined to a 16 byte boundary. The cropped
dimensions are added as new fields.
This fixes a nasty visual pulse when switching between scaled and
unscaled frame dimensions due to a mismatch between the scaling
ratio and the 16-byte aligned sizes.
Change-Id: Id4a3f6aea6b9b9ae38bdfa1b87b7eb2cfcdd57b6
This is like VP8_COPY_REFERENCE, but returns a pointer to the reference
frame rather than a copy of it. This is useful when the application
doesn't know what the size of the reference is, as is the case when
scaling is in effect.
Change-Id: I63667109f65510364d0e397ebe56217140772085
Enable entropy coding of motion vectors up to +/-2048. Also
extend the motion search range accordingly.
Change-Id: Iac2bb015e8934521cef83a19edbe967d9f097436
Removing redundant code, introducing new functions for better
decomposition, adding 'clamp' function to vp9_common.h.
Change-Id: Ic3b8ca13bbc38f60f0c9c43910b5802005e31aaf
Remove the temporary branch count arrays and build the adapted probabilities
while walking the tree. Gives an additional 1.5% or so on CIF.
Change-Id: I875d61e5e0ec778e5d2f7f9d0837b989a91cf3a3
The previous implementation visited each node in the tree multiple times
because it used each symbol's encoding to revisit the branches taken and
increment its count. Instead, we can traverse the tree depth first and
calculate the probabilities and branch counts as we walk back up. The
complexity goes from somewhere between O(nlogn) and O(n^2) (depending on
how balanced the tree is) to O(n).
Only tested one clip (256kbps, CIF), saw 13% decoding perf improvement.
Note that this optimization should port trivially to VP8 as well. In VP8,
the decoder doesn't use this function, but it does routinely show up
on the profile for realtime encoding.
Change-Id: I4f2848e4f41dc9a7694f73f3e75034bce08d1b12
Adds probability updates for extra bits for the nzcs, code for
getting nzc stats, plus some minor cleanups and fixes.
Change-Id: If2814e7f04fb52f5025ad9f400f3e6c50a00b543
Added SSE2 idct4_1d which is called by vp9_short_iht4x4. Also,
modified the parameter type passed to vp9_short_iht functions to
make it work with rtcd prototype.
Change-Id: I81ba7cb4db6738f1923383b52a06deb760923ffe
Increase the motion search range by 4x. Change MV_CLASS tree of the
entropy coding to allow two additional mv classes to cover the
extended motion vector limit. The codec determines the effective
motion search range conditioned on the actual frame dimension.
It provides coding gains:
stdhd 0.39%
yt 0.56%
hd 0.47%
Major coding performance gains are packed in several sequences with
intense motion activities, e.g., ped_1080p gains 7% at high bit-rates,
and on average 3%.
TODO: Need to further tune the rate control and motion search units.
Change-Id: Ib842540a6796fbee5a797809433ef6a477c6d78d
This also changes the RD search to take account of the correct block
index when searching (this is required for ADST positioning to work
correctly in combination with tx_select).
Change-Id: Ie50d05b3a024a64ecd0b376887aa38ac5f7b6af6
This patch revamps the entropy coding of coefficients to code first
a non-zero count per coded block and correspondingly remove the EOB
token from the token set.
STATUS:
Main encode/decode code achieving encode/decode sync - done.
Forward and backward probability updates to the nzcs - done.
Rd costing updates for nzcs - done.
Note: The dynamic progrmaming apporach used in trellis quantization
is not exactly compatible with nzcs. A suboptimal approach has been
used instead where branch costs are updated to account for changes
in the nzcs.
TODO:
Training the default probs/counts for nzcs
Change-Id: I951bc1e22f47885077a7453a09b0493daa77883d
Split macroblock and superblock tokenization and detokenization
functions and coefficient-related data structs so that the bitstream
layout and related code of superblock coefficients looks less like it's
a hack to fit macroblocks in superblocks.
In addition, unify chroma transform size selection from luma transform
size (i.e. always use the same size, as long as it fits the predictor);
in practice, this means 32x32 and 64x64 superblocks using the 16x16 luma
transform will now use the 16x16 (instead of the 8x8) chroma transform,
and 64x64 superblocks using the 32x32 luma transform will now use the
32x32 (instead of the 16x16) chroma transform.
Lastly, add a trellis optimize function for 32x32 transform blocks.
HD gains about 0.3%, STDHD about 0.15% and derf about 0.1%. There's
a few negative points here and there that I might want to analyze
a little closer.
Change-Id: Ibad7c3ddfe1acfc52771dfc27c03e9783e054430
Fixed a couple of variable/function definitions, as well as header
handling to support 16K sequence coding at high bit-rates.
The width and height are each specified by two bytes in the header.
Use an extra byte to explicitly indicate the scaling factors in
both directions, each ranging from 0 to 15.
Tested coding up to 16400x16400 dimension.
Change-Id: Ibc2225c6036620270f2c0cf5172d1760aaec10ec
Simplified idct32x32 calculation when there are only 10 or less
non-zero coefficients in 32x32 block. This helps the decoder
performance.
Change-Id: If7f8893d27b64a9892b4b2621a37fdf4ac0c2a6d
The ABOVESPREFMV experiment uses four pixels to the left of the
current block, which don't exist for the left-most column.
Change-Id: I4cf0b42ae8f54c0b3e7b1ed8755704b74fafc39c
sse4_1 code used uint16_t for returning sad, but that
won't work for 32x32 or 64x64. This code fixes the
assembly for those and also reenables sse4_1 on linux
Change-Id: I5ce7288d581db870a148e5f7c5092826f59edd81
Fixing code style, using array lookup instead of switch statements for
forward hybrid transforms (in the same way as for their inverses).
Consistent usage of ROUND_POWER_OF_TWO macro in appropriate places.
Change-Id: I0d3822ae11f928905fdbfbe4158f91d97c71015f
This function was part of an optimization used in VP8 that required
caching two macroblocks. This is unused in VP9, and might not
survive refactoring to support superblocks, so removing it for now.
Change-Id: I744e585206ccc1ef9a402665c33863fc9fb46f0d
This avoids duplicating all the filters twice. Includes fixups to the
convolve routines and associated tests to make this work.
Change-Id: I922f86021594e55072ddb63b42b2313605db6e00
This patch extends the previous support for using references of a
different resolution in ZEROMV mode to all inter prediction modes.
Subpixel based best-mv scoring is disabled when the reference frame
differs in resolution from the current frame.
Change-Id: Id4dc3e5e6692de98d9857fd56bfad3ac57e944ac
This commit updates the 4x4 prediction to consistently use the
build_2x1_inter_predictor() method. That function is updated to
calculate the scale offset, rather than relying on the caller
to calculate it. In the case that the 2x1 prediction can not
be used, the scale offset is recalculated for each 1x1 block.
The idea here is that the offsets are calculated before each
call to vp9_build_scaled_inter_predictor().
Change-Id: I0ac3343dd54e2846efa3c4195fcd328b709ca04d
This patch allows coding frames using references of different
resolution, in ZEROMV mode. For compound prediction, either
reference may be scaled.
To test, I use the resize_test and enable WRITE_RECON_BUFFER
in vp9_onyxd_if.c. It's also useful to apply this patch to
test/i420_video_source.h:
--- a/test/i420_video_source.h
+++ b/test/i420_video_source.h
@@ -93,6 +93,7 @@ class I420VideoSource : public VideoSource {
virtual void FillFrame() {
// Read a frame from input_file.
+ if (frame_ != 3)
if (fread(img_->img_data, raw_sz_, 1, input_file_) == 0) {
limit_ = frame_;
}
This forces the frame that the resolution changes on to be coded
with no motion, only scaling, and improves the quality of the
result.
Change-Id: I1ee75d19a437ff801192f767fd02a36bcbd1d496
Wrote SSE2 version of vp9_dc_only_idct_add_c function. In order to
improve performance, clipped the absolute diff values to [0, 255].
This allowed us to keep the additions/subtractions in 8 bits.
Test showed an over 2% decoder performance increase.
Change-Id: Ie1a236d23d207e4ffcd1fc9f3d77462a9c7fe09d
Ensure that all inter prediction goes through a common code path
that takes scaling into account. Removes a bunch of duplicate
1st/2nd predictor code. Also introduces a 16x8 mode for 8x8
MVs, similar to the 8x4 trick we were doing before. This has an
unexpected effect with EIGHTTAP_SMOOTH, so it's disabled in that
case for now.
Change-Id: Ia053e823a8bc616a988a0af30452e1e75a739cba
The commit improves the 32x32 forward dct implementation:
1. change to use same constants and rounding as other forward dcts
2. select rounding to specifically minimize the roundtrip error, which
improved average 19/block to .77/block using 100000 random input.
Test showed a small but consistent gain on all test sets, about .15%
Change-Id: If0afd6a71880a522f60c1c234be0462092c2eb53
Removing switch statements for inverse hybrid transforms. Making code style
consistent for all similar transform implementations. Renaming shortpitch
and short_pitch variables to half_pitch.
Change-Id: I875f7a82aae4e8063a58777bf1cc3f1e67b48582
Rebased.
Remove the old matrix multiplication transform computation. The 16x16
ADST/DCT can be switched on/off and evaluated by setting ACTIVE_HT16
300/0 in vp9/common/vp9_blockd.h.
Change-Id: Icab2dbd18538987e1dc4e88c45abfc4cfc6e133f
This patch alters the balance of context between the
coefficient bands (reflecting the position of coefficients
within a transform blocks) and the energy of the previous
token (or tokens) within a block.
In this case the number of coefficient bands is reduced
but more previous token energy bands are supported.
Some initial rebalancing of the default tables has been
by running multiple derf clips at multiple data rates using
the ENTOPY_STATS macro. Further balancing needs to be
done using larger image formatsd especially in regard to
the bigger transform sizes which are not as well represented
in encodings of smaller image formats.
Change-Id: If9736e95c391e711b04aef6393d26f60f36e1f8a
Removing redundant 'extern' keywords and parentheses, fixing indentation,
making variable names lower case, using short expressions x *= c
instead of x = x * c, minor code simplifications.
Change-Id: If6a25fcf306d1db26e90d27e3c24a32735c607de
This patch includes 4x4, 8x8, and 16x16 forward butterfly ADST/DCT
hybrid transform. The kernel of 4x4 ADST is sin((2k+1)*(n+1)/(2N+1)).
The kernel of 8x8/16x16 ADST is of the form sin((2k+1)*(2n+1)/4N).
Change-Id: I8f1ab3843ce32eb287ab766f92e0611e1c5cb4c1
Refactors the switchable filter search in the rd loop to
improve encode speed.
Uses a piecewise approximation to a closed form expression to estimate
rd cost for a Laplacian source with a given variance and quantization
step-size.
About 40% encode time reduction is achieved.
Results (on a feb 12 baseline) show a slight drop:
derf: -0.019%
yt: +0.010%
std-hd: -0.162%
hd: -0.050%
Change-Id: Ie861badf5bba1e3b1052e29a0ef1b7e256edbcd0
- Using multiplication and shifting instead of division in
intra prediction.
- Maximum absolute difference is 1 for division statements
in d45, d27, d63 prediction modes. However, errors can
cumulate for large block sizes when using already predicted
values.
- Maximum number of non-matching result values in loops using
division are:
4x4 0/16
8x8 0/64
16x16 10/256
32x32 13/1024
64x64 122/4096
Overall PSNR
derf: 0.005
yt: -0.022
std-hd: 0.021
hd: -0.006
Change-Id: I3979a02eb6351636442c1af1e23d6c4e6ec1d01d
rebased.
This patch includes 16x16 butterfly inverse ADST/DCT hybrid
transform. It uses the variant ADST of kernel
sin((2k+1)*(2n+1)/4N),
which allows a butterfly implementation.
The coding gains as compared to DCT 16x16 are about 0.1% for
both derf and std-hd. It is noteworthy that for std-hd sets
many sequences gains about 0.5%, some 0.2%. There are also few
points that provides -1% to -3% performance. Hence the average
goes to about 0.1%.
Change-Id: Ie80ac84cf403390f6e5d282caa58723739e5ec17
The commit changes the coding mode to lossless whenever the lowest
quantizer is choosen.
As expected, test results showed no difference for cif and std-hd
set where Q0 is rarely used. For yt and yt-hd set, Q0 is used for
a number of clips, where this commit helped a lot in the high end.
Average over all clips in the sets:
yt: 2.391% 1.017% 1.066%
hd: 1.937% .764% .787%
Change-Id: I9fa9df8646fd70cb09ffe9e4202b86b67da16765
Since there is no Y2, these values are always zero. This changes the
bitstream results slightly, hence a separate commit.
Change-Id: I2f838f184341868f35113ec77ca89da53c4644e0
These allow sending partial bitstream packets over the network before
encoding a complete frame is completed, thus lowering end-to-end
latency. The tile-rows are not independent.
Change-Id: I99986595cbcbff9153e2a14f49b4aa7dee4768e2
Since addition of the larger-scale transforms (16x16, 32x32), these
don't give a benefit at macroblock-sizes anymore. At superblock-sizes,
2nd-order transform was never used over the larger transforms. Future
work should test whether there is a benefit for that use case.
Change-Id: I90cadfc42befaf201de3eb0c4f7330c56e33330a
This patch abstracts the selection of the coefficient band
context into a function as a precursor to further experiments
with the coefficient context.
It also removes the large per TX size coefficient band structures
and uses a single matrix for all block sizes within the test function.
This may have an impact on quality (results to follow) but is only an
intermediate step in the process of redefining the context. Also the
quality impact will be larger initially because the default tables will
be out of step with the new banding.
In particular the 4x4 will in this case only use 7 bands. If needed we
can add back block size dependency localized within the function, but
this can follow on after the other changes to the definition of the
context.
Change-Id: Id7009c2f4f9bb1d02b861af85fd8223d4285bde5
This is an initial step to facilitate experimentation
with changes to the prior token context used to code
coefficients to take better account of the energy of
preceding tokens.
This patch merely abstracts the selection of context into
two functions and does not alter the output.
Change-Id: I117fff0b49c61da83aed641e36620442f86def86
1. Added a bit in frame header to to indicate if a frame is encoded
in lossless mode, so decoder does not make the decision based on Q0
2. Minor changes to make sure that lossy coding works same as when
the lossless experiment is not enabled.
3. Renamed function pointers for transforms to be consistent, using
prefix fwd_txm and inv_txm for forward and inverse respectively
To encode in lossless mode, using "--lossless=1 --min-q=0 --max-q=0"
with vpxenc.
Change-Id: Ifae53b26d2ffbe378d707e29d96817b8a5e6c068
Initial ssse3 convolve avg functions and is one step closer
to using x86inc.asm. The decoder performance improved by 8% for
the test clip used. This should be revisited later to see if
averaging outside the loop is better than having many similar
filter functions.
Change-Id: Ice3fafb423b02710b0448ffca18b296bcac649e9
Removal of the NEWCOEFCONTEXT experiment to
reduce code clutter and make it easier to experiment with
some other changes to the coefficient coding context.
Change-Id: Icd17b421384c354df6117cc714747647c5eb7e98
This is after discussion with the hardware team. Update the unit test
to take these sizes into account. Split out some duplicate code into
a separate file so it can be shared.
Change-Id: I8311d11b0191d8bb37e8eb4ac962beb217e1bff5
Implement convolve8_avg using common functions which are already optimized
instead of using more obscure ones which have only C versions. Encoder
overall speed-up of about 12%.
Change-Id: I8c57aa76936c8a48f22b115f19f61d9f2ae1e4b6
fixed format issues.
Implement the inverse 4x4 ADST using 9 multiplications. For this
particular dimension, the original ADST transform can be
factorized into simpler operations, hence is retained.
Change-Id: Ie5d9749942468df299ab74e90d92cd899569e960
A 16 bit overflow condition occurs when using the EIGHTTAP_SMOOTH filters.
(vp9_sub_pel_filters_8lp) Changed the order of the adds to fix this problem.
Also added ssse3 support for 4x4 subpixel filtering.
Change-Id: I475eaadae920794c2de5e01e9735c059a856518e
Replace as_mv.{first, second} with a two element array, so that they
can easily be processed with an index variable.
Change-Id: I1e429155544d2a94a5b72a5b467c53d8b8728190
Also port the 4x4, 16x16, 8x16 and 16x8 versions to x86inc.asm; this
makes them all slightly faster, particularly on x86-64. Remove SSE3
sad16x16 version, since the SSE2 version is now faster.
About 1.5% overall encoding speedup.
Change-Id: Id4011a78cce7839f554b301d0800d5ca021af797
Pass the current mb row and column around rather than the
recon_yoffset and recon_uvoffset, since those offsets will
change from predictor to predictor, based on the reference
frame selection.
Change-Id: If3f9df059e00f5048ca729d3d083ff428e1859c1
This commit adds the 8 tap SSSE3 subpixel filters back into the code
underneath the convolve API. The C code is still called for 4x4
blocks, as well as compound prediction modes. This restores the
encode performance to be within about 8% of the baseline.
Change-Id: Ife0d81477075ae33c05b53c65003951efdc8b09c
Tweak to default mode context to account for the fact
that when there are no non zero motion candidates
Nearest is now the preferred mode for coding a 0,0
vector.
Also resolve duplicate function name and typos.
Change-Id: I76802788d46c84e3d1c771be216a537ab7b12817
Refactor the 8x8 inverse hybrid transform. It is now consistent
with the new inverse DCT. Overall performance loss (due to the
use of this variant ADST, and the rounding errors in the butterfly
implementation) for std-hd is -0.02.
Fixed BUILD warning.
Devise a variant of the original ADST, which allows butterfly
computation structure. This new transform has kernel of the
form: sin((2k+1)*(2n+1) / (4N)). One of its butterfly structures
using floating-point multiplications was reported in Z. Wang,
"Fast algorithms for the discrete W transform and for the discrete
Fourier transform", IEEE Trans. on ASSP, 1984.
This patch includes the butterfly implementation of the inverse
ADST/DCT hybrid transform of dimension 8x8.
Change-Id: I3533cb715f749343a80b9087ce34b3e776d1581d
This patch adds column-based tiling. The idea is to make each tile
independently decodable (after reading the common frame header) and
also independendly encodable (minus within-frame cost adjustments in
the RD loop) to speed-up hardware & software en/decoders if they used
multi-threading. Column-based tiling has the added advantage (over
other tiling methods) that it minimizes realtime use-case latency,
since all threads can start encoding data as soon as the first SB-row
worth of data is available to the encoder.
There is some test code that does random tile ordering in the decoder,
to confirm that each tile is indeed independently decodable from other
tiles in the same frame. At tile edges, all contexts assume default
values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode),
and motion vector search and ordering do not cross tiles in the same
frame.
t log
Tile independence is not maintained between frames ATM, i.e. tile 0 of
frame 1 is free to use motion vectors that point into any tile of frame
0. We support 1 (i.e. no tiling), 2 or 4 column-tiles.
The loopfilter crosses tile boundaries. I discussed this briefly with Aki
and he says that's OK. An in-loop loopfilter would need to do some sync
between tile threads, but that shouldn't be a big issue.
Resuls: with tiling disabled, we go up slightly because of improved edge
use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf,
~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5%
on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is
concentrated in the low-bitrate end of clips, and most of it is because
of the loss of edges at tile boundaries and the resulting loss of intra
predictors.
TODO:
- more tiles (perhaps allow row-based tiling also, and max. 8 tiles)?
- maybe optionally (for EC purposes), motion vectors themselves
should not cross tile edges, or we should emulate such borders as
if they were off-frame, to limit error propagation to within one
tile only. This doesn't have to be the default behaviour but could
be an optional bitstream flag.
Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
Update the code to call the new convolution functions to do subpixel
prediction rather than the existing functions. Remove the old C and
assembly code, since it is unused. This causes a 50% performance
reduction on the decoder, but that will be resolved when the asm for
the new functions is available.
There is no consensus for whether 6-tap or 2-tap predictors will be
supported in the final codec, so these filters are implemented in
terms of the 8-tap code, so that quality testing of these modes
can continue. Implementing the lower complexity algorithms is a
simple exercise, should it be necessary.
This code produces slightly better results in the EIGHTTAP_SMOOTH
case, since the filter is now applied in only one direction when
the subpel motion is only in one direction. Like the previous code,
the filtering is skipped entirely on full-pel MVs. This combination
seems to give the best quality gains, but this may be indicative of a
bug in the encoder's filter selection, since the encoder could
achieve the result of skipping the filtering on full-pel by selecting
one of the other filters. This should be revisited.
Quality gains on derf positive on almost all clips. The only clip
that seemed to be hurt at all datarates was football
(-0.115% PSNR average, -0.587% min). Overall averages 0.375% PSNR,
0.347% SSIM.
Change-Id: I7d469716091b1d89b4b08adde5863999319d69ff
This commit introduces a new convolution function which will be used to
replace the existing subpixel interpolation functions. It is much the
same as the existing functions, but allows for changing the filter
kernel on a per-pixel basis, and doesn't bake in knowledge of the
filter to be applied or the size of the resulting block into the
function name.
Replacing the existing subpel filters will come in a later commit.
Change-Id: Ic9a5615f2f456cb77f96741856fc650d6d78bb91
This commit changes the 4x4 iDCT to use same algorithm & constants as
other iDCTs. The 4x4 fDCT is also changed to be based on the new iDCT.
Change-Id: Ib1a902693228af903862e1f5a08078c36f2089b0
This commit makes the NearestMV match the chosen
best reference MV. It can be a 0,0 or non zero vector
which means the the compound nearest mv mode can
combine a 0,0 and a non zero vector.
Change-Id: I2213d09996ae2916e53e6458d7d110350dcffd7a
and called this function in vp9_dequant_idct_add_32x32_c when
eob == 1. For the test clip used, the decoder performance improved
by 21+%. Based on Yaowu's 16 point idct work.
Change-Id: Ib579a90fed531d45777980e04bf0c9b23c093c43
fixed a function prototypes to prevent compiler warnings;
removed a function not in use;
un-capitialize "Refstride" to ref_stride
Change-Id: Ib4472b6084f357d96328c6a06e795b6813a9edba
This commit changes the inverse 16 point dct to use the same algorithm
as the one for 32 point idct. In fact, now 16 point dct uses the exact
version of the souce code for even portion of the 32 point idct.
Tests showed current implementation has significant better accuracy
than the previous version. With this implementation and the minor bug
fix on forward 16 point dct, encoding tests showed about 0.2% better
compression of CIF set, test results on std-hd setting pending.
Change-Id: I68224b60c816ba03434e9f08bee147c7e344fb63
This commit changes the 32x32 idct to use integer only. The algorithm
was taken directly from "A Fast Computational Algorithm for the
Discrete Cosine Tranform" by W. Chen, et al., which was published in
IEEE Transaction on Communication Vol. Com.-25 No. 9, 1977. The signal
flow graph in the original paper is for a 32 point forward dct, the
current implementation of inverse DCT was done by follow the graph in
reversed direction.
With this implementation, the 32 point inverse dct contains a 16 point
inverse dct in its even portion, similarly the 16 point idct further
contains 8 point and 4 point inverse dcts.
As of patch 4, encoding tests showed there is no compression loss when
compared against the floating point baseline. Numbers even showed very
small postives. (cif: .01%, std-hd: .05%).
Change-Id: I2d2d17a424b0b04b42422ef33ec53f5802b0f378
Adds a special combination mode specific to intra prediciton
mode D45.
Current results with the compound inter/intra experiment:
derf: 0.2%
yt: 0.55%
std-hd: 0.75%
hd: 0.74%
Change-Id: I8976bdf3b9b0b66ab8c5c628bbc62c14fc72ca86
