This commit moves token_cache buffer into macroblock struct, instead
of defining as a local variable in cost_coeffs. This avoids repeatedly
re-allocating memory space in the rate-distortion optimization loop.
The runtime at speed 0 reduces:
bus 2000kbps, 161692ms to 159951ms
football 600kbps, 229505ms to 225821ms
Change-Id: If7da6b0b6d8c5138a16271a33c4548fba33d8840
Values of MODE_UPDATE_PROB and VP9_COEF_UPDATE_PROB are equal, so replacing
them with one constant. Inlining appropriate arguments for functions:
vp9_cond_prob_diff_update (encoder)
vp9_diff_update_prob (decoder)
Change-Id: I1255a1cb477743b799b3bfbbcd8de6b32b067338
Converts the constant rddiv parameter to 128 (from 100) and
implements RDCOST with bit-shift rather than multiplication.
Other parameters are also adjusted to roughly keep the same
balance between Rate and Distortion.
There is a slight speed-up of about 0.5-1% (at speed 0) as
testted on football_cif.
There is a slight change in performance due to small change
in the parameters.
derfraw300: +0.033%
stdhdraw250; +0.102%
Change-Id: I70ac69f58fa71c83108f68fe41796cd19d1fc760
The commit changes to mask available intra prediction modes for test
based on prediction block size.
With this patch, encoding time of CpuUsed 2 reduces from 10% to 20% for
HD clips with a compression drop of 0.2%
Change-Id: I65f320f1237c0f5ae3a355bf7caf447f55625455
When the codec in VBR (or cq) mode hits its max q limits and is
struggling to hit a target bandwidth, the bit target per frame collapses.
In the first instance normal frames cap out at the maximum allowed
Q and then the ARF and GFs do the same. This latter behavior is not
generally desirable as GFs and ARFs are only effective from a quality
and data rate perspective if they have at lease some level of -Q delta
compared to the surrounding frames.
In this patch I define a separate max Q for GFs and ARFs that is
derived from but somewhat lower than that defined for normal frames.
In effect there is a minimum Q delta that will always be available for
GFs and ARFs regardless of the target rate and MAXQ setting.
This may of course mean that the absolute lowest rate obtainable for
a given clip is somewhat higher.
Change-Id: I268868b28401900d0cd87e51e609cd3b784ab54a
Renames:
fdct4_1d -> fdct4
fadst4_1d -> fadst4
fdct8_1d -> fdct8
fadst8_1d -> fadst8
fdct16_1d -> fdct16
fadst16_1d -> fadst16
"_1d" suffix is redundant, so removing it. The same will happen with idct
in the next change sets.
Change-Id: Ibf421cd2f569146c6079269df7a31819c098265e
This commit re-designs the per transformed block rate-distortion
costs tracking buffers. It removes redundant buffer usage, makes
the needed context memory allocation per VP9_COMP instance and
reuses the same buffer sets inside the rate-distortion optimization
search loop, thereby avoiding repeatedly requiring memory space.
It reduces speed 0 runtime:
bus at 2000 kbps from 166763ms to 158967ms,
football at 600 kbps from 246614ms to 234257ms.
Both about 5% speed-up. Local tests suggest about 2% to 5% speed-up
for speed 1 and 2 settings. This does not change compression
performance.
Change-Id: I363514c5276b5cf9a38c7251088ffc6ab7f9a4c3
Increases these parameters.
There is a small efficiency gain.
Change-Id: Ie5f0ddb39c907d335e0dafa5eb112365a81f4542
derfraw300: +0.091%
stdhdraw250: +0.238%
The intra mode distortion adjustment for skip_encode feature was
broken in the refactoring cc91851. This commit fixes it and tunes
the distortion models used therein.
Change-Id: I0d676e82f8e855536a90cf9b3e3fdefafcd886c6
Use b_mode_info to store the inter prediction mode of sub8x8 block,
in replacement of the use of partition_info. Remove redundant buffer
update for partition_info. For bus_cif at 2000 kbps, this seem to make
speed 0 about 1% faster.
Change-Id: Id1b3be45e75a24fb4b42335ac480c23e440978f6
We already have itxm_add member in MACROBLOCKD structure. Both
inv_txm4x4_1_add and inv_txm4x4_add are just its special cases for
different eob values. But eob logic is already implemented in
vp9_iwht4x4_add and vp9_idct4x4_add (that's why also removing
inverse_transform_b_4x4_add).
Change-Id: I80bec9b6f7d40c5e5033c613faca5c819c3e6326
For CpuUsed 1 & 2, this commit allow to skip retangular partition check
when NONE is better than SPLIT. It also changed to allow such logic
on alt ref frame coding rather than use square partition all them. The
change has gain compressio about .3% on yt and ythd for both 1&2, It
helped .6% compression on cif and stdhd for both CpuUsed 1&2.
Change-Id: I814b653baf89f59acd20e042629a12938a1bd4e5
This commit allows sub8x8 intra modes test in the rate-distortion
loop for hd sequences in speed 1 and 2.
For sequence y90n of hd set at 8000 kbps, speed 2 runtime goes
from 207s to 210s. For ped_1080p at 3000 kbps, speed 2 runtim goes
from 336s to 337s. Both are running with 300 frames.
This improves compression performance by 0.24% for stdhd and 0.32%
for hd.
Change-Id: I173ca38a6411565ae6cfadd184c42b2070c5de1f
The idea is to have the following names for each transform size:
vp9_idct4x4_add
vp9_idct4x4_1_add
vp9_idct4x4_10_add
vp9_idct4x4_16_add
vp9_idct8x8_add
vp9_idct8x8_1_add
vp9_idct8x8_10_add
vp9_idct8x8_64_add
etc for 16x16, 32x32
The actual list of renames in this patch:
vp9_idct_add_lossless -> vp9_iwht4x4_add
vp9_short_iwalsh4x4_add -> vp9_iwht4x4_16_add
vp9_short_iwalsh4x4_1_add -> vp9_iwht4x4_1_add
vp9_idct_add -> vp9_idct4x4_add
vp9_short_idct4x4_add -> vp9_idct4x4_16_add
vp9_short_idct4x4_1_add -> vp9_idct4x4_1_add
Change-Id: I6f43f7437c68dd30cdd05d72e213765578ed30b1
Speed 4 still does not give a big gain over speed 3.
This just cleans it up a little from the last patch and comments
out features that do not seem to be giving much benefit.
Change-Id: I5f366e6160e1dbe5dc45cf5eb90cc02712baa1b6
Allow selective masking of individual split modes rather than
just a single on / off flag.
For speed 2 recovers the large speed loss seen for some derf
clips in change Ie6bdfa0a370148dd60bd800961077f7e97e67dd4
and a small quality gain.
For speed 1 10 % speed increase observed locally on some derf clips
for minimal quality change.
Change-Id: If86191087b93cbc05351c26c60c7933e2149e485
This commit removes the redundant second reference frame check in
the rate-distortion optimization loop for sub8x8 blocks.
Change-Id: I13a57a6f624c4a9bcef02ff2a867fa30d8b44a93
This commit defines b_mode_info as a struct type. This will allow
us to further remove the use of PARTITION_INFO in the encoding process.
Change-Id: I975b0f7d557b5e0f66545a61b472def76b671cce
This commit separates the rate-distortion optimization loop of
superblocks from that of sub8x8 blocks. This allows better design
rate-distortion optimization search loop for each setting. It also
removes the use of SPLITMV and I4X4_PRED therein.
No performance change in speed 0 settings. For bus@CIF at 2000kbps,
the speed 1 runtime goes from 48009ms to 43894ms (about 10% faster).
The overall compression performance on derf changed by -0.021%.
Speed 2 runtime goes from 27114ms to 28700ms (6% slower), while the
overall coding efficiency goes up by 1.629% for derf, 1.236% for yt.
Change-Id: Ie6bdfa0a370148dd60bd800961077f7e97e67dd4
Substantial reworking of the speed vs quality trade offs for
speed 1 and 2.
In this patch I am attempting to freeze the "quality" meaning of
speeds 1 and 2 relative to speed 0 so that in future we can
better evaluate progress.
I am targeting :
Speed 1 quality ~-5% vs speed 0.
Speed 2 quality ~-10% vs speed 0
It is inevitable that quality will still fluctuate a little as we adjust
settings and add new features, but we will attempt to keep as
close as possible to these values. Above speed 2 things will remain
a bit more fluid for now.
In this patch speed 1 is approximately 4-5x as fast as speed 0. This
is similar to before but the quality hit is a lot less. Likewise speed 2
is approximately 2x as fast as speed 1 but is similar in quality to the
previous speed 1 configuration.
Also slight change to behavior of FLAG_EARLY_TERMINATE to insure
all reference frames get at least one rd test. Important for very low
variance regions.
WIP :- Added a new speed level with old speed 4 becoming speed 5.
Speed 3 and 4 tradeoffs still WIP
Change-Id: Ic7a38dd7b5b63ab1501f9352411972f480ac6264
This commit causes use last partition to consider whether a 64x64 has
motion that might make a new partitioning worth while.
Change-Id: I3a57bedef4f3cd961fadbfa96651c206fa36da4a
Simplify the k_cvtlo_epi16 and k_cvthi_epi16 to only two
instructions. Then inlined them.
quoting from intel MMX_App_Compute_16bit_Vector.pdf
"The PMADDWD instruction multiplies four
pairs of 16-bit numbers and produces partial sums of the results
and can do so once per clock (with a three-clock latency)."
so I am assuming that there will be three clock overhead after the
last _mm_madd_pi16 command.
Even with the overhead the number of clocks in general should be
smaller. I am not sure though becasue I could not find information
about number of clocks required for instructions in k_cvtlo_epi16
and k_cvthi_epi16. I will run a test and compare the execution time.
Change-Id: Ieda4aa338f69ad3dd196ac6e7892da3cf1b47ea7
Moving functions from vp9_idct_blk to vp9_idct because these functions are
used from both encoder and decoder. Removing duplicated code from
vp9_encodemb.c and reusing existing functions.
Change-Id: Ia0a6782f8c4c409efb891651b871dd4bf22d5fe8
Make encoder skip rectangular partition check in speed 1 and above,
when early termination was triggered in partition split.
Thanks Guillaume (gmartres@) for catching this issue.
This change makes bus_cif at 2000kbps speed 1 runtime goes down from
25612ms to 23438ms (about 9% speed-up), at the expense of -0.235%
performance down.
Change-Id: I98613fad081a261d30d5fa206f934ca70601c180
The code now takes into account temporal and spatial
information to determine the partition size range, but the
frequency counts have been removed.
The net effect is similar in quality but about 10% faster.
Change-Id: I39a513fb79cec9177b73b2a7218f0da70963ae95
This patch deletes the variance based speed three partitioning.
Speed 3 now uses the same partitioning method as speed 2
but with some stricter conditions.
The speed and quality are now somewhere between speeds 2 and 4
whereas before it was worse in both than speed 4.
Change-Id: Ia142e7007299d79db3ceee6ca8670540db6f7a41
Some small changes to the quantizer mapping functions.
Also includes some cleanups.
Change-Id: I9dea29b24015f6e6697012a0e4d8983049d8e5c7
Results:
derfraw300: +0.106%
stdhdraw250: +0.139%
Don't divide RDMULT and RDDIV by 100 when RDMULT > 1000. This was
probably done to avoid overflow when the rd cost was stored in a 32 bits
integer but this is not the case anymore. This change will make it easier
to support multiple quantizers per frame.
derf compression gain at speed 0: 0.037%
Change-Id: Ibeeb9b7cfa1a132a7af41bc90fc07a3bba0857f6
Both first pass and mbgraph search use block size 16x16 for motion
estimation. This commit put a limit of motion vector range. The
effective range allows the entire 16x16 with required subpel
interpolation input to be completely outside image border, but
not any further away from image border.
Change-Id: Id70a5ed08be49e70959f064859d72adc7d775d08
INT64_MAX may be assigned as RDCOST when RDCSOST computation is skipped
for speed, this commit to prevent INT64_MAX from being used as real
RDCOST in transform size decision.
Change-Id: I89a945134191bbdea1f1431ade70424ac079eaac
After change of MI context storage , mi_8x8[] pointer may be null for
a block outside of image border. The commit changes to access the data
only after validation of mi_row and mi_col.
Change-Id: I039c4eb486a228ea9d8e5f35ab9ae6717d718bf3
39c7b01d accidently reverted the row/col initialization, which broke
mv clamps, which is dependent on the sites for valid motion vector
range. This commit fixed the issue.
Change-Id: Ibcce0226e0360b1ef483fe760b2e33f1af4bf494
This commit enables forcing all coefficients zero per transformed
block, when its rate-distortion cost is lower than regular coeff
quantization.
The overall performance improvement (including its parent patch on
calculating rd cost per transformed block) at speed 1:
derf: 0.298%
yt: 0.452%
hd: 0.741%
stdhd: 0.006%
Change-Id: I66005fe0fd7af192c3eba32e02fd6d77952accb5
Adds modeled functions to decide the qp for altref frames in constant q
mode similar to other functions in use in bitrate mode.
Also turns on the constrained quality mode (end-usage=2) option which
was turned off before. Basic testing shows the mode works in principle,
to cap bitrate to the target-bitrate specified, while allowing lower
bitrate depending on the cq-level specified. The mode will need to be
improved over time.
Results for constant quality vs bitrate control mode:
derfraw300/fullderfraw: +3.0% at constant quality over bitrate control.
fullstdhdraw: +4.341%
stdhdraw250: +5.361%
Change-Id: If5027c9ec66c8e88d33e47062c6cb84a07b1cda9
This commit makes the rate-distortion optimization loop evaluate
the rd costs of regular quantization and all zero coeffs, per
transformed block. It improves speed 1 compression performance:
derf: 0.245%
yt: 0.515%
For a large partition that consists multiple transformed blocks,
this allows more flexibility to selectively force a portion of
them coded as all zero coeffs, as well be continued in the next
patches.
Change-Id: I211518be4179747b57375696f017d1160cc91851
The sub8x8 blocks has its own motion vector reference scheme. The
mv_pred is only used blocks of sizes 8x8 and above, to find the
starting point for motion search.
This change does not change any coding behavior. It makes the
encoding process slightly faster. (0.5% speed-up for local test on
speed 1.)
Change-Id: I746ee6ef0eac19aa3621be014afa12be8d82cbb9
The fake token EOSB may cause invaild memory read in pack token, this
commit reworked the loop to avoid such invalid read.
Change-Id: I37fdfce869b44a7f90003f82a02f84c45472a457
This commit cleans up the second reference check in the
rate-distortion optimization loop of sub8x8 blocks.
Change-Id: Ife68feaa4cddbfad2878c9b44d3012788d634f97
The commit added reset of pred_mv at the beginning of each SB64x64
partition mv search, also limited the usage of pred_mv only when
search on the largest partition is already done. This is to fix
a crash at speed 1/2 encoder where an invalid mv is used in mv
search.
Change-Id: I39010177da76d054e3c90b7899a44feb2e3a5b1b
This commit enables adaptive constraint on motion search range for
smaller partitions, given the motion vectors of collocated larger
partition as a candidate initial search point.
It makes speed 0 runtime of bus at CIF and 2000 kbps goes from
167s down to 162s (3% speed-up), at 0.01dB performance gains. In
the settings of speed 1, this makes the runtime goes from 33687 ms
to 32142 ms (4.5% speed-up), at 0.03dB performance gains.
Compression performance wise, it gains at speed 1:
derf 0.118%
yt 0.237%
hd 0.203%
stdhd 0.438%
Change-Id: Ic8b34c67810d9504a9579bef2825d3fa54b69454
Mis-merge of the following change managed to break mode order
and delete two mode options (new alt ref and near alt ref)
It also created a situation where we could test two undefined
modes off the end of the VP9_mode_order[] data structure.
"clang warnings : remove split and i4x4_pred fake modes"
"Change Id: I8ef3c*"
Initial testing on Akiyo at speed 2.
101.35 44.567 44.447 improves to
96.82 44.915 44.815
Approx 0.3-0.4db gain and 2.5% size reduction
Change-Id: Icff813e7c0778d140ad4f0eea18cf1ed203c4e34
Removes this speed feature since it is very slow and unlikely
to be used in practice. This cleanup removes a bunch of unnecessary
complications in the outer encode loop.
Change-Id: I3c66ef1ca924fbfad7dadff297c9e7f652d308a1
Propose some changes to the speed 2 settings to improve quality.
In particular, turns off the adjust_thresholds_by_speed feature
which improves results by 6%. Also removes the code for
adjust_thresholds_by_speed since it conflicts with the adaptive
rd thresh feature.
Overall, with this change speed 2 is -15.2% from speed 0 settings,
on derf, which is significantly better than -21.6% down before.
Change-Id: I6e90a563470979eb0c258ec32d6183ed7ce9a505
mode_info_context was stored as a grid of MODE_INFO structs.
The grid now constists of pointers to MODE_INFO structs. The
MODE_INFO structs are now stored as a stream (decoder only),
eliminating unnecessary copies and is a little more cache
friendly.
Change-Id: I031d376284c6eb98a38ad5595b797f048a6cfc0d
The c code implementation of 32x32 quantization does the zbin check
of all coefficients prior to the quant/dequant loop, hence removing
the redundant zbin check inside the loop. This only affects the
c code version. SSSE3 version does not separate the zbin check out.
Change-Id: Ic197a7d61d0b25fcac3cc092987651378cb56e4e
If the current obtained distortion is very small, which happens
for static image case, we pick the current partition type without
further split checking.
This won't affect regular videos. For static videos, we got 10%~12%
encoding speed gain. PSNR was better for some clips, and worse for
others. Overall it was even.
Change-Id: If787a57bedf46fc595ca4f5ded2b0c0a69e9fdef
Thank Paul for the suggestions. While turning on static-thresh
for static-image videos, a big jump on bitrate was seen. In this
patch, we detected static frames in the video using first-pass
stats. For different cases, disable encode breakout or reduce
encode breakout threshold to limit the skipping.
More modification need be done to break incorrect partition
picking pattern for static frames while skipping happens.
Change-Id: Ia25f47041af0f04e229c70a0185e12b0ffa6047f
A previous speed feature skipped modes not used in earlier
partitions but this not longer worked as intended following
changes to the partition coding order and in conjunction
with some other speed features (Especially speed 2 and above).
This modified mode skip feature sets a mask after the first X
modes have been tested in each partition depending on the
reference frame of the current best case.
This patch also makes some changes to the order modes are
tested to fit better with this skip functionality.
Initial testing suggests speed and rd hit count improvements
of up to 20% at speed 1. Quality results. (derf -1.9%, std hd +0.23%).
Change-Id: Idd8efa656cbc0c28f06d09690984c1f18b1115e1
Sample app: vp9_spatial_scalable_encoder
vpx_codec_control extensions:
VP9E_SET_SVC
VP9E_SET_WIDTH, VP9E_SET_HEIGHT, VP9E_SET_LAYER
VP9E_SET_MIN_Q, VP9E_SET_MAX_Q
expanded buffer size for vp9_convolve
modified setting of initial width in vp9_onyx_if.c so that layer size
can be set prior to initial encode
Default number of layers set to 3 (VPX_SS_DEFAULT_LAYERS)
Number of layers set explicitly in vpx_codec_enc_cfg.ss_number_layers
Change-Id: I2c7a6fe6d665113671337032f7ad032430ac4197
The 16x16 transform unit test suggested that the peak coefficient
value can reach 32639. This could cause potential overflow issue
in the SSSE3 implmentation of 16x16 block quantization. This commit
fixes this issue by replacing addition with saturated addition.
Change-Id: I6d5bb7c5faad4a927be53292324bd2728690717e
Adds a new end-usage option for constant quality encoding in vpx. This
first version implemented for VP9, encodes all regular inter frames
using the quality specified in the --cq-level= option, while encoding
all key frames and golden/altref frames at a quality better than that.
The current performance on derfraw300 is +0.910% up from bitrate control,
but achieved without multiple recode loops per frame.
The decision for qp for each altref/golden/key frame will be improved
in subsequent patches based on better use of stats from the first pass.
Further, the qp for regular inter frames may also be varied around the
provided cq-level.
Change-Id: I6c4a2a68563679d60e0616ebcb11698578615fb3
mode_info_context was stored as a grid of MODE_INFO structs.
The grid now constists of a pointer to a MODE_INFO struct and
a "in the image" flag. The MODE_INFO structs are now stored
as a stream, eliminating unnecessary copies and is a little
more cache friendly.
For the test clips used, the decoder performance improved
by ~4.3% (1080p) and ~9.7% (720p).
Patch Set 2: Re-encoded clips with latest. Now ~1.7% (1080p)
and 5.9% (720p).
Change-Id: I846f29e88610fce2523ca697a9a9ef2a182e9256
The 32x32 forward transform can potentially reach peak coefficient
value close to 32700, while the rounding factor can go upto 610.
This could cause overflow issue in the SSSE3 implementation of 32x32
quantization process.
This commit resolves this issue by replacing the addition operations
with saturated addition operations in 32x32 block quantization.
Change-Id: Id6b98996458e16c5b6241338ca113c332bef6e70
Moves counting of mv branches to where we have a new mv, instead of after
the whole frame is summed.
Change-Id: I945d9f6d9199ba2443fe816c92d5849340d17bbd
