Avoid mm_set1 operation when it's possible to create the constant with
one bit-shift operation from another instead. Thanks Intel for
3-operand instruction encoding!
AVX2 support on a processor should always imply BMI and ABM support.
The lzcnt and tzcnt instructions have more suitable semantics in the
corner case that source word is 0, and allow us to even handle that
scenario without a branch. Apparently Visual Studio will already
include this support when building with AVX2 enabled, so only the
automake files need to be tweaked.
Adds command line option --fast-residual-cost=<limit>. When QP is below
the limit, estimates the cost of coding the residual coefficients from
the sum of absolute coefficients. Skipping CABAC is not worth it with
high QPs because there are fewer coefficients so CABAC is not as slow.
PowerPC target may lack or have non-standard FPU:
$ cc -dumpmachine
powerpcspe-undermydesk-freebsd
$ cc -c -maltivec -Isrc src/strategies/altivec/picture-altivec.c
src/strategies/altivec/picture-altivec.c:1: error: AltiVec and E500 instructions cannot coexist
All intra coding is a special case so it can be permitted even though
Kvazaar normally only supports intra periods that are divisible by the
GOP length.
- Replaces lcu_dqp_enabled with max_qp_delta_depth in encoder_control_t.
- Fixes set_cu_qps so that it can handle quantization groups of
arbitrary size.
- Fixes computation of QP predictors so that it works for quantization
groups of arbitrary size.
Changes asserts on the reference count of objects to assert the value
after KVZ_ATOMIC_INC instead of directly checking the value. Fixes some
data races detected by TSan.
- Adds missing default values.
- Adds help for --crypto and --key.
- Adds help for --rd=3.
- Adds help for --sao options.
- Some changes to help wording.
The transform skip flag in cu_info_t was stored under the intra
substruct even though transform skip can be used for inter as well. This
caused bitstream errors. Fixed by moving the flag out of the substruct.
Adds an additional grid search step that starts from the zero motion
vector after the normal grid search. The search range for this step is
half of the normal range.
- Changes TZ grid search and star refinement to keep the origin constant
instead of moving to the best position after each iteration.
- Changes star refinement to loop until there is no more improvement,
instead of running the step only once.
When the starting point for integer motion estimation was selected among
the merge candidates, the candidate motion vectors were always rounded
down. This commit changes the rounding so that they are rounded to the
nearest integer MV instead.
Changes reference index list creation so that the negative references
are added to L1 in addition to L0 when biprediction is enabled and no
reordering of pictures is done. Biprediction can now be used with the
low-delay GOP structure.
Changes the crypto module to use unsigned char instead of byte. The byte
typedef is no longer included in the global namespace in crypto++ 6.0.0.
See https://github.com/weidai11/cryptopp/issues/442.
Fixes#184.
Checks the cost of flushing all coefficients of an inter block to zero.
This is much faster than doing full RDOQ but can still reduce bitrate
significantly. Encoding speed is increased since fewer coefficient bits
have to be coded with CABAC.
Moves duplicate code for writing the MVD of a single motion vector from
kvz_get_mvd_coding_cost_cabac and encoder_inter_prediction_unit to a new
function.
SMP/AMP search is performed using a lower work tree level than the
normal inter search so the prediction info must be copied up if an
SMP/AMP mode is chosen. Previously pixels and coefficient were copied as
well. Changed to only copy prediction info.
DST function was returned for inter luma transform blocks of size 4x4
even though they must use DCT. Fixed by checking the prediction mode of
the block in addition to whether it is chroma or luma.
When 4x4 intra blocks are enabled and inter search is limited to 16x16
and larger blocks, it is possible that inter search is accidentally done
for 4x4 blocks. Fixed by checking that block size is at least 8x8 before
doing inter search.
Overrides the linkers used for kvazaar, libkvazaar.la and kvazaar_tests.
When crypto++ is enabled, the C++ linker is used and when it is
disabled, the C linker is used.
This removes the need to explicitly specify -lstdc++ in configure when
crypto++ is used and fixes the build with crypto++ when libstd++ is not
installed.
- Fixes two errors in calculating the POC for the reference frame for
temporal candidate MV scaling.
- Fixes using the MV for the wrong direction when the temporal MV
predictor block uses bi-prediction.
Fixes#160.
Changes handling of intra pictures for --gop=8 so that every picture
with POC divisible by the intra period is intra. The first picture is
IDR and the rest of the intra pictures are CRA. POC is not reset at CRA
pictures. The leading pictures that follow the CRA picture are changed
to RASL so they are allowed to refer to pictures before the CRA picture.
Changes inter slice types to P when the L1 reference list is empty and
to B otherwise.
In all-intra, all pictures are now IDR pictures with POC zero.
When using --gop=8 with an intra period greater than one, a single POC
would be skipped before every intra frame. This commit fixes the problem
by turning the intra frames into BLA frames with leading pictures when
using --gop=8.
Changes thread queue so that only the jobs that are ready to run are
stored in the queue. Other jobs are kept track of by pointers in the
reverse dependency lists of other jobs. When a job is ready to run it is
appended to the queue. The job queue is stored as a linked list.
The definitions of threadqueue_queue_t and threadqueue_job_t are moved
to the .c file, turning them into opaque structs.
Makes thread queue code simpler. Fixes some TSan errors.
Adds struct inter_search_info_t for holding the parameters that are used
by most function related to inter search. Passing the parameters in
a single struct greatly reduces the number of parameters for many
functions.
Functions kvz_sao_reconstruct and encoder_sao_reconstruct used
frame->width as the stride instead of frame->rec->stride when accessing
frame->rec->data. This caused errors when using tiles and SAO.
Changes the work_tree parameter in search.c functions from an array to
a pointer. Fixes "formal parameter with requested alignment of 8 won't
be aligned" errors.
Changes field state->tile->frame->cu_array->data to point to the CU
array in the main encoder state. Removes the need to copy the CU array
to the main CU array after search.
Inter costs are computed using SAD except when fractional motion
estimation or bi-prediction is enabled. This commit changes
search_pu_inter_ref to recalculate the cost with SATD. Fixes inter/intra
cost comparisons since intra costs are always SATD costs.
Changes function kvz_get_coeff_cost to only copy the CABAC contexts and
not the whole encoder state.
Other threads could be simultaneously using the other parts of the
encoder state. Only copying the CABAC fixes a TSan data race warning.
Adds alignment attribute to lcu_coeff_t. The coefficients are sometimes
handled as 64-bit integers containing four coefficients so the arrays
should be aligned to 8 bytes.
Fixes a UBSan error about misaligned reads.
Changes OWF selection so that it is chosen based on the maximum number
of parallel CTUs. Number of threads is limited to prevent overhead from
extra threads.
Drop pthread_cond_broadcast on threadqueue->cond in function
kvz_threadqueue_waitfor. The broadcast caused threads to be woken up
more often than necessary.
Changes encoder_state_init_new_frame to only call normalize_lcu_weights
when the weights have been written to the array and rate control is
enabled. When rate control is disabled, the weights are not used.
Adds field max_inter_ref_lcu to encoder_control_t. It is used to set up
inter-LCU dependencies in encoder_state_encode_leaf and restrict motion
vectors in fracmv_within_tile.
When using WPP and OWF, the first CTU of a row depends on the last CTU
of the row below in the reference frame. This is necessary when SAO is
enabled since we currently do SAO for a whole CTU row at a time. When
SAO is disabled, however, it is unnecessary to wait for the whole row.
Changes CTUs to depend only on the CTU below in the reference frame
instead of the whole row when WPP and OWF are enabled and SAO disabled.
Gives a significant speedup when running on a machine with many CPU
cores.
Moves SAO reconstruction into encoder_state_worker_encode_lcu instead of
doing it in a separate step for the whole CTU row. Reconstruction of the
rightmost 10 pixels and bottommost 10 pixels of a CTU is delayed until
the neighboring CTU has been deblocked.
Doing SAO for the whole CTU row at a time caused unnecessary inter-CTU
dependencies when using WPP and OWF. The first CTU of a row would need
to wait until SAO was done for the row below in the previous frame.
Moving SAO reconstruction to immediately after deblocking each CTU fixes
this problem.
Adds width and height parameters to function kvz_sao_reconstruct and
changes it to take coordinates in units of pixels. This will be useful
for doing SAO for areas smaller than a whole CTU.
AVX2 filter functions read pixels in chunks of 8 or 16 bytes. At the end
of the block, the read goes out of the bounds of the pixels array. The
extra pixels do not affect the result.
Fixes valgrind complaining about the invalid reads by allocating 5 extra
pixels in kvz_get_extended_block_avx2
With low delay GOP structure, it is possible to use an intra period that
is not a multiple of the GOP structure length. Commit 00c9f52 changed
encoder_state_init_new_frame to reset POC on intra frames. GOP offset,
however, was not reset, resulting in invalid POCs and references for the
following frames.
This commit changes function kvz_encoder_feed_frame so that GOP offset
is correctly reset on intra frames.
When closing the encoder, the pictures stored in the input frame buffer
are freed by repeatedly calling kvz_encoder_feed_frame. If the encoder
was closed immediately after opening it, kvz_encoder_feed_frame would be
called with an unprepared encoder state. This would trigger an assert.
Fixed by changing kvz_encoder_feed_frame so that it does not require the
encoder state to be prepared.
Sets max_transform_hierarchy_depth_inter to 0 in SPS. This saves some
bits because split_transform_flag does not need to be coded for inter
blocks.
When SMP and AMP blocks are enabled the depth is set to 1 instead.
Otherwise inter split flag would default to 1 for SMP and AMP blocks,
resulting in an unnecessary transform split.
Changes kvazaar_close to stop all threads before freeing encoder states.
Fixes a crash when the encoder is closed before all pictures have been
encoded.
Changes kvazaar_close to free the remaining pictures in the the input
frame buffer. Fixes a memory leak when the encoder is closed while there
are pictures left in the buffer.
Replaces function kvz_intra_recon_lcu_luma and
kvz_intra_recon_lcu_chroma in intra.c with function kvz_intra_recon_cu.
The new function can handle reconstruction for both luma and chroma.
Removes some duplicated code.
Moves code for intra leaf transform block reconstruction from functions
kvz_intra_recon_lcu_luma and kvz_intra_recon_lcu_chroma to a new
function intra_recon_tb_leaf. Removes some duplicated code.
Changes functions encode_intra_coding_unit and encode_coding_tree to
take coordinate arguments in units of luma pixels instead of 8 px
blocks. This should make the code easier to understand.
Changes functions encode_transform_unit and encode_transform_coeff to
take coordinate arguments in units of luma pixels instead of 4 px
blocks. This should make the code easier to understand.
When kvazaar was built with crypto++ but running without using
encryption features, kvazaar attempted to delete an uninitialized crypto
handle. Fixed by setting the handle to NULL in kvz_encoder_state_init.
A CryptoPP::CFB_Mode<CryptoPP::AES>::Encryption was allocated at the
beginning of encoder_state_encode_leaf and was never freed. This commit
changes encoder_state_worker_encode_lcu to delete the CFB_Mode. Also
moves crypto handle from encoder_state_config_tile_t to encoder_state_t
so that it can be safely deleted without affecting other threads in the
same tile.
Moves the inline declaration of intra_mode_encryption before the type
and changes it to use the INLINE macro. Inline declaration after type
triggered a warning on GCC.
Fixes the order of acquiring locks for the job and its dependency in
kvz_threadqueue_job_dep_add. The dependency is locked before the job
that depends on it. This is the same order as in threadqueue_worker.
Acquiring the locks in different order in kvz_threadqueue_job_dep_add
and threadqueue_worker would sometimes result in a deadlock.
When using --me-early-termination=sensitive, early termination of inter
search used to always return the starting point if no tested motion
vector was good enough to continue the search. This commit changes
early_termination to always return the best motion vector and cost
found.
Changes kvz_config_validate to output an error if the given QP is out of
range and changes kvz_set_picture_lambda_and_qp to clip the QP to the
valid range if is too large after applying QP offset from GOP structure.
Replaces functions kvz_quantize_lcu_luma_residual and
kvz_quantize_lcu_chroma_residual in transform.c with function
kvz_quantize_lcu_residual. The new function can handle any of the YUV
colors. Removes some duplicated code.
Changes the way coefficients are passed from kvz_search_lcu to
kvz_encode_coding_tree. Drops fields coeff_y, coeff_u and coeff_v in
videoframe_t and instead passes them through field coeff in
endoder_state_t.
Adds option --erp-aqp for enabling adaptive QP for 360 degree video with
equirectangular projection. When projected into a spherical surface,
the middle part of the video covers relatively larger area than the top
and bottom parts. Enabling --erp-aqp sets up a ROI delta QP array which
uses higher QPs for the top and bottom of the video and lower QPs for
the middle part.
Delta QPs for LCUs are enabled when either ROI coding or rate control is
enabled. Having a single field is simpler than always checking whether
ROI or rate control is enabled.
When video width or height was not a multiple of the smallest CU size,
no prediction would be performed at the border CUs. Kvazaar would later
crash at an assertion failure when attempting to write the bitstream for
the CU.
Fixed by permitting inter and intra prediction when the CU split is
forced, even if CUs of that size would otherwise be disabled.
When WPP is enabled, a reference to SAO reconstruction job is copied
from the wavefront to the main encoder state. However, when threads are
disabled, the job is a null pointer and dereferencing it crashes the
encoder. Fixed by adding a null pointer check.
Both the thread queue and the encoder states hold pointers to the thread
queue jobs. It is possible that a job is removed from the thread queue
and freed while the encoder state is still using it. This commit adds
reference counting to threadqueue_job_t in order to fix the problem.
Fixes#161.
Pthread mutexes were used for synchronization in encmain by locking and
unlocking them from different threads. However, according to the POSIX
standard, unlocking a mutex from a different thread is undefined
behaviour. This commit replaces the mutexes by semaphores which can be
used from different threads.
Following warning was given and is false positive
error: 'memset' used with length equal to number of elements without multiplication by element size [-Werror=memset-elt-size]
When the decoding and presentation orders of pictures are different
(with GOP), the frames in YUV debug output would be in the decoding
order. This commit changes the kvazaar command line program to store the
reconstructed pictures in a buffer so that they can be output in the
presentation order.
Fixes#101.
Adds struct merge_candidates_t for holding the spatial and temporal
merge candidates. Changes functions with separate parameters for each
candidate to use the struct instead.
Limits video size so that the number of luma and chroma pixels can be
stored in an int. Fixes some integer overflows that resulted in
segmentation faults.
Finalization functions for frame and tile encoder states accessed the
frame and tile fields of the encoder state even though they might be
NULL. This is the case when the initialization of an encoder state
fails. Fixed by adding NULL checks.
Some of the fields in encoder_control_t were simply copies of the
corresponding fields in kvz_config. This commit drops the copied fields
in favor of using the fields in encoder_control_t.cfg directly.
The kvz_config struct is created by the user but kvazaar keeps a pointer
to it. It is easy to break things by modifying the configuration outside
kvazaar. In addition, kvazaar modifies the struct even though it is has
a const modifier.
This commit changes the field cfg in encoder_control_t to be a copy of
the kvz_config struct instead of a pointer, removing modifications to
the const struct and allowing users to do whatever they want with it
after opening the encoder.
The end of slice was being calculated incorrectly, which led to no tile
being created inside the slice, which led to an assert triggering.
This fixes the wrong end of slice calculation, but also disallows
wavefront rows from being created, if there would be only one.
The wavefront initialization code assumes there are always more than
one row, so the inter-frame dependency doesn't get added properly.
Fixes#153.
Main thread was stuck looping on pthread_cond_timedwait because
the abs time given on OS-X had already passed and the wait
returned immediately without releasing the mutex to allow worker
threads to proceed.
Fix was to use the gettimeofday, which returns real time instead
of monotonic, which is what pthread_cond_timedwait wants.
rdo.c:475:25: warning: absolute value function 'abs' given an argument of type 'int64_t' (aka 'long long') but has parameter
of type 'int' which may cause truncation of value [-Wabsolute-value]
current.cost = -abs(quant_cost_in_bits) + (bits << PRECISION_INC);
^
rdo.c:475:25: note: use function 'llabs' instead
current.cost = -abs(quant_cost_in_bits) + (bits << PRECISION_INC);
cfg.c:1024:74: warning: format specifies type 'size_t' (aka 'unsigned
long') but the argument has type 'unsigned long long'
[-Wformat]
fprintf(stderr, "Too large ROI size: %llu (maximum %zu).\n", size, SIZE_MAX);
This encapsulates tiles or WPP rows into their own slices, making
it possible to send them as soon as they are done, instead of waiting
for the other substreams to finish and coding the substream offsets
in the slice header.
These fixes allow more than one slice to be used to code a picture.
- Use correct number of bits to code the slice segment address.
- Don't offset_len_minus1 for slices without substreams.
Appending to the child stream doesn't work is the child is a leaf
slice state.
Simplifies flow by removing distinction between tile and slice. Now
that slice headers are written in the parent stream, there is zero
difference between tiles and slices from bitstream point of view.
Having some of the termination bits in the LCU coding and some in the
substream finalization was needlessly confusing. Doing substream
finalization directly after LCU coding makes it easy to verify that the
finalization is done correctly.
Removes one job per WPP row from the job queue.
Removes kvz_cabac_flush, because I don't like bits being put into the
bitstream implicitly. Better to have it all in the open.
When the number of merge candidates was five, biprediction search would
read past the bounds of the priority list arrays. Fixed to limit the
search to the first four candidates.
- Checks the return value of fopen when opening the ROI file. Fixes
a segfault when the file cannot be opened.
- Check that the width and height are positive. Fixes reading past the
end of the delta QP array in kvz_set_lcu_lambda_and_qp.
- Check for overflow in width * height. Fixes an overflow resulting in
a segfault.
- Properly check that fscanf succeeds. Fixes silently accepting ROI
files that are too short.
- Properly close the FILE pointer.
The function search_pu_inter_ref incorrectly rounded the coordinates of
the block to down to a multiple 8 pixels. Small SMP and AMP blocks may
start at coordinates that are not multiples of 8. Fixed by removing the
rounding.
Fixes a failing assert when --mv-constraint is used with --smp or --amp.
Stops assuming that having cfg->gop_lowdelay set means that GOP
structure is used since it is possible that cfg->gop_lowdelay is true
but cfg->gop_len is zero. Adds checks for cfg->gop_len where needed.
Fixes a possible division by zero in kvz_encoder_feed_frame.
Drop a conditional for depth > MAX_DEPTH in search_cu. The depth cannot
be greater than MAX_DEPTH (== 3) since an earlier if-clause checks that
it is less than MAX_PU_DEPTH (== 4).
Subpixel motion estimation return 0-vector when no subpixel vector is
within the constraint. Fix is to not call subpixel motion estimation
when the integer vector is not within the constraint.
Rename and reorder everything to make more sense.
- Moved input tables into their own struct and renamed them to what
they actually represent.
- Renamed pretty much every variable to comform to our style and
to make sense.
- Removed the lastCG stuff, as the function already gets passed the
last coeff anyway. (it was named width, what the hell?)
This is to prepare for changing the code using the floating point table
to use the fixed point table instead.
This also allows reducing the size of the fractional part, which was
useful for finding every place where the the fixed point presentation
is relied upon.
Changes luma deblocking to use gather and scatter instead of reading
to and writing from here and there in memory. Should make them
faster and easier to vectorize, or at least cleaner.
Splits strong and weak luma deblocking to two functions, as they have
almost nothing in common.
Adds field lcu_stats to encoder_state_config_frame_t. The following data
is recorded for each LCU:
- number of bits
- squared cost
- used lambda value
- alpha parameter used for rate control
- beta parameter used for rate control
When rate control is enabled, enable cu_qp_delta_enabled_flag in PPS
with diff_cu_qp_delta_depth set to 0. Also adds code for writing the QP
deltas and a new cabac context.
Adds fields lambda, lambda_sqrt and qp to encoder_state_t. Drops field
cur_lambda_cost_sqrt from encoder_state_config_frame_t and renames
cur_lambda_cost to lambda.
- Defines MIN_LAMBDA and MAX_LAMBDA constants.
- Moves resetting state->frame->cur_gop_bits_coded to rate_control.c.
- Changes gop_allocate_bits to return the number of bits allocated like
pic_allocate_bits does.
When --threads=auto was given on the command line, cfg->threads was
actually set to zero, disabling threads altogether. Fixed to set
cfg->threads to -1, so that the number of threads is chosen
automatically.
The CABAC engine only writes to the bitstream when it has a full byte.
These writes are also always byte-aligned, so there is no need to even
check for stream alignment.
Speedup was around 3% with ultrafast and low QP.
Enforce bit depth promised by --input-bitdepth to avoid crashes when
larger values are provided.
Do endianess byte swap for all bytes when the buffer gets extended
to multiple of 8 pixels, and not just the number of input pixels.
Don't swap bytes on a little-endian system.
- Reduce indentation to 6 spaces
- Word wrap everything to under 80 characters
- Remove defaults from options covered by presets
- Add a dash in front of argument descriptions
- Add --(no-) to names of parameters that accept it and remove mention
of enabling or disabling
- Add executable and scripts as a dependancy to make docs
This value is not represented in the HEVC bitstream, which is why it
was not set previously. FFmpeg sets and needs it however, so make the
CLI set it as well to make sure we handle it correctly.
The rd-complexity of slow presets is better with a less agressive GOP.
Adding the GOP as part of the preset improved BDRate enough, that it
didn't make sense anymore to have a veryslow target the best BDRate.
Instead, push that responsibility to placebo by making it a little bit
faster.
GOPs with depth 1 had the same structure as those with depth 2:
g4d3t1 = 3 2 3 1
g4d2t1 = 2 2 2 1
g4d1t1 = 2 2 2 1
It now results in the correct:
g4d1t1 = 1 1 1 1
Coding inter without GOP of any kind really isn't a very sensible
default. Defaulting to B-GOP of some kind would be more better,
but lp-gop is more robust for now.
Use the vectorized general SSE41 inter SAD in AVX reg_sad for shapes
for which we don't have AVX versions yet.
Also improves speed of --smp and --amp a lot. Got a 1.25x speedup for:
--preset=ultrafast -q 27 --gop=lp-g4d3r3t1 --me-early-termination=on --rd=1 --pu-depth-inter=1-3 --smp --amp
* Suite speed_tests:
-PASS inter_sad: 0.898M x reg_sad(64x63):x86_asm_avx (1000 ticks, 1.000 sec)
+PASS inter_sad: 2.503M x reg_sad(64x63):x86_asm_avx (1000 ticks, 1.000 sec)
-PASS inter_sad: 115.054M x reg_sad(1x1):x86_asm_avx (1000 ticks, 1.000 sec)
+PASS inter_sad: 133.577M x reg_sad(1x1):x86_asm_avx (1000 ticks, 1.000 sec)
Add implementations for these functions that process the image line by
line instead of using the 16x16 function to process block by block.
The 32x32 is around 30% faster, and 64x64 is around 15% faster,
on Haswell.
PASS inter_sad: 28.744M x reg_sad(32x32):x86_asm_avx (1014 ticks, 1.014 sec)
PASS inter_sad: 7.882M x reg_sad(64x64):x86_asm_avx (1014 ticks, 1.014 sec)
to
PASS inter_sad: 37.828M x reg_sad(32x32):x86_asm_avx (1014 ticks, 1.014 sec)
PASS inter_sad: 9.081M x reg_sad(64x64):x86_asm_avx (1014 ticks, 1.014 sec)
Arrange the decision tree such that there is only 3 branches on the
most common paths and the more likely branch is always fall-through.
A profile guided optimization pass would probably do something similar.
A lot of time is being taken up by this function on ultrafast, and it
doesn't do a very good job. This change aims to both simplify the
logic and make the estimate better.
The logic is simplified by using a look up for the step mvd bit cost
step function instead of mimicking the binarization process. The
estimation is made better by checking fractional cabac bit costs.
The new function returns the same results as
kvz_get_mvd_coding_cost_cabac, but is also faster than the old
function.
Write bitstream without chroma when encoding with --input-format=P400.
This reduces bitstream size by 0-1 %, compared to coding monochrome in
420 format, and speeds up encoding slightly due to not processing
chroma.
Changes encoder_set_source_picture to set the reconstructed picture to
a copy of the source picture instead of allocating a new picture when
lossless coding is used.
- Moves allocation of the reconstructed picture after the source picture
is set.
- Extracts main state initialization to a separate function from
encoder_state_new_frame.
- Changes kvz_encoder_feed_frame to return the frame.
- Renames some functions to better match their purpose.
When --lossless is given, set cu_transquant_bypass_flag for every CU and
bypass transform and quantization by directly copying reference pixels
to reconstruction and the residual to coefficients.
When a list does not have space for the new element, its size is
doubled. If the size of the list is zero, it would not be resized. Fixed
to always resize the list so that the new element can be added.
Enables search for 2NxN and Nx2N partition modes for 8x8 CUs and 2NxnU,
2NxnD, nLx2N and nRx2N partition modes for 16x16 CUs.
Changes the loop for copying reconstructed luma pixels in
kvz_inter_recon_lcu to use 4 byte chunks instead of 8 byte chunks since
it is now possible to have 4 pixel wide blocks.
The first frame was always qp51 due to gop_offset being -1 for the
first frame. This fix makes it so that bits are allocated as if it was
the last (high quality) frame from the previous GOP.
When using ratecontrol with lowdelay-P, this improves BDRate by 1-25%.
Strongest effect is when using 4 layers and multiple references.
Also allow using 1 or 2 layers with ratecontrol.
This problem resulted in an illegal bitstream with --gop=lp, because it
uses IDR's. The --gop=8 would not code IDR pictures, even when told to
with -p, which masked this problem.
This fix solves the problem with --gop=lp and also prevents references
across the intra picture in --gop=8. The intra pictures should be set
to IDR in a later fix, or an alternate method of differentiating
between IDR and non-IDR intra should be made.
The includes should make more sense now and not just happen to compile
due to headers included from other headers.
Used a modified version of IWYU. Modifications were to attribute int8_t
and so on to stdint.h instead of sys/types.h and immintrin.h instead of
more specific headers.
include-what-you-use 0.7 (git:b70df35)
based on clang version 3.9.0 (trunk 264728)
While these are only used for strategies, it's non-intuitive to have
to include strategyselector.h in every file under strategies before
including anything else.
The main thread has to wait for the worker threads to finish. The
pthread_cond_timedwait call used to accomplish this was given
a relative instead of absolute time, which resulted in the call
returning immediately, because the time had already passed.
This removes the now unnecessary sleeps and fixes the time given to
the pthread_cond_timedwait such that it now waits until a job finishes
or 100ms have passed.
The OWF wpp limit code assumed square blocks, and as such did not work
correctly when height != width. This changes the relevant code to consider
both height and width.
Add md5 through extras/libmd5 taken from HM with BSD license. It's
implemented as a generic strategy using the same interface as checksum,
so we can write a SIMD version if it seems necessary.
The previous reasoning used deblocking and fractional motion estimation
together to arrive at a margin of 4 pixels. This was wrong, and with
either of these off, half pixel chroma interpolation could use pixels
outside the intended region.
Deblocking does not currently affect the margin needed.
I was a bit unclear about exactly what happens and when regarding SAO
and deblocking when we do frame-parallel WPP parallelism, so I checked
and commented the bits that were unclear to me.
The check was done in regard to the wrong dimension, allowing the
access to unfinished parts of the frame when coding multiple frames
at the same time.
Add new parameter --tiles that accept only uniform split. I considered
supporting the syntax of --tiles-width-split for this, but writing
--tiles=u2xu2 is just not as intuitive as --tiles=2x2, and there is
hardly ever any reason to use anything but uniform split. The more
cumbersome --tiles-width-split and --tiles-height-split parameters
are still there to allow finer control.
There was an off by one error in the dependance setting code, which
resulted in dependencies not being set resulting in checksum errors.
For example if ref_neg=1 and owf=1.
Earlier fix that fixed the supply side of the cu_array to take tile
coordinates into account should have been accompanied with this one
that does the same thing to demand side.
Moves sao search from function encoder_state_worker_encode_lcu in
encoderstate.c to function kvz_sao_search_lcu in sao.c. Makes functions
kvz_init_sao_info, kvz_sao_search_chroma and kvz_sao_search_luma static
since they are no longer used outside sao.c.
If 0,0 vector is illegal, it's possible that no legal movement vector,
is found, in which case a large cost is returned instead. The cost
overflowed and there is all sorts of silliness with converting from
double to int, but I'm not going to fix all of it because when we
remove the doubles it will all get fixed.
An incorrect frame boundary check caused a checksum error, because the
chroma reconstruction of the encoder was wrong. The encoder treated
horizontal tile boundaries as frame boundaries when the vertical
component of the movement vector was a multiple of 8.
CU data was being copied to the wrong place in the reference frames
cu_array, which led to uninitialized data being used as a starting
point for motion vector search.
Fixes#99.
A 32 bit int overflowed after 2^31 bits (2Gb). It will still overflow
eventually, after 500 years of outputting 1Gb/s, but by that time,
I recon we will have fixed this properly and it's time to upgrade.
Changes communication between the input thread and main thread in
encmain.c so that only one of them uses img_in and retval at a time.
Fixes a race condition which would sometimes result in a deadlock.
Add dependency to the reference frame instead of the previous frame,
in order to allow more frames to be encoded in parallel when temporal
stepping >1 in LP-gop (such as --gop=lp-g8d4r1t2).
This moves the interlacing from CLI code to api->encoder_encode, in
order to make it possible to use field coding through the lib API.
The field order is now determined per frame, as FFmpeg gives it per
frame and it's signaled per frame.
As a side effect, the CLI also now prints info from frames instead of
fields. While we might want to extend the API in the future to allow
printing of more detailed information about fields, for now it's
more important that the CLI uses the real lib API.
PSNR calculation for interlaced frames disabled until we have a way to
avoid deinterlacing the frame when it's not necessary.
Prevents a conflict with config.h and src/config.h so that the config.h
generated by configure is included in global.h. Fixes problems with
large input files on 32-bit systems.
- Handle input processing in a separate thread to allow main thread more time with thread handling etc
- Significant speedup can be seen when run on ultrafast settings and on a system with great number of cores
Option -mavx2 was omitted when compiling AVX2 strategies. This commit
moves strategies to convenience libraries so that their compilation
flags can be easily set and adds -mavx2 to CFLAGS of the AVX2 library.
We have soname versioning now, so we should focus on getting that right
instead. This also serves as an example of correctly incrementing the
lib-version.
Now that we put the timing info into the bitstream, the time base must
be precisely known. Represent framerate as a fraction and add timing
info only if the old floating point framerate was not used.
Deprecate cfg->framerate so it can be removed once we get patches to
FFmpeg and libav.
Add support for (num)/(denom) format to --input-fps.
Also moves CLI stuff under CLI project, so they are compiled as their
own lib just like when the Makefile is used.
The file interface_main.c was an artifact from a bygone era and should have
been deleted long ago.
Add module information to all header files.
Update all header file documentations to briefly say what they are, and
to use the javadoc format so the brief actually gets included into the
doxygen documentation.
Remove \file from implementation files, in order to not repeat the info
from the header files.
Add files under strategies and tools to Doxygen and update the Doxygen
settings to be just plain better.
Make README be the main page of Doxygen documentation.
Bits were being added to rate distortion without being multiplied by
lambda in a few places. Fixing this bug also finally allows us to remove
the magic bits from the Coding Unit split decision.
I tried to find new optimum value for CU_COST and it turned out to be 2
for veryslow and 0 for superfast. The difference between 0 and 2 on
veryslow was only 0.1% however, so I don't think this parameter is
needed any longer. Before this fix the effect of removing CU_COST would
have been 0.8%.
- Updates function search_mv_full so that it compiles and handles
non-square blocks.
- Enables compilation of search_mv_full.
- Sets full search radius to 32.
- Enables selecting full mv search with "--me full".
Changes search_frac and kvz_search_cu_iter to use kvz_satd_any_size for
computing the SATDs instead of getting the SATD function with
kvz_pixels_get_satd_func.
Adds strategy satd_any_size for generic and AVX2. The satd_any_size
functions are implemented with macro SATD_ANY_SIZE defined in
strategies-picture.h.
- Adds function is_pu_boundary.
- Moves code for filtering an edge of a single PU or TU to a new
function filter_deblock_unit.
- Replaces recursive CU tree traversal in filter_deblock_cu with
a simple loop and renames it to filter_deblock_lcu_inside.
Adds parameter block_height to functions inter_recon_frac_luma,
inter_recon_14bit_frac_luma and inter_recon_14bit_frac_chroma so that
they can handle SMP blocks.
Makes the following functions static since they are not used outside
inter.c:
- kvz_inter_recon_frac_luma
- kvz_inter_recon_14bit_frac_luma
- kvz_inter_recon_frac_chroma
- kvz_inter_recon_14bit_frac_chroma
The buffers allocated in functions kvz_get_extended_block_avx2 and
kvz_get_extended_block_generic were too small when the width of the
block was less than its height. Fixed to allocate correctly sized
buffers.
Adds parameter height to functions kvz_inter_recon_lcu and
kvz_inter_recon_lcu_bipred and makes them work on non-square sizes.
Fractional reconstruction functions do not handle non-square blocks yet.
- Moves SIZE_* definitions to cu.h.
- Adds constant arrays kvz_part_mode_num_parts, kvz_part_mode_offsets
and kvz_part_mode_sizes for storing the number of PUs, PU offsets and
PU sizes.
- Adds macros PU_GET_X, PU_GET_Y, PU_GET_W and PU_GET_H for getting the
location and size of a PU.
Moves checks for motion vector prediction and merge candidate block
types (inter/intra) from functions kvz_inter_get_mv_cand and
kvz_inter_get_merge_cand to kvz_inter_get_spatial_merge_candidates.
Remove the need to count the coefficients by populating the significant
coefficient group map first and finding the last coefficient from the
last group afterward. The speedup is about 2% on ultrafast.
The previous version of this patch was reverted due to a bug, which
has now been fixed.
This reverts commit 25462124f8.
That commit broke the bitstream. If it's not good enough to push on Friday
night, it's probably not good enough on Monday morning either.
Remove the need to count the coefficients by populating the significant
coefficient group map first and finding the last coefficient from the
last group afterward.
Increases the MV safety margin of OWF from 2 to 3 when deblocking
is used and 4 when both deblocking and FME are used.
Fractional pixel motion estimation can move the vector one more pixel
down causing checksum error. This fixes that error by increasing the
OWF safety margin and changes the interface, so that different margin
can be used when FME or deblocking are not in use.
Replaces repetitive calls to kvz_filter_deblock_luma and
kvz_filter_deblock_chroma with loops in functions
filter_deblock_edge_luma and filter_deblock_edge_chroma.
Marks the following functions static and removes them from filter.h
since they are not used outside the filter module.
- kvz_filter_deblock_luma
- kvz_filter_deblock_chroma
- kvz_filter_deblock_edge_luma
- kvz_filter_deblock_edge_chroma
- kvz_filter_deblock_cu
- Replace parameter depth in kvz_filter_deblock_edge_{luma,chroma} with
length.
- Move checking whether an edge needs to be filtered from functions
kvz_filter_deblock_edge_{luma,chroma} to functions
kvz_filter_deblock_{cu,lcu}.
- Use pixel coordinates instead of 8-pixel block coordinates in
kvz_filter_deblock_cu.
- Add comments.
These changes should make it easier to modify the deblocking filter to
handle SMP and AMP blocks.
The mac version of KVZ_GET_TIME macro has many statements, which
prevented it being used inside a for loop statement. Added brackets
to all versions to prevent this issue arising in the future.
Fixes#115.
Giving a single number as an argument to --deblock option would enable
deblocking and set both beta and tc to that value. This commit changes
a single number argument to be interpreted as a boolean specifying
whether to enable deblocking or not. As a result, "--deblock 0" can be
now used to disable deblocking.
This fixes deblocking being enabled in all presets.
Just along side for now to help with debugging.
The main difference with the new versions is that they take and output
width**2 blocks and two width*2+1 arrays of reference samples,
instead of the (2*width+8)**2 blocks the old ones do. This should make
the interface clearer and the memory footprint smaller.
Also commented the shit out of angular prediction, so hopefully Ari L.
will have an easier time with a SIMD implementation.
Changes main function to compute frame PSNR by calling
kvz_videoframe_compute_psnr directly with the source and reconstructed
pictures returned from encoder_encode.
The code for building the reference picture lists was duplicated in
functions encoder_state_ref_sort and print_frame_info. This commit moves
it to a new function kvz_encoder_get_ref_lists. Also makes
encoder_ref_insertion_sort static since it is not used outside the
encoderstate module any more.
Adds a new output parameter info_out to encoder_encode. It returns
a struct containing information about the encoded frame, including POC,
QP and slice type.
This bug caused a single tiles worth of lcu_info_t structs to be copied
unnecessarily for every LCU in the frame. This obviously caused huge
memory bandwidth issues when coding large frames without tiles. The
effect was minimized somewhat with a large number of tiles, because
only the current tile was copied.
From context it is clear that this piece of code was supposed to copy
a single tile or frame, once the frame was done, but because it was
placed in a function which is called for every LCU, it copied the data
for the LCU, but also lots of extra stuff.
The fix is to copy only the current LCU instead of the whole tile.
A call to kvz_threadqueue_waitfor caused the tqj_bitstream_written field
of the previous encoder state to become a dangling pointer, subsequently
causing an assertion to fail. This would only occur when the encoder
state used for a new frame was not the last finished one.
Fixed by setting tqj_bitstream_written to NULL after the job is done and
removing unnecessary calls to kvz_threadqueue_waitfor.
Changes the Makefile to use the static library and additional object
files as linker input instead of all the object files when linking the
command line program.
Now outputs a pointer to a block with guaranteed padding for filtering.
Only generate extra pixels if samples are needed out of bounds.
Use memcpy otherwise.
- Added 64x64 version for completeness.
- With the exception of 16x16, these were all slightly slower than the ASM
versions, as measured by "kvazaar_test -s speed -t intra_sad", but now they
are on par or slightly faster.
- None of these actually use any AVX2 intrinsics, and probably never will,
unless someone adds an interface for doing more than one block at a time,
in which case the non-destructive versions might come in handy.
- Having more than one rule in a pattern rule means that both of those files
are created at the same time with the rule. This only worked for debug,
because debug build was never done in the same invocation as release build.
- Always use the compiler to invoke the linker. Clang will give additional
parameters to the linker when compiled with -flto.
- Giving a different optimization level to linker did not make any difference
in gcc-5.1.1.
- It's required for .so and .dylib, but not for .dll or the executable.
- It might be better to use libtool for this, but I'm not ready to go that
far yet.
