error_per_bit and sad_per_bit were designed as estimates of a bit worth
of sum squared error and sum absolute difference respectively. Under
this assumption, error_per_bit should be used in combination with 2nd
order errors (variance or sum squared error) while sad_per_bit should
be used in combination with 1st order SADs in motion estimation. There
were a few places where sad_per_bit has been misused with variances,
this commit changes to use error_per_bit for those places, also changes
parameter names to properly indicate which constant is being used.
On cif set, the change has a universal gain by all metrics: 0.13% by
average/overall psnr and 0.1% by ssim.
Change-Id: I4850fdcc3fd6886b30f784bd843f13dd401215fb
While profile=3, there is no sub-pixel search. Distortion and SSE
have to calculated using get_inter_mbpred_error().
Change-Id: Ifb36e17eef7750af93efa7d0e2870142ef540184
Declared the bmi in MODE_INFO as a union instead of B_MODE_INFO.
This reduced the memory footprint by 518,400 bytes for 1080
resolutions. The decoder performance improved by ~4% for the
clip used and the encoder showed very small improvements. (0.5%)
This reduction was first mentioned to me by John K. and in a
later discussion by Yaowu.
This is WIP.
Change-Id: I8e175fdbc46d28c35277302a04bee4540efc8d29
fixed a bug where active_worst_quality could be set
below active_best_quality which could result in an
infinite loop.
Change-Id: I93c229c3bc5bff2a82b4c33f41f8acf4dd194039
This patch collects the twopass specific memebers of VP8_COMP into a
dedicated struct. This is a first step towards isolating the two pass
rate control and aids readability by decorating these variables with
the 'twopass.' namespace. This makes it clear to the reader in what
contexts the variable will be valid, and is a hint that a section of
code might be a good candidate to move to firstpass.c in later
refactoring. There likely will be other rate control modes that need
their own specific data as well.
This notation is probably overly verbose in firstpass.c, so an
alternative would be to access this struct through a pointer like
'rc->' instead of 'cpi->firstpass.' in that file. Feel free to make
a review comment to that effect if you prefer.
Change-Id: I0ab8254647cb4b493a77c16b5d236d0d4a94ca4d
The partition_info struct contains info just for SPLITMV,
so it should be used instead of BLOCKD. Eventually, I want
to reduce the size of B_MODE_INFO struct found in BLOCKD, so
this is the first step toward that goal.
Also, since SPLITMV is not supported in vp8_pick_inter_mode(),
the unnecessary mem copies and checks were removed. For rt
encodes, this gave a slight performance improvement.
Change-Id: I5585c98fa9d5acbde1c7e0f452a01d9ecc080574
This commit restructures the mb activity masking code
to better facilitate experimentation using different metrics
etc. and also allows for adjustment of the zero bin either
for encode only or both the encode and mode selection
stages
It also uses information from the current frame rather than
the previous frame and the default strength has been
reduced.
Change-Id: Id39b19eace37574dc429f25aae810c203709629b
This patch improves the accuracy of frame rate estimation by using a
larger, 1 second window. It also more quickly adapts to step changes
in the input frame rate (ie 30fps to 15fps)
Change-Id: I39e48a8f5ac880b4c4b2ebd81049259b81a0218e
The compiler produces better assembly when using int_mv
for assignments. The compiler shifts and ors the two 16bit
values when assigning MV.
Change-Id: I52ce4bc2bfbfaf3f1151204b2f21e1e0654f960f
Further modification and wrong implementation fix which caused
refining_search and refining_searchx4 result mismatching.
Change-Id: I80cb3a44bf5824413fd50c972e383eebb75f9b6f
This is to reflect the RD improvement in the encoder. The change has a
small positive impact on quality (0.25% by VPXSSIM and 0.05% by PSNR)
Change-Id: Ic66ffc19b10870645088c0624c85556f009fd210
The variable is introduced in commit 2e53e9e53 to make more use of
trellis quantization, but this is no longer necessary after RDMULT
was made adaptive in a number of later commits.
Change-Id: I7420522ec7723f38cf77033466c25afb405d52ae
In NEWMV mode, currently, full search is used as the refining search
after n-step search. By replacing it with an iterative diamond search
of radius 1 largely reduced the computation complexity, but still
maintained the same encoding quality since the refining search is
done for every macroblock instead of only a small precentage of
macroblocks while using full search.
Tests on the test set showed a 3.4% encoding speed increase with none
psnr & ssim loss.
Change-Id: Ife907d7eb9544d15c34f17dc6e4cfd97cb743d41
Paul pointed out that the pointer to the gf_active_flags is not being
properly incremented in multithreaded encoder. This commit fixes the
issue by making sure the gf_active_ptr points to the starting of next
group of mb rows.
Change-Id: I3246e657d23beabb614dfb880733a68a5fd7e34c
Commit db5057c introduced a bug in that the active_worst_quality
selected by the 2 pass rate controller was being overridden for key
frames, causing a severe quality loss.
Change-Id: I4865a6fbe3e94e9b4fb9271c7dd68b455d7b371d
vp8_fast_quantize_b_neon function updated and further optimized.
- match current C implementation of fast quantizer
- updated to use asm_enc_offsets for structure members
- updated ads2gas scripts to handle alignment issues
Change-Id: I5cbad9c460ad8ddb35d2970a8684cc620711c56d
The existing emulation of posix semaphores on Windows uses SetEvent()
and WaitForSingleObject(), which implements a binary semaphore, not a
counting semaphore as implemented by posix. This causes deadlock when
used with the expected posix semantics. Instead, this patch uses the
CreateSemaphore() and ReleaseSemaphore() calls (introduced in Windows
2000) which have the expected behavior.
This patch also reverts commit eb16f00, which split a semaphore that
was being used with counting semantics into two binary semaphores.
That commit is unnecessary with corrected emulation.
Change-Id: If400771536a27af4b0c3a31aa4c4e9ced89ce6a0
This patch is to fix a rare hang in multi-thread encoder that was
only seen on Windows. Thanks for John's help in debugging the
problem. More test is needed.
Change-Id: Idb11c6d344c2082362a032b34c5a602a1eea62fc
Changed 8-neighbor searching to 4-neighour searching, and continued
searching until the center point is the best match.
Test on test set showed 1.3% encoding speed improvement as well as
0.1% PSNR and SSIM improvement at speed=-5 (rt mode).
Will continue to improve it.
Change-Id: If4993b1907dd742b906fd3f86fee77cc5932ee9a
The commit also removed the slow ssim calculation that uses a 7x7
kernel, and revised the comments to better describe how sample ssim
values are computed and averaged
Change-Id: I1d874073cddca00f3c997f4b9a9a3db0aa212276
Renamed configure option "enable-psnr" to "enable-internal-stats" to
better reflect the purpose of the option and eliminate the confusion
reported in http://code.google.com/p/webm/issues/detail?id=35
Change-Id: If72df6fdb9f1e33dab1329240ba4d8911d2f1f7a
Insertion sort performs better for sorting small arrays. In real-
time encoding (speed=-5), test on test set showed 1.7% performance
gain with 0% PSNR change in average.
Change-Id: Ie02eaa6fed662866a937299194c590d41b25bc3d
Combine calc_iframe_target_size, previously only used for forced
keyframes, with calc_auto_iframe_target_size, which handled most
keyframes.
Change-Id: I227051361cf46727caa5cd2b155752d2c9789364
This is a first step in cleaning up the redundancies between
vp8_calc_{auto_,}iframe_target_size. The pick_frame_size() function is
moved to ratectrl.c, and made to be the primary interface. This means
that the various calc_*_target_size functions can be made private.
Change-Id: I66a9a62a5f9c23c818015e03f92f3757bf3bb5c8
the decision to run the regular or simple loopfilter is made outside the
function and managed with pointers
stop tracking the option in two places. use filter_type exclusively
Change-Id: I39d7b5d1352885efc632c0a94aaf56b72cc2fe15
Rather than using a default size of 1/2 or 3/2 seconds for the first
frame, use a fraction of the initial buffer level to give the
application some control.
This will likely undergo further refinement as size limits on key
frames are currently under discussion on codec-devel@, but this gives
much better behavior for small buffer sizes as a starting point.
Change-Id: Ieba55b86517b81e51e6f0a9fe27aabba295acab0
vp8_adjust_key_frame_context() divides by
estimate_keyframe_frequency() which can
return 0 in the case where --kf-max-dist=0.
Change-Id: Idfc59653478a0073187cd2aa420e98a321103daa
The accumulator array is an integer array, so use paddd instead of paddw
to add values to it. Fixes overflows when using large --arnr-maxframes
(>8) values.
Change-Id: Iad83794caa02400a65f3ab5760f2517e082d66ae
The arguments to these fprintfs are int not long int so
the format specifier should be "%d" and not "%ld". This
was writing garbage in the linux build.
Change-Id: I3d2aa8a448d52e6dc08858d825bf394929b47cf3
add an sse4 quantizer so we can use pinsrw/pextrw and keep values in xmm
registers instead of proxying through the stack. and as long as we're
bumping up, use some ssse3 instructions in the EOB detection (see ssse3
fast quantizer)
pick up about a percent on 32bit and about two on 64bit.
Change-Id: If15abba0e8b037a1d231c0edf33501545c9d9363
The dc_diff flag is used to skip loopfiltering. Instead
of setting this flag in the decoder/encoder, we now check
for this condition in the loopfilter.
Change-Id: Ie2b9cdf9e0f4e8b932bbd36e0878c05bffd28931
Code cleanup. The build inter predictor functions are
redundantly checking the mode_info_context for either
INTRA_FRAME or SPLITMV.
Change-Id: I4d58c3a5192a4c2cec5c24ab1caf608bf13aebfb
Golden and ALT reference buffers were refreshed by copying from
the new buffer. Replaced this by index manipulation.
Also moved all the reference frame updates to one function for
easier tracking.
Change-Id: Icd3e534e7e2c8c5567168d222e6a64a96aae24a1
Remove tot_key_frame_bits and prior_key_frame_size[] as they were
tracked but never used. Remove intra_frame_target, as it was only
used to initialize prior_key_frame_size.
Refactor vp8_adjust_key_frame_context() some to remove unnecessary
calculations.
Change-Id: Icbc2c83d2b90e184be03e6f9679e678f3a4bce8f
the win64 abi requires saving and restoring xmm6:xmm15. currently
SAVE_XMM and RESTORE XMM only allow for saving xmm6:xmm7. allow
specifying the highest register used and if the stack is unaligned.
Change-Id: Ica5699622ffe3346d3a486f48eef0206c51cf867
Went through the code and fixed it. Verified on Windows.
Where possible, remove dependencies on xmm[67]
Current code relies on pushing rbp to the stack to get 16 byte
alignment. This broke when rbp wasn't pushed
(vp8/encoder/x86/sad_sse3.asm). Work around this by using unaligned
memory accesses. Revisit this and the offsets in
vp8/encoder/x86/sad_sse3.asm in another change to SAVE_XMM.
Change-Id: I5f940994d3ebfd977c3d68446cef20fd78b07877
Passed SSE from sub-pixel search back to pick_inter_mode
function, which is compared with the encode_breakout to
see if we could skip evaluating the remaining modes.
Change-Id: I4a86442834f0d1b880a19e21ea52d17d505f941d
In vp8_pick_inter_mode(), for NEWMV mode, use the error result got
from motion search as distortion. This helps performance in real-
time mode.
Change-Id: I398c4e46cc5381f7d874e748cf78827ef0e0860c
The value of distortion2 returned by vp8_pick_intra4x4mby_modes
was being overwritten by the value returned by get16x16prederror
before it was tested.
Change-Id: If00e80332b272c5545c3a7e381c8041e8319b41a
Index i is used to detect early breakout from the first loop, but
its value is lost due to reuse in the second for loop. I moved
the position of the second loop and did some format cleanup.
Change-Id: I02780eae1bd89df4b6c000fb8a018b0837aac2e5
This patch cleans up the source buffer storage and copy mechanism to
allow access through a standard push/pop/peek interface. This approach
also avoids an extra copy in the case where the source is not a
multiple of 16, fixing issue #102.
Change-Id: I05808c39f5743625cb4c7af54cc841b9b10fdbd9
in encodframe.c, quant_shift is set to 0 or 1 in vp8cx_invert_quant
only use 8 bits to store this, instead of 16. will allow saving an
xmm register in an updated version of the regular quantize
Change-Id: Ie88c47fe2aff5af0283dab1147fb2791e4b12f90
This patch changes the rc_undershoot_pct and rc_overshoot_pct controls
to set the "aggressiveness" of rate adaptation, by limiting the
amount of difference between the target buffer level and the actual
buffer level which is applied to the target frame rate for this frame.
This patch was initially provided by arosenberg at logitech.com as
an attachment to issue #270. It was modified to separate these controls
from the other unrelated modifications in that patch, as well as to
use the pre-existing variables rather than introducing new ones.
Change-Id: Id542e3f5667dd92d857d5eabf29878f2fd730a62
Previous to commit de4e9e3, there was an early return in the alt-ref
case that was inadvertantly removed when the function was refactored
to return void. This patch restores the prior behavior.
Change-Id: I783ffd594a4690297e2742f99526fd7ad67698b2
The error accumulator stats values cpi->prediction_error and
cpi->intra_error were being populated with rd values not
distortion values.
These are only "currently" used in a limited way for RT compress
key frame detection.
Change-Id: I2702ba1cab6e49ab8dc096ba75b6b34ab3573021
This commit fixed an overflow in ssim calculation, added register
save and restore to make sure assembly code working for x64 platform.
It also changed the sampling points to every 4x4 instead of 8x8 and
adjusted the constants in SSIM calculation to match the scale of
previous VPXSSIM.
Change-Id: Ia4dbb8c69eac55812f4662c88ab4653b6720537b
on the same order as the sse2 fast quantize change: ~2%
except for 32bit. only a slight improvment there.
Change-Id: Iff80e5f1ce7e646eebfdc8871405458ff911986b
MV sad cost error is only used in full-pixel motion search,
which only need full-pixel resolution instead of quarter-pixel
resolution. This change reduced mvsadcost table size, and
removed unneccessary pamameter passing since this table is
constant once it is generated.
Change-Id: I9f931e55f6abc3c99011321f1dfb2f3562e6f6b0
rather than look up rc in the zig zag table, embed it in the macro. this
also allows us to shuffle some values in the macro and keep *d in rsi
gains of about the same order as the obj_int_extract implementation: ~2%
Change-Id: Ib7252dd10eee66e0af8b0e567426122781dc053d
Address calculations moved from encodemb_arm.c file to neon
optimized assembly function to save cycles in function calls.
- vp8_subtract_b_neon_func replaced with vp8_subtract_b_neon
that contains all needed address calculations
- unnecessary file encodemb_arm.c removed
- consistent with ARMv6 optimized version
Change-Id: I6cbc1a2670b56c2077f59995fcf8f70786b4990b
Detect the number of available cores and limit the thread allocation
accordingly. On decoder side limit the number of threads to the max
number of token partition.
Core detetction works on Windows and
Posix platforms, which define _SC_NPROCESSORS_ONLN or _SC_NPROC_ONLN.
Change-Id: I76cbe37c18d3b8035e508b7a1795577674efc078
Adds following ARMv6 optimized functions to encoder:
- vp8_subtract_b_armv6
- vp8_subtract_mby_armv6
- vp8_subtract_mbuv_armv6
Gives 1-5% speed-up depending on input sequence and encoding
parameters. Functions have one stall cycle inside the loop body
on Cortex pipeline.
Change-Id: I19cca5408b9861b96f378e818eefeb3855238639
Half pixel interpolations optimized in variance calculations. Separate
function calls to vp8_filter_block2d_bil_x_pass_armv6 are avoided.On
average, performance improvement is 6-7% for VGA@30fps sequences.
Change-Id: Idb5f118a9d51548e824719d2cfe5be0fa6996628
remove helper function and avoid shadowing all the arguments to the
stack on 64bit systems
when running with --good --cpu-used=0:
~2% on linux x86 and x86_64
~2% on win32 x86 msys and visual studio
more on darwin10 x86_64
significantly more on
x86_64-win64-vs9
Change-Id: Ib7be12edf511fbf2922f191afd5b33b19a0c4ae6
This declaration did not match the prototype_sad() prototype, but was
unused in this translation unit, so it is removed instead. Fixes
issue 290.
Change-Id: I168854f88a85f73ca9aaf61d1e5dc0f43fc3fdb3
Optimized fdct4x4 (8x4) for ARMv6 instruction set.
- No interlocks in Cortex-A8 pipeline
- One interlock cycle in ARM11 pipeline
- About 2.16 times faster than current C-code compiled with -O3
Change-Id: I60484ecd144365da45bb68a960d30196b59952b8
Thread synchronization was not correct when frame width was 1 MB.
Number of allocated encoding threads is limited by the sync_range.
There is no point having more because each thread lags sync_range MBs
behind the thread processing the row above.
http://code.google.com/p/webm/issues/detail?id=302
Change-Id: Icaf67a883beecc5ebf2f11e9be47b6997fdf6f26
A large number of functions were defined with external linkage, even
though they were only used from within one file. This patch changes
their linkage to static and removes the vp8_ prefix from their names,
which should make it more obvious to the reader that the function is
contained within the current translation unit. Functions that were
not referenced were removed.
These symbols were identified by:
$ nm -A libvpx.a | sort -k3 | uniq -c -f2 | grep ' [A-Z] ' \
| sort | grep '^ *1 '
Change-Id: I59609f58ab65312012c047036ae1e0634f795779
The vp8_build_intra_predictors_mby and vp8_build_intra_predictors_mby_s
functions had global function pointers rather than using the RTCD
framework. This can show up as a potential data race with tools such as
helgrind. See https://bugzilla.mozilla.org/show_bug.cgi?id=640935
for an example.
Change-Id: I29c407f828ac2bddfc039f852f138de5de888534
Clean up vp8_init_config() a bit and remove null pointer case,
as this code can't be called any more and is not an adequate
trap anyway, as a null pointer would cause exceptions before
hitting the test.
Change-Id: I937c00167cc039b3aa3f645f29c319d58ae8d3ee
Issue 291 highlighted the fact that CQ mode was not working
as expected in 1 pass mode,
This commit fixes that specific problem but in so doing I also
uncovered an overflow issue in the VBR code for 1 pass and
some data values not being correctly initialized.
For some clips (particularly short clips), the resulting
improvement is dramatic.
Change-Id: Ieefd6c6e4776eb8f1b0550dbfdfb72f86b33c960
In multithreaded mode the loopfilter is running in its own thread (filter level
calculation and frame filtering). Filtering is mostly done in parallel with the
bitstream packing. Before starting the packing the loopfilter level has
to be calculated. Also any needed reference frame copying is done in the
filter thread.
Currently the encoder will create n+1 threads, where n > 1 is the number of
threads specified by application and 1 is the extra filter thread. With n = 1
the encoder runs in single thread mode. There will never be more than n threads
running concurrently.
Change-Id: I4fb29b559a40275d6d3babb8727245c40fba931b
The firstpass motion map consists of an 8-bit flag for
each MB indicating how strongly the firstpass code
believes it should be filtered during the second pass
ARNR filtering.
For long or large format material the motion map can
become extremely large and hamper the operation of
the encoding process.
This change removes the motion map altogether, leaving
the second pass to rely on the magnitude of the motion
compensated error to determine the filter weight to
use for the MB during ARNR filtering.
Tests on the derf set indicate that the effect of this
change is neutral, with some small wins and losses. The
motion map has therefore been removed based on
a cost/benefit evaluation.
Change-Id: I53e07d236f5ce09a6f0c54e7c4ffbb490fb870f6
The previous calculation of macroblock count (w*h)/256
is not correct when the width/height are not multiples of
16. Use the precalculated macroblock count from
cpi->common instead. This manifested itself as a divide
by zero when the number of pixels was less than 256.
num_mbs updated in estimate_max_q, estimate_q,
estimate_kf_group_q, and estimate_cq
Change-Id: I92ff98587864c801b1ee5485cfead964673a9973
1. Process 16 pixels at one time instead of 8.
2. Add check for both xoffset =0 and yoffset=0, which happens
during motion search.
This change gave encoder 1%~3% performance gain.
Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e
GCC 4.5 and 4.6 both issue a warning about the multi-line format
string introduced in bc9c30a0, which also changed the whitespace
in the associated stt file by line-wrapping the long format string.
Instead, use multiple string constants, which the compiler will
concatenate. This maintains the original formatting, but remains
legible within the standard line length.
Change-Id: I27c9f92d46be82d408105a3a4091f145f677e00e
Disable zbin boost in SPLITMV mode as intended. Was incorrectly looking
at vp8_ref_frame_order instead of vp8_mode_order when comparing against
SPLITMV. This condition should have always been false, as SPLITMV is
not in the range of valid reference frames.
Change-Id: I0408cc7595eff68f00efef6d008e79f5b60d14bf
In some cases where clips have been encoded with
borders (eg. some wide-screen content where there is a
border top and bottom and slide shows containing portrait
format photographs (border left and right)) key frames were
not being correctly detected.
The new code looks to measure cases where a portion of
the image can be coded equally easily using intra or inter
modes and where the resulting error score is also very low.
These "neutral" areas are then discounted in the key frame
detection code.
Change-Id: I00c3e8230772b8213cdc08020e1990cf83b780d8
This code extends what was previously done for GFs, to pick
cases where insertion of a key frame after a fade (or other
transition or complex motion) followed by a still section, will
be beneficial and will reduce the number of forced key frames.
Change-Id: If8bca00457f0d5f83dc3318a587f61c17d90f135
When the modified_error_left accumulator exceeds INT_MAX, an incorrect
cast to int resulted in a negative value, causing the rate control to
allocate no bits to that keyframe group, leading to severe undershoot
and subsequent poor quality.
This error was exposed by the recent change to the rolling target and
actual spend accumulators in commit 305be4e4 which fixed them to
actually calculate the average value rather than be re-initialized
on every frame to the average per-frame bitrate. When this bug was
triggered, the target bitrate could be 0, so the rolling target
becomes small, which causes the undershoot. The code prior to 305be4e4
did not exhibit this behavior because the rolling target was always
set to a reasonable value and was independent of the actual target
bitrate. With this patch, the actual target bitrate is calculated
correctly, and the rate control tracks as expected.
This cast was likely added to silence a compiler warning on a comparison
between a double (modified_error_left) and an int (0). Instead, this
patch removes the cast and changes the comparison to be against 0.0,
which should prevent the warning from reoccuring.
This fixes issue #289. Special thanks to gnafu for his efforts in
reporting and debugging this fix.
Change-Id: Ie5cc1a7b516c578a76c3a50c892a6f04a11621fe
This fixes an overflow problem in the frame error accumulators.
The overflow condition is extreme but did trigger when Frank B.
coded some high motion interlaced HD content.
The observed effect was a catastrophic breakdown of the rate
control leading to massive undershoot and poor bit allocation.
All the error values should really be unsigned but I will look at this
separately.
Change-Id: I9745f5c5ca2783620426b66b568b2088b579151f
Currently, when the video frame width is not multiples of 16, the
source buffer has a stride of non-multiples of 16, which forces
an unaligned load in SAD function and hurts the performance. To
avoid that, this change allocates source buffers to be multiples
of 16.
Change-Id: Ib7506e3eb2cea06657d56be5a899f38dfe3eeb39
In real-time mode, vp8_sad16x16 function is called heavily in
motion search part. Improvement of this function gives 1.2%
encoding performance gain (real-time mode, tulip clip).
Change-Id: I23c401fc40c061f732a9767e8d383737a179bd58
checks added to make sure that cpi->tplist
is freed correctly in vp8_dealloc_compressor_data
and vp8_alloc_compressor_data.
Change-Id: I66149dbbd25c958800ad94f4379d723191d9680d
Eliminated unnecessary calculations. Improved performance
by 10% on keyframes and 1.6% overall for the test clip used.
Change-Id: I87671b26af5e2cc439e81d0fee3b15c7cd2a3309
As mentioned in check-in "Improve motion search in real-time mode",
MV prediction calculation causes speed loss for speed 7 and above.
This change added a flag to turn off this calculation for speed>6
in real-time mode.
Change-Id: I9f4ae5a8bf449222d1784b54e7d315fc8347b2d1
Applied better MV prediction in real-time mode, which improves
the encoding quality.
Used quarter-pixel search instead of iterative sub-pixel search
for speed >=5 to improve encoding performance.
Tests on the test set showed:
1. For speed=-5, quality improvement: 1.7% on AvgPSNR and 2.1%
on SSIM, performance improvement: 3.6% (This counts in the
performance lose caused by MV prediction calculation in "Improve
MV prediction in vp8_pick_inter_mode() for speed>3").
2. For speed=-8, quality improvement: 2.1% on AvgPSNR and 2.5%
on SSIM. but, 6.9% performance decrease because of MV prediction
calculation. This should be improved later.
Change-Id: I349a96c452bd691081d8c8e3e54419e7f477bebd
Created a new speed 1 which is in the middle of the old
speed 0 and speed 1. (for both quality and performance)
Change-Id: I4802133cdb43f359ca787646c090899679dd5d84
Use 0xFFF0 vice 240 (0xF0) for determining whether the sometimes
implicit bit 3 will be transmitted. This is consistent with the decoder
and encode_mvcomponent().
Change-Id: Ic1304d0ab56844bed8236edd1c5243a6767fc6b1
it's difficult to mux the *_offsets.c files because of header conflicts.
make three instead, name them consistently and partititon the contents
to allow building them as required.
Change-Id: I8f9768c09279f934f44b6c5b0ec363f7943bb796
The encoder was not correctly catching transitions in the quantizer
deltas. If a delta_q was set, then the quantizer would be reinitialized
on every frame, but if they transitioned to 0, the quantizer would
not be reinitialized, leading to a encode-decode mismatch.
This bug was triggered by commit 999e155, which sets a Y2 delta Q
for very low base Q levels.
Change-Id: Ia6733464a55ee4ff2edbb82c0873980d345446f5
Whe auto keyframe insertion is enabled and conditions are right (scene change)
the encoder can decide to insert a key frame and does a re-encoding. This can
introduce extra latency. In RT mode we do not do the re-encoding of the current
frame but force the next frame to key frame.
Change-Id: I15c175fa845ac4c1a1f18bea3676e154669522a7
Reduce the number of sync points by letting each thread
continue imediatly with a new MB row.
Better multicore scaling, improves performance by 5-20% on ARM multicore.
Change-Id: Ic97e4d1c4886a842c85dd3539a93cb217188ed1b
from vp8cx_encode_intra_macro_block. prediction_error is used when
deciding if a frame should be a keyframe. After reviewing this with
Yaowu, it was pointed out that vp8cx_encode_intra_macro_block
is only called for keyframes, so the accumulation is unnecessary.
Change-Id: Id79dc81b80d4f5d124f3a0dba1b923887e2e1ec8
vp8_pick_intra4x4mby_modes uses the passed in distortion
for an early breakout. The best distortion was never saved
and the distortion for TM_PRED was always used.
Change-Id: Idbaf73027408a4bba26601713725191a5d7b325e
Previously, the DC check is to make sure there is no code-able
DC shift for quantizer Q0, which has been verified rather
conservative. This commit changes the criteria to have two
components, DC and AC, to address the conservativeness. First,
it checks if all AC energy is enough to contribute a single
non-zero quantized AC coefficient. Second, for DC, the decision
to skip further considers two possible scenarios: 1. There is
no code-able 2nd order DC coefficient at all; 2 The residue is
relatively flat, but the uniform DC change is very small, i.e.
less than 1/2 gray level per pixel.
Comparing to previous criteria, the new criteria is about 10%
to 15% faster in encoding time with a very small quality loss.
(threshold ~1000 and quality range 33db-45db)
It should be noted that this commit enables "automatic" static
threshold for encodebreakout if a non-zero small value is passed
in to encoder.
Change-Id: I0f77719a1ac2c2dfddbd950d84920df374515ce3
The condition for using RD when selecting the intra coding mode
for a MB is that the RD flag is set AND we're not in real-time
mode.
Previously the code used RD if either the RD flag was set OR
we were not using real-time mode.
Change-Id: Ic711151298468a3f99babad39ba8375f66d55a08
This function was using a variance metric compared to and SSE metric in
other places (eg. vp8_rd_inter_uv)
Change-Id: I9109fcc5a13bca9db1d7ead500fe14999ab233eb
Adds following targets to configure script to support RVCT compilation
without operating system support (for Profiler or bare metal images).
- armv5te-none-rvct
- armv6-none-rvct
- armv7-none-rvct
To strip OS specific parts from the code "os_support"-config was added
to script and CONFIG_OS_SUPPORT flag is used in the code to exclude OS
specific parts such as OS specific includes and function calls for
timers and threads etc. This was done to enable RVCT compilation for
profiling purposes or running the image on bare metal target with
Lauterbach.
Removed separate AREA directives for READONLY data in armv6 and neon
assembly files to fix the RVCT compilation. Otherwise
"ldr <reg>, =label" syntax would have been needed to prevent linker
errors. This syntax is not supported by older gnu assemblers.
Change-Id: I14f4c68529e8c27397502fbc3010a54e505ddb43
vp8/encoder/rdopt.c:728: warning: pointer targets in passing argument 3
of 'macro_block_yrd' differ in signedness
vp8/encoder/rdopt.c:541: note: expected 'int *' but argument is of type
'unsigned int *'
distortion is signed when calling macro_block_yrd is both other cases,
as well as for RDCOST
Change-Id: I5e22358b7da76a116f498793253aac8099cb3461
Change-Id: I6ca2d89f355839c4c770773c09fc69dcea7c1406
warning: implicit declaration of function
'vp8_variance_halfpixvar16x16_[h|v|hv]_neon'
'vp8_sub_pixel_variance16x16_neon_func'
Improved the performance of the first pass only
(~6% on 720p test clip) by making use of LUT instead of the
float calculations. Might try a SIMD version later.
Also started to make use of int_mv instead of
MV.
Change-Id: If2a217c7d6b59cd2c25c5553e0ca7e0502403af8
Use the function macro_block_yrd() to calculate error and distortion
in keeping with what is done for inter frames.
The old code was using a variance metric for once case and an
SSE function for measuring distortion in the other case.
The function vp8_encode_intra16x16mbyrd() is no longer used.
Change-Id: Ic228cb00a78ff637f4365b43f58fbe5a9273d36f
The code previously tested cpi->common.refresh_alt_ref_frame
but there are situations where this flag may be set for viewable frames.
The correct test should be !cm->show_frame.
Change-Id: Ia1a600622992a4a68fe1d38ac23bf6b34b133688
This commit also removes artificial RDMULT cap for low quantizers.
The intention is to address some abnormal behavior of mode selections
at the low quantizer end, where many macroblocks were coded with
SPLITMV with all partitions using same motion vector including (0,0).
This change improves the compression quality substantially for high
quality encodings in both PSNR and SSIM terms. Overall effect on
mid/low rate range is also positive for all metrics, but smaller
in magnitude.
Change-Id: I864b29c4bd9ff610d2545fa94a19cc7e80c02667
Commit 336aa0b7da incorrectly
declared current_pos as and int, when it should have been
a FIRSTPASS_STATS pointer.
Change-Id: I0a51c7a86ebba8546c95dd5d9d1c1143d4613e40
Adjust checking points in motion vector prediction to better cover
possible movements, and get a better prediction. Tests on test
clips showed a 0.1% improvement in SSIM, and no change in PSNR
and performance.
Change-Id: Ifdab05d35e10faea1445c61bb73debf888c9d2f8
The old 2 pass code estimated error distribution when coding a
forced (by interval) key frame. The result of this was that in some
cases, when allocating bits at the GF group level within a KF
group there was either a glut of bits or starvation of bits at the end
of the KF group.
Added code to rescan and get the correct data once the position of
a forced key frame has been determined.
Change-Id: I0c811675ef3f9e4109d14bd049d7641682ffcf11
vp8cx_mb_init_quantizer was being called for every mode checked
in vp8_rd_pick_inter_mode. zbin_extra is the only value that
really needs to be recalculated. This calculation is disabled
when using the fast quantizer for mode selection.
This gave a small performance boost (~.5% to 1%).
Note: This needs to be verified with segmentation_enabled.
Change-Id: I62716a870b3c82b4a998bdf95130ff0b02106f1e
In sub-pixel calculation, xoffset and yoffset mostly take some
specific values. Modified sub-pixel filter functions according to
these possible values to improve performance.
Change-Id: I83083570af8b00ff65093467914fbb97a4e9ea21
Added code to scan ahead a few frames when we see what
we think is a static scene in the two pass GF loop to see if the
conditions persist.
Moved calculation of decay rate out into a fuunction.
Change-Id: I6e9c67e01ec9f555144deafc8ae67ef25bffb449
These changes are specifically targeted at fade transitions to
static scenes. Here we want to place a GF/ARF immediately
after the fade and prevent an ARF just before the fade.
Also some code lines and comment lines shortened to 80 chars
while I was there.
Change-Id: Iefdc09a4fa7b265048fc017246b73e138693950f
In both vp8_find_next_key_frame and define_gf_group,
motion_pct was initialised at the top of the loop before
next_frame stats had been read in.
This fix sets motion_pct after next_frame stats have
been read.
Change-Id: I8c0bebf372ef8aa97b97fd35b42973d1d831ee73
Incorrect value loop_decay_rate used in GF loop.
The intent was to test the cumulative value decay_accumulator.
Change-Id: I62928c63eb09f4f6936a45ebd1c23784d1c9681b
In vp8_find_best_sub_pixel_step_iteratively(), many times xoffset
and yoffset are specific values - (4,0) (0,4) and (4,4). Modified
code to call simplified NEON version at these specific offsets to
help with the performance.
Change-Id: Iaf896a0f7aae4697bd36a49e182525dd1ef1ab4d
This code fixes a bug in the calculation of
the minimum Q for alt ref frames.
It also allows an extended gf/arf interval for sections
of clips that completely static (or nearly so).
Change-Id: I1a21aaa16d4f0578e5f99b13bebd78d59403c73b
Remove allocation/deallocation of stats storage.
Remove full search functions in machine specific encoder inits.
Remove last pass validation in validate_config.
Change-Id: I7f29be69273981a4fef6e80ecdb6217c68cbad4e
The CQ level was not using the q_trans[] array to convert
to a 0-127 range as per min and maxq
Experimental change to try and match the reconstruction
error for forced key frames approximately to that of the
previous frame by means of the recode loop. Though this
may cause extra recodes and the recode behavior has not
been optimized, it can only happen on forced key frames.
Change-Id: I1f7e42d526f1b1cb556dd461eff1a692bd1b5b2f
Previously when a frame was being overlaid on a previously coded
alt ref frame we only checked the alt ref 0,0 mode. Where there is
a possibility that the alt ref buffer is a filtered frame we should allow
the other prediction modes as normal or at the least allow use of
the last frame buffer.
Change-Id: I4d6227223d125c96b4f3066ec6ec9484fee7768c
In cases where the frame width is not a multiple of 16 the
ARNR filter would go wrong.
In vp8_temporal_filter_iterate_c when updating pointers
at the end of a row of MBs, the image size was
incorrectly used rather than using Num_MBs_In_Row
times 16 (Y) or 8 (U,V).
This worked when width is multiple of 16 but failed
otherwise.
Change-Id: I008919062715bd3d17c7aa2562ab58d1cb37053a
This change is designed to try and reduce pulsing effects when moving
with a complex transition like a fade, into an easy or static section in
an otherwise difficult clip in CQ mode.
The active CQ level is relaxed down to the user entered level for frames that
are generating less than the passed in minimum bandwidth.
Change-Id: Id6d8b551daad4f489c087bd742bc95418a95f3f0
Fixed discrepancy cpi->ni_frames vs cm->current_video_frame > 150.
Make one pass path explicit.
There is still scope for some odd behaviour around the transition
point at cpi->ni_frames > 150.
Change-Id: Icdee130fe6e2a832206d30e45bf65963edd7a74d
Where a key frame occurs because of a minimum interval
selected by the user, then these forced key frames ideally need
to be more closely matched in quality to the surrounding frame.
Change-Id: Ia55b1f047e77dc7fbd78379c45869554f25b3df7
Add a flag to always enable block4x4 search for speed=0 (good
quality) to guarantee no quality loss for speed0.
Change-Id: Ie04bbc25f7e6a33a7bfa30e05775d33148731c81
The maximum possible MV in 1/8 pel units is (1<<11), which could
cause mvcost out of its range that is 1023. Change maximum
possible MV in 1/8 pel units to (1<<11)-8 will fix this problem.
Change-Id: I5788ed1de773f66658c14f225fb4ab5b1679b74b
Further experiment with restriction of the Q range.
This uses the average non KF/GF/ARF quantizer, instead
of just relying on the initial value. It is not such a strong constraint
but there may be a reduced risk of rate misses.
Change-Id: I424fe782a37a2f4e18c70805e240db55bfaa25ec
The merge includes hooks to for CQ mode and other code
changes merged from the test branch.
CQ mode attempts to maintain a more stable quantizer within a clip
whilst also trying to adhere to a guidline maximum bitrate.
The existing target data rate parameter is used to specify the
guideline maximum bitrate.
A new parameter allows the user to specify a target CQ level.
For normal (non kf/gf/arf) frames, the quantizer will not drop BELOW the
user specified value (0-63). However, in some cases the encoder may
choose to impose a target CQ that is above that specified by the user,
if it estimates that consistent use of the target value is not compatible
with guideline maximum bitrate.
Change-Id: I2221f9eecae8cc3c431d36caf83503941b25e4c1
In two pass encoding each frame is given an active
Q range to work with. This change limits how much this
Q range can be altered over time from the initial estimate
made for the clip as a whole.
There is some danger this could lead to overshoot or undershoot
in some corner cases but it helps considerably in regard to
clips where either there is a glut or famine of bits in some sections,
particularly near the end of a clip.
Change-Id: I34fcd1af31d2ee3d5444f93e334645254043026e
cpi->target_bits_per_mb is currently not being used,
so delete it. Also removed other unused code in rdopt.c.
Change-Id: I98449f9030bcd2f15451d9b7a3b9b93dd1409923
count can be reduced to short because the max number of filtered frames
is set to 15. the max value for any frame is 32 (modifier = 16,
filter_weight = 2). 15*32 = 480 which requires 9 bits
this function goes from about 7000 us / 1000 iterations for the C code
to < 275 us / 1000 iterations for sse2 for block_size = 16 and from
about 1800 us / 1000 iters to < 100 us / 1000 iters for block_size = 8
Change-Id: I64a32607f58a2d33c39286f468b04ccd457d9e6e
The following features don't make sense for the first
pass in its current form and have a significant impact on its
speed (up to 50%).
Slow quantizer, slow dct and trellis optimization.
Change-Id: Id9943f6765ffbd71fc0084ec7dfbc9d376fd6fcd
Calculate a minimum intra value to be used in determining the
IIratio scores used in two pass, second pass.
This is to make sure sections that are low complexity" in the
intra domain are still boosted appropriately for KF/GF/ARF.
For now I have commented out the Q based adjustment of
KF boost.
Change-Id: I15deb09c5bd9b53180a2ddd3e5f575b2aba244b3
Scott pointed out that last_frame_type only gets updated while
loopfilter exists. Since last_frame_type is also needed in
motion search now, it needs to be updated every frame.
Change-Id: I9203532fd67361588d4024628d9ddb8e391ad912
Use the fast quantizer for inter mode selection and the
regular quantizer for the rest of the encode for good quality,
speed 1. Both performance and quality were improved. The
quality gains will make up for the quality loss mentioned in
I9dc089007ca08129fb6c11fe7692777ebb8647b0.
Change-Id: Ia90bc9cf326a7c65d60d31fa32f6465ab6984d21
1. Search for block8x16/block16x8 uses block8x8's search results.
2. Check block4x4 only if block8x8 is chosen. (This hurts quality,
which will be improved in another check-in.)
3. In block4x4 search, the previous block's result is used as
MV predictor for next block.
This change improves performance.
Change-Id: I9dc089007ca08129fb6c11fe7692777ebb8647b0
Re-calibrated sad_per_bit16 and sad_per_bit4 tables to linearly
correlated to quantizer values, these two variables are used in
motion search for costing motion vectors. This change has an small
positive effect on compression.
Change-Id: Ic9b5ea6fb8d5078ef663ba4899db019cc51f4166
the lookup table is based on floating point calculations (see source)
by moving the *3 before the downshift and adding the rounding bit, the
delta (LUT - integer) goes from:
______________________________________
__ 1__ 1______________________________
__ 1__ 1______________________________
____ 1______ 1________________________
____ 1 2__ 2 1________________________
______ 1 1 2__ 2__ 2__ 2 1 1__________
________ 1 1 2 2__ 1 2 3 1 2__ 2__ 2__
to:
__-1__-1______________________________
______________________________________
____-1______-1________________________
______________________________________
________-1______________-1____________
______________________________________
it's important to be able to use the integer version because the LUT
more or less precludes SIMD optimizations
Change-Id: I45a81127dc7b72a06fba951649135d9d918386c0
allow for optimized versions of apply_temporal_filter
(now vp8_apply_temporal_filter_c)
the function was previously declared as static and appears to have been
inlined. with this change, that's no longer possible. performance takes
a small hit.
the declaration for vp8_cx_temp_filter_c was moved to onyx_if.c because
of a circular dependency. for rtcd, temporal_filter.h holds the
definition for the rtcd table, so it needs to be included by onyx_int.h.
however, onyx_int.h holds the definition for VP8_COMP which is needed
for the function prototype. blah.
Change-Id: I499c055fdc652ac4659c21c5a55fe10ceb7e95e3
Add a new encoder control, VP8E_SET_TUNING, to allow the application
to inform the encoder that the material will benefit from certain
tuning. Expose this control as the --tune option to vpxenc. The args
helper is expanded to support enumerated arguments by name or value.
Two tunings are provided by this patch, PSNR (default) and SSIM.
Activity masking is made dependent on setting --tune=ssim, as the
current implementation hurts speed (10%) and PSNR (2.7% avg,
10% peak) too much for it to be a default yet.
Change-Id: I110d969381c4805347ff5a0ffaf1a14ca1965257
In SPLITMV, the 8x8 segment will be checked first. If the 8x8 rd
is better than the best, we check the other segments. Otherwise
bail. Adjustments to the thresh_mult were necessary to make
up for the initial quality loss.
The performance improved by 20% (average) for good quality,
speed 0 and speed 1, while the overall quality remained the same.
Change-Id: I717aef401323c8a254fba3e9777d2a316c774cc3
vp8_rd_pick_best_mbsegmentation looks at y only. The new
breakout does not include the frame cost, the prob_skip_false
cost, or the uv rate. Performance improved by a few percent
and the quality remained the same.
Change-Id: I94ff013998ac51e8ecce7130870f7b6600758e15
This fix added MV range checks for NEWMV mode as suggested by Jim.
To reduce unnecessary MV range checks, I tried Yaowu's suggestion.
Update UMV borders in NEWMV mode to also cover MV range check.
Also, in this way, every MV that is valid gets checked in diamond
search function.
Change-Id: I95a89ce0daf6f178c454448f13d4249f19b30f3a
The MV's range is 256. Since the new motion search uses a different
starting MV than the center ref MV, a MV range checking needs to
be done to avoid corruption.
Change-Id: I8ae0721d1bd203639e13891e2e54a2e87276f306
Per John's previous change, shrink TOKENEXTRA from 20 to 8 bytes
original: b7b1e6fb
reverted: 41f4458a
Also drop unused field from vp8_extra_bit_struct
Update ARM ASM to deal with this change. In particular, Extra is signed
and needs to be sign-extended when loaded.
Change-Id: Ibd0ddc058432bc7bb09222d6ce4ef77e93a30b41
This code is unused, as the current preproc implementation uses the
same spatial filter that postproc uses.
Change-Id: Ia06d5664917d67283f279e2480016bebed602ea7
Change the size of structure elements to reduce memory utilization.
Removed the 'section' member entirely, as it is set but never read.
Change-Id: Iad043830392fb4168cb3cd6075fb0eb70c7f691c
Added the initialization of the pointer to active map. Also added the
same logic for cyclic refresh in mbrow encoding threads.
Change-Id: Ic48d0849dc706b27fba72d07dcc498075725663d
Changed the end of block computation to use pmaxw. Removed
additional pushing and popping of registers that was not needed.
Change-Id: I08cb9b424513cd8a2c7ad8cea53b4e2adc66ef98
Change I3430820 performed an uninitialized read when
encode_breakout == 0, since the sum and sse wouldn't be set:
if(x->encode_breakout)
VARIANCE_INVOKE(..., get16x16var)(..., &sum, &sse);
if (cpi->active_map_enabled && x->active_ptr[0] == 0) {
...
} else if (sse < x->encode_breakout)
Change-Id: I915eb76d1227b4b6d1137a0dedf2c143860098a2
Corrected the initial Q range limits for the recode loop
to reflect the current allowed range for the frame.
In experimental work on constrained quality this bug was
causing unnecessary recodes.
Change-Id: I7e256fbfa681293b0223fe21ec329933d76c229f
This patch adds a weighting factor on RDMULT for UV blocks. The change
has an overall gain about 0.5% based on ssim, between 0.1 and 0.2% by
psnr numbers.
Change-Id: I97781b077ce3bb7e34241b03268491917e8d1d72
Deallocating the buffers before re-allocating them.
The fix passed James Berry's test program for memory
leak check.
Change-Id: I18c3cf665412c0e313a523e3d435106c03ca438d
Moved the code from the segmentation loop into a function
which is now called for each segment. This will allow us
to change the segment order checking more easily.
Change-Id: I9510d26f0acae5a73043fcca8f1984b121d3e052
When auto_golden wasn't set it forced all frames to be a golden
frame. Now the manual configured frequency is adhered to.
Change-Id: I360acac9bc487db0d9c4d4da6ee41f70c227c539
The inter_minq table controls the range of quantizers available
for a particular frame in two pass relative to a max Q value.
The changes reduces the range somewhat. The effect of this
was a small increase (0.3% average) in psnr for the test set
but it should also help encode speed somewhat for higher
quality modes as it will reduce the number of iterations in the
recode loop.
The change damps the range of quantizers available locally
within a section of a clip and should therefore help keep quality
more uniform. If there is systematic overshoot or undershoot the
range can shift gradually to accommodate. However, there is
some increased risk of overshoot or undershoot against the target
bit rate in VBR mode and this risk will be more pronounced for short
clips.
The change damps the range of quantizers available locally
within a section of a clip and should therefore help keep quality
more uniform. If there is systematic overshoot or undershoot the
range can shift gradually to accommodate. However, there is
some increased risk of overshoot or undershoot against the
target bit rate in VBR mode and this risk will be more
pronounced for short clips.
Change-Id: I84465567d49ae767c6c73ff2a2aac30c895adb52
Add vp8_mv_pred() to better predict starting MV for NEWMV
mode in vp8_rd_pick_inter_mode(). Set different search
ranges according to MV prediction accuracy, which improves
encoder performance without hurting the quality. Also,
as Yaowu suggested, using diamond search result as full
search starting point and therefore adjusting(reducing)
full search range helps the performance.
Change-Id: Ie4a3c8df87e697c1f4f6e2ddb693766bba1b77b6
On a keyframe alt ref and golden are refreshed. The flag was
not being set and so on the frame after a keyframe, motion
search would occur on the alt ref frame. This is not necessary
because the alt ref frame identical to the last frame in this
scenario.
Handle corner case where a forward alt-ref frame is put
directly after a keyframe.
Change-Id: I9be4cf290d694f8cf2f9a31852014b5ccf1504d3
The baseline bits per MB prediction tables have been
re calibrated based on the assumption that bits per mb
is inversely proportional to the quantizer level.
Change-Id: Ibd355c7acac4b8053dda1baf1032fe35f11da7f7
Replaced existing code to decide if a frame recode is required
with a function call. This is to simplify addition of extra clauses
that may be needed for the planned constrained quality mode.
Also fixed a bug where by alt ref not considered in the test.
Change-Id: I3d40bb21abe3e19f8456761e6849deb171738b60
x86-64 passes arguments in registers. There is no need to push
them to the stack before using them.
This fixes 15acc84f10 where ebx
was not getting preserved on x86.
Change-Id: I1214b5f818a0201f75ab6ad7d5c6f448e09b16c2
The use of incorrect mv costing tables in the ARNR sub-pel
filtering code led to corruption of the altref buffer in some cases,
particularly at low data rates.
The average gain from this fix is about 0.3% but there are a few
extreme cases where nasty and visible artifacts manifested and
for these few data points the improvement is > 10%.
PGW and AWG
Change-Id: I95cc02b196a433e71d0d2bd2b933fe68ed31e796
This intends to correct the tendency that VP8 aggressively favors rate
on intra coded frames. Experiments tested different numbers in [0, 1]
and found 9/16 overall provided about 2-4% gains for all-intra coded
clips based on vpx-ssim metric. The impact on regular encoded clips
is much smaller but positive overall. Overall impact on psnr is also
positive even though very small.
Change-Id: If808553aaaa87fdd44691f9787820ac9856d9f8a
The fast quantizer assembly code has not been updated to match the new
exact quantizer, which was made the default in commit 6adbe09.
Specifically, they are not aware of the potential for the coefficient
to be scaled, which results in the quantized result exceeding the range
of the DCT. This patch restores the previous behavior of using the
non-shifted coefficients when in the fast quantizer code path, but
unfortunately requires rebuilding the tables when switching between the
two.
Change-Id: I0a33f5b3850335011a06906f49fafed54dda9546
