Commit Graph

1829 Commits

Author SHA1 Message Date
James Zern
e7c5954c10 dec_neon: remove returns from void functions
Change-Id: I3c66a5dfe3de2bb3653cbbf1b92b0328aba62881
2015-01-09 18:08:05 -08:00
James Zern
f79c163bbf anim_encode: fix type conversion warnings
fixes:
C4267: '=' : conversion from 'size_t' to 'int', possible loss of data

Change-Id: Ie8e0bbd6f19fde21b2dbbd2a92cc99e76502dfed
2015-01-09 17:12:06 -08:00
Djordje Pesut
cbcbedd0de move rescaler functions to rescaler* files in src/dsp/
Change-Id: I906add1b1010a59ebfcc2dd81e15745433cc206b
2015-01-09 16:47:09 +01:00
James Zern
e8694d4dc3 mux: remove experimental FRGM parsing
fragment references remain: to be removed in a future commit

Change-Id: I02974c8a709cfe16dce72568639c8b912859de8e
2015-01-08 20:02:40 -08:00
Urvang Joshi
9e92b6eac6 AnimEncoder API: Optimize single-frame animated images
Try converting them to a non-animated image and pick that one if it's smaller
in size.

Change-Id: Ib97438fd2a95b1bfa9b7526a0938a9d85df33a57
2015-01-08 12:30:46 -08:00
Djordje Pesut
a28c4b363d MIPS: move WORK_AROUND_GCC define to appropriate place
Change-Id: I3055eca57dc4e9d39533a5b8170bbf7af9cd818f
2015-01-08 15:55:41 +01:00
Djordje Pesut
012d2c60fa MIPS: dspr2: added optimization for functions SSEAxB
list of optimized functions: SSE16x16, SSE8x8, SSE16x8, SSE4x4

Change-Id: Ie99e7cdd73b0d4ff855977315a5d0db9ffaa5f04
2015-01-08 13:49:17 +01:00
Djordje Pesut
9241ecf45d MIPS: dspr2: added optimization for function Average
Change-Id: I7ca316bc3f5fbdaf8dcaf9a2d2227a5134bf4f63
2015-01-08 11:46:15 +01:00
pascal massimino
9422211d5f Merge "Tune BackwardReferencesLz77 for low_effort (m=0)." 2015-01-08 00:46:51 -08:00
pascal massimino
df40057b21 Merge "Speedup VP8LGetHistoImageSymbols for low effort (m=0) mode." 2015-01-08 00:46:43 -08:00
Vikas Arora
ea08466d34 Tune BackwardReferencesLz77 for low_effort (m=0).
- Lower the threshold parameters for HashChainFindCopy.

For 1000 image PNG corpus (m=0), this change yields speedup of 15-20% at
lower quality range (0.25% drop in compression density) and about 10%
for higher quality range without any drop in the compression density.
Following is the compression stats (before/after) for method = 0:
         Before           After
         bpp/MPs          bpp/MPs
q=0      2.8615/18.000    2.8651/18.631
q=5      2.8615/18.216    2.8650/20.517
q=10     2.8572/18.070    2.8650/21.992
q=15     2.8519/18.371    2.8584/21.747
q=20     2.8454/18.975    2.8515/20.448
q=25     2.8230/8.531     2.8253/9.585
// Compression density remains same for q-range [30-100]
q=30     2.7310/7.706     2.7310/8.028
q=35     2.7253/6.855     2.7253/7.184
q=40     2.7231/6.364     2.7231/6.604
q=45     2.7216/5.844     2.7216/6.223
q=50     2.7196/5.210     2.7196/5.731
q=55     2.7208/4.766     2.7208/4.970
q=60     2.7195/4.495     2.7195/4.602
q=65     2.7185/4.024     2.7185/4.236
q=70     2.7174/3.699     2.7174/3.861
q=75     2.7164/3.449     2.7164/3.605
q=80     2.7161/3.222     2.7161/3.038
q=85     2.7153/2.919     2.7153/2.946
q=90     2.7145/2.766     2.7145/2.771
q=95     2.7124/2.548     2.7124/2.575
q=100    2.6873/2.253     2.6873/2.335

Change-Id: I0e17581fb71f6094032ad06c6203350bd502f9a1
2015-01-08 00:30:21 -08:00
Vikas Arora
b0b973c39b Speedup VP8LGetHistoImageSymbols for low effort (m=0) mode.
- Do light weight entropy based histogram combine and leave out CPU
  intensive stochastic and greedy heuristics for combining the
  histograms.

For 1000 image PNG corpus (m=0), this change yields speedup of 10% at
lower quality range (1% drop in compression density) and about 5% for
higher quality range (1% drop in compression density). Following is the
compression stats (before/after) for method = 0:
         Before           After
         bpp/MPs          bpp/MPs
q=0      2.8336/16.577    2.8615/18.000
q=5      2.8336/16.504    2.8615/18.216
q=10     2.8293/16.419    2.8572/18.070
q=15     2.8242/17.582    2.8519/18.371
q=20     2.8182/16.131    2.8454/18.975
q=25     2.7924/7.670     2.8230/8.531
q=30     2.7078/6.635     2.7310/7.706
q=35     2.7028/6.203     2.7253/6.855
q=40     2.7005/6.198     2.7231/6.364
q=45     2.6989/5.570     2.7216/5.844
q=50     2.6970/5.087     2.7196/5.210
q=55     2.6963/4.589     2.7208/4.766
q=60     2.6949/4.292     2.7195/4.495
q=65     2.6940/3.970     2.7185/4.024
q=70     2.6929/3.698     2.7174/3.699
q=75     2.6919/3.427     2.7164/3.449
q=80     2.6918/3.106     2.7161/3.222
q=85     2.6909/2.856     2.7153/2.919
q=90     2.6902/2.695     2.7145/2.766
q=95     2.6881/2.499     2.7124/2.548
q=100    2.6873/2.253     2.6873/2.285

Change-Id: I0567945068f8dc7888041e93d872f9def91f50ba
2015-01-08 00:29:57 -08:00
pascal massimino
c6d3292738 argb_sse2: cosmetics
clarify some variable names in PackARGB() + add some comments

Change-Id: I2bb91d6c52dcbcdebe0f92d5f2136c2d7d11af2a
2015-01-08 00:18:54 -08:00
James Zern
67f601cd46 make the 'last_cpuinfo_used' variable names unique
allows the sources to be #include'd in some hackish builds (don't do
that!)

Change-Id: I0c7a43acbebd0e2d5068845e6daa8ce47361cd91
2015-01-07 23:38:53 -08:00
Urvang Joshi
b9489861a3 AnimEncoder API: Init method for default options.
Change-Id: I3ccd7fe782e10c51986b55fc1a515d958ff70752
2015-01-07 14:32:11 -08:00
pascal massimino
856f8ec1fd Merge "AnimEncoder API: Remove AnimEncoderFrameOptions." 2015-01-07 13:39:10 -08:00
pascal massimino
c537514d46 Merge "AnimEncoder API: GenerateCandidates bugfix." 2015-01-07 13:38:24 -08:00
pascal massimino
dc0ce039f3 Merge "AnimEncoder API: Compute change rectangle for first frame too." 2015-01-07 13:37:27 -08:00
pascal massimino
f00b639b96 Merge "AnimEncoder API: In Assemble(), always set animation parameters." 2015-01-07 13:36:53 -08:00
pascal massimino
29ed796c17 Merge "AnimEncoder lib cleanup: prev to prev canvas not needed." 2015-01-07 13:36:21 -08:00
pascal massimino
9f0dd6e539 Merge "WebPAnimEncoder API: Header and implementation" 2015-01-07 13:35:49 -08:00
Urvang Joshi
5e56bbe09a AnimEncoder API: Remove AnimEncoderFrameOptions.
We only need config now, so this struct is not needed.

Change-Id: I5139956d13c36ceb4871d52122f248fe70f40c4b
2015-01-07 13:34:02 -08:00
Urvang Joshi
b902c3ea50 AnimEncoder API: GenerateCandidates bugfix.
As 'curr_canvas_mod' is being modified during calls to IncreaseTransparency()
and FlattenSimilarBlocks(), GetSubRect() should get the sub-frame from
'curr_canvas_mod' as well.

Earlier, GetSubRect() was computed from 'curr_canvas', so modifying
'curr_canvas_mod' had no effect on encoding.

Change-Id: Ia847503007b66364817fe57def5a9e3c37d1b3cc
2015-01-07 11:48:55 -08:00
Urvang Joshi
ef3c39bbd2 AnimEncoder API: Compute change rectangle for first frame too.
Earlier, we were always using full canvas for first frame.

Change-Id: Ib8d32961682c4b07010ea559a71dd59ab9ec0157
2015-01-07 11:26:27 -08:00
Urvang Joshi
eec423abe9 AnimEncoder API: In Assemble(), always set animation parameters.
We set the parameters even if there is just one frame. This is to make sure
assembly is correct even if single frame animation is NOT converted to a full
frame later.

Change-Id: If79e6aa5e2575cb0f3cd229f16c655b3663c35b0
2015-01-07 11:20:24 -08:00
Urvang Joshi
ae1c046e12 AnimEncoder lib cleanup: prev to prev canvas not needed.
Given that we decided to only handle frame disposal for output WebP
internally,
only current and previous canvas need to be maintained.

Change-Id: I625293bed5aeb5aabf4eca779f6ec3ee84c9ff2a
2015-01-07 11:17:40 -08:00
Urvang Joshi
4b997ae46d WebPAnimEncoder API: Header and implementation
A separate API to generate animated WebP images.
It will eventually replace the internal gif2webp_util methods.

Also: update makefiles.

Change-Id: Idf61dfc1016c10b24fea70425d1a2323cffba515
2015-01-07 10:42:02 -08:00
Pascal Massimino
9592053859 Merge "multi-thread fix: lock each entry points with a static var" 2015-01-07 00:03:51 -08:00
Pascal Massimino
4c1b300ada Merge "SSE2 implementation of VP8PackARGB" 2015-01-06 23:53:50 -08:00
James Zern
04c20e75ea Merge "MIPS: dspr2: added optimization for function Intra4Preds" 2015-01-06 16:15:10 -08:00
Pascal Massimino
a437694a17 multi-thread fix: lock each entry points with a static var
we compare the current VP8GetCPUInfo pointer to the last used.
This is less code overall and each implementation is still
testable separately (by just changing VP8GetCPUInfo, but not
a separate threads!)

Change-Id: Ia13fa8ffc4561a884508f6ab71ed0d1b9f1ce59b
2015-01-05 07:48:49 -08:00
Pascal Massimino
ca7f60db5f SSE2 implementation of VP8PackARGB
Change-Id: I40c0e26a6a2701216e4ddebcf793aa535677f437
2015-01-05 05:17:51 -08:00
Pascal Massimino
72d573f693 simplify the PackARGB signature
Change-Id: I51570e362126b2681f93211a4f59a3fedb5fd4b5
2015-01-05 02:10:04 -08:00
James Zern
4e2589ff81 demux: restore strict fragment flag check
inadvertently removed in:
demux: remove experimental FRGM parsing

Change-Id: Ia9bb8211e2153df51e7a01cabe8552524b8ed218
2014-12-23 12:47:06 -05:00
James Zern
e752f0a673 Merge "demux: remove experimental FRGM parsing" 2014-12-23 09:12:15 -08:00
James Zern
f8abb112f2 Merge changes I109ec4d9,I73fe7743
* changes:
  dec_neon: add DC8uvNoTop / DC8uvNoLeft
  dec_neon: add DC8uv
2014-12-23 09:11:22 -08:00
Djordje Pesut
ae2188a435 MIPS: dspr2: added optimization for function Intra4Preds
Change-Id: Ie2a23c356a8715817b020fbee2b40e878e2946de
2014-12-23 17:32:27 +01:00
Pascal Massimino
1f4b8642e8 move VP8EncDspARGBInit() call closer to where it's needed
Change-Id: I0d5121b456918f0ee6646903a8d71d4384deafe3
2014-12-23 16:04:14 +01:00
James Zern
14108d7878 dec_neon: add DC8uvNoTop / DC8uvNoLeft
adds do_top/do_left flags to DC8uv; ~88% / ~92% faster respectively
no change in DC8uv speed.

Change-Id: I109ec4d9ad13c9db64516e98ed4693a21a3e9b54
2014-12-22 15:47:38 -05:00
James Zern
d8340da756 dec_neon: add DC8uv
~87% faster.

Change-Id: I73fe77437792f1361ce8ab0b411132c6ec0fa021
2014-12-22 14:36:45 -05:00
Djordje Pesut
7ce8788b06 MIPS: dspr2: added optimization for function MakeARGB32
inline function MakeARGB32 calls changed to call
via pointers to functions which make (a)rgb for
entire row

Change-Id: Ia4bd4be171a46c1e1821e408b073ff5791c587a9
2014-12-22 12:31:36 +01:00
James Zern
012e623ddd demux: remove experimental FRGM parsing
references to fragments remain, along with some superfluous checks; these
will be removed in a future commit.

Change-Id: I39fe9314900ecbc5d60e5065b65fa1b4c668af63
2014-12-19 19:03:17 -08:00
Pascal Massimino
87c3d53180 method=0: Don't evaluate any predictor
and apply Paeth predictor (predictor#11) for the low effort (m=0) mode.

For 1000 image PNG corpus (m=0), this change yields speedup of 25% at lower quality
range and about 10% for higher quality range.

Change-Id: I0f036b8ffe45c241e63a067cbf01527b13d8de93
2014-12-17 18:41:08 +01:00
Djordje Pesut
6f4fcb983e Merge "MIPS: dspr2: added optimization for function ImportRow" 2014-12-17 09:36:02 -08:00
Pascal Massimino
24284459c7 replace unneeded calls to HistogramCopy() by swaps
most of the time, we don't need to actually move the
data.

Compression is randomly slightly different, because HistogramCompactBins() changed.
Timing is about the same.

Change-Id: Ia6af8e9780581014d6860f2b546189ac817cfad1
2014-12-17 15:32:36 +01:00
Djordje Pesut
bdf7b40c5c MIPS: dspr2: added optimization for function ImportRow
Change-Id: I8205b551755ee51f5efd0c54d64c8b09771786b1
2014-12-17 15:24:41 +01:00
pascal massimino
e66a9225f3 Merge "MIPS: dspr2: added optimization for function ExportRowC" 2014-12-17 04:36:25 -08:00
Djordje Pesut
c279fec192 MIPS: dspr2: added optimization for function ExportRowC
Change-Id: Ie1a303089eb64736f8bc7573819a8219aa7528a3
2014-12-17 12:01:48 +01:00
Pascal Massimino
31a9cf6417 Speedup WebP lossless compression for low effort (m=0) mode with following:
- Disable Cross-Color transform.
- Evaluate predictors #11 (paeth), #12 and #13 only.

Change-Id: I857264c85c61c3957d4fb45ae32d261d947c8bed
2014-12-17 11:52:11 +01:00
Djordje Pesut
9275d91c79 MIPS: dspr2: added optimization for function TrueMotion
Change-Id: Id006d9591c0c922e28f7f4c01e4006f0f07bdd56
2014-12-12 14:38:55 +01:00
pascal massimino
26106d662e Merge "enc_neon: fix building with non-Xcode clang (iOS)" 2014-12-11 02:25:25 -08:00
Pascal Massimino
1c4e3efea0 unroll the kBands[] indirection to remove a dereference in GetCoeffs()
speed-up is small but visible.

Change-Id: Icff546adc3276f3c3d46b147c4a735b5eb8ff22e
2014-12-11 08:06:20 +01:00
James Zern
a3946b8956 enc_neon: fix building with non-Xcode clang (iOS)
check for __apple_build_version__ to distinguish the two; a version
check could work as Apple bumped Xcode's to 5.x/6.x, but it's unclear
how upstream will deal with their versioning as they go 3.6+, so avoid
it for now.

Change-Id: I67cda67c4f68e262a92d805a63cc1496374be063
2014-12-10 15:50:26 -08:00
Pascal Massimino
8ed9c00d5e Merge "simplify the Histogram struct, to only store max_value and last_nz" 2014-12-10 02:02:05 -08:00
Pascal Massimino
bad775715a simplify the Histogram struct, to only store max_value and last_nz
we don't need to store the whole distribution in order to compute the alpha

Later, we can incorporate the max_value / last_non_zero bookkeeping
in SSE2 directly.

Change-Id: I748ccea4ac17965d7afcab91845ef01be3aa3e15
2014-12-10 10:44:57 +01:00
Djordje Pesut
3cca0dc7f0 MIPS: dspr2: Added optimization for DCMode function
Change-Id: I8ea31907c1ea1259ec4db8cee1a479bd13a025a1
2014-12-09 13:58:39 +01:00
Djordje Pesut
37e395fd1c MIPS: fix functions to use generic BPS istead of hardcoded value
Change-Id: I2d68abef886eff7f8df230f155b758dccd7d04fd
2014-12-05 15:55:47 +01:00
James Zern
9475bef4d7 PickBestUV: fix VP8Copy16x8 invocation
param order is src, dst

broken in:
66ad372 factorize BPS definition in dsp.h and add VP8Copy16x8

Change-Id: I761f618e3fe31ae7f58953256381f4f16bdb238e
2014-12-04 23:12:30 -08:00
James Zern
441f273f19 Merge changes I55f8da52,Id73a1e96
* changes:
  cosmetics: add some missing != NULL comparisons
  factorize BPS definition in dsp.h and add VP8Copy16x8
2014-12-04 20:46:29 -08:00
Pascal Massimino
4a279a680e cosmetics: add some missing != NULL comparisons
Change-Id: I55f8da527e5e8ee4b49c7e7aa0d61ea4a6c80904
2014-12-04 14:54:11 +01:00
Pascal Massimino
66ad372500 factorize BPS definition in dsp.h and add VP8Copy16x8
Change-Id: Id73a1e968c96455808755df4d131d74e3e2e135d
2014-12-04 13:45:14 +01:00
Pascal Massimino
432e5b550e make ALIGN_xxx naming consistent
(potentially for future factorization between enc/ and dec/)

Change-Id: Ibf6670e21433a6a6a7202dcbe76f7efc8493b8cf
2014-12-04 13:32:10 +01:00
Pascal Massimino
57606047ec encoder: switch BPS to 32 instead of 16
this is a first step to unifying encoding/decoding cache stride
and possibly sharing the prediction functions in dsp/

With this layout, there's a little (~7%) space lost with unused samples.
But no speed change was observed.

Change-Id: I016df8cad41bde5088df3579e6ad65d884ee711e
2014-12-04 09:17:18 +01:00
Djordje Pesut
1b66bbe998 MIPS: dspr2: added optimization for function TransformColor_C
Change-Id: Idbf5cecf6775340585b0fd7e6ddcb29c2fcbea36
2014-12-01 15:46:06 +01:00
James Zern
c6d0f9e758 histogram: cosmetics
fix indent + other minor spelling / whitespace changes

Change-Id: I6e4462b75c98994e3c53c115de07047dbe71ce3c
2014-11-25 15:53:19 -08:00
James Zern
f399d30764 Merge changes I6eac17e5,I32d2b514
* changes:
  dec_neon: add TM8uv
  dsp: initialize VP8PredChroma8 in VP8DspInit()
2014-11-25 15:32:14 -08:00
James Zern
9de9074c92 dec_neon: add TM8uv
~68% faster

reuses TM4() adding support for the additional rows, the columns were
already being done.

Change-Id: I6eac17e58cd1c636082bf7281f70f884ec399a6b
2014-11-25 14:40:17 -08:00
James Zern
8e517eca68 bit_reader/kVP8NewRange: range_t -> uint8_t
decreases the size of each entry from 4 bytes to 1.

Change-Id: I3e6a50bcbc279e5edfa411edb97b04300dedc7ae
2014-11-24 22:16:26 -08:00
James Zern
e18571393d dsp: initialize VP8PredChroma8 in VP8DspInit()
the table becomes non-const to allow for platform-specific optimizations

Change-Id: I32d2b51480020dc653ecfafd20b6b0f096af349f
2014-11-24 22:12:42 -08:00
Vikas Arora
e0c809ad23 Move Entropy methods to lossless.c
Move all the Entropy evaluation methods to lossless.c (from histogram.c).
There's slight difference in the way entropy is computed for evaluating
entropy in prediction methods and histogram (literal) for huffman trees.
Plan (later) to merge few (static) methods and reduce the code size.

This change has no impact on the compression speed/density.

Change-Id: Ife3d96a3c4a8d78a91723d9e0a8d1b78c0256a15
2014-11-20 13:48:05 -08:00
Vikas Arora
a0df55104e Remove handling for WEBP_HINT_GRAPH
Remove handling for WEBP_HINT_GRAPH w.r.t use_palette flag.

The WEBP_HINT_GRAPH is now used at one place, to set the initial size of the
Bit Writer as bpp for photo images are generally larger than the graphical
images.

Change-Id: I1b9c4436c85a8f69da74c0dbcd292397323f2696
2014-11-13 15:49:23 -08:00
Vikas Arora
413dfc0c4b Move static method definition before its usage.
Change-Id: Id766c2bea92e7ebf0de65046f73429b74b4fdda4
2014-11-13 13:18:30 -08:00
Vikas Arora
0f23566558 Update BackwardRefsWithLocalCache.
Update BackwardRefsWithLocalCache to do in-place update of backward
references w.r.t local color cache index.

No impact on the compression density or compression speed.

Change-Id: Ie066251464c3928c044e037b43df3af28b48ca30
2014-11-13 11:54:26 -08:00
Vikas Arora
d69e36ec59 Remove TODOs from lossless encoder code.
histogram.c:
 - Verified (earlier) that there's low correlation between Red & Blue colors
   (particularly after applying Cross-color transform). The Bin based histogram
   merge, bins on three entropies viz literal, red & blue symbols. Removing
   either of blue or red increases the compression density. So keeping the bins
   for red & blue sybmols.
 - Keeping the compact bins method as-is. This way it's simpler to read.
huffman_encode.h: Added field comments for struct HuffmanTree and removed the TODO.

Change-Id: Ia76f7bc730079d1b3b644038c5d9931db3797f0e
2014-11-12 16:10:16 -08:00
Vikas Arora
fdaac8e0ca Optmize VP8LGetBackwardReferences LZ77 references.
Use the refs_lz77 computed (with cache_bits=0) in the method 'CalculateBestCacheSize'
to regenerate the LZ77 references corresponding to the optimum cache_bits and avoid
calling costly 'BackwardReferencesLz77' one extra time.

This change leaves the compression density unchanged and speeds up compression
by 10-15%.

Change-Id: I5a92e11788d3c3f656aa7e1fba54fb5d96ee0027
2014-11-12 14:50:04 -08:00
Djordje Pesut
2f0e2ba826 MIPS: dspr2: added optimization for function Select
Change-Id: I22470d8b9ab8c5e90c5330ff12c9852676da1a3d
2014-11-07 09:44:16 +01:00
pascal massimino
a3e79a46f6 Merge "WebPEncode: Support encoding same pic twice (even if modified)" 2014-11-06 22:20:01 -08:00
Urvang Joshi
e4f4dddba3 WebPEncode: Support encoding same pic twice (even if modified)
This wasn't working for this specific scenario:
- Encode an RGBA 'pic' (with trivial alpha) using lossy encoding.
(so that pic->a == NULL after import happens).
- Modify the 'pic->argb' so that it has non-trivial alpha.
- Encode the same 'pic' again.
This used to fail to encode alpha data as pic->a == NULL.

Change-Id: Ieaaa7bd09825c42f54fbd99e6781d98f0b19cc0c
2014-11-06 13:52:48 -08:00
pascal massimino
cbc3fbb4d7 Merge "Updated VP8LGetBackwardReferences and color cache." 2014-11-06 13:47:21 -08:00
Vikas Arora
95a9bd85c4 Updated VP8LGetBackwardReferences and color cache.
- The optimal cache bits is evaluated inside the method 'VP8LGetBackwardReferences'.
- The input cache_bits to 'VP8LGetBackwardReferences' sets the maximum cache
  bits to use (passing 0 implies disabling the local color cache).
- The local color cache is disabled for lowerf (<= 25) quality levels (as before).
- Enabled local color cache for palette images as well. This saves additional
  0.017% bytes with a slight (2-3%) improvement in the compression speed.
- Removed 'use_2d_locality' parameter from method VP8LGetBackwardReferences, as
  this option is not an option now (after we freeze the lossless bit-stream).

Change-Id: I33430401e465474fa1be899f330387cd2b466280
2014-11-06 13:14:05 -08:00
Djordje Pesut
54f2c14cce MIPS: dspr2: added optimization for function FTransform
Change-Id: Ib5850edbc2a586ec9781f494b2337f024e22af78
2014-11-06 14:21:33 +01:00
Djordje Pesut
aa42f4231f MIPS: dspr2: Added optimization for function VP8LSubtractGreenFromBlueAndRed
Change-Id: I683c73cceee4a40ca810deba15e54fbf7dbe8918
2014-11-06 10:56:18 +01:00
Djordje Pesut
95ca44a718 MIPS: dspr2: added optimization for Disto4x4
enc/dec common macros moved to mips_macro.h

Change-Id: I38d491e772554ac663dd5eb4d15485c0343f23b1
2014-11-05 12:06:15 +01:00
James Zern
4171b6724e backward_references.c: reindent after c8581b0
Change-Id: Icfc0fe8e266c0f67a70b8cb095e5aaee155290b6
2014-11-04 17:40:04 +01:00
Vikas Arora
c8581b06e1 Optimize BackwardReferences for RLE encoding.
Updated BackwardReferencesRle method by utilizing the local color cache.
Also changed the name of method BackwardReferencesHashChain to
BackwardReferencesLz77 to reflect the LZ77 coding.

For the 1000 image corpus, this change saves 0.2% bytes
(at default settings) and is 2-5% faster to encode.

Change-Id: Ic3f288253b3bbb101a69945a80994c3fd0917f8b
2014-11-04 08:12:07 -08:00
Djordje Pesut
5798eee6be MIPS: dspr2: unfilters bugfix (Ie7b7387478a6b5c3f08691628ae00f059cf6d899)
Change-Id: I78d97960efbd1ec1af51a5426e38dc01bdb48140
2014-11-03 15:39:00 +01:00
Vikas Arora
4167a3f5f7 Optimize backwardreferences
Optimize backwardreferences (about 0.1% byte savings) with almost same
compression speed (3% faster on defaut compression settings).
1.) Simplified iteration logic for HashChainFindCopy.
    - Remapped the iter_max constant.
2.) Simplified main for loop for BackwardReferencesHashChain
    - Removed 'if' conditions for corner cases in the main loop.
    - Refactored the method(AddSingleLiteral) for adding one pixel.

Change-Id: I1bc44832fd81f11e714868a13e606c8f83157e64
2014-10-31 18:08:38 -07:00
James Zern
d18554c30d Merge "webp/types.h: use inline for clang++/-std=c++11" 2014-10-31 03:53:06 -07:00
Vikas Arora
77bdddf016 Speed up BackwardReferences
Speed up BackwardReferencesHashChainDistanceOnly method by:
1.) Remove for loop for shortmax code path.
2.) Execute the shortmax code path after regular call to
HashChainFindCopy, only if HashChainFindCopy() returns length > 2 (MIN_LENGTH).
3.) Also for shortmax, call method HashChainFindOffset (for length = 2),
instead of expensive method HashChainFindCopy().
4.) Handling first pixel (i==0) outside main loop and removing one if
condition (i > 0) per pixel.
5.) Handle the last pixel outside the main 'for' loop.

Overall compression speedup observed is around 5% (+/- noise).

Change-Id: Ifa30c4035f8d26e6e43e3c4881244d777961c22b
2014-10-30 10:58:24 -07:00
James Zern
6638710b9e webp/types.h: use inline for clang++/-std=c++11
at least clang 3.[45] in c++ mode with -std=c++11 define __STRICT_ANSI__
this change set WEBP_INLINE to inline for c++/non-strict-ansi/> c99

fixes crbug.com/428383

Change-Id: Ief2b934353c336a75865c73c90cc3dc5e4f83913
2014-10-30 15:25:27 +01:00
Vikas Arora
abf04205b3 Enable entropy based merge histo for (q<100)
Enable bin-partition entropy based heuristic for merging histograms
for higher (q >= 90) qualities as well. Keep the old behavior at the
maximum quality level (q==100).

This speeds up the compression between Q=90-99 (method=4) by factor 5-7X
and with loss of 0.5-0.8% in the compression density.

Change-Id: I011182cb8ae5403c565a150362bc302630b3f330
2014-10-30 03:59:36 -07:00
James Zern
572022a350 filters_mips_dsp_r2.c: disable unfilters
the output does not match the C-code.

Change-Id: Ie7b7387478a6b5c3f08691628ae00f059cf6d899
2014-10-30 11:10:11 +01:00
Djordje Pesut
a28e21b141 MIPS: dspr2: Added optimization for function ClampedAddSubtractFull
Change-Id: Iee98eaf007158f44a299dd5ba8d972d0d4108380
2014-10-29 13:08:06 +01:00
Djordje Pesut
18d5a1efa8 MIPS: dspr2: added optimization for function ClampedAddSubtractHalf
Change-Id: Iec22e897a4f56e79c18ec00f8caa9cefac67f186
2014-10-29 11:08:37 +01:00
Djordje Pesut
829a8c19a0 MIPS: dspr2: added optimization for ITransform
Change-Id: I3534fca143535c53d18a3749b3a1b0c8a7563463
2014-10-28 14:28:14 +01:00
Vikas Arora
653ace55c3 Increase the MAX_COLOR_CACHE_BITS from 9 to 10.
The Maximum allowed limit is 11.
The Q=25 and below is not impacted as cache bits are forced to 0.
This saves 0.05% - 0.1% bytes for other quality with almost same compression
speed (+/- 2-3%, that's more of a noise).

Change-Id: Icf972a98f298c89e140e37a627baf709134be9a0
2014-10-27 14:19:04 -07:00
Vikas Arora
919220c7e6 Change the logic adjusting the Histogram bits.
Updated the logic to limit the Histogram size to a constant, instead of
computing the same based on the Histogram size (that's variable size based on
the cache bits) for the maximum possible cache bits. The actual cache bits may
be lower than the maximum.
Note: The constant 2600 is 16MB/Sizeof(HistogramSize(MAX_COLOR_CACHE_BITS)).

The compression density remains the same with this change, with little faster
compression speed.

Change-Id: I3149894962852e9dad2501b9aa16bb847a20fd86
2014-10-27 09:57:17 -07:00
pascal massimino
53b096c0d7 Merge "Fix bug in VP8LCalculateEstimateForCacheSize." 2014-10-27 02:31:10 -07:00
Vikas Arora
e912bd55be Fix bug in VP8LCalculateEstimateForCacheSize.
The method VP8LCalculateEstimateForCacheSize is not evaluating the all possible
range for cache_bits.
Also added a small penality for choosing the larger cache-size. This is done to
strike a balance between additional memory/CPU cost (with larger cache-size) and
byte savings from smaller WebP lossless files.

This change saves about 0.07% bytes and speeds up compression by 8% (default
settings). There's small speedup at Q=50 along with byte savings as well.
Compression at Quality=25 is not effected by this change.

Change-Id: Id8f87dee6b5bccb2baa6dbdee479ee9cda8f4f77
2014-10-26 20:05:48 -07:00
James Zern
22881c999e dec_neon: add RD4 intra predictor
based on the SSE2 version; a bit rough around the loads, but still ~38%
faster.

Change-Id: I22426d939a7354cbc9a85ca8c68235d6081b882f
2014-10-24 21:22:07 +02:00
James Zern
1304eb3418 Merge "dec_neon: DC4: use pair-wise adds for top row" 2014-10-23 08:08:34 -07:00
pascal massimino
7083006b61 Merge "dsp/dec_{neon,sse2}: VE4: normalize variable names" 2014-10-23 07:29:27 -07:00
James Zern
0db9031c79 dsp/dec_{neon,sse2}: VE4: normalize variable names
use '0' rather than '_' when dealing with variables that result from a
shift

Change-Id: I29280c0dead645ce39dc4bb42c3e19929b302fd4
2014-10-23 16:04:13 +02:00
James Zern
b5bc15305b dec_neon: DC4: use pair-wise adds for top row
reduces load count, slightly faster

Change-Id: I880340ef8ef75ce4ce321c330f56f86b758bda08
2014-10-23 15:48:49 +02:00
Pascal Massimino
5b90d8fe42 Unify the API between VP8BitWriter and VP8LBitWriter
BitReader will be next...

Change-Id: Icd9e7ab2e3890131e664c0523627d9b8c5399a74
2014-10-23 15:35:16 +02:00
pascal massimino
f7ada560ce Merge changes I2e06907b,Ia9ed4ca6,I782282ff
* changes:
  dec_neon: add DC4 intra predictor
  dec_neon: add TM4 intra predictor
  dec_neon: add LD4 intra predictor
2014-10-23 06:31:54 -07:00
pascal massimino
5beb6bf070 Merge "dec_neon: add VE4 intra predictor" 2014-10-23 05:38:41 -07:00
James Zern
eba6ce06c3 dec_neon: add DC4 intra predictor
~70% faster

Change-Id: I2e06907b8d69be71a8c5581832c931923c24bab0
2014-10-23 14:21:08 +02:00
James Zern
79abfbd9df dec_neon: add TM4 intra predictor
~21% faster

Change-Id: Ia9ed4ca650f9d544821fa1faf3173611806a272a
2014-10-23 14:21:08 +02:00
James Zern
fe395f0e4d dec_neon: add LD4 intra predictor
based on SSE2 version, ~55% faster

Change-Id: I782282ffc31dcf238890b3ba0decccf1d793dad0
2014-10-23 14:20:47 +02:00
James Zern
32de385eca dec_neon: add VE4 intra predictor
based on SSE2 version, ~59% faster

Change-Id: Iaa2181eb51bd975de0e9fe5c7b66ed18188f0e3b
2014-10-23 11:46:08 +02:00
Vikas Arora
c2b5a0396a Modify CostModel to allocate optimal memory.
Change-Id: I7d52675d28bfc109d4e901581fc24cd36fcb79ee
2014-10-22 13:30:33 -07:00
Pascal Massimino
b7a33d7e91 implement VE4/HE4/RD4/... in SSE2
(30% faster prediction functions, but overall speed-up is ~1% only)

Change-Id: I2c6e7074aa26a2359c9198a9015e5cbe143c2765
2014-10-22 18:25:36 +02:00
Pascal Massimino
97c76f1f30 make VP8PredLuma4[] non-const and initialize array in VP8DspInit()
also convert 'type *dst' to 'type* dst'

Change-Id: I41ab66ad15b548cc45d1cb8b10bbca4fe1528cae
2014-10-22 18:14:20 +02:00
pascal massimino
0ea8c6c219 Merge "PrintReg: output to stderr" 2014-10-22 08:55:10 -07:00
James Zern
f85ec712b0 PrintReg: output to stderr
allows use of '-o -' while testing

Change-Id: Ibc02d7cede2df4eb8be0a28c0ca4bf5e91864191
2014-10-22 17:28:19 +02:00
Vikas Arora
139142e440 Optimize BackwardReferenceHashChainFollowPath.
Instead of calling HashChainFindMethod, call a new (subset) method
HashChainFindOffset to get the offset/distance for a given length.

The encoding is tad faster at default compression

                       Before              After
                     bpp/rate            bpp/rate
442 Palette     0.2720/5.270 MP/s      0.2720/5.790 MP/s
558 non-palette 3.7607/0.797 MP/s      3.7607/0.816 MP/s

Change-Id: If4041a9c18f7e972f49fcbab8c3e2f013d8bf1cf
2014-10-21 10:04:27 -07:00
James Zern
5f36b68d22 enc/backward_references.c: fix indent
reindent after c24f895

Change-Id: I55adcbef21ea3fdaded84b138745515596191a09
2014-10-20 11:35:20 +02:00
James Zern
e0e9960dd1 Merge "sync version numbers to 0.4.2 release" 2014-10-17 11:47:30 -07:00
James Zern
64ac51446d sync version numbers to 0.4.2 release
libwebp{,decoder} - 0.4.2
libwebp libtool - 5.2.0
libwebpdecoder libtool - 1.2.0

mux/demux - 0.2.2
libtool - 1.2.0

(cherry picked from commit eec5f5f121)
(cherry picked from commit 857578a811)

Change-Id: Ie9d10c68e28083674a8865ad8447b1a70dcea95d
2014-10-17 19:50:21 +02:00
Vikas Arora
c24f8954be Simplify and speedup Backward refs computation.
Updated VP8LGetBackwardReferences and HashChainFindCopy method with following:
- Remove the recursive CostModelBuild.
- Reuse the lz77 backward refs in CostModelBuild, instead of evaluating it
  again (as it was done for recursion_level=0).
- Consolidated the Match-length logic inside FindMatchLength method.
- Removed the logic for altering best_length/val based on the 2D distance.
  The additional 162 value (+= 9 * 9 + 9 * 9 - y * y - x * x) can't change the
  best_val eval computation to choose a different curr_length, as best_val was
  set to 'curr_length << 16'.

  Following is the impact on the compression speed/density at default & max
  quality, overall this speeds up compression by 5-15% (q=100 -> 75) with a tad
  drop (0.02-0.03%) in compression density for the non-palette images.

                  Before                After
                bpp/Rate(MP/s)        bpp/Rate(MP/s)
q=75 (def)
All 1000        2.4492/1.049 MP/s     2.4498/1.230 MP/s
Palette         0.2719/5.060 MP/s     0.2719/6.110 MP/s
non-Palette     3.7597/0.732 MP/s     3.7607/0.840 MP/s

q=100
All 1000        2.4134/0.125 MP/s     2.4142/0.131 MP/s
Palette         0.2692/2.585 MP/s     0.2692/2.885 MP/s
non-Palette     3.7040/0.079 MP/s     3.7053/0.083 MP/s

Change-Id: I27a5eff3356d876c3e949fd32262244b25678b7a
2014-10-17 09:21:30 -07:00
James Zern
d1c359ef29 fix shared object build with -fvisibility=hidden
set WEBP_EXTERN to visibility=default
+ explicitly mark VP8GetCPUInfo as it's referenced within the examples

Change-Id: Ie3d2b15088e888f0b55203b205993eba75899d99
2014-10-17 11:50:52 +02:00
James Zern
a4c3a31b8f WEBP_TSAN_IGNORE_FUNCTION: fix gcc compat warning
move the attribute to the front of the function to quiet clang warning:
GCC does not allow no_sanitize_thread attribute in this position on a
function definition

Change-Id: Ie4cc6e35a07bd00eab67d9cd6801bd2be9cfe676
2014-10-16 18:06:43 +02:00
Pascal Massimino
80247291c6 mark some init function as being safe for thread_sanitizer.
introduces the macro WEBP_TSAN_IGNORE_FUNCTION

Change-Id: I3de2b6c1a2076fba4da7ae50322551e026b2082b
2014-10-16 16:34:07 +02:00
James Zern
79b5bdbfde bit_reader.h: cosmetics: fix a typo
Change-Id: I1ba09124700b3120f18eb3705eb5ba805feb2ca0
2014-10-16 10:52:47 +02:00
Pascal Massimino
6c6736816c Improved near-lossless mode.
Compared to previous mode it gives another 10-30% improvement in compression keeping comparable PSNR on corresponding quality settings.

Still protected by the WEBP_EXPERIMENTAL_FEATURES flag.

Change-Id: I4821815b9a508f4f38c98821acaddb74c73c60ac
2014-10-15 10:57:21 -07:00
James Zern
0ce27e715e enc_mips32: workaround gcc-4.9 bug
avoids an ICE with NDK r10b + NDK_TOOLCHAIN_VERSION=4.9

In function 'SSE16x16':
enc_mips32.c (684) internal compiler error: Segmentation fault

Change-Id: I1a3d33c0a9534c97633ab93bcdf9bf59d3a7e473
2014-10-15 19:14:04 +02:00
James Zern
aca1b98f52 enc/vp8l.c: fix indent
reindent after ca00502

Change-Id: I8c88dbc11dc96c117531b17682b764a235ef23bb
2014-10-13 11:33:23 +02:00
Vikas Arora
ca00502788 Evaluate non-palette compression for palette image
Evaluate if for Palette images (num_colors <= 256), non-palette
compression path (Subtract green, predictor transform etc) yield an
optimal compression density.

This change reduces the WebP file (for palette images) size by 0.4% with
drop of 3-5% in compression speed.

Change-Id: I1ad66fa94db4fd7ba7bc215763791ef662cd4f42
2014-10-10 11:55:45 -07:00
James Zern
c8a87bb62d AssignSegments: quiet -Warray-bounds warning
the number of segments are previously validated, but an explicit check
is needed to avoid a warning under gcc-4.9

Change-Id: Ifa7c0dd7f3f075b3860fa8ec176d2c98ff54fcea
2014-10-10 17:18:39 +02:00
pascal massimino
32f67e309f Merge "enc_neon: initialize vectors w/vdup_n_u32" 2014-10-09 12:23:18 -07:00
Pascal Massimino
fabc65da32 1-3% faster encoding optimizing SSE_NxN functions
got rid of the |a-b|^|b-a| method and went back
to just (a-b)^2 instead.

quality | size(bytes) after/before | time (ms) after/before

Change-Id: Ia3e0e6507b3f903deb1e182f78dad6df07380fd0
2014-10-09 07:20:00 -07:00
James Zern
7534d71640 enc_neon: initialize vectors w/vdup_n_u32
replaces {} initialization gnu-ism

Change-Id: I5a7b2d4246f0205e4bfb7f4b77d720c47d8674ec
2014-10-09 12:35:41 +02:00
Pascal Massimino
5f81391263 Merge "Fix return code of EncodeImageInternal()" 2014-10-07 23:49:29 -07:00
Pascal Massimino
e321abe43d Fix return code of EncodeImageInternal()
It was returning 'VP8_ENC_OK' in case of memory error.

Change-Id: I184a3e29c9f1b863637cacbe389b058d75c3dbf8
2014-10-08 08:48:53 +02:00
Pascal Massimino
f82cb06afb optimize palette ordering
We compact the palette by weighted distance, favoring the green channel.

Average gain on paletted file is ~0.5%, with gain up to 6-7% on some favorable cases.
Encoding speed is unaffected.

Disabled for alpha (or any single-channel input)

Also: always use quality=20 for EncodePalette() since it
doesn't make any real difference.

Change-Id: I19fb14316a366f139a941b45aef5663a33c905e1
2014-10-08 08:42:36 +02:00
Pascal Massimino
f545feee64 don't set the alpha value for histogram index image
This leads to tiny extra compression (~few bytes per file) for free

Change-Id: Ia4d8cef3de4365e32eacefd69a57689c80042a23
2014-10-08 08:24:19 +02:00
Pascal Massimino
2d9b0a4472 add WebPDispatchAlphaToGreen() to dsp
SSE2 version is 2.1x faster

This is used to transfer the alpha plane to green channel before lossless compression.

Change-Id: I01d9df0051c183b1ff5d6eb69961d4f43e33141a
2014-10-06 23:15:44 +02:00
Vikas Arora
d5e498d47f Change Entropy based Histogram Combine heuristic.
Don't combine the Histograms that have trivial (single valued A, R & B)
  symbols.
Following is the compression savings data along with compression time (before
& after) per image.
                     Before             After
                     bpp, rate(MP/s)    bpp, rate(MP/s)
Q=25, method = 4     2.508, 1.807       2.499, 1.916
Q=50, method = 4     2.460, 1.488       2.456, 1.512
Q=75, method = 4     2.452, 1.078       2.450, 1.092
Q=25, method = 5     2.505, 1.398       2.496, 1.383
Q=50, method = 5     2.458, 1.170       2.453, 1.143
Q=75, method = 5     2.453, 0.886       2.450, 0.855

This change provides 0.1-0.4% compression gains and speeds up the lossless
compression for the default method=4 (the drop in compression speed is between 1-3.5% for method=5).

Change-Id: Idfd88c2092f37afacd26a97097b3053f8183953a
2014-09-30 13:41:39 -07:00
Pascal Massimino
47a2d8e1d9 fix MSVC float->int conversion warning
+ add a clarifying comment

Change-Id: I8ac1df1de2e5277f2d968dec489546e680bb5e0c
2014-09-27 00:36:01 -07:00
James Zern
35ad48b848 HistoHeapInit: correct positions allocation size
Change-Id: I1879fd48bee3aea6f0504926d7030b504dd9be07
2014-09-26 11:21:19 -07:00
Pascal Massimino
45d9635fd3 lossless: entropy clustering for high qualities.
Tested on 1000 pngs corpus with quality 90-100 it gives ~0.15% improvement
in compression density and ~7% speed up.

Change-Id: I460f56c96707edb3c1f0b51a024e5122e10458df
2014-09-26 15:26:56 +02:00
Pascal Massimino
dc37df8c7a fix type warning for VS9_x64
Error report was:
src\utils\color_cache.c(48) : warning C4334: '<<' : result of 32-bit shift implicitly converted to 64 bits (was 64-bit shift intended?)

Change-Id: I93463ba7cd94faf1cf04986acbfaa06b62700d26
2014-09-25 23:47:06 -07:00
Vikas Arora
fdd6528ba2 Remove unused VP8LDecoder member variable
Remove the unused VP8LDecoder member variable (last_cached_)

Change-Id: I4a7d2f1b72d166efb978850e061dc69c8509e224
2014-09-24 11:59:51 -07:00
James Zern
ea3bba5a66 Merge "rewrite Disto4x4 in enc_neon.c with intrinsic" 2014-09-24 10:51:47 -07:00
Pascal Massimino
f060dfc422 add lossless incremental decoding support
* We don't need to change DecodeAlpha, since incremental
decoding is not useful for Alpha (we already decode
progressively along the RGB)
* Similarly, we don't do incremental decoding for level>0 planes:
   the metadata don't turn into visible pixel (only the ones in level0), so...
(No visible speed change)

Change-Id: I2fd4b9ba227561a7dbede647686584b752be7baa
2014-09-24 09:55:01 +02:00
Yang Zhang
ab70794ddb rewrite Disto4x4 in enc_neon.c with intrinsic
Performance test:
Platform: A9
Input data: bryce.yuv  11158x2156
performance of assembly is the base. Less ratio is better.
|toolchain |assembly |intrinsic |
|gcc4.6    |100%     |97.15%    |
|gcc4.8    |100%     |95.51     |

Change-Id: Idc2446685acdeb58a4dbdcdae533c68a83a1b879
2014-09-23 18:28:36 -07:00
Djordje Pesut
d4471637ef MIPS: dspr2: added optimization for function FilterLoop24
affected functions: VFilter16i, HFilter16i, VFilter8i and HFilter8i

Change-Id: I5d2bc7716e60e048a33d630fe4a86011bfb6d42e
2014-09-23 10:32:55 +02:00
skal
2aef54d429 Merge "prepare VP8LDecodeImage for incremental decode" 2014-09-23 00:31:27 -07:00
pascal massimino
aed0f5a231 Merge "MIPS: dspr2: added optimization for function FilterLoop26" 2014-09-23 00:17:25 -07:00
skal
286306853e prepare VP8LDecodeImage for incremental decode
- don't call VP8LClear() when there's no error (and let the caller do it)
- only initialize output once if state_ is not READ_DATA
- don't over-set dec->status_ = READ_DATA
- don't re-set dec->status_ if DecodeImageStream() fails
- remove unneeded dec->action_ field
- make ReadImageInfo() check br->eos_
- use ErrorStatusLossless() more consistently

Change-Id: Ica6e4b1c82e3fce8b1ce0274def551a886b73b0b
2014-09-23 00:13:52 -07:00
skal
248f3aed22 remove br->error_ field
it's somewhat redundant with br->eos_

also make the status-check coherent.

Change-Id: I98e755e037d45acb0760baf2344bf11fb5fb5cda
2014-09-23 00:04:58 -07:00
Djordje Pesut
49e15044ef MIPS: dspr2: added optimization for function FilterLoop26
affected functions: VFilter16, HFilter16, VFilter8 and HFilter8

Change-Id: Ib2fc41aaa00b10c2906d689bdc5a10f4568e70a8
2014-09-23 08:46:05 +02:00
skal
c792d4129a Premultiply with alpha during U/V downsampling
This prevents the 'alpha-leak' reported in issue #220

Speed-diff is kept minimal.

Change-Id: I1976de5e6de7cfcec89a54df9233c1a6586a5846
2014-09-18 23:40:34 -07:00
Vikas Arora
b901416b90 Record the lossless size stats.
Record and show the lossless header and image data sizes in the cwebp.

Change-Id: I08f19693cb7a756b6fdce5b55d71f5367b5f02fc
2014-09-17 15:16:05 -07:00
Pascal Massimino
cddd334050 Add a WebPExtractAlpha function to dsp
This is the opposite of WebPDispatchAlpha

+ Implement the SSE2 version

Change-Id: I0c297309255f508c5261da8aad01f7e57f924d6c
2014-09-15 08:12:03 +02:00
Pascal Massimino
0716a98eb3 fix indent after I0204949917836f74c0eb4ba5a7f4052a4797833b
Change-Id: I5d9e5d0a2ad2cefd8c539571d2eaee948da60ad5
2014-09-12 19:59:53 +02:00
Vikas Arora
f9ced95a9b Optimize lossless decoding for trivial(ARB) codes.
Optimize the decoding for region that have trivial literal codes.
The trivial literal is defined as huffman image with Red, Blue and Alpha
huffman trees with only single code values.
This speeds up lossless decoding by 3%

Change-Id: I0204949917836f74c0eb4ba5a7f4052a4797833b
2014-09-12 09:08:08 -07:00
Pascal Massimino
690b491af1 fix loop bug in DispatchAlpha()
* We were re-doing most of the work in plain-C as 'left-over'.
* we were always returning has_alpha = true because of a bad mask all_0xff

These bugs were conservative and silent, in the sense that we were 'just' doing
more work than necessary.

Now, the SSE2 version is really 2x faster than the C version.

Change-Id: I6c8132a267fe3c7a3d1fa70e7a5fcd10719543fa
2014-09-11 22:35:08 +02:00
Djordje Pesut
3101f53720 MIPS: dspr2: added optimization for TransformOne
added macros for TransformOne, TransformAC3 and TransfromDC

Change-Id: I4341450f443cf46dcf91c0db17bde63c8fb8afee
2014-09-11 17:02:02 +02:00
Pascal Massimino
a6bb9b17d8 SSE2 for inverse Mult(ARGB)Row and ApplyAlphaMultiply
Change-Id: Iab5c0e4a4d2b31f86736a9b277e62b6e28c3d2b4
WebPMultRow: ~7x faster
WebPMultARGBRow: ~3x faster
ApplyAlphaMultiply: 60% faster
2014-09-11 07:58:42 +02:00
Vikas Arora
d84a8ffdf7 Remove default initialization of decoder status.
emove the default initialization of decoder status in the method
VP8LDecodeImage().

Change-Id: Ie6b949606349f4e937c4c1dd2c02ff2a4f86870f
2014-09-10 14:55:46 -07:00
Vikas Arora
e0a9932161 Rectify bug in lossless incremental decoding.
Handle the corner case when VP8LDecodeImage() method is called with an invalid
header data. The lossless decoding doesn't support incremental mode yet.
Return the error status as BITSTREAM error in case not all pixels are decoded
with the provided bit-stream. Also added asserts in the VP8LDecodeImage() method
to validate the decoder header with appropriate/valid data for huffman trees
(htree_groups_ etc).

Change-Id: Ibac9fcfc4bd0a2c5f624bb9d4a2b9f6459aa19ea
2014-09-09 15:34:16 -07:00
Djordje Pesut
e2502a97c1 MIPS: dspr2: added optimization for TransformAC3
Change-Id: Icd789ee5f6d764297e7dc0a0f8a3bc47ab92ac65
2014-09-09 14:53:36 +02:00
Djordje Pesut
24e1072aac MIPS: dspr2: added optimization for TransformDC
Change-Id: Iee69758f6442ea9c80ddaa32cea8d00dda4c6252
2014-09-09 14:15:04 +02:00
Pascal Massimino
c0e84df8e8 Merge "Slightly faster lossless decoding (1%)" 2014-09-09 03:55:00 -07:00
Pascal Massimino
8dd28bb560 Slightly faster lossless decoding (1%)
-> introduce special case 64b pattern-copy, similar to the 8b one for alpha.
-> use mempcy() for non-overlapping areas
+ cosmetics and homogenezation of the code

Change-Id: I0e65e04b96fec94c009a4614137dfba2a0f98561
2014-09-09 11:18:30 +02:00
Djordje Pesut
f0103595dd MIPS: dspr2: added optimization for ColorIndexInverseTransforms
Change-Id: I5b6094ce489d4f896bc4b8f575142eb3c5054beb
2014-09-08 17:22:59 +02:00
Pascal Massimino
d3242aee16 make VP8LSetBitPos() set br->eos_ flag
ReadSymbol() finishes with a VP8LSetBitPos() call only and could miss an eos_ during the decode loop.

Things are faster because of inlining too.

Change-Id: I2d2a275f38834ba005bc767d45c5de72d032103e
2014-09-06 08:40:20 +02:00
Pascal Massimino
a9decb5584 Lossless decoding: fix eos_ flag condition
eos_ needs to be set only when superfluous bits have actually
been requested.
Earlier, we were assuming pre-mature end-of-stream to be an error.
Now, more precisely, we mark error when we have encountered end-of-stream *and*
we attempt to read more bits after that.

This handles cases where image data requires no bits to be read

Change-Id: I628e2c39c64f10c443fb51f86b1f5919cc9fd299
2014-09-05 20:21:50 +02:00
Pascal Massimino
3fea6a28da fix erroneous dec->status_ setting
We only need to set BITSTREAM_ERROR if !ok.

Change-Id: I5bd13e64797e8bc509477edb29158abb39cb0ee1
2014-09-05 19:48:11 +02:00
Djordje Pesut
80b8099fd8 MIPS: dspr2: add some specific mips code to commit I2c3f2b12f8df15b785fad5a9c56316e954ae0c53
added some C-code tuning also

Change-Id: I67ce70a063ef6b5821b9158a4defd6987eccbb9a
2014-09-04 13:42:39 +02:00
skal
e564062522 Merge "further refine the COPY_PATTERN optim for DecodeAlpha" 2014-09-04 03:43:55 -07:00
James Zern
854509fec0 enc/histogram.c: reindent after f4059d0
fixes indent in HistogramRemap after:
f4059d0 Code cleanup for HistogramRemap.

Change-Id: I9f53a088749e9100a70331bda1662488666c5156
2014-09-03 16:58:49 -07:00
skal
344219645b Merge "~3-5% faster encoding optimizing PickBestIntra*()" 2014-09-03 15:53:32 -07:00
skal
865069c12e further refine the COPY_PATTERN optim for DecodeAlpha
* use functions instead of MACRO
* adjust var's name

Overall, same speed, with more readible code

Change-Id: I2c3f2b12f8df15b785fad5a9c56316e954ae0c53
2014-09-04 00:25:27 +02:00
Djordje Pesut
a59562283f added C-level optimization for DecodeAlphaData function
Copies with short distances of 1,2 and 4 are specialized.

up to 10-14% faster alpha decoding.

Change-Id: I9708e98193910bfaf8ef43091f3fdea73b63896d
2014-09-03 16:49:17 +02:00
skal
187d379db6 add a fallback to ALPHA_NO_COMPRESSION
if ALPHA_LOSSLESS_COMPRESSION produces a too big file (very rare!),
we fall-back to no-compression automatically.

Change-Id: I5f3f509c635ce43a5e7c23f5d0f0c8329a5f24b7
2014-09-02 21:55:13 +02:00
skal
a48a2d7635 ~3-5% faster encoding optimizing PickBestIntra*()
* Add early-out check for Intra16
* replace some memcpy() by pointer swap

Change-Id: I5edc5f7fbc8e39984deb48e6c045c97c61418589
2014-09-01 14:40:25 +02:00
skal
77d4c7e337 address cosmetic comments from patch #71380
Change-Id: Iaba301b9e77aa4febe0efe1e6016fab42d5589f3
2014-08-28 18:08:00 -07:00
skal
f75dfbf23d Speed up Huffman decoding for lossless
speed-up is ~1.6% for photographic image to 10% for graphical image
(1000 images corpus was sped up by 5.8 %)

Code by akramarz@google.com and jyrki@google.com

Change-Id: Iceb2e50e6cc761b9315a3865d22ec9d19b8011c6
2014-08-28 12:28:04 -07:00
James Zern
637b388809 dsp/lossless: workaround gcc-4.9 bug on arm
force Sub3() to not be inlined, otherwise the code in Select() will be
incorrect.
https://android-review.googlesource.com/#/c/102511

Change-Id: I90ae58bf3e6cc92ca9897f69974733d562e29aaf
2014-08-27 20:31:21 -07:00
James Zern
8323a9038d dsp.h: collect gcc/clang version test macros
endian_inl.h already relies on dsp.h, grab the definitions from there.

Change-Id: I445f7d0631723043c55da1070498f89965bec7b1
2014-08-27 19:33:09 -07:00
skal
e6c4b52f28 move static initialization of WebPYUV444Converters[] to the Init function.
Split initialization of YUV444Converters[] out of Upsamplers init.

update test for NULL function pointers

Change-Id: I9603f54250f90c85a12ffbecfd6c59e9b06c47e0
2014-08-27 11:36:37 -07:00
skal
49911d4df2 Merge "fix indentation" 2014-08-27 07:52:36 -07:00
Vikas Arora
f4059d0c7d Code cleanup for HistogramRemap.
Avoid call to HistogramAddThresh when there's only one Histogram image.
Change-Id: I43b09e8e2d218c95969567034224777dcce37ab9
2014-08-26 15:45:22 -07:00
skal
e632b0929b fix indentation
Change-Id: I2294a6c83e5f345f64bd5120b91532e00ed6c543
2014-08-25 23:52:09 -07:00
skal
f5c04d64b7 Merge "add a DispatchAlpha() for SSE2 that handles 8 pixels at a time" 2014-08-25 22:43:42 -07:00
skal
fc98edd936 add a DispatchAlpha() for SSE2 that handles 8 pixels at a time
Only slightly faster.

Change-Id: Ie2e57e6a0950166124cf1075c6c9b45b7abdad8c
2014-08-25 21:03:03 -07:00
skal
73d361dd5f introduce VP8EncQuantize2Blocks to quantize two blocks at a time
No speed diff for now. We might reorder better the instructions later,
to speed things up.

Change-Id: I1949525a0b329c7fd861b8dbea7db4b23d37709c
2014-08-25 20:21:42 -07:00
Djordje Pesut
0b21c30b1a MIPS: dspr2: added optimization for EmitAlphaRGB
New dsp function: WebPDispatchAlpha()

Change-Id: I48e539d22471279ec75185759bc68d18b127f716
2014-08-21 20:39:35 -07:00
James Zern
953acd56a4 enc_neon: enable QuantizeBlock for aarch64
vtbl4_u8 is available everywhere except iOS arm64: use vtbl2q_u8 there
with a corresponding change in the load.

Change-Id: Ib84212dda3c7875348282726c29e3b79b78b0eac
2014-08-20 11:48:25 -07:00
Djordje Pesut
f4ae143720 MIPS: mips32: code rebase
mips code rebased to be same as C code
from commit I8c29a8a0285076cb3423b01ffae9fcc465da6a81

Change-Id: I3848f4ce43387c3a62b336606498779f7b07ec44
2014-08-19 15:13:16 +02:00
Djordje Pesut
569771549a MIPS: dspr2: added optimizations for VP8YuvTo*
VP8YuvToRgb
VP8YuvToBgr
VP8YuvToRgb565
VP8YuvToRgba4444
VP8YuvToArgb
VP8YuvToBgra
VP8YuvToRgba

Change-Id: I22212a125d890e1fd28388fec906a1a5c07ff386
2014-08-19 14:29:32 +02:00
skal
2523aa73cb SmartRGBYUV: fix odd-width problem with pixel replication
rightmost pixel was missing a copy, which could lead to invalid read.

Also added a lower dimension of 4, below which we use the regular conversion.
This is to prevent corner cases, in addition to not being overkill.

Change-Id: Iac12e7a3d74590f12fe8eeb1830b9891e61439f6
2014-08-18 15:58:36 -07:00
Pascal Massimino
ee52dc4e54 fix some MSVC64 warning about float conversion
Change-Id: I27ab27fc15033d27d0505729f6275fb542c8d473
2014-08-16 00:15:29 -07:00
James Zern
3fca851a20 cpu: check for _MSC_VER before using msvc inline asm
_M_IX86 will be defined in mingw builds after including windows.h. as
the gcc inline asm is first, this missing check would only have caused
an error if the code was reorganized.

Change-Id: I395679bcfc43e94d308d1ceb0c0fbf932b2c378c
2014-08-15 15:11:40 -07:00
skal
e2a83d7109 faster RGB->YUV conversion function (~7% speedup)
with a special case for dithering==0., it gets somewhat faster on x86
thanks to inlining.

Also, less macros.

Change-Id: Ic2f2bf6718310743bb40cef2104fa759a073e6d5
2014-08-15 11:13:25 -07:00
skal
de2d03e12f Merge "Add smart RGB->YUV conversion option -pre 4" 2014-08-15 11:07:49 -07:00
skal
3fc4c539aa Add smart RGB->YUV conversion option -pre 4
New function: WebPPictureSmartARGBToYUVA()
This implement smart RGB->YUV conversion.

This is rather undocumented for now, and is triggered using '-pre 4'
preprocessing option.

This is slow-ish and use quite some memory, but should be improvable.
This is somehow a usable beta version.

Change-Id: Ia50a8c30134e4cab8a7d3eb70aef13ce1f6187a1
2014-08-15 10:55:09 -07:00
Djordje Pesut
b4dc4069a2 MIPS: dspr2: added optimization for (un)filters
HorizontalFilter
VerticalFilter
GradientFilter
HorizontalUnfilter
VerticalUnfilter
GradientUnfilter

Change-Id: I54055b4767c37719691811072e95bf79c1f627b1
2014-08-14 11:55:19 -07:00
Djordje Pesut
b61c9ceca8 MIPS: dspr2: Optimization of some simple point-sampling functions
Change-Id: I6a4ab29bd0cc5a2951a8882cf9997032dc38bd79
2014-08-13 17:18:49 +02:00
Djordje Pesut
98c54107df MIPS: mips32r2: added optimization for BSwap32
gcc < 4.8.3 doesn't translate bswap optimally.
use optimized version always

Change-Id: I979ea26ad6dc0166d3d2f39c4148eb8adfb7ddec
2014-08-12 09:29:13 +02:00
Djordje Pesut
b7e5a5c451 MIPS: detect mips32r6 and disable mips32r1 code
Change-Id: Id1325c789a990c9a8704e84e99a22d580303eb8a
2014-08-08 17:29:31 +02:00
pascal massimino
bb07022b66 Merge "cosmetics" 2014-08-06 12:30:08 -07:00
James Zern
e300c9d819 cosmetics
fix some indent/whitespace, remove a few duplicate includes, extra
semi-colons

Change-Id: If937182b40a21e0f2028496e7b4b06c6e8a41352
2014-08-06 12:10:59 -07:00
pascal massimino
0e519eea8e Merge "cosmetics: remove some extraneous 'extern's" 2014-08-05 23:00:04 -07:00
pascal massimino
3ef0f08af5 Merge "vp8enci.h: cosmetics: fix '*' placement" 2014-08-05 22:34:13 -07:00
James Zern
4c6dde37b9 bit_writer: cosmetics: rename kFlush() -> Flush()
Change-Id: I8907927974188bee85ffade1d75d2e50817aa115
2014-08-05 22:14:29 -07:00
James Zern
f7b4c48bba cosmetics: remove some extraneous 'extern's
Change-Id: Ib3f0cff37120c51633387dd1c46592c53ab0ba6d
2014-08-05 22:14:24 -07:00
James Zern
b47fb00ac0 vp8enci.h: cosmetics: fix '*' placement
associate with the type

Change-Id: Icf94f11bf79f6ccee3150e27b228755f8f3f0f37
2014-08-05 22:14:12 -07:00
skal
b5a36cc9ad add -near_lossless [0..100] experimental option
This compresses the uimage using lossless compression and controlable
decimating pre-process.
Code is under WEBP_EXPERIMENTAL_FEATURE while it's being experimented with.

Change-Id: I8b7f4cfcc3c6afc52a556102842bdbb045ed5ee8
2014-08-05 19:17:10 +02:00
James Zern
0524d9e5e8 dsp: detect mips64 & disable mips32 code
Change-Id: Icf68dafd5cf0614ca25b36a0252caa1784ac8059
2014-08-01 21:18:53 -07:00
James Zern
29a9fe222a libwebp 0.4.1
- 7/24/14: version 0.4.1
   This is a binary compatible release.
   * AArch64 (arm64) & MIPS support/optimizations
   * NEON assembly additions:
     - ~25% faster lossy decode / encode (-m 4)
     - ~10% faster lossless decode
     - ~5-10% faster lossless encode (-m 3/4)
   * dwebp/vwebp can read from stdin
   * cwebp/gif2webp can write to stdout
   * cwebp can read webp files; useful if storing sources as webp lossless
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.4.11 (GNU/Linux)
 
 iQIcBAABAgAGBQJT1xp9AAoJEPnD1r24IytdjDEP/3ZOnrWG0OIThlGE6bqgO3oy
 Y5O7RrvzFuPdGEZ1Kl9jDXjzsYY018/+HJmOD3kf+Qt/+F/8hpGH520VuEiJdVIW
 UcvoYaYq9xrmKNqEJx910Vh8TP7wE2T62OJcqKWg2JEczfUWn8WOKjmM5c8N1kJ2
 q6EbpCdWlxcD49L/MavJ5Yfw9jSZAjKzOIxxz0C294iMTK4IcSmeVvdqhkdyh96E
 CABw3o8sJfqB6p+KXjweXcE2KOhvzAWqTRcIogDC0jV/PgOlindf6k0am2FJHvMM
 A+sf/pmD0YKI1vEaXW+Vs6cz6LzvwbIkJSwuzBA7FYHAG5yqTSkQDxTSttw/RwiW
 fUScqHjQVBUqkM5bdOsdYBSDutQKDF2+WfcK5jXFdnydkQi59HKHV2R0K5cXYqfN
 Tu7aMBqFcfGunLlzfKCJcz8SElEmUjG6oAzRZYcdM9dmnR7ypQK17A/GbaysKKOE
 HMmep7uNX25w+6AL7zExnmPPPtSz+kj1SXt9fgldkelDhg1faAgfwXb/N4E+00lA
 1+aJD3gHcR4QnDI4gnKBKHyIktQPfNKMQ6xuL0oyvsalQ/loz08wu0aACcGDFrg4
 uOVVxTqU+pEITuwGcNk228+O2EbMWzzi3+Vhi1v3Gg3jJ3TRB3QN6NohmrsIackL
 4W2V5NoX5i2VizGfLy2g
 =GWd5
 -----END PGP SIGNATURE-----

Merge tag 'v0.4.1'

libwebp 0.4.1
- 7/24/14: version 0.4.1
  This is a binary compatible release.
  * AArch64 (arm64) & MIPS support/optimizations
  * NEON assembly additions:
    - ~25% faster lossy decode / encode (-m 4)
    - ~10% faster lossless decode
    - ~5-10% faster lossless encode (-m 3/4)
  * dwebp/vwebp can read from stdin
  * cwebp/gif2webp can write to stdout
  * cwebp can read webp files; useful if storing sources as webp lossless

* tag 'v0.4.1':
  update ChangeLog
  iosbuild.sh: specify optimization flags
  update ChangeLog
  makefile.unix: add vwebp.1 to the dist target
  update ChangeLog
  gif2webp: dust up the help message
  remove -noalphadither option from README/vwebp.1
  update NEWS for the next release
  update AUTHORS
  bump version to 0.4.1
  restore mux API compatibility
  remove the !WEBP_REFERENCE_IMPLEMENTATION tweak in Put8x8uv
  restore encode API compatibility
  restore decode API compatibility
  gif2webp: fix compile with giflib 5.1.0
  gif2webp: simplify giflib version checking

Change-Id: Icf599f29bc6c0db757bc133aaddb3dbbbc316e08
2014-07-29 18:06:58 -07:00
James Zern
85213b9bbe bump version to 0.4.1
libwebp{,decoder} - 0.4.1
libwebp libtool - 5.1.0
libwebpdecoder libtool - 1.1.0

mux/demux - 0.2.1
libtool - 1.1.0

Change-Id: If593a198f802fd68c7dbbdbe0fc2612dbc44e2df
2014-07-23 17:17:25 -07:00
James Zern
695f80ae25 Merge "restore mux API compatibility" into 0.4.1 2014-07-23 17:11:33 -07:00
James Zern
862d296cf9 restore mux API compatibility
protect WebPMuxSetCanvasSize w/a WEBP_MUX_ABI_VERSION check

Change-Id: I6b01af55ebb4cc4c860d3cbf43be722077896748
2014-07-23 16:13:56 -07:00
skal
8f6f8c5dde remove the !WEBP_REFERENCE_IMPLEMENTATION tweak in Put8x8uv
There's no speed diff, so better remove it altogether

Reported in https://code.google.com/p/webp/issues/detail?id=215

Change-Id: I991330de18bec340029d6df5fed0dfb4337e4662
2014-07-23 14:15:40 -07:00
James Zern
c2fc52e4ec restore encode API compatibility
protect WebPConfigLosslessPreset/WebPMemoryWriterClear w/a
WEBP_ENCODER_ABI_VERSION check

Change-Id: If4debc15fee172a3f18079bc2bd29eb8447bc14b
2014-07-22 22:19:55 -07:00
James Zern
793368e8c6 restore decode API compatibility
protect flip/alpha_dither w/a WEBP_DECODER_ABI_VERSION check

Change-Id: I437a5d5f78800f71b7e7e323faa321f946bf9515
2014-07-22 20:03:52 -07:00
Vikas Arora
d2cc61b7dd Extend MakeARGB32() to accept Alpha channel.
Change-Id: I31b8e2d085000e2e3687a373401e4f655f11fc42
2014-07-21 14:49:38 -07:00
skal
3398d81ac3 Actuate memory stats for PRINT_MEMORY_INFO
Change-Id: If7eac591b5205990ca452ca02b084a908482850a
2014-07-21 13:16:18 -07:00
skal
6c347bbb0c move WebPPictureInit to picture.c
Change-Id: I4b8c352cfd47256d0c3827334a6942c1caf742f6
2014-07-21 14:16:19 +02:00
Pascal Massimino
1549d62067 reorder the YUVA->ARGB and ARGB->YUVA functions correctly
+ rework few loops
+ consolidate few error-checks / error-reporting
+ don't modify picture->colorspace in Import() for ARGB output

Change-Id: Iae6da9b50acc738c59b85c3ee64efbaf6af8bffc
2014-07-18 07:15:54 -07:00
Pascal Massimino
736f2a175e extract colorspace code from picture.c into picture_csp.c
had to refactor few functions here and there.

Change-Id: I86fde6fec7c2fc7eb48f0ecf327dbbd2bd40b9d4
2014-07-16 16:37:26 -07:00
Pascal Massimino
fbadb48026 split monolithic picture.c into picture_{tools,psnr,rescale}.c
Change-Id: Ia5eb5496e4337e5bac8203872c5b014cad21c4f9
2014-07-12 09:13:33 -07:00
James Zern
c76f07ecc2 dec_neon/TransformAC3: initialize vector w/vcreate
replaces {} initialization gnu-ism

Change-Id: I5bedcba1a9c21883207301f07456cc6a843199a0
2014-07-11 15:56:53 -07:00
Urvang Joshi
bb4fc051bf gif2webp: Allow single-frame animations
Some single-frame GIF images have a canvas larger than the frame rectangle. For
such images, we retain the ANMF, ANIM and VP8X chunks in the output WebP file.
This ensures that the full canvas width/height and frame offsets are retained.

Change-Id: I3ebae4893f953984de4072fda0938411de787a29
2014-07-10 15:21:05 -07:00
James Zern
46fd44c104 thread: remove harmless race on status_ in End()
if a thread was still doing work when End() was called there'd be a race
on worker->status_. in these cases, however, the specific value is
meaningless as it would be >= OK and the thread would have been shut
down properly, but we'll check 'impl_' instead to avoid any potential
TSan/DRD reports.

Change-Id: Ib93cbc226a099f07761f7bad765549dffb8054b1
2014-07-08 20:32:29 -07:00
James Zern
6781423b7d configure: check for __builtin_bswapXX()
defines HAVE_BUILTIN_BSWAP16/32/64
updated endian_inl.h to have a non-configure fallback for gcc and clang
BSwap16() now uses __builtin_bswap16 if available

Change-Id: Ia04ee07b39303c4b247df96d84f298fb8a81f389
2014-07-05 12:35:13 -07:00
James Zern
6422e683af VP8LFillBitWindow: enable fast path for 32-bit builds
also reduce the load size from 64 to 32 bits as the top 32 bits are
being shifted away in the operation.

the change is neutral speed-wise on x86_64 as is the change in load size
on x86, but it gives a slight improvement on 32-bit arm.
x86 is improved ~13%, 32-bit arm ~3.7%
aarch64 is untested but will likely benefit as well.

Change-Id: Ibcb02a70f46f2651105d7ab571afe352673bef48
2014-07-04 14:42:47 -07:00
James Zern
4f7f52b2a1 VP8LFillBitWindow: respect WEBP_FORCE_ALIGNED
Change-Id: I23eddf01590de002efc21d8c7acc545a08fc3e48
2014-07-04 13:53:29 -07:00
James Zern
e458badcc3 endian_inl.h: implement htoleXX with BSwapXX
+ s/htole(16|32)/HToLE$1/ to avoid any name conflicts

Change-Id: Ic1c84711557e50f73d83ca5aa2b3992ac6738216
2014-07-04 12:16:36 -07:00
James Zern
f2664d1aab endian_inl.h: add BSwap16
+ use it in VP8LoadNewBytes()

Change-Id: I701d3652dc0cbd553852978702ef68c2657bca1c
2014-07-04 12:16:28 -07:00
James Zern
6fbf5345e3 Merge "configure: add --enable-aligned" 2014-07-04 00:12:28 -07:00
James Zern
dc0f479d6a configure: add --enable-aligned
forces aligned memory reads (via memcpy) in the VP8 bit reader, useful
for platforms that don't support unaligned loads.

Change-Id: Ifa44a9a1677fbdc6a929520f9340b7e3fcbd6692
2014-07-03 23:30:09 -07:00
Pascal Massimino
257adfb0be remove experimental YUV444 YUV422 and YUV400 code
(never used)

Change-Id: I12a886703592133939607df05132e9b498f916c1
2014-07-03 22:26:25 -07:00
James Zern
380cca4f2c configure.ac: add AC_C_BIGENDIAN
this defines WORDS_BIGENDIAN, replacing uses of
__BIG_ENDIAN__/__BYTE_ORDER__ with it

+ fixes lossless BGRA output with big-endian toolchains
  that do not define __BIG_ENDIAN__ (codesourcery mips gcc)

Change-Id: Ieaccd623292d235343b5e34b7a720fc251c432d7
2014-07-03 18:15:50 -07:00
James Zern
ee70a90187 endian_inl.h: add BSwap64
Change-Id: I66672b770500294b8f4ee8fa4bf1dfff1119dbe6
2014-07-03 13:35:34 -07:00
James Zern
47779d46c8 endian_inl.h: add BSwap32
Change-Id: I96e3ae49659307024415d64587e6312888a0070f
2014-07-03 13:28:13 -07:00
James Zern
d5104b1ff6 utils: add endian_inl.h
moves the following to this header:
- htole*() definitions from bit_writer.c
- __BIG_ENDIAN__ fallback define from bit_reader_inl.h

Change-Id: I7fff59543f08a70bf8f9ddac849b72ed290471b1
2014-07-03 13:07:14 -07:00
Vikas Arora
516971b136 lossless: Remove unaligned read warning
(typecast uint32 pointer to uint64).
The proposed change is little (0.05%) slower but avoids uint32 to uint64
pointer conversion.

Change-Id: I6b8828077ea1324fabd04bfa7e7439e324776250
2014-07-02 20:55:27 -07:00
James Zern
e59f53600f neon: normalize vdup_n_* usage
with constants, prefer this over vmov_n_* or vcreate_*

Change-Id: Ia84b2a82faea58e2626211a7e2257e0ba4af358a
2014-07-01 00:55:05 -07:00
James Zern
6ee7160dd2 Merge changes I0da7b3d3,Idad2f278,I4accc305
* changes:
  neon: add INIT_VECTOR4
  neon: add INIT_VECTOR3
  neon: add INIT_VECTOR2
2014-07-01 00:48:38 -07:00
skal
abc02f2492 Merge "fix (uncompiled) typo" 2014-07-01 00:29:19 -07:00
James Zern
bc03670f01 neon: add INIT_VECTOR4
used to initialize NxMx4 vector types
replaces initialization via '{{ }}' gnu-ism.

Change-Id: I0da7b3d321f3d48579b7863fb2e4d3f449ae7f5e
2014-07-01 00:18:23 -07:00
James Zern
6c1c632b03 neon: add INIT_VECTOR3
used to initialize NxMx3 vector types
replaces initialization via '{{ }}' gnu-ism.

Change-Id: Idad2f278ab104cf2cc650517194258ce3cfb37b4
2014-06-30 23:53:23 -07:00
James Zern
dc7687e51b neon: add INIT_VECTOR2
used to initialize NxMx2 vector types
replaces initialization via '{{ }}' gnu-ism.

Change-Id: I4accc305c7dd4c886b63c22e38890b629bffb139
2014-06-30 23:52:42 -07:00
skal
4536e7c49c add WebPMuxSetCanvasSize() to the mux API
previously, the final canvas size was adjusted tightly from the
animation frames. Now, it can be specified separately (to be larger, in particular).

calling WebPMuxSetCanvasSize(mux, 0, 0) triggers the 'adjust tightly' behaviour.
This can be useful after calling WebPMuxCreate() if further image addition
is expected.

-> Fixed gif2webp accordingly.

also: made WebPMuxAssemble() more robust by systematically zero-ing WebPData.

Change-Id: Ib4f7eac372cf9dbf6e25cd686a77960e386a0b7f
2014-06-30 07:00:49 +02:00
skal
824eab1086 fix (uncompiled) typo
output was not properly descaled when compiled
without USE_GAMMA_COMPRESSION

Change-Id: I5282c9573cfcfbef4be22b6b3cdec7e16cad9c1c
2014-06-27 17:10:25 +02:00
Pascal Massimino
1f3e5f1e60 remove unused 'shift' argument and QFIX2 define
this will remove a warning about the shift amount not being
an immediate (=constant).

Change-Id: Ie9a00fefdb9a07ec8994fb113f24234518bc878a
Also: fix the NULL sharpen argument mismatch.
2014-06-26 00:44:12 -07:00
James Zern
1da3d46138 VP8LoadNewBytes: use __builtin_bswap32 if available
mostly to balance the use of bswap64, some gcc platforms are already
interpreting the default case the same

Change-Id: Icf860f55b3f16bea349a7d721e6d6abeeb4e5cf3
2014-06-24 23:24:36 -07:00
Pascal Massimino
25aaddc84c rename interface -> winterface
to avoid name clash on win32

Change-Id: Ia4ad3d4528df4652bab803f9a9544e21e6c4b177
2014-06-23 14:21:22 -07:00
skal
5584d9d2fc make WebPSetWorkerInterface() check its arguments
Change-Id: I522c58cfe05e864a50cacb58bdfa14d5369c6d60
2014-06-23 07:39:56 +02:00
James Zern
a9cf31913c cosmetics: update thread.h comments
WebPWorker*() are now part of WebPWorkerInterface; refer to them with
unadorned names.

Change-Id: Iae1dd59f1e545cba6dd8c18f26ba60eb9a84419b
2014-06-19 19:31:10 -07:00
levytamar82
27bfeee43a QuantizeBlock SSE2 Optimization:
Another store to load forward block was detected coming from the function
FTransform.
FTransform save the output data 4 times 8 bytes each. when this data is
later being loaded by the QuantizeBlock function in one chunk of 16 bytes
that caused a store to load forward block.
The fix was done in the FTransform function where each two consecutive 8 bytes
were merged into one 16 bytes register and saved into the memory.
This fix gives ~21% function level gain and 1.6% user level gain.

Change-Id: Idc27c307d5083f3ebe206d3ca19059e5bd465992
2014-06-18 16:22:00 -07:00
skal
c5c6b408b1 Merge "add alpha dithering for lossy" 2014-06-18 07:32:24 -07:00
James Zern
6e93317f5b muxread: fix out of bounds read
ChunkVerifyAndAssign() expects to have at least 8 bytes to work with,
but was only checking for the presence of 4.

Change-Id: I8456b15d872de24a90c1e8fbfba463391ced5c7f
2014-06-17 12:53:28 -07:00
skal
bbe32df1e3 add alpha dithering for lossy
new options:
 dwebp -alpha_dither
 vwebp -noalphadither

When the source was marked as quantized, we use a threshold-averaging
filter to smooth the decoded alpha plane.
Note: this option forces the decoding of alpha data in one pass, and
might slow the decoding a bit.

The new field in WebPDecoderOptions struct is 'alpha_dithering_strength'
(0 by default, means: off). Max strength value is '100'.

Change-Id: I218e21af96360d4781587fede95f8ea4e2b7287a
2014-06-14 00:06:16 +02:00
skal
790207679d Merge "make error-code reporting consistent upon malloc failure" 2014-06-13 00:25:30 -07:00
skal
77bf4410f7 make error-code reporting consistent upon malloc failure
Sometimes, the error-code was not set correctly.
We now return OUT_OF_MEMORY everytimes it's appropriate
(tested using MALLOC_FAIL_AT mechanism)

Took the opportunity to clean-up the code and dust the error
code returned (some were erroneously set to INVALID_CONFIGURATION)

Change-Id: I56f7331e2447557b3dd038e245daace4fc82214c
2014-06-13 08:45:12 +02:00
James Zern
7a93c000ee **/Makefile.am: remove unused AM_CPPFLAGS
only 1 of <lib>_CPPFLAGS and AM_CPPFLAGS is used, with the former
getting precedence when it's defined. configure's DEFAULT_INCLUDES is
covering what's necessary given the include paths are all source
relative.

Change-Id: I7d14076acd266b28a88a3d92bcc3d7165284d5f3
2014-06-12 11:59:05 -07:00
skal
24e3080571 Add an interface abstraction to the WebP worker thread implementation
This allows custom implementations of threading mecanism.

Patch by Leonhard Gruenschloss.

Change-Id: Id8ea5917acd2f24fa8bce79748d1747de2751614
2014-06-12 11:35:44 +02:00
James Zern
32b3137936 configure: move config.h to src/webp/config.h
this change has the side-effect of using directory names in the
include, silencing a lint warning.

Change-Id: Ib91cf63a90534e32fadfa5c2372bfdb29f854d02
2014-06-10 23:42:00 -07:00
James Zern
90090d99b5 Merge changes I7c675e51,I84f7d785
* changes:
  configure: test for -msse2
  rename upsampling_mips32.c to yuv_mips32.c
2014-06-10 16:15:21 -07:00
skal
69fce2ea78 remove the special casing for res->first in VP8SetResidualCoeffs
if res->first = 1, coeffs[0]=0 because of quant.c:749 and line
added at quant.c:744
So, no need for the extra case.
Going forward, TrellisQuantizeBlock() should also be calling
a variant of VP8SetResidualCoeffs() to set the 'last' field.

also: fixes a warning for win64
    + slight speed-up

Change-Id: Ib24b611f7396d24aeb5b56dc74d5c39160f048f0
2014-06-08 06:40:22 +02:00
James Zern
6e61a3a905 configure: test for -msse2
+ add a WEBP_HAVE_SSE2 to dsp.h

not all 32-bit toolchain configurations will have sse2 enabled by
default

Change-Id: I7c675e511581f93cf55c79f960fa7efa2df4987e
2014-06-07 19:44:08 -07:00
James Zern
b9d2efc629 rename upsampling_mips32.c to yuv_mips32.c
matches yuv_sse2 added in;
bdfeeba dsp/yuv: move sse2 functions to yuv_sse2.c

Change-Id: I84f7d7858ca6851c956e8366a7c76b45070dcbc3
2014-06-07 12:35:47 -07:00
James Zern
bdfeebaa01 dsp/yuv: move sse2 functions to yuv_sse2.c
Change-Id: I2f037ff18e7cf07e8801f49b3a89c1e36ef73000
2014-06-05 23:52:54 -07:00
pascal massimino
46b32e861a Merge "configure: set WEBP_HAVE_AVX2 when available" 2014-06-05 02:57:42 -07:00
pascal massimino
88305db4fc Merge "VP8RandomBits2: prevent signed int overflow" 2014-06-05 01:46:42 -07:00
James Zern
73fee88c4a VP8RandomBits2: prevent signed int overflow
'diff' at its largest may be INT_MAX; << 1 of anything at or above
1 << 30 will overflow.

Change-Id: Idb2b5a9b55acc2f6d5e32be8baaebee3f89919ad
2014-06-04 23:19:03 -07:00
James Zern
db4860b355 enc_sse2: prevent signed int overflow
_mm_movemask_epi8 returns a 16-bit mask; << 16 can overflow a signed
int.

Change-Id: Ia0bb0804fe548fb9b0edb3695e82727506066cda
2014-06-04 23:18:22 -07:00
James Zern
230a055501 configure: set WEBP_HAVE_AVX2 when available
this is used to set WEBP_USE_AVX2 in files where the build flag won't be
used, i.e., dsp/enc.c, which enables VP8EncDspInitAVX2() to be called

Change-Id: I362f4ba39ca40d3e07a081292d5f743c649d9d7f
2014-06-03 23:29:23 -07:00
skal
a2ac8a420e restore original value_/range_ field order
no speed change, just for coherency

Change-Id: Iaa395bca24f33a14b68ba6920b838ef87d0d0db6
2014-06-03 09:36:56 +02:00
James Zern
5e2ee56fdd Merge "remove libwebpdspdecode dep on libwebpdsp_avx2" 2014-06-03 00:28:54 -07:00
James Zern
61362db57c remove libwebpdspdecode dep on libwebpdsp_avx2
it's encode only, libwebpdecoder doesn't need the symbols

Change-Id: I5633dd2017a96e60068ae5384f1ba27898d29f83
2014-06-03 00:05:56 -07:00
Pascal Massimino
42c447aeb0 Merge "lossy bit-reader clean-up:" 2014-06-02 23:53:00 -07:00
James Zern
479ffd8b5d Merge "remove unused #include's" 2014-06-02 23:07:20 -07:00
James Zern
9754d39a4e Merge "strong filtering speed-up (~2-3% x86, ~1-2% for NEON)" 2014-06-02 23:06:18 -07:00
skal
158aff9bb9 remove unused #include's
Change-Id: Icd91a4b6a0bde49145f57e3e74a997822c45792c
2014-06-03 08:01:05 +02:00
Pascal Massimino
09545eeadc lossy bit-reader clean-up:
* remove LEFT/RIGHT_JUSTIFY distinction. It's all RIGHT_JUSTIFY now.
* simplify VP8GetSigned(), and add some masking branch-less code. Much
  faster on ARM (~13% speed-up). 8% on x86-64, 5% on MacBook.
* split critical implementation into separate bit_reader_inl.h file that
  is only included where needed (vp8.c / tree.c / bit_reader.c)
* bumped BITS value from 16 to 24 for x86-32b too, since it's a bit faster.

Change-Id: If41ca1da3e5c3dadacf2379d1ba419b151e7fce8
2014-06-03 07:46:55 +02:00
skal
ea8b0a171d strong filtering speed-up (~2-3% x86, ~1-2% for NEON)
Extract loop invariant and avoid storing/loading samples
if they can be re-used. This is particularly interesting when
a transpose is involved (HFilter16i).

Change-Id: I93274620f6da220a35025ff8708ff0c9ee8c4139
2014-06-03 07:14:23 +02:00
skal
6679f8996f Optimize VP8SetResidualCoeffs.
Brings down WebP lossy encoding timings by 5%

Change-Id: Ia4a2fab0a887aaaf7841ce6d9ee16270d3e15489
2014-06-03 06:44:04 +02:00
James Zern
4dfa86b29c dsp/cpu: NaCl has no support for xgetbv
or the raw opcode; fixes:
  934ed4: unrecognized instruction

Change-Id: I981870baf0e8b03bf40144ea8ec25eff140d5bc3
2014-05-29 23:02:23 -07:00
skal
c9b340a279 fix missing WebPInitAlphaProcessing call for premultiplied colorspace output
(lossless only)

Change-Id: Ic2d01c8cf9bc1082f07f348733461eb2ee30288a
2014-05-28 10:44:05 +02:00
pascal massimino
57897bae09 Merge "lossless_neon: use vcreate_*() where appropriate" 2014-05-28 01:36:13 -07:00
pascal massimino
6aa4777b39 Merge "(enc|dec)_neon: use vcreate_*() where appropriate" 2014-05-28 01:34:56 -07:00
skal
0d346e418d Always reinit VP8TransformWHT instead of hard-coding
Change-Id: I2012749ed29bd166d2a96555372f0d9baa784385
2014-05-28 10:21:07 +02:00
James Zern
bf0e003067 lossless_neon: use vcreate_*() where appropriate
this is more portable than {} initialization.
more involved cases are left for a follow-up.

Change-Id: If7e111864f287ea0a5de6311454aeda37afbb52a
2014-05-27 16:27:46 -07:00
James Zern
9251c2f6d2 (enc|dec)_neon: use vcreate_*() where appropriate
this is more portable than {} initialization.
more involved cases are left for a follow-up.

Change-Id: If8783423d17e90694b168a64ba313ed62ce2cc17
2014-05-27 16:26:56 -07:00
skal
399b916d27 lossy decoding: correct alpha-rescaling for YUVA format
The luminance needs to be pre- and post- multiplied by
the alpha value in case of rescaling, for proper averaging.

Also:
- removed util/alpha_processing and moved it to dsp/
- removed WebPInitPremultiply() which was mostly useless
and merged it with the new function WebPInitAlphaProcessing()

Change-Id: If089cefd4ec53f6880a791c476fb1c7f7c5a8e60
2014-05-27 15:27:13 -07:00
Pascal Massimino
fdbcd44dd3 simplify VP8LInitBitReader()
gcc was generating very complex code, one for each case of br->len_ values!

also, pretty-fy the mask constants

Change-Id: If62b1e8266f3fe5334517305113038d2ea8a6b42
2014-05-22 21:44:16 -07:00
James Zern
515e35cfb1 Merge "add stub dsp/enc_avx2.c" 2014-05-22 18:28:38 -07:00
skal
a05dc1402c SSE2: yuv->rgb speed-up for point-sampling
- use statically initialized tables (if WEBP_YUV_USE_SSE2_TABLES is defined)
 - use SSE2 row conversion for yuv->ARGB / RGBA / ABGR / RGB / BGR
 - clean-up and harmonize the WebpUpsamplers[] usage.

Change-Id: Ic5f3659a995927bd7363defac99c1fc03a85a47d
2014-05-22 09:56:47 +02:00
James Zern
178e9a69ae add stub dsp/enc_avx2.c
VP8EncDspInitAVX2 is included in sse2 builds for now, later a configure
flag should be added to avoid the stub when avx2 is unavailable/disabled

Change-Id: I6127b687c273f46f41652aaf8e3b86ae3cfb8108
2014-05-22 00:31:46 -07:00
James Zern
e46a247c87 cpu: fix check for __cpuidex availability
__cpuidex was added in VS2008 /SP1/

Change-Id: Ie49b00b0246bd6537c0ed583412f17d6fd135baa
2014-05-21 22:59:47 -07:00
skal
176fda2650 fix the bit-writer for lossless in 32bit mode
Sometimes, we can write 18bit or more at time, and it would
overflow the 32bit accumulator.

Also clarified the num-bits limitations (and exposed
VP8L_MAX_NUM_BIT_READ in bit_reader.h)

fixes http://code.google.com/p/webp/issues/detail?id=200

Seems a bit faster (use of local fields for bits_ / used_)

also: added the __QNX__ bswap while at it.

Change-Id: I876db93a931db15b083cf1d838c70105effa7167
2014-05-22 07:19:22 +02:00
James Zern
541784c710 dsp.h: add a check for AVX2 / define WEBP_USE_AVX2
Change-Id: I90cc870f0bb4426af701779c367587dc2ae79c8b
2014-05-21 20:46:28 -07:00
James Zern
bdb151ee80 dsp/cpu: add AVX2 detection
currently unused.
https://software.intel.com/en-us/articles/how-to-detect-new-instruction-support-in-the-4th-generation-intel-core-processor-family
http://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-optimization-manual.pdf

Change-Id: I314200f890c58b9a587b902b214f90deb95f0579
2014-05-20 22:48:54 -07:00