Commit Graph

42 Commits

Author SHA1 Message Date
Vikas Arora
5f0cfa80ff Do a binary search to get the optimum cache bits.
This speeds up the lossless encoder by a bit (1-2%), without impacting the
compression density.

Change-Id: Ied6fb38fab58eef9ded078697e0463fe7c560b26
2014-03-13 10:30:32 -07:00
Vikas Arora
b33e8a05ee Refactor code for HistogramCombine.
Refactor code for HistogramCombine and optimize the code by calculating
the combined entropy and avoid un-necessary Histogram merges.

This speeds up lossless encoding by 1-2% and almost no impact on compression
density.

Change-Id: Iedfcf4c1f3e88077bc77fc7b8c780c4cd5d6362b
2014-03-03 13:50:42 -08:00
James Zern
a42ea9742a cosmetics: backward_references.c: reindent after a7d2ee3
a7d2ee3 Optimize cache estimate logic.

Change-Id: I81dd1eea49f603465dc5f3afae8a101e5205e963
2014-02-11 15:52:22 -08:00
Vikas Arora
a7d2ee39be Optimize cache estimate logic.
Optimize 'VP8LCalculateEstimateForCacheSize' for lower quality ranges (Q < 50).
The entropy is generally lower for higher cache_bits, so start searching from
higher cache_bits and settle for a local minima, instead of evaluating all
values.

This speeds up the lossless encoding at lower qualities by 10-15%.

Change-Id: I33c1e958515a2549f2e6f64b1aab3f128660dcec
2014-02-11 10:59:01 -08:00
Scott Talbot
391316fee2 Don't dereference NULL, ensure HashChain fully initialized
Found by clang's static analyzer, they look validly uninitialized
to me.

Change-Id: I650250f516cdf6081b35cdfe92288c20a3036ac8
2014-02-03 21:16:59 -08:00
James Zern
4931c3294b cosmetics: fix some typos
Change-Id: I0d6efebd817815139db5ae87236fd8911df4d53c
2013-11-26 19:21:14 -08:00
skal
ff379db317 few % speedup of lossless encoding
mostly visible for method 4 and up

Change-Id: I1561d871bc055ec5f7998eb193d927927d3f2add
2013-11-12 00:09:45 +01:00
Vikas Arora
e081f2f359 Pack code & extra_bits to Struct (VP8LPrefixCode).
Also created variant VP8LPrefixEncodeBits that returns the
code & extra_bits only.
There's no impact on compression density and compression speed.

Change-Id: I2cafdd3438ac9270cd72ad9d57b383cdddfdfa4c
2013-08-12 11:56:42 -07:00
Vikas Arora
69257f70df Create LUT for PrefixEncode.
This speeds up lossless compression by 5%.
Change-Id: Ifd114b1d9850dc3aac74593809e7d48529d35e3d
2013-08-05 10:20:18 -07:00
Vikas Arora
7d60bbc6d9 Speed up HashChainFindCopy function.
Speed up HashChainFindCopy by optimizing on number of calls to
FindMatchLength method.
This change speeds up the lossless & lossy (Alpha) encoding by 20%
without loss of compression density.
At method=3, lossy (Alpha) compression speed (and density) remains
unchanged, as at that settings, the costly Backward Refs method is not
called

Change-Id: Ia1797148e9e4ee2787011837fa248afbae2242cb
2013-07-16 19:58:18 -07:00
Vikas Arora
6674014077 Speedup Alpha plane encoding.
Disable costly 'BackwardReferencesTraceBackwards' for encoding Alpha plane.
Increase the threshold for triggering 'BackwardReferencesTraceBackwards' to
quality 25 and above. Also lower the Alpha quality (at method 3) to be
lesser than this threshold (25).

Change-Id: Ic29fb2e6943472c564223df9fe099b19ccda0f31
2013-07-12 11:02:13 -07:00
James Zern
d640614d54 update copyright text
rather than symlink the webm/vpx terms, use the same header as libvpx to
reference in-tree files

based on the discussion in:
https://codereview.chromium.org/12771026/

Change-Id: Ia3067ecddefaa7ee01550136e00f7b3f086d4af4
2013-06-06 23:09:14 -07:00
Vikas Arora
8eae188a62 WebP-Lossless encoding improvements.
Lossy (with Alpha) image compression gets 2.3X speedup.
Compressing lossless images is 20%-40% faster now.

Change-Id: I41f0225838b48ae5c60b1effd1b0de72fecb3ae6
2013-05-08 17:22:11 -07:00
Vikas Arora
0478b5d214 Tune Lossless compression for lower qualities.
This is required for WebP lossy+Alpha images, where Alpha channel is taking
60-70% of the compression (CPU) cycles.

Also evaluated on 1000 PNG corpus and overall compression speed
is 15-40% better for lossy (PNG+Alpha) compression.
The pure lossless compression numbers are almost same (or little
better) with this change.

Change-Id: I9e5ae7372ed6227a9a5b64cd9cff84c747195a57
2013-03-12 14:17:28 -07:00
skal
23c0f354a6 fix missing intptr_t->int cast for MSVC
Change-Id: I39acdfbe287ef7b7f615d59e0729aab161189bf9
2013-02-06 17:51:10 +01:00
skal
043076e2ef Merge "speed-up lossless in BackwardTrace" 2013-02-05 10:46:06 -08:00
skal
f3a44dcd83 remove one malloc from TraceBackwards()
Change-Id: I4f77c0240ca2ece86e8beab11d02c74277409921
2013-02-05 19:43:43 +01:00
skal
0fc1a3a072 speed-up lossless in BackwardTrace
we special-case code=2 (with a later TODO to adapt this on quality)

Change-Id: I93d43f5b3f8f1ef9f211cce253bb4b415918ee57
2013-02-05 19:42:23 +01:00
skal
0d19fbff51 remove some -Wshadow warnings
these are quite noisy, but it's not a big deal to remove
them.

Change-Id: I5deb08f10263feb77e2cc8a70be44ad4f725febd
2013-01-22 23:06:28 +01:00
vikas arora
ab5ea217f7 Tune Lossless encoder
- Changed the dynamic range where more aggressive
  (BackwardReferencesTraceBackward) heuristic is run from quality > 10
  (instead of quality > 25).
- Limit the backward-ref Window size to 16*width & 256*width for lower
  qualities ([0, 25[ & [25, 50[) respectively, instead of 1M window.
- Evaluate the params for HashChainFindCopy outside this function call
  and pass it, instead of recomputing them for every call.

Change-Id: If9eedfc14b978e7632d7cf69c96186e2910b0554
2012-10-30 17:30:17 -07:00
James Zern
e855208c16 fix double to float conversion warning
introduced in:
 a792b91 fix the -g/O3 discrepancy for 32bit compile

Change-Id: I6a77223f237527eda4ee1d6eaa993351bd74f1d6
2012-10-08 18:27:27 -07:00
Vikas Arora
7b3eb372ad Tune lossless compression to get better gains.
Tune compression heuristics to get better gains across wide quality range.

Change-Id: Ic342d4dbcf83fe2086a34e5c184aef0714109430
2012-10-05 09:58:29 -07:00
skal
a792b913bd fix the -g/O3 discrepancy for 32bit compile
in debug mode, some float operations see their intermediate
values stored in memory rather than staying in the FPU (which
is 80bit precision).

Several fixes are possible (breaking long calculations into
atomic steps for instance), but simpler of all is just about
turning the cost[] array into float* instead of double*.

The code is a tad faster, and i didn't see any major output
size difference.

Change-Id: Icf1f833e15f8ee4ecc7f9a521d07fdc96ef711aa
2012-10-03 15:15:58 +02:00
Pascal Massimino
323dc4d9b9 remove use of log2(). Use VP8LFastLog2() instead.
Order-by-cost mostly unchanged (up to a scaling constant 1/log(2))
(except for few minor diff in < 2% of cases)

+ remove unused field cost_mode->cache_bits_

Change-Id: I714f8ab12f49a23f5d499a64c741382c9b489a3e
2012-08-02 00:08:58 -07:00
Pascal Massimino
bff34ac1ca harness some malloc/calloc to use WebPSafeMalloc and WebPSafeCalloc
quite a large security sweep.

Change-Id: If150dfbb46e6e9b56210473a109c8ad6ccd0cea4
2012-08-01 12:06:04 -07:00
Pascal Massimino
906be65744 rationalize use of color-cache
* ~1-4% faster
* if it's not used, don't use it
* remove the special handling of cache_bits = 0
* remove some tests in the loops

Change-Id: I19d87c3ca731052ff532ea8b2d8e89816507b75f
2012-08-01 00:32:12 -07:00
Pascal Massimino
cdf885c67c shuffle variables in HashChainFindCopy
might lead to observable speed-up on some compiler/arch

Change-Id: I6c89ec9cd4f490e7e22b790785f80097a47d27b5
2012-07-16 19:08:59 -07:00
Pascal Massimino
78f3e34504 Enable lossless encoder code
Remove USE_LOSSLESS_ENCODER compile flag
Update Makefile.am and makefile.unix

Change-Id: If7080c4d8f37994c7c784730c5e547bb0a851455
2012-06-13 00:26:58 -07:00
Vikas Arora
7c6898051e Fix asserts in Palette and BackwardReference code.
Fix inequality assertion on number of palette colors.
Fix inequality assertion test in BackwardReferencesHashChainFollowChosenPath.

Change-Id: Ie3242f1bbeaf96db91b839b6732ccce2634cebf3
2012-06-11 12:27:59 +05:30
Vikas Arora
0424b1ef3b Rebase default encoding settings.
Updated histo_bits to 3 from 4 and changed the quality threshold for inner loop for HashChainFindCopy.
Impact: 0.5%-0.8% better bpp with 15%-20% hit on encoding throughput
at default encoding settings.

Change-Id: I316ef88403148b1e19036fa0817d944eb0301255
2012-06-08 09:18:15 +05:30
Pascal Massimino
1875d926e7 trap two unchecked error conditions
CostModelBuild() and TrackBackwards() returns weren't checked

+ code clean-up
+ de-inline VP8LBackwardRefs non-critical methods
+ shuffle the .h around to group things together
+ extract some constants as #define's
+ fixed the "if (!(cc_init = ...)) {...}" constructs
+ removed some unneeded VP8L prefixes

Change-Id: Ic634cb87bc6b2033242d3e8e8731fab4c134f327
2012-05-25 02:52:44 -07:00
Urvang Joshi
e75dc80516 Move some more defines to format_constants.h
Also remove some duplicate const/defines.

Change-Id: I0ec48866b874f546022d72e938fb65669b0b3211
2012-05-24 17:46:01 +05:30
James Zern
37a77a6bf4 remove some variable shadowing
Change-Id: I4348253ec6b50639095b22c4745dc26da0904466
2012-05-15 14:04:24 -07:00
Vikas Arora
e7167a2b95 Provide one entry point for backward references.
The new method VP8LGetBackwardReferences hides internal
heuristics used for choosing RLE or LZ77 based refs.
- Tuned VP8LHashChainFindCopy for better compression at higher Q.
- Refactored code.
- Removed the unused method VP8LVerifyBackwardReferences.

Change-Id: Ibb7bb072bab5a49a001577a20d88226f52e6c663
2012-05-09 12:13:06 +05:30
Pascal Massimino
1a210ef1a9 big code clean-up and refactoring and optimization
* de-inline some function
* make VP8LBackwardRefs be more like a vectorwith max capacity
* add bit_cost_ field to VP8LHistogram
* general code simplifications
* remove some memmov() from HistogramRefine
* simplify HistogramDistance()
...

Change-Id: I16904d9fa2380e1cf4a3fdddf56ed1fcadfa25dc
2012-05-07 14:25:29 -07:00
Vikas Arora
b39e7487a7 Reducing emerging palette size from 11 to 9 bits.
This is required to reduce memory used to construct histo_image.

Change-Id: I491a06e10a3e3f3d8a00ecec286394378283ffea
2012-05-07 14:25:19 -07:00
Urvang Joshi
e491729905 A quick pass of cleanup in backward reference code
const correctness, renaming, cosmetics etc.

Change-Id: I432befbb22f0eafd9a613f5f632398b6ef03c0f6
2012-05-07 14:25:05 -07:00
Vikas Arora
c6ac4dfbb4 Run TraceBackwards for higher qualities.
Also reduce the iteration count in function VP8LHashChain_FindCopy.

Change-Id: I73e3811e142e81314515587fd655ab3bfa74d099
2012-05-07 14:25:00 -07:00
Vikas Arora
84547f540c Add EncodeImageInternal() method.
Most of changes in enc/vp8l.c is cherry-picked from src/lossless/encode.c

Change-Id: I27938cb2590eccbfe1db0a454343e856bd483e75
2012-05-07 14:24:25 -07:00
Urvang Joshi
6b38378acb Guard the lossless encoder (in flux) under a flag
Change-Id: I6dd8fd17089c199001c06b1afde14233dc3e3234
2012-05-07 14:24:23 -07:00
Vikas Arora
32714ce3be Add VP8L prefix to backward ref & histogram methods.
Change-Id: I8c14fb219a1d7830d3244aa780c91c9964867330
2012-05-07 14:24:17 -07:00
Vikas Arora
bc7037465d Add backward_ref, histogram & huffman encode modules from lossless.
Change-Id: Iac056d27972956782defa182caa3ea400cdb77f8
2012-05-07 14:24:14 -07:00