and 15% faster MultARGBRow()
by switching to formulae:
X / 255 = (X + 1 + (X >> 8)) >> 8 for any 16bit value X.
(X / 255 + .5) = (XX + (XX >> 8)) >> 8, with XX = X + 128
Change-Id: Ia4a7408aee74d7f61b58f5dff304d05546c04e81
after:
fbba5bc optimize predictor #1 in plain-C For some reason, gcc has hard
time inlining this one...
Change-Id: I2e2416593acd4c9d14958d8757bfd284d999100b
For some reason, gcc has hard time inlining this one...
Also optimize predictor #0 and #1 for encoding, so we don't have to
call the generic pointers VP8LPredictors[...]
Change-Id: I1ff31e3b83874b53f84fe23487f644619fd61db9
Average3 created a slowdown of 1-2% in lossless decoding.
Average4 created a slowdown of 2-3% in lossless decoding.
Change-Id: Ic2e62cdd83fc897887ec2bf41ea7cadbada84fe5
...instead of the pointers stored in the array.
Should be faster (inlined) and safer.
Also: suffix explicitly the functions with _SSE2
Change-Id: Ie7de4b8876caea15067fdbe44abfedd72b299a90
Before, a first thread could enter VP8LDspInitSSE2, set
VP8LPredictorsAdd to an SSE2 version BEFORE another thread
would do the memcpy from VP8LPredictorsAdd to VP8LPredictorsAdd_C
thus leading to a C version actually being the SSE2 one (which
would then create an infinite recursion in the SSE2 predictors
at execution).
Change-Id: I224f4ceab31d38f77a1375a7e2636a6014080e3a
Benchmarks from vrabaud@:
8BIT/GRAY corpus speed: faster: -4.3 % , corpus size: unchanged
skal/sources_png_skal corpus speed: faster: -5.2 % , corpus size: unchanged
images/png_rgb corpus speed: faster: -5.1 % , corpus size: unchanged
images/lpcb corpus speed: unchanged, corpus size: unchanged
images/png_big corpus speed: faster: -1.7 % , corpus size: unchanged
images/png_doc corpus speed: unchanged, corpus size: unchanged
images/png_1bit corpus speed: faster: -1.2 % , corpus size: unchanged
images/jpeg_small corpus speed: unchanged, corpus size: unchanged
images/icip_core1 corpus speed: unchanged, corpus size: unchanged
images/png_gray corpus speed: faster: -2.5 % , corpus size: unchanged
images/jpeg_high_quality corpus speed: faster: -4.0 % , corpus size: unchanged
images/jpeg corpus speed: faster: -2.3 % , corpus size: unchanged
images/png_translucent corpus speed: faster: -2.8 % , corpus size: unchanged
images/gif corpus speed: faster: -1.4 % , corpus size: unchanged
images/png_opaque corpus speed: faster: -2.8 % , corpus size: unchanged
images/png_rgb_opaque corpus speed: unchanged, corpus size: unchanged
images/png_indexed corpus speed: faster: -2.0 % , corpus size: unchanged
images/all corpus speed: faster: -1.5 % , corpus size: unchanged
images/png_small corpus speed: unchanged, corpus size: unchanged
images/png corpus speed: unchanged, corpus size: unchanged
images/gif_still corpus speed: faster: -1.6 % , corpus size: unchanged
Change-Id: I69fe11baa188c5d32cbc77a84b8c0deae13d792b
Roughly, if both the source and the reference areas are
darker too dark (R/G/B <= ~6), they are ignored.
One caveat: SSIM calculation won't work for U/V planes,
which are 128-centered and not related to luminance.
But WebPPlaneDistortion() enforces the conversion to RGB,
if needed.
Change-Id: I586c2579c475583b8c90c5baefd766b1d5aea591
* prevent 64bit overflow by controlling the 32b->64b conversions
and preventively descaling by 8bit before the final multiply
* adjust the threshold constants C1 and C2 to de-emphasis the dark
areas
* use a hat-like filter instead of box-filtering to avoid blockiness
during averaging
SSIM distortion calc is actually *faster* now in SSE2, because of the
unrolling during the function rewrite.
The C-version is quite slower because still un-optimized.
Change-Id: I96e2715827f79d26faae354cc28c7406c6800c90
SSIM results are incompatible with previous version!
We're now averaging the SSIM value for each pixels instead of
printing a frame-level global SSIM value.
* Got rid of some old code
* switched to uint32_t for accumulation
* refactoring
SSIM calculation is ~4x faster now.
Change-Id: I48d838e66aef5199b9b5cd5cddef6a98411f5673
Having it architecture dependent resulted in an extra
function call of an extern function, hence no inlining and
a 5-10% impact on performance.
Change-Id: I0ff40d2d881edc76d3594213a64ee53097d42450
-print_psnr is now much faster because it doesn't use the SSIM code.
The SSIM speed-up and re-write will come later.
Change-Id: Iabf565e0a8b41651d8164df1266cfeded4ab4823
We add the following MSA optimized rescaling functions:
- RescalerExportRowExpand
- RescalerExportRowShrink
Change-Id: Ic1c76065423b02617db94cf0c22bb564219b36e6
We add the following MSA optimized color transform functions:
- TransformColor
- SubtractGreenFromBlueAndRed
Change-Id: Ib182d2b5faa7191f503ce70f0dfde0ac89402fd3
The decision is based on the variance between DC values of each
sub-4x4 block. This heuristic is rather ok for predicting whether
the 2nd transform (intra-16) is going to help or not.
The decision threshold varies with quality (=quantization).
It's only used for -m 0 and -m 1, where no full RD-opt is performed.
It actually makes these modes quite faster, with RD curve much
closer to the -m 2 mode.
Change-Id: I15f972db97ba4082cbd1dfd16bee3eb2eca701a8
We add the following MSA optimized encoder quantization functions:
- QuantizeBlock
- Quantize2Blocks
Change-Id: Ie32b442afa99eee62d2ef48942b41116a4e157d3
- 6/14/2016: version 0.5.1
This is a binary compatible release.
* miscellaneous bug fixes (issues #280, #289)
* reverted alpha plane encoding with color cache for compatibility with
libwebp 0.4.0->0.4.3 (issues #291, #298)
* lossless encoding performance improvements
* memory reduction in both lossless encoding and decoding
* force mux output to be in the extended format (VP8X) when undefined chunks
are present (issue #294)
* gradle, cmake build support
* workaround for compiler bug causing 64-bit decode failures on android
devices using clang-3.8 in the r11c NDK
* various WebPAnimEncoder improvements
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=Mchr
-----END PGP SIGNATURE-----
Merge tag 'v0.5.1'
libwebp-0.5.1
- 6/14/2016: version 0.5.1
This is a binary compatible release.
* miscellaneous bug fixes (issues #280, #289)
* reverted alpha plane encoding with color cache for compatibility with
libwebp 0.4.0->0.4.3 (issues #291, #298)
* lossless encoding performance improvements
* memory reduction in both lossless encoding and decoding
* force mux output to be in the extended format (VP8X) when undefined chunks
are present (issue #294)
* gradle, cmake build support
* workaround for compiler bug causing 64-bit decode failures on android
devices using clang-3.8 in the r11c NDK
* various WebPAnimEncoder improvements
* tag 'v0.5.1': (30 commits)
update ChangeLog
Clarify the expected 'config' lifespan in WebPIDecode()
update ChangeLog
Fix corner case in CostManagerInit.
gif2webp: normalize the number of .'s in the help message
vwebp: normalize the number of .'s in the help message
cwebp: normalize the number of .'s in the help message
fix rescaling bug: alpha plane wasn't filled with 0xff
Improve lossless compression.
'our bug tracker' -> 'the bug tracker'
normalize the number of .'s in the help message
pngdec,ReadFunc: throw an error on invalid read
decode.h,WebPGetInfo: normalize function comment
Inline GetResidual for speed.
Speed-up uniform-region processing.
free -> WebPSafeFree()
DecodeImageData(): change the incorrect assert
Fix a boundary case in BackwardReferencesHashChainDistanceOnly.
Make sure to consider small distances in LZ77.
add some asserts to delimit the perimeter of CostManager's operation
...
Change-Id: I44cee79fddd43527062ea9d83be67da42484ebfc
We add the following MSA optimized color transform functions:
- AddGreenToBlueAndRed
- TransformColorInverse
Change-Id: Iceab3813905955aa8b811253df9188512fc7de3f
The old implementation in enc/near_lossless.c performing a separate
preprocessing step is used only when a prediction filter is not used,
otherwise a new implementation integrated into lossless_enc.c is used.
It retains the same logic for converting near lossless quality into max
number of bits dropped, and for adjusting the number of bits based on
the smoothness of the image at a given pixel. As before, borders are not
changed.
Then, instead of quantizing raw component values, the residual after
subtract green and after prediction is quantized according to the
resulting number of bits, taking care to not cross the boundary between
255 and 0 after decoding. Ties are resolved by moving closer to the
prediction instead of by bankers’ rounding.
This results in about 15% size decrease for the same quality.
Change-Id: If3e9c388158c2e3e75ef88876703f40b932f671f
copy and paste error in the previous commit, change
no_sanitize("unsigned-integer-overflow") from WEBP_UBSAN_IGNORE_UNDEF ->
WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
Change-Id: Id178ee14df1f2c4923a91ce423241e26b60b5d32
include utils.h directly where needed to allow utils.h to rely on
defines from dsp.h in a follow-up.
Change-Id: I32e26aaeb0b04ba60b3332f685f9a2be5a0a8d3d
configure gets 2 new options:
--enable-neon / --enable-neon-rtcd
the NEON modules are split to their own convenience lib and built with
auto-detected flags if none are given via CFLAGS.
the /proc/cpuinfo check will only be used for armv7 targets whose
toolchain does not enable NEON by default or didn't have NEON forced by
the CFLAGS from the environment.
Change-Id: I2755bc1d065d5d6ee6143b44978c2082f8bef1c5
This will allow to work in-place on cropped area later.
Also sped up the inverse gradient filtering in SSE2 (~4%)
Change-Id: I463149eee95d36984328f163a1e17f8cabd87441
This is in preparation for some SSE2 code.
And generally speaking, the whole SSIM code needs some
revamp: we're not averaging the SSIM value at each pixels
but just computing the overall SSIM value once, for the whole
plane. The former might be better than the latter.
Change-Id: I935784a917f84a18ef08dc5ec9a7b528abea46a5
based on the sse2 change in:
9960c31 Remove an unnecessary transposition in TTransform.
~9-10.5% faster at the function-level, < 1% overall
Change-Id: I44413369b230b250fb0dbc51ff2f17cfeda609b7
SSE4.1 is slower than the SSE2 implementation and this seems to
be due to a slow _mm_loadl_epi64 implementation by gcc
(hence a bug with my gcc 4.8) and a very slow _mm_hadd_epi32. Both
got confirmed by IACA and experiments.
Change-Id: I05607f66b7ccd8f4f42e000693aea583ffd5768f
The transpose refactoring will help removing a transpose in a
later CL.
The horizontal add function helps removing a _mm_sad_epu8 in DC8uv
=> the latency/throughput went from 29/25 to 23/19
Change-Id: I5f3dfd4aad614eb079b1e83631e6a7cef49a3766
'implicit conversion from 'int' to 'short' changes value from 33050 to
-32486'
original patch:
https://codereview.chromium.org/1657313003/
Make libwebp build with -Wconstant-conversion from newer clangs.
After http://llvm.org/viewvc/llvm-project?rev=259271&view=rev, clang
points out that _mm_set1_epi16(33050) causes an overflow in the short
argument to _mm_set1_epi16(). Since there's no version that takes an
unsigned short, add an explicit cast to tell the compiler that this is
intentional.
No behavior change.
Change-Id: I6b4e3401b15cfbcc895f9e81b5c2dc59d43ffb9b
The code and logic is unified when computing bit entropy + Huffman cost.
Speed-wise, we gain 8% for lossless encoding.
Logic-wise, the beginning/end of the distributions are handled properly
and the compression ratio does not change much.
Change-Id: Ifa91d7d3e667c9a9a421faec4e845ecb6479a633
The functions containing magic constants are moved out of ./dsp .
VP8LPopulationCost got put back in ./enc
VP8LGetCombinedEntropy is now unrefined (refinement happening in ./enc)
VP8LBitsEntropy is now unrefined (refinement happening in ./enc)
VP8LHistogramEstimateBits got put back in ./enc
VP8LHistogramEstimateBitsBulk got deleted.
Change-Id: I09c4101eebbc6f174403157026fe4a23a5316beb
This implementation brings:
- an SSE implementation of packing / unpacking
- bigger buffers processed at the same time
The speedup is of 4% on lossy decoding (YUV to RGB), 0.5% on
lossy encoding (RGB to YUV was already optimized).
Change-Id: Iec677ee17f91c08614d1adab67c6df551925767f
INFO: From Compiling src/dsp/cpu.c:
src/dsp/cpu.c: In function 'x86CPUInfo':
src/dsp/cpu.c:36:3: inconsistent operand constraints in an 'asm'
With PIC and mcmodel=medium, the %rbx register must be saved and
restored which causes this problem. This was also solved in GCC-4.9 with
this patch:
https://gcc.gnu.org/ml/gcc-patches/2012-12/msg01484.html
Tested:
Builds fine with this change.
Change-Id: Icca8eea7bf5af3ef9f17f6ae2886e3430143febf
CombinedShannonEntropy takes 30% for lossless compression.
This implementation speeds up the overall process by 2 to 3 %.
Change-Id: I04a71743284c38814fd0726034d51a02b1b6ba8f
* Precision is slightly different
* also implemented in SSE2 the missing WebPUpsamplers for MODE_ARGB, MODE_Argb, MODE_RGB565, etc.
* removing yuv_tables_sse2.h saved ~8k of binary size
* the mips32/mips_dsp_r2 code is disabled for now, since it has drifted away
* the NEON code is somewhat tricky
Change-Id: Icf205faa62cf46c2825d79f3af6725dc1ec7f052
Gives 0.9% smaller (2.4% compared to before alpha cleanup) size on the 1000 PNGs dataset:
Alpha cleanup before: 18856614
Alpha cleanup after: 18685802
For reference, with no alpha cleanup: 19159992
Note: WebPCleanupTransparentArea is still also called in WebPEncode. This cleanup still helps
preprocessing in the encoder, and the cases when the prediction transform is not used.
Change-Id: I63e69f48af6ddeb9804e2e603c59dde2718c6c28
for more speed.
This gives a roughly a 1% speedup for low_effort. But actually this is a
preparation for the upcoming CL that changes RGB values of transparent pixels
based on prediction, which should not be done for low_effort because that would
slightly hurt its performance.
On 1000 PNGs, with quality 0, method 0:
Before:
Compression (output/input): 2.9120/3.2667 bpp, Encode rate (raw data): 36.034 MP/s
After:
Compression (output/input): 2.9120/3.2667 bpp, Encode rate (raw data): 36.428 MP/s
Change-Id: I5ed9f599bbf908a917723f3c780551ceb7fd724d
The same computation was done for both values: go over two buffers,
sum them up, and take a decision on the sum at each iteration.
MIPS32 code has been disabled for now, pending a code update.
Change-Id: I997984326f7092b3dbb8cfa1e524bd8132b2ab9d
instead of per block. This prepares for a next CL that can make the
predictors alter RGB value behind transparent pixels for denser
encoding. Some predictors depend on the top-right pixel, and it must
have been already processed to know its new RGB value, so requires per
scanline instead of per block.
Running the encode speed test on 1000 PNGs 10 times with default
settings:
Before:
Compression (output/input): 2.3745/3.2667 bpp, Encode rate (raw data): 1.497 MP/s
After:
Compression (output/input): 2.3745/3.2667 bpp, Encode rate (raw data): 1.501 MP/s
Same but with quality 0, method 0 and 30 iterations:
Before:
Compression (output/input): 2.9120/3.2667 bpp, Encode rate (raw data): 36.379 MP/s
After:
Compression (output/input): 2.9120/3.2667 bpp, Encode rate (raw data): 36.462 MP/s
No effect on compressed size, this produces exactly same files. No
significant measured effect on speed. Expected faster speed from better
memory layout with scanline processing but slower speed due to needing
to get predictor mode per pixel, may compensate each other.
Change-Id: I40f766f1c1c19f87b62c1e2a1c4cd7627a2c3334
the problem was the incorporation of the extra constant 1<<16 in the kC1
constant, to emulate the addition. It's now removed and the addition is
performed explicitly.
No real speed difference observed.
cf. issue #278
Change-Id: I2c6499031571d98afff392fb5ebe21a5fa60722d
global effect is ~2% faster encoding from JPG source
and ~8% faster lossless-webp source decoding to PGM (e.g.)
Also revamped the YUVA case to first accumulate R/G/B value into 16b
temporary buffer, and then doing the UV conversion.
-> New function: WebPConvertRGBA32ToUV
Change-Id: I1d7d0c4003aa02966ad33490ce0fcdc7925cf9f5
Just for RGB24/BGR24 for now, which are the hard-to-optimize ones.
SSE2 implementation coming next.
ConvertRowToY() should go into dsp/ too, at some point.
Change-Id: Ibc705ede5cbf674deefd0d9332cd82f618bc2425
also switch to using ExtractAlpha() instead of hard-coding the loop.
The ARGBToY/UV functions are rather easy to port to SSE2 / NEON.
Change-Id: I8f1346a9ca427a36ce2d6c848369ca7964d8b3c7
use 'u' rather than the unnecessary 'l' as a suffix. this prevents a
conversion warning with some toolchains
Change-Id: I21c33ce08819b3c839c75e03a8f7f3a6041d0695
same functionality, but better code layout.
What changed:
* don't trash the palette_[] in EncodePalette(), so it can be re-used
* split generation of image from bit-stream coding
* move all the delta-palette code to delta_palettization.c, and only have 1 entry point there WebPSearchOptimalDeltaPalette()
* minimize the number of "#ifdef WEBP_EXPERIMENTAL_FEATURES" in vp8l.c
* clarify the TransformBuffer stuff. more clean-up to come here...
This should make experimenting with delta-palettization easier and more compartimentalized.
Change-Id: Iadaa90e6c5b9dabc7791aec2530e18c973a94610
some limitations: only for RGBA output,
and if reduction factor is not too small (dst_width > src_width / 128)
20-25% faster, ~4-6% global improvement total decoding.
Change-Id: I95366ddaa4a38e0a96bed754dfe790126f7bb84a
It's better to stay with a 32b fixed-point precision overall, otherwise
the C-version on ARM gets *slower*.
Actually, gcc ARM compiler optimizes some instructions pretty
well when WEBP_RESCALER_FIX is exactly 32, even in C.
Change-Id: I0eea97f7db5947470f5af355dee098eca81e178d
The rounding and arithmetic is not the same as previously, to prevent overflow cases for large upscale factors.
We still rely on 32b x 32b -> 64b multiplies. Raised the fixed-point precision to 32b
so that we have some nice shifts from epi64 to epi32.
Changed rescaler_t type to 'uint32_t' in order to squeeze in all the precision required.
The MIPS code has been disabled because it's now out-of-sync. Will be fixed in
a subsequent CL when the dust settles.
~30-35% faster
Change-Id: I32e4ddc00933f1b1aa3463403086199fd5dad07b
* vertical expansion now uses bilinear interpolation
* heavily assumes that the alpha plane is decoded in full, not row-by-row
* split the RescalerExportRow and RescalerImportRow methods into Shrink
and Expand variants.
* MIPS implementation of ExportRowExpand is missing.
There's room for extra speed optim and code re-org, but let's keep that for later patches.
addresses https://code.google.com/p/webp/issues/detail?id=254
Change-Id: I8f12b855342bf07dd467fe85e4fde5fd814effdb
the x_add/x_sub increments were wrong for u/v in the upscaling case.
They shouldn't be left to the caller's discretion, but set up by
WebPRescalerInit to their exact necessary values.
-> Cleaned-up WebPRescalerInit() param list.
-> added safety asserts
-> removed the mips32/mips_r2 variant of "ImportRow" which were buggy prior
Change-Id: I347c75804d835811e7025de92a0758d7929dfc09
this moves the function outside the WEBP_USE_INTRINSICS check.
there's no alternative version and it's ~54% faster at the
function level and mildly faster overall
Change-Id: Ibc648e9ee35021d48901e05aa596aa01067796a2
share the computation between different modes
3-5 % speedup for lossless alpha
1 % for lossy alpha
no change in compression density
Change-Id: I5e31413b3efcd4319121587da8320ac4f14550b2
After several trials at re-organizing the main loop and accumulation scheme,
this is apparently the faster variant.
removed the SSE41 version, which is no longer faster now.
For some reason, the AVX variant seems to benefit most for the change.
Change-Id: Ib11ee18dbb69596cee1a3a289af8e2b4253de7b5
this moves the function outside the WEBP_USE_INTRINSICS check.
there's no alternative version and it's ~70% faster at the
function level and 1-2% faster overall
Change-Id: I59fb4918ec86b1ac3a47cbd5d05ce62f007461cb
Changed the code (again) to process 4 pixels at a time. Loop is more
involved, but overall it's faster.
Removed the SSE4.1 implementation which is now slower than SSE2.
Change-Id: I7734e371033ad8929ace7f7e1373ba930d9bb5f1
New implementations: SubtractGreenFromBlueAndRed and TransformColor
around 1-2% faster lossless encoding.
Change-Id: I1668e36fdc316ba55b3b798b91b4a3e36ce62861
DispatchAlpha* functions are hard to speed up, compared to SSE2.
ExtractAlpha sees a ~15% speed-up though.
Change-Id: I8715c2defecbc832f469eed7e6ffd012146b52de
over a 1000 image corpus
Single photograph benchmark:
Before:
Q=20: 2.560 MP/s
Q=40: 2.593 MP/s
Q=60: 1.795 MP/s
Q=80: 1.603 MP/s
Q=99: 1.122 MP/s
After:
Q=20: 3.334 MP/s
Q=40: 2.464 MP/s
Q=60: 2.009 MP/s
Q=80: 1.871 MP/s
Q=99: 1.163 MP/s
This CL allows for some further improvements that would not be possible
otherwise.
Change-Id: I61ba154beca2266cb96469281cf96e84a4412586
use vld1_dup_u8() rather than a separate ld+dup after the values were
zero extended; mildly faster at the function level
Change-Id: I1b3666a6aeb465722a1214dbc6d71c27689a7f89
VP8EncPredChroma8 improvements over ~20M pixels
left/top: ~67%
left-only: ~52%
top-only: ~57%
none: ~61%
based on dec_sse2 versions with minor changes to benefit from the linear
storage of the left boundary
Change-Id: Iee7e387fb2570b4eb5af5bfd123e9c2e9ea49c76
VP8EncPredLuma16 improvements over ~20M pixels
left/top: ~75%
left-only: ~47%
top-only: ~59%
none: ~63%
based on dec_sse2 versions with minor changes to benefit from the linear
storage of the left boundary
Change-Id: I7548be7214fa85c38fd11d30f5b8b271f437657d
structured extended feature flags require eax = 7; avoids incorrectly
detecting avx2 on some older processors that support avx.
for completeness also check for value=1 support used by the other
checks.
from [1]:
INPUT EAX = 0: Returns CPUID’s Highest Value for Basic Processor
Information and the Vendor Identification String
[1]
http://www.intel.com/content/www/us/en/processors/processor-identification-cpuid-instruction-note.html
Change-Id: I60b20d661a978d551614dbf7acdc25db19cb6046
use psadbw to perform top row summation; left remains in C as repacking
it into a vector to apply the same operation is too costly.
DC8uv: ~19% faster
DC8uvNoLeft: ~12% faster
Change-Id: I707c4f6177a65b5d1f2d3deeca87d2bb740185e2
use psadbw to perform top row summation; left remains in C as repacking
it into a vector to apply the same operation is too costly.
DC16: ~20% faster
DC16NoLeft: ~14% faster
Change-Id: I7ec3f8a6e5923f88a530f79fceb88d5001bef691
generates a stub function when the specific architecture is not enabled,
exposing a symbol in the module, avoiding a compiler warning
Change-Id: Ia9336e57466a9b5241b85c1c95838e91c9283147
Visible speed-up, thanks to pshufb and pabsw and psignw use.
had to tweak configure.ac to make "smmintri.h" presence correctly
detected (we need to set the CPPFLAGS instead of the CFLAGS!)
Change-Id: I2ab99e16a27a64fdf1f09b2b4e30a5e74ccca080
allows the former to be inlined; negligible speed-up in most cases,
however this is structure is consistent with the rest of the optimized
modules
Change-Id: Ib080240b06f7a995b47f1906627850c355b82901
