I was profiling and remembered GN hasn't turned on much CPU-specific code. I looked at zlib, then libpng, then libjpeg-turbo:
- zlib was easy but not useful. No routine we use in decoding changes significantly.
- libpng was easy and useful, and we were already using NEON filters on ARM.
- libjpeg-turbo requires yasm and is annoying.
BUG=skia:5875
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4531
Change-Id: Ie072a457da41ee6538eebacb5eb5dbe5a6eb585e
Reviewed-on: https://skia-review.googlesource.com/4531
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Matt Sarett <msarett@google.com>
This extends the pattern in freetype2 to expat, icu, libjpeg-turbo, libpng, libwebp, and zlib, and gives all these an arg to control which to use.
Homebrew doesn't have dng_sdk, piex, or sftnly, or I'd have done the same for them too.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4260
DOCS_PREVIEW= https://skia.org/?cl=4260
Change-Id: I82e780502bf2217336e791787f172a6fc8f55460
Reviewed-on: https://skia-review.googlesource.com/4260
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Hal Canary <halcanary@google.com>
Once you have downloaded an android NDK, you can set the ndk GN arg to use it.
E.g. my gn.args looks like:
is_debug = false
ndk = "/opt/android-ndk"
This should be enough to get you going for an arm64 build. You ought to be able to tweak that to other architectures by changing target_cpu to "arm", "x86", "x86-64", etc. That won't quite work until I follow this up a bit, but the skeleton is there.
This is enough to get me compiled, linked, and running to completion on my N5x.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2275983004
Review-Url: https://codereview.chromium.org/2275983004
This builds, links, and runs on Linux. Have not tried Mac.
I've tested is_debug={true,false} and is_component_build.
It's neat that the component build DM works, but it's also an indication I've missed an essential flag or two... it shouldn't work. :)
The GPU backend isn't working yet, but all the software configurations I've tried look good.
This fleshes out all the other parts of SkCodec too... I noticed we weren't able to decode gifs or webp.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2188643002
Review-Url: https://codereview.chromium.org/2188643002
- Make fiddle build on Mac (skipping GL).
- Now that we're building in SkCodec, we depend on libpng and libjpeg-turbo unconditionally, not just on Linux.
- Re-arrange third_party a bit so that our targets are Fuchsia/Chrome compatible.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2184133002
NOTREECHECKS=true
This doesn't affect Chrome/Blink, so landing through the closed tree.
Review-Url: https://codereview.chromium.org/2184133002
With this change, the CMake build, which does not use DEPS to sync
external projects, is able to build and use the same version of libpng
that is used in other builds.
This will allow all platforms (including Google3 CMake build) to test on
the same version of libpng, so we do not need to make SkPngCodec support
all versions of libpng.
- Update CMakeLists.txt to use the checked in libpng.
- Check in libpng version 1.6.22rc01
- Update README.google
- Replace our old LICENSE file with the latest one from libpng
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2033063003
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot,Build-Mac-Clang-x86_64-Release-CMake-Trybot
Review-Url: https://codereview.chromium.org/2033063003
The version of libpng we were using has some security vulnerabilities.
Update to a newer version - 1.6.20 (047737496a/).
Update pnglibconf.h. The new version is a copy of the newer version of
the file, with our /* custom settings */ tacked onto the end. (The
custom settings are unchanged from the last version.)
Update the LICENSE. The new version is a copy of the LICENSE that
libpng now includes as a separate file.
Update the README file to reflect the change in source of the LICENSE
file.
BUG=skia:4573
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1453373003
Review URL: https://codereview.chromium.org/1453373003
Local timing says this 4-byte Paeth function takes about 0.3x the time the serial libpng code does, dropping from ~10 cycles per byte to ~2.9.
bpp=4 is mainly an easy demo. This approach can work for any bpp up to 16, 1 pixel at a time, at roughly the same cost per pixel. Doing more than 1 pixel at a time is a tricky math problem I have yet to attempt to solve.
Everything here can be trivially downgraded to MMX, supporting bpp up to 8. It seems to be a little slower (~3.5 cycles per byte), but it would make the code compatible with every x86 that can still power on.
I've tried four approaches:
- this way;
- doing things naively in 16-bit;
- a 16-bit version that requires division by 3 (i.e. mulhi_epu16(..., 0x5580) );
- a mostly 8-bit version of the same.
They're all fine, but this one is consistently the fastest I've measured.
I'd be happy to settle on the naive 16-bit version too, which would have a very clear implementation that's only minorly slower than this version. The other two are way more complicated, and would require us to draw some serious ASCII diagrams to explain.
I have learned that the .skp serialization tests (serialize-8888) have a nice side effect of testing the correctness of these filters!
(Since writing the description above, I've bumped things up to {Paeth,Sub,Avg} x { 3 bpp, 4 bpp }.)
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1573943002
Review URL: https://codereview.chromium.org/1573943002
All platforms except android are configured to use the statically linked copy
of libpng. Android uses the system provided dynamic copy for SkImageDecoder
and the static copy for SkCodec. The exception being android framework builds
that currently use the dynamic copy everywhere.
This CL also enables NEON optimizations for libpng.
Review URL: https://codereview.chromium.org/1058823002
Reason for revert:
breaking the nexus_9 and ios builds.
Original issue's description:
> Enable both static and dynamically linked libpng
>
> All platforms except android are configured to use the statically linked copy of libpng. Android uses the system provided dynamic copy for SkImageDecoder and the static copy for SkCodec. The exception being android framework builds that currently use the dynamic copy everywhere.
>
> This CL also enables NEON optimizations for libpng.
>
> Committed: https://skia.googlesource.com/skia/+/2469c999518e7b0063d35e9e2eb074a0477c21acTBR=scroggo@google.com,msarett@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1050183002
All platforms except android are configured to use the statically linked copy of libpng. Android uses the system provided dynamic copy for SkImageDecoder and the static copy for SkCodec. The exception being android framework builds that currently use the dynamic copy everywhere.
This CL also enables NEON optimizations for libpng.
Review URL: https://codereview.chromium.org/1032253003
DM:
Add a flag to use SkCodec instead of SkImageDecoder.
SkCodec:
Base class for codecs, allowing creation from an SkStream or an SkData.
An SkCodec, on creation, knows properties of the data like its width and height. Further calls can be used to generate the image.
TODO: Add scanline iterator
SkPngCodec:
New decoder for png. Wraps libpng. The code has been repurposed from SkImageDecoder_libpng.
TODO: Handle other destination colortypes
TODO: Substitute the transpose color
TODO: Allow silencing warnings
TODO: Use RGB instead of filler?
TODO: sRGB
SkSwizzler:
Simplified version of SkScaledSampler. Unlike the sampler, this object does no sampling.
TODO: Implement other swizzles.
Requires a gclient sync to pull down libpng.
BUG=skia:3257
Committed: https://skia.googlesource.com/skia/+/ca358852b4fed656d11107b2aaf28318a4518b49
(and then reverted)
Review URL: https://codereview.chromium.org/930283002
