- build_stages.py uses gobjdump for both architectures, minor formatting changes
- let byte arrays be written inline in splice()
Change-Id: I84bd47c18e5ae0b34b35f8c2f0a329fb1ea58f60
Reviewed-on: https://skia-review.googlesource.com/6833
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
This should be a no-op on x86 (unified icache and dcache), but is required on
ARM because the icache and dcache are separate.
The Pixel C bots are crashing at head, and so is my local Pixel C, but not
after this CL. Interestingly, my Pixel not-C (phone) has always been fine.
I wonder if it has a unifed cache?
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Android-Clang-PixelC-CPU-TegraX1-arm64-Release-Android
Change-Id: I8f9d53729613c51aabfeb49495aac38ce2801a84
Reviewed-on: https://skia-review.googlesource.com/6847
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This is still just linear (non-sRGB), but adding sRGB will
be the next step. I've verified that this is really making
R8 textures when uploading Gray8 bitmaps. Tests pass, and
the all_bitmap_configs GM still renders correctly (unlike
when we just mapped Gray8 to Alpha8).
This adds another pixel config, which could grow our cache
footprint, but the benefits of not using 4bpp for 1bpp data
should outweigh that?
Re-land of https://skia-review.googlesource.com/c/6817/,
with fixes for Vulkan.
BUG=skia:6110
Change-Id: Ia763c276808be28027ed0005ee4b88637306583f
Reviewed-on: https://skia-review.googlesource.com/6839
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
set_k() is simulating the effect of loading a pointer constant into the register used to pass the 4th function argument. An easier way to do this is to pass that pointer constant as the 4th function argument... duh.
Change-Id: I5604d6bbadd86eaaa82f8c4391080f6489b1927f
Reviewed-on: https://skia-review.googlesource.com/6843
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Not sure what to do about this long term.
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Debug-MSAN
Change-Id: If9e86c285914ce2d8255ac25197845728d7c5d49
Reviewed-on: https://skia-review.googlesource.com/6842
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
I also updated the dump feature to work with aarch64, and included comments on how to disassemble an aarch64 dump.
Looking at an aarch64 dump made it immediately obvious that the jump offset was off by 1.
Change-Id: I17fa6ee44779e8be69ab4582e338c88212aba36c
Reviewed-on: https://skia-review.googlesource.com/6841
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
The new test would fail without the the change in SkSplicer.cpp to call fSpliced(x,x+body) instead of fSpliced(x,body). The rest of the changes are cosmetic, mostly renaming n to limit.
Change-Id: Iae28802d0adb91e962ed3ee60fa5a4334bd140f9
Reviewed-on: https://skia-review.googlesource.com/6837
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit f2956459f7.
Reason for revert: GM and image failures on some bots (rendering red, not gray).
Original change's description:
> Add Gray8 pixel config
>
> This is still just linear (non-sRGB), but adding sRGB will
> be the next step. I've verified that this is really making
> R8 textures when uploading Gray8 bitmaps. Tests pass, and
> the all_bitmap_configs GM still renders correctly (unlike
> when we just mapped Gray8 to Alpha8).
>
> This adds another pixel config, which could grow our cache
> footprint, but the benefits of not using 4bpp for 1bpp data
> should outweigh that?
>
> BUG=skia:6110
>
> Change-Id: I4fc4c2479fc25f1d278e174a9bb5b542a0cb184c
> Reviewed-on: https://skia-review.googlesource.com/6817
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com,reviews@skia.org
BUG=skia:6110
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I95a4fc0450a569d5791f6bceb7fae61c7e5eba61
Reviewed-on: https://skia-review.googlesource.com/6838
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Will follow up later with (is_android && target_cpu == "arm64"). Just being cautious.
Change-Id: I32bf48dec55633d6386c66fd0f11fc5616596477
Reviewed-on: https://skia-review.googlesource.com/6834
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Recent versions of ANGLE require this. See:
https://skia-review.googlesource.com/c/6698/
This also updates us to the latest version of the Windows SDK,
2015 Update 3 (Windows 10 SDK).
BUG=skia:
CQ_INCLUDE_TRYBOTS=skia.primary:Perf-Win2k8-MSVC-GCE-CPU-AVX2-x86_64-Debug-GDI,Test-Win10-MSVC-ShuttleA-GPU-GTX660-x86_64-Debug-Vulkan,Test-Win2k8-MSVC-GCE-CPU-AVX2-x86_64-Release
Change-Id: I42c8cabc87717f8695763f2c5573b27ab8ab65be
Reviewed-on: https://skia-review.googlesource.com/6801
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This only changes behavior when the input SkBitmap/SkPixmap is
tagged with a non-null SkColorSpace. Android tags their bitmaps
as sRGB when linear blending is enabled. So this only changes
behavior in Android when linear blending is turned on.
*If linear blending is turned on, this will do a color correct
encode (which is the desired behavior).
*If linear blending is turned off, this will do a legacy encode.
TODO: Add support for F16.
TODO: Add color space support to WEBP.
TODO: Tag encoded images with ICC profiles (when it makes sense).
BUG=skia:
Change-Id: Idd8a2836371d24a453d953e6fe2e76a87751be96
Reviewed-on: https://skia-review.googlesource.com/6498
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>
SkSplicer is better.
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: I014ec0e9fb00a8a4694d442e672c65402621dc67
Reviewed-on: https://skia-review.googlesource.com/6830
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
BUG=skia:
Change-Id: I52504b21313717bf8321ec3f1df770773703a1a1
Reviewed-on: https://skia-review.googlesource.com/6831
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>
This method and its associated globals are no longer used by any user
and can now be removed.
BUG=skia:2817
Change-Id: Ia627e2494f568a733999b980aec9284467d2640d
Reviewed-on: https://skia-review.googlesource.com/6821
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
This is still just linear (non-sRGB), but adding sRGB will
be the next step. I've verified that this is really making
R8 textures when uploading Gray8 bitmaps. Tests pass, and
the all_bitmap_configs GM still renders correctly (unlike
when we just mapped Gray8 to Alpha8).
This adds another pixel config, which could grow our cache
footprint, but the benefits of not using 4bpp for 1bpp data
should outweigh that?
BUG=skia:6110
Change-Id: I4fc4c2479fc25f1d278e174a9bb5b542a0cb184c
Reviewed-on: https://skia-review.googlesource.com/6817
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The idea here is that we will pass GrTextureProxys in (where we're currently passing GrTextures) and defer the normalization until the texture is actually instantiated (and possibly move it to the GPU entirely)
This CL does (intentionally) change the texturedomaineffect GM but I believe the new behavior is more correct.
Change-Id: I4e0510b3dfb65ff0d0ee5921f9a6f94151e602d3
Reviewed-on: https://skia-review.googlesource.com/6807
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Seems to be working. The jump to loop_start might be a little off, but not by much. Correctness is really still a big TODO.
$ adb shell 'cd /data/local/tmp; ./monobench SkRasterPipeline 200'
SkRasterPipeline_…
200 …f16_compile 1x …f16_run 1.42x …srgb_compile 2.21x …srgb_run 2.59x⏎
Change-Id: I0e1acc6404cf3ce8084d9ef8011cbe0b5f1fd6e3
Reviewed-on: https://skia-review.googlesource.com/6811
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit d4ed326d6f.
Reason for revert: reverting temporarily to get us rolling into Chrome again.
Must be this CL right?
https://storage.googleapis.com/chromium-layout-test-archives/linux_trusty_blink_rel/3364/layout-test-results/results.html
Original change's description:
> Improve quad edges' smoothness in non-AA cases
>
> Previously, non-AA quad edges only have an accuracy about 1/2 pixel
> (while the AA quad edges have an accuracy about 1/8 pixel). Now, we
> increase non-AA quad edges' accuracy to 1/8 pixel as well.
>
> The difference is very significant for rotating non-AA filled circles.
> For example, run `./out/Debug/SampleApp --slide GM:fillcircle` with AA
> turned off (by pressing b).
>
> The benchmark added reveals that increasing quad accuracy from 1/2 to
> 1/8 doesn't affect the performance significantly. The following is the
> 1/2-accuracy performance versus 1/8-accuracy performance:
>
> curr/maxrss loops min median mean max stddev samples
> config bench
> 7/17 MB 19 2.43µs 2.57µs 2.81µs 10.5µs 22% 16119
> 8888 path_fill_big_nonaacircle
> 7/17 MB 17 1.38µs 1.42µs 1.52µs 13µs 20% 21409
> 8888 path_fill_small_nonaacircle
>
> curr/maxrss loops min median mean max stddev samples
> config bench
> 7/17 MB 71 2.52µs 2.59µs 2.79µs 7.67µs 19% 7557
> 8888 path_fill_big_nonaacircle
> 7/17 MB 64 1.45µs 1.49µs 1.51µs 2.39µs 5% 12704
> 8888 path_fill_small_nonaacircle
>
>
> BUG=skia:
>
> Change-Id: I3482098aeafcc6f2ec9aa3382977c0dc1b650964
> Reviewed-on: https://skia-review.googlesource.com/6699
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Yuqian Li <liyuqian@google.com>
>
TBR=caryclark@google.com,liyuqian@google.com,reed@google.com,djsollen@google.com,bungeman@google.com,reviews@skia.org,fmalita@chromium.org
BUG=skia:
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I5bc4596ab506f6f61ac2da91a07cf51d61114f31
Reviewed-on: https://skia-review.googlesource.com/6829
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
A few methods in PathOpsExtendedTest were changed to write to SkStrings
instead of SkStreams. However, the names of these functions confusingly
still have "Stream" in their names. This CL is just a static function
rename and some small clean-up.
Change-Id: Idf21b2aba28a2f984ef30cb5c18e26a43a9c7201
Reviewed-on: https://skia-review.googlesource.com/6819
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
ffd8f62236..08fd250e1a
This picks up a number of fixes and an implementation of
FT_Get_Var_Design_Coordinates.
Change-Id: Idac2b3b5d2b0684fa2c13f4f2484c09f39a4eced
Reviewed-on: https://skia-review.googlesource.com/6815
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Change-Id: Id1ecf20bb785e5979afa9e07ca4bb04a7d839fb7
Reviewed-on: https://skia-review.googlesource.com/6806
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
BUG=skia:6101
Change-Id: Iee798dd2d9dcc4521f643b814e65029b9383cc6f
Reviewed-on: https://skia-review.googlesource.com/6696
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Previously, non-AA quad edges only have an accuracy about 1/2 pixel
(while the AA quad edges have an accuracy about 1/8 pixel). Now, we
increase non-AA quad edges' accuracy to 1/8 pixel as well.
The difference is very significant for rotating non-AA filled circles.
For example, run `./out/Debug/SampleApp --slide GM:fillcircle` with AA
turned off (by pressing b).
The benchmark added reveals that increasing quad accuracy from 1/2 to
1/8 doesn't affect the performance significantly. The following is the
1/2-accuracy performance versus 1/8-accuracy performance:
curr/maxrss loops min median mean max stddev samples
config bench
7/17 MB 19 2.43µs 2.57µs 2.81µs 10.5µs 22% 16119
8888 path_fill_big_nonaacircle
7/17 MB 17 1.38µs 1.42µs 1.52µs 13µs 20% 21409
8888 path_fill_small_nonaacircle
curr/maxrss loops min median mean max stddev samples
config bench
7/17 MB 71 2.52µs 2.59µs 2.79µs 7.67µs 19% 7557
8888 path_fill_big_nonaacircle
7/17 MB 64 1.45µs 1.49µs 1.51µs 2.39µs 5% 12704
8888 path_fill_small_nonaacircle
BUG=skia:
Change-Id: I3482098aeafcc6f2ec9aa3382977c0dc1b650964
Reviewed-on: https://skia-review.googlesource.com/6699
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
BUG=skia:
Change-Id: I1f0df58ffe79e439f92a8c1cd3c39812c32d039f
Reviewed-on: https://skia-review.googlesource.com/6743
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Stephan White <senorblanco@chromium.org>
BUG=skia:5876,skia:6082,skia:6086,skia:6097
Change-Id: I1f302c7bb0d147d976f320e01f969a1bd693c78c
Reviewed-on: https://skia-review.googlesource.com/6804
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
I think I may have cracked the compile-ahead-of-time-splice-at-runtime nut.
This compiles stages ahead of time using clang, then splices them together at runtime. This means the stages can be written in simple C++, with some mild restrictions.
This performs identically to our Xbyak experiment, and already supports more stages. As written this stands alone from SkRasterPipeline_opts.h, but I'm fairly confident that the bulk (the STAGE implementations) can ultimately be shared.
As of PS 25 or so, this also supports all the stages used by bench/SkRasterPipelineBench.cpp:
SkRasterPipeline_…
400 …f16_compile 1x …f16_run 1.38x …srgb_compile 1.89x …srgb_run 2.21x
That is, ~30% faster than baseline for f16, ~15% faster for sRGB.
Change-Id: I1ec7dcb769613713ce56978c58038f606f87d63d
Reviewed-on: https://skia-review.googlesource.com/6733
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit a8f80de2bc.
Reason for revert: nanobench failing on windows bots, possibly others
Change-Id: Iacb8c650064a28654c165665be057377ffb02ba5
Reviewed-on: https://skia-review.googlesource.com/6802
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
All GrXPFactory instances are static constexpr.
Change-Id: If1086b08534166201e53b3fd9379104e361eb5e6
Reviewed-on: https://skia-review.googlesource.com/6701
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Change-Id: I08e907cd5051d9f8cd97cdd773f9ff326cc6a0d8
Reviewed-on: https://skia-review.googlesource.com/6739
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
xbyak hasn't changed in a couple weeks, so this pinned version is likely the version all bots are already on.
Change-Id: I050e508f601838015edcb9890214bd7ee1ac1e59
Reviewed-on: https://skia-review.googlesource.com/6737
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
