To do this with a clean conscience, I needed to convert the unroll-
counting logic from a linear time algorithm to constant-time. Getting
all the edge cases correct requires a lot of care, and there are now
plenty of unit tests.
Change-Id: I620909d069ac425b7310e345bf80ec844fe035f8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445643
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
The motivation for this is that with the new backend info structs the
sample count will be stored on non backend specific classes
(e.g. GrAttachment) and will not be part of the GrVkSurfaceSpec struct.
Thus the info will not be available from just a GrVkImage (or at least
until GrVkImage and GrVkAttachment get merged down the road).
This also removes some uses of GrVkImage::imageInfo() getter since that
struct will also be going away.
Bug: skia:12402 skia:10727
Change-Id: Ie90870e1d834e6ea6ba029d0b480641fb8d2cfa0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/446177
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This allows us to get rid of a lot of .c_str()'s.
Change-Id: I09102f90d69620614dc5a7a2ebc64bd3e9b1c437
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445816
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Bug: 12145
Test: imagedecoder_png_fuzzer
Test: Codec_PngRoundTrip
If a PNG's sBIT specifies that the significant bits are 565, SkPngCodec
treats the image as defaulting to kRGB_565_SkColorType, rather than the
typical kN32_SkColorType. This feature is likely rarely used in the
wild, but it is used by Skia to encode kRGB_565_SkColorType to a PNG and
then recreate the original SkPixmap. These Skia-created PNGs always use
PNG_COLOR_TYPE_RGB, so only respect this sBIT with PNG_COLOR_TYPE_RGB.
Further, if the PNG is PNG_COLOR_TYPE_PALETTE, treating it as 565 means
that we read it incorrectly.
Change-Id: I81dcce771c8492cdcf5677d3c13167a1231acfe4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445963
Reviewed-by: Heather Miller <hcm@google.com>
Commit-Queue: Derek Sollenberger <djsollen@google.com>
* Sets up the resolve path so we don't necessarily store the MSAA result.
* Modifies copies so that in this case we blit copy from the resolve
attachment rather than trying to do a resolve copy from the MSAA
attachment.
Bug: skia:12086
Change-Id: Ie9cb71879d714676922e378ab74e32263b8dc50a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445111
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
The fuzzer has discovered that it can trigger integer overflow in SkVM
via loops that perform arithmetic that can be constant-evaluated after
the loop is unrolled. GLSL doesn't define the results of this operation,
and SkVM doesn't need to define it either; we will accept whatever
platform-specific behavior we get if our shader's ints overflow.
Change-Id: Ib424bd3b0e5768b7b174ac8a52e126794962f6cb
Bug: oss-fuzz:37916
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445971
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This is a reland of 9f82158f17
Original change's description:
> Metal: add loadMSAAFromResolve support.
>
> Also fixes some minor warnings.
>
> Bug: skia:12086
> Change-Id: Ia476a7a196b490022978761e92f935a6d668136a
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/441797
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Jim Van Verth <jvanverth@google.com>
Bug: skia:12086
Change-Id: I1d9dd36b3f8f19a0ab4a98ccaab18146566c1885
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/446077
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Change-Id: I3ae83fbdcb4b3b13150df8b6a6461c36cbeaa55e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445641
Reviewed-by: Ravi Mistry <rmistry@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:12402
Change-Id: Ifd077add34bb7d299654b29f5e8fbdb746cffed1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445968
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
`layerHeight` is only used in an assert, so in release builds, it's not
actually used for anything.
Change-Id: I6fd802dd4a4f29bd01c469b80697366d4f4bb7d6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445966
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
An additional layer is required to composite correctly. PaintComposite
blends/composites between source and destination. The result is
then drawn on top of previous paint results. Previously, the source
layer was just drawn atop previous paints, then the destination
layer was composited down with composite mode, which was incorrect.
More details in [1].
Add tests for composite modes, compare [2]
[1] https://github.com/googlefonts/nanoemoji/issues/332
[2] https://github.com/googlefonts/color-fonts/pull/73
Cq-Include-Trybots: luci.skia.skia.primary:Test-Android-Clang-GalaxyS6-GPU-MaliT760-arm64-Release-All-Android_NativeFonts,FM-Debian10-Clang-GCE-CPU-AVX2-x86_64-Release-All-MSAN
Bug: skia:12424
Change-Id: I8ff69022ec09758af7021756b0fb26443216fcca
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445936
Commit-Queue: Dominik Röttsches <drott@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Change-Id: I291dedc14041647c32fcc95f0b96cdd91cafc3d8
Bug: skia:12426
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445969
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Several fuzzer issues, and one Chromium issue that's blocking the roll.
Bug: chromium:1246795
Bug: skia:12423
Change-Id: I00370b74569b447e543d9a1f22c588eb493063da
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445960
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
An ES2-compatible for loop supports six separate rel-ops:
< <= > >= != ==
Each rel-op, in addition to its expected usage, is also able to
represent a loop which never terminates, as well as a loop which
terminates instantly. Since SkVM unrolls these loops, we should make
sure we do it properly. We now have unit tests for all of these cases.
Change-Id: Icae04d48bc158bf8c0c98db97f76756a1a29110c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445756
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Intrinsic-call optimization can be triggered during inlining. In this
case, inlining turned `normalize(x)` into `normalize(constant-value)`.
DSL is used to implement optimizations for a handful of intrinsic calls,
including `normalize`, which internally relies on `length`.
The DSL expects that it can use the IRGenerator to handle function
calls. This was not working because we were finished with the initial
compilation pass, and the IRGenerator's symbol table is removed when
finish() was called.
We now temporarily give a symbol table back to the IRGenerator while
the inliner runs. We remove it again as soon as inlining is complete.
Change-Id: I6da98788d93749ffeb008c1f4c3f72b436e8ceeb
Bug: oss-fuzz:37994
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445956
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
By processing the instructions in reverse order, we can avoid the need
for a worklist vector.
Change-Id: Id81673b41ac6f201bac3da87328c9fbdde35846a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445757
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
fontMgr doesn't export the function named `makeTypefaceFromData`.
it has `**M**akeTypefaceFromData`.
This is an imported pull request from
https://github.com/google/skia/pull/87
GitOrigin-RevId: 7a80a1a7776d0b20810510665f6a3727100bef97
Change-Id: I02eb6a5f406d4448f61266016f9c0882eb59c871
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445804
Reviewed-by: Kevin Lubick <kjlubick@google.com>
When profiling with a large `--ms` value, the sample count could become
very large. We would still tell the Stats object that we wanted the plot
string (even though we weren't going to print it). Sadly, creation of
that string is O(n^2), thanks to SkString's allocation behavior.
Change-Id: I20ed61a49ab2827f3561673f086e7c5c298eaf86
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445645
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
https://skia-review.googlesource.com/c/skia/+/443897 regressed
the motionmarksuits.skp (180ms to 202ms on my machine). With this change
performance was brought back to 180ms.
Bug: skia:12398, chromium:1245485
Change-Id: Ib7e5776f1f9ab37f8681425e3972680bc547ed99
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445236
Auto-Submit: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Florin Malita <fmalita@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
- Use .eval() in all example fiddles
- Add some more explanation of how the parts of a draw/paint contribute
to the GPU fragment shader.
Bug: skia:12302
Change-Id: Ib69b9af39368c980e1aa9206af585f26498d083e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445640
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Change-Id: Id894eb70273454716eb33c85dff2056333e90cdd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445281
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Bug: skia:12302
Change-Id: I4ff394f1f9d93d2def19a9f9d49cb208651aff10
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445639
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
de09f8db31..3271dea167
2021-09-02 cnorthrop@google.com FrameCapture: Check EXT before limiting tex buffer
2021-09-02 cnorthrop@google.com GetTexImage: Fix image initialization
2021-09-02 lexa.knyazev@gmail.com Vulkan: Suppress VUID-vkCmdDraw-blendEnable
If this roll has caused a breakage, revert this CL and stop the roller
using the controls here:
https://autoroll.skia.org/r/angle-skia-autoroll
Please CC fmalita@google.com on the revert to ensure that a human
is aware of the problem.
To file a bug in ANGLE: https://bugs.chromium.org/p/angleproject/issues/entry
To file a bug in Skia: https://bugs.chromium.org/p/skia/issues/entry
To report a problem with the AutoRoller itself, please file a bug:
https://bugs.chromium.org/p/skia/issues/entry?template=Autoroller+Bug
Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md
Cq-Include-Trybots: skia/skia.primary:Build-Debian10-Clang-x86_64-Release-ANGLE;skia/skia.primary:Test-Win10-Clang-AlphaR2-GPU-RadeonR9M470X-x86_64-Debug-All-ANGLE;skia/skia.primary:Test-Win10-Clang-Golo-GPU-QuadroP400-x86_64-Debug-All-ANGLE;skia/skia.primary:Test-Win10-Clang-NUC5i7RYH-GPU-IntelIris6100-x86_64-Debug-All-ANGLE;skia/skia.primary:Test-Win10-Clang-NUC6i5SYK-GPU-IntelIris540-x86_64-Debug-All-ANGLE;skia/skia.primary:Test-Win10-Clang-NUC8i5BEK-GPU-IntelIris655-x86_64-Debug-All-ANGLE;skia/skia.primary:Test-Win10-Clang-NUCD34010WYKH-GPU-IntelHD4400-x86_64-Debug-All-ANGLE
Tbr: fmalita@google.com
Test: Test: Fortnite MEC
Test: Test: FrameCapture on Pixel 4
Change-Id: Ie2d4f12605163fcf2f307d3f1b2f71d4954223e9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445559
Commit-Queue: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Bot-Commit: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Also update RELEASE_NOTES to describe new syntax.
Change-Id: I2666551b98f80b61ae3a48c92a9e306cdc7242b0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/444735
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This analysis pass assumes we have a program with a valid structure--all
loops must be ES2-compliant, and all function-calls must reference real
functions that exist. If we detected an error during compilation, our
program might not meet these criteria.
Change-Id: I4c7aefb3221438643614f1e0cbc2bad40b94b161
Bug: skia:12396
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/444982
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Calling `visitProgramElement` on a FunctionDefinition should always set
fFunctionSize on any path through the code. However, this is difficult
to confirm by inspection, since there are several exit conditions. It's
safest to initialize the variables at construction time, just in case.
Change-Id: I8bfda5fc5c4d71f5593ea2469586d668be762449
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445636
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Includes .eval() as a new alternative syntax for sample/shade.
Bug: skia:12302
Change-Id: Ia921c141b3eeeba7e5309d921967b941f9cd055e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/444756
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This moves the swizzle domain test into the path used by the DSL so that
the error check benefits both sides. To make this possible we need to
be able to distinguish between equivalent swizzle components like x and
r, so they aren't collapsed down to the same component until the very
end.
Change-Id: I48f2582886391eabd7ce6eae949babdeead6051e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445280
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Previously, deeply-nested loops could lead to stack overflow during SkVM
dead-code elimination. The DCE algorithm in SkVM now stores its worklist
on the heap and does not rely on recursion. Since we now enforce an
upper bound on maximum program size within SkSL, this worklist's size
should always be reasonable and bounded.
Change-Id: I8b7955298b2540a9f600fc5d94d3bc804c68632c
Bug: skia:12396, oss-fuzz:37827, oss-fuzz:37837
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445597
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The previous values are about to be removed from Dawn.
Bug: chromium:1199057
Change-Id: Iec6c4e4d641a99bceae975e68ab4b53ccdc75ef0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445436
Auto-Submit: Brandon Jones <bajones@chromium.org>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Our program-size analysis pass needs to recurse into called functions;
depending on the exact order of functions in the program, this recursion
can hypothetically be as deep as the deepest function-call chain. Set an
upper bound on recursion here, so we don't overflow the stack while
trying to check the program size. In practice, 50 frames is far deeper
than a regular shader should ever go.
Change-Id: I733ee48dad6f8053facdfd9f6d8a2b9b2a4af188
Bug: skia:12396
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445279
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
AfterEffects and Bodymovin support an "AllCaps" text flag which forces
text capitalization.
* add toUpper() bindings to SkUnicode/SkICU
* add capitalization options to SkottieShaper
* plumb existing Lottie 'ca' (AllCaps) prop
* also fix a couple of unrelated whoopsies
Change-Id: I8e80921b66530e9830938004946082c6e450b04b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445104
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Florin Malita <fmalita@google.com>
