I noticed it was broken and untested when writing docs.
Change-Id: I5254b7bd50fde8a361c72ccbd7380809d31ccc08
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321464
Reviewed-by: Nathaniel Nifong <nifong@google.com>
We don't believe this is needed anymore and was mostly here to support
old desktop GL
Bug: skia:10727
Change-Id: Iab35ec7d6ed64d817aed57c38864932cc417c68e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321541
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This reverts commit 3e40ce0937.
Reason for revert: Lots of bad Vk images from Pixel2/3
Original change's description:
> Convert sksl_frag.sksl to an IRIntrinsicMap
>
> Did some related cleanup:
> - We were setting the IRGenerator's fIntrinsics to the (empty) GPU map
> while converting the GPU/frag/vert source. Make the IR generator
> support a null intrinsic map, so we can (more correctly) structure the
> compiler's constructor.
> - Use explicit types on all calls to findAndInclude.
> - Move the assert back into grab_intrinsics - every converted include
> only contains supported elements (and will continue to do so).
>
> Change-Id: I80ebb247107dde656946858bf2cd1f50a03f67d3
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321496
> Reviewed-by: John Stiles <johnstiles@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I6f11366971bfd252f11d30c48e2e776d30458933
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321683
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
b09f16ee7a..1fd3e5d89a
2020-10-02 jmadill@chromium.org Test Runner: Fix race in watchdog timeouts.
2020-10-02 syoussefi@chromium.org Vulkan: Fix UtilsVk clear in non-zero subpass
2020-10-02 jmadill@chromium.org Revert "Vulkan:Add debug labels for OpenGL calls"
2020-10-02 jmadill@chromium.org Add a standalone GN isolate map.
2020-10-01 jonahr@google.com Suppress MultisampledRenderToTextureBenchmark on Pixel2/GLES
2020-10-01 syoussefi@chromium.org Skip MSRTT tests on Nvidia/Windows7
2020-10-01 b.schade@samsung.com Vulkan: Change sampleCoverage calculation
2020-10-01 lehoangq@gmail.com Metal: Implement transform feedback
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 brianosman@google.com on the revert to ensure that a human
is aware of the problem.
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/master/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: brianosman@google.com
Test: Test: dEQP-GLES3.functional.multisample.fbo*sample_coverage*
Change-Id: I85897c5a1daa356f38a61a748e51ac100d987302
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321647
Reviewed-by: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Commit-Queue: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Did some related cleanup:
- We were setting the IRGenerator's fIntrinsics to the (empty) GPU map
while converting the GPU/frag/vert source. Make the IR generator
support a null intrinsic map, so we can (more correctly) structure the
compiler's constructor.
- Use explicit types on all calls to findAndInclude.
- Move the assert back into grab_intrinsics - every converted include
only contains supported elements (and will continue to do so).
Change-Id: I80ebb247107dde656946858bf2cd1f50a03f67d3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321496
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
As a first step, convert sksl_pipeline.sksl to an IntrinsicMap (rather
than inherited element list). This makes the new code operate on
sk_FragCoord (which was previously being shared by all runtime effect
programs).
The new unit test angered TSAN, and now runs without complaint.
Also finish converting the .fp intrinsics over, so those don't need an
inherited element list either. And while doing that, refactor that
parsing to match all of the others. FP was uniquely implementing
processIncludeFile itself, rather than reusing the pattern of other
pre-include parsing.
The meat of the CL is the subtle changes in Compiler, and the logic in
cloneBuiltinVariables. Note that we need to clone the global variable
declaration element (because one of the goals is to get rid of shared
and inherited program elements), but also the variable itself (and the
new copy needs to live in the program's symbol table).
Bug: skia:10589
Bug: skia:10679
Bug: skia:10680
Change-Id: Ied352f8434dac2b8eacb4e515b014b6af7b57d20
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319023
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
We don't use the entrance ID values anywhere; we only check to see if
the block is reachable or not.
Change-Id: I0988e6c1999936de25dd04404409b736d8ad14f5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321540
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
The inliner generally only cares whether a function has 50 nodes or
less. Once we hit the 50th node, we can stop counting.
Change-Id: I92918989a9b3b5b73c1d1f13fa25c9adfa5b7e40
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321198
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
fa082bb58b..b09f16ee7a
2020-10-01 jmadill@chromium.org Don't pull in Vulkan headers in Skia standalone.
2020-10-01 cnorthrop@google.com Capture/Replay: Minor fixes for MEC
2020-10-01 angle-autoroll@skia-public.iam.gserviceaccount.com Roll SPIRV-Tools from fcb22ecf0f7e to 615fbe6cbc78 (6 revisions)
2020-10-01 angle-autoroll@skia-public.iam.gserviceaccount.com Roll SwiftShader from b042f4e70879 to 5e947adaf26e (4 revisions)
2020-10-01 angle-autoroll@skia-public.iam.gserviceaccount.com Roll Vulkan-ValidationLayers from e2b329a75b6a to 7a78cc629cca (3 revisions)
2020-10-01 angle-autoroll@skia-public.iam.gserviceaccount.com Roll Chromium from 18cdc3179c58 to 5efa84e265f7 (627 revisions)
2020-10-01 jmadill@chromium.org Integrate Skia Gold with restricted trace tests.
2020-10-01 jmadill@chromium.org Drop isolated perf output arg in test runner.
2020-10-01 timvp@google.com Vulkan: Always reset during createPipelineLayout()
2020-10-01 ianelliott@google.com Vulkan:Add debug labels for OpenGL calls
2020-09-30 courtneygo@google.com Move LayoutCaches to ShareGroup
2020-09-30 cclao@google.com Vulkan: disable glFlush deferral optimization for QualComm GPUs.
2020-09-30 jmadill@chromium.org Fix info collection on Android without Vulkan.
2020-09-30 jmadill@chromium.org Additional perf test cleanup & functionality.
2020-09-30 courtneygo@google.com Vulkan: Check that its okay to add commands
2020-09-30 syoussefi@chromium.org Vulkan: Constants for unpacked D/S attachment indices
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 brianosman@google.com on the revert to ensure that a human
is aware of the problem.
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/master/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: brianosman@google.com
Test: Test: Manually verify Cut the Rope doesn't crashTest: Test: angle_perftests --gtest_filter="*Trace*"
Change-Id: I07c6099e1cf1506ed9fb5cd85b4b470386db2c20
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321479
Reviewed-by: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Commit-Queue: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Change-Id: I8ab093c68e0c70d78722589bcfd28934f2cf5cb9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321517
Commit-Queue: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The SkRasterClip on DeviceCM appears to be unused and just increased the
size of that struct. Now the clip is tracked on the MCRec or within
SkBitmapDevice itself.
This adjusts the reserved size for DeviceCM to match (as reported on my
machine), and moves it into an anonymous namespace.
Change-Id: Ifa8635f839fc3d75339a64c9378ee054fe0cc071
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321250
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Avoids double call of release proc on failure.
Bug: skia:9832
Change-Id: I517c010376ba73f6d98c95f3453c8b0f9a5a299a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321461
Commit-Queue: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Auto-Submit: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Change-Id: Ib97cbf2e20580d3b9e54a5ffe793013ec97d4892
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321539
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Part of this change is to move some of this static format information
off of GrCaps since it is not cap dependent in anyway. This allows us
to the need for caps in many places. Also changes the low level format
query to be based off of bytes per block so it can be shared for
compressed and non compressed formats.
This change will also make it easier to add stencil/depth formats in
follow on change since we don't have to fill in a whole caps
FormatInfo block just so we can get the bytesPerPixel which is all
they need.
Bug: skia:10727
Change-Id: I2e6fdabf3ed699b4145ef9e6f0a73078d32a0444
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321463
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This reverts commit 135e237656.
No-Tree-Checks: true
Change-Id: I7e9dd2148f7b2a8dee1e49a9a9cc593e0d7ceb6e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321460
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Instead of being passed separately.
Check sample count in various onWrap methods in GrMtlGpu.
Bug: skia:9832
Cq-Include-Trybots: luci.skia.skia.primary:Test-iOS-Clang-iPhone8-GPU-AppleA11-arm64-Release-All-Metal,Test-iOS-Clang-iPhone7-GPU-PowerVRGT7600-arm64-Release-All-Metal,Test-iOS-Clang-iPadPro-GPU-PowerVRGT7800-arm64-Release-All-Metal,Test-iOS-Clang-iPhone11-GPU-AppleA13-arm64-Release-All-Metal,Test-iOS-Clang-iPhone6-GPU-PowerVRGX6450-arm64-Release-All-Metal,Test-Mac10.15-Clang-MacBookAir7.2-GPU-IntelHD6000-x86_64-Release-All-Metal
Change-Id: If5813db95b07f5d272e80920486f461cc5a587fc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320956
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Maybe a hair faster for OOP-R/DDL-style rendering of the shadow_utils GM.
Bug: 1108408
Change-Id: If8827f7a765b8fc231f7a37178b7a12dc78357ea
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321159
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
No discernable perf improvement for OOP-R/DDL-style rendering of the matrixconvolution* GMs.
Bug: 1108408
Change-Id: Iea016feb2a8a0e5303a64ee8166c3ef089bb2649
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321137
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Using return statements generates better code when there are no early
exits.
Change-Id: I18edd49991324aaf1473b614261862d273eac541
Bug: skia:10549
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321237
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Similar to the same field on Enum and FunctionDeclaration, will be used
to facilitate cloning builtin variables into Programs that use them.
Bug: skia:10589
Change-Id: Ic63701c61ee4658a5ec72adb506cc96aa0b2836f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321196
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Makes it simpler to call methods that update the internal ref counted
state.
Change-Id: Iecefab2c14ff13eae2fa34eafdcc7e031b3c85b0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320957
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
By using `return` directly instead of creating a temporary `_output`
variable, we generate less code and simplify the final GLSL output.
In the case of OverrideInputFragmentProcessor, the ternary expression
is performed on a `layout(key)` bool and will always be optimized away.
Before:
vec4 _10_ConstColorProcessor_Stage1_c0;
{
vec4 _12_output;
_12_output = ucolor_Stage1_c0;
_10_ConstColorProcessor_Stage1_c0 = _12_output;
}
After:
vec4 _8_ConstColorProcessor_Stage1_c0;
{
_8_ConstColorProcessor_Stage1_c0 = ucolor_Stage1_c0;
}
Change-Id: Ic3fd446345aa73e3936d1759443b80b41f5468f0
Bug: skia:10549
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320962
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
For mipmaps, the texture view we need depends on the mipmap mode. In
particular, for mipmap "none" mode, we need a view which ignores
everything but mip 0. So remove the persistent one.
Change-Id: I37a5e8a299edf423b67fd85135d6fbd9d26e62a7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320636
Commit-Queue: Stephen White <senorblanco@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Android is turning on a flag which zero's all malloc calls for security
purposes. Since allocations with sk_malloc_flags are potentially very
large, and this flag is known to regress some HWUI usecases, disabling
this initialization for graphics buffers (which we know will get filled
anyway).
Bug: 131355925
Test: removes large hwuimacro regressions w/ heap initialiation
Change-Id: I7ae11b075f88d03c0a9709928ad7bbbc3b04f73f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320266
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Auto-Submit: Steven Moreland <smoreland@google.com>
Commit-Queue: Derek Sollenberger <djsollen@google.com>
Change-Id: If2b1e8f89730bfff9e08c1ff5f5cb02c16088d86
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321117
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This, maybe, yields a slight perf improvement for the blur_matrix_rect
GM but, regardless, is good form.
Bug: 1108408
Change-Id: I6d589cc0f08f4264419ab33f27ac5a503cd036e8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320933
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The X and Y values of `clampedCoord` were being calculated and stored
separately, even in cases where work could easily be done in tandem.
Updated the code so that we use .xy when it makes sense to do so.
This CL is a partial re-attempt of http://review.skia.org/318436 but
changing only one variable instead of three.
Change-Id: I8cab3c2225e1d95bc77371f85d46ab45a184e145
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321116
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
We don't need to create a temporary variable for expressions like
`half3(x)`.
Change-Id: Ie0fa6a6dfb3d77d4372f96c676d3081f7e278852
Bug: skia:10786
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320960
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Previously, this would only return kClippedOut if the draw didn't
intersect with the device bounds. Now takes into account the bounds of
any identified single rrect/rect clip element and reports kClippedOut
if the draw does not intersect the element.
This fixes the bug identified in flutter where draws would appear outside
the clip. Within the problematic drawPoints call, each segment would turn
into a separate rect to draw, and if the rect was fully outside the clip,
it would incorrectly pass preApply and then not get cropped properly since
it didn't actually intersect the crop rect.
This only presented as a bug when SK_DISABLE_NEW_GR_CLIP_STACK was defined
(the current behavior for flutter). The new GrClipStack preApply already
reported kClippedOut in this scenario.
Bug: https://github.com/flutter/flutter/issues/66719
Change-Id: Idd810efcce4dfb6b51745d4a7a2ea66de9d4f46b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320965
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
This allows us to traverse a program's hierarchy and make changes (as
long as the structure remains intact). It's the caller's responsibility
to make sure they don't invalidate any iterators of the ProgramWriter.
Change-Id: Icfc651134d916e19b92004c92fe09880bb96600b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320717
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Since we are now referencing the exact methods instead of calling
GetProcAddress, there is no need to cast the function pointers. This
also catches a bug where the emscripten signatures for glWaitSync and
glClientWaitSync did not match the OpenGL API.
Change-Id: If365a8f6090961cb4b44146d87e102430bc5b30f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320659
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
The "sksl_huge" test has been renamed to "sksl_medium."
Added a new "sksl_large" test which contains a dithered gradient shader.
Change-Id: Ia8488460022008ae4efecb5593f01db06c1c283f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320827
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit dd33b3ea90.
Change-Id: I348b2b5976966a7451d88bd7f96ce17ce1702b79
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320826
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Remove separate constructor arg from GrBackendRenderTarget.
Add appropriate checks in onWrap functions in GrD3DGpu to limit to
previous functionality.
Bug: skia:9832
Cq-Include-Trybots: luci.skia.skia.primary:Test-Win10-MSVC-Golo-GPU-QuadroP400-x86_64-Debug-All-Direct3D,Test-Win10-Clang-ShuttleA-GPU-RadeonHD7770-x86_64-Debug-All-Direct3D
Change-Id: Ieba398d4bc00d18fdb0adffcda81f1a166db24b3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320757
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
