Change-Id: I0af800900a7fbd9d16af0058ee0754358ebc3875
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293562
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Change-Id: I0ff358665c77b9826ee76befc4490b2204f2fa4a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294698
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
We are updating FPs to receive their input via a child FP where
possible, instead of relying on the input color.
Change-Id: I41cd5cba7225c7b99c656d075145588b12b49ffc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294604
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
There can be only one SkFontMgr::Factory, so each should go in its own
translation unit. In addition, the name SkFontHost_mac is now rather out
of date and should now be SkFontMgr_mac_ct. Since the additional burden
of additional build changes after the first is minimal, also split out
SkTypeface_mac_ct and SkScalerContext_mac_ct. The original
SkFontHost_mac.cpp is kept as a shell which #includes the cpp files
which are replacing it. Once references to it are removed from all
builds it can be removed.
This is intended to be a reorganization without much code change. Most
changes are simple renaming of functions which are now shared between
translation units. However, there are a few behavior changes here.
* Drop SkTypefaceCache global for SkTypeface_Mac 'local' global.
* SkCTFontCTWidthForCSSWidth returns CGFloat instead of 'int'.
* SkFontMgr_New_CoreText takes a CTFontCollectionRef.
Change-Id: Iaf58a0371667f266ada20c918941fab6bc27d9df
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294456
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
An 'optional' target my wish to define public includes. When the target
is not enabled those public includes should not be present.
Change-Id: I1d0fc4553df100adf9178ad3f57bc0f5bb8f0fcb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294658
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Dawn does not support writePixels to a TextureAsRenderTarget, so update its caps to reflect that.
Also update the WritePixelsNonTexture_Gpu test to check the caps.
Bug: skia:10333
Change-Id: I4dcf2e0fecc34cba9586e5cab71739e427301efb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294597
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
This will allow multi-stage draw operations which use these FPs to
honor the input color.
Change-Id: I96ba3ff8720ed9fbd58b07bbe6315ffbbc7cf5b6
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294656
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Like Vulkan, Dawn requires "early abandon", so that the wgpu::Device is
freed before the dawn_native::Instance which created it.
Bug: skia:10334
Change-Id: I72346b724868790ede69c3cb09641e9054ef738e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294602
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
The SkAtlasTextTarget forced W for drawing using 3D vertices. Since it
is gone, the force W system is not needed.
Change-Id: Ic70b2f9aa5fd845dcfa7b06b5905afeb72f96441
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294600
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Our "header" reading and writing code didn't agree, so we always failed
to recognize cached program binaries. The asserts in the testing sink
failed to notice, because we did get a 100% cache hit rate, but we
immediately discarded the data we received.
We now also check that we didn't insert anything into the cache, as a
proxy for doing any shader compile work. That change, plus the tweak to
set cached=false when the header fields are invalid (like we do if we
encounter problems further in the blob) detected the problem. Adding the
version tag to the start of the encoded blob fixes the test, and means
that program binary caching is actually working again.
This code still looks (and is) fragile. The next CL is going to rewrite
things to use SkReadBuffer and SkWriteBuffer, make the parsing code less
brittle, and give us a more robust way to detect failure anywhere in the
stream.
Bug: skia:9402
Change-Id: I0329f088e0afce3998494d91ef2206e5eb9cac42
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294599
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Deletes support for the "side" temp command buffer
Bug: chromium:1087124
Change-Id: I97cda4e98faddd2d65f257613e19a825f52402c1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294518
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Change-Id: Ifbe0f3ae6e01d65f18351903da8aef63170ce6c7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294457
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Auto-Submit: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Adds a method to determine the worst-case number of tessellated line
segments that a path might require, and disables hardware tessellation
if it is more segments than are supported (falling back on indirect
draw shaders).
If the path requires even more segments than are supported by the
indirect draw shaders (1024), we crop the path to the viewport. The
required number of segments is proportional to the square root of the
bounding box's diagonal, so we won't start cropping paths until their
device-space bounding box diagonal is nearly 175,000 pixels long.
Change-Id: I8a9435e70bb93dda3464cc11a3e44fbe511744ae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293691
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Change-Id: I792f79eb99106b36b8a53af01ee55898bd2041ac
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294566
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Herb Derby <herb@google.com>
My experience porting the old opts over to the current SKX opts setup
was so bad that I don't want to try any more, and think it's probably
safer to port the SKX code to the old setup.
Need to hook up dependents (Chromium and Google3 I think) before we can
move the actual SXK opts code over.
Change-Id: Ibb8bc4a083cb104cd39f27cbfbc16e9eedd9bd46
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294495
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
The AbandonedContextImage test destroys the GrContextFactory (and the
DawnTestContext) before destroying the GrDawnGpu, which leaves its wgpu::Device
alive after the dawn_native::Instance which created it is freed.
The fix is to drop the ref to the Queue and Device in GrDawnGpu::disconnect().
Bug: skia:10332
Change-Id: I757d0f6e902f700d67d1d3e0d5b2d28487528b6e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294517
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
The bug here is very subtle, as is the mitigation.
Quick background on WASM memory, there is an object
called wasmMemory (which might be hoisted into scope for
CanvasKit's pre-js functions), of type WebAssembly.Memory
which is a resizable ArrayBuffer. Emscripten provides the
JS code to initialize this and handle size increases.
Emscripten also provides TypedArray "views" into this buffer.
These are called CanvasKit.HEAPU8, CanvasKit.HEAPF32, etc.
When there is a call to CanvasKit._malloc, wasmMemory may
be resized. If that happens, the previous TypedArray views
become invalid. However, in the same call to _malloc,
emscripten will refresh the views [1]. So, dealing with
CanvasKit.HEAPU8 directly (quick aside, we never expect clients
to mess with these views, only us in our glue JS code
[e.g. interface.js]), should always be safe because if they
were to be invalidated in a call to _malloc, the views would
be refreshed before _malloc continues.
The problem that existed before was when we were passing
CanvasKit.HEAP* as a parameter to a function, in which the
function would call _malloc before using the typed array
parameter:
//... let us suppose wasmMemory is backed by ArrayBuffer D
copy1dArray(arr, HEAPU32);
// The HEAPU32 TypedArray (backed by ArrayBuffer D) is stored
// to a function parameter "dest"
function copy1dArray(arr, dest, ptr) {
// ...
if (!ptr) {
ptr = CanvasKit._malloc(arr.length * dest.BYTES_PER_ELEMENT);
// Suppose _malloc needs to resize wasmMemory and is
// now backed by ArrayBuffer E.
// Note: The field CanvasKit.HEAPU32 is correctly backed
// by ArrayBuffer E, but variable dest still points to a
// TypedArray backed by ArrayBuffer D.
}
// dest.set will fail with a "neutered ArrayBuffer" error
// because ArrayBuffer D is effectively gone (replaced by E).
dest.set(arr, ptr / dest.BYTES_PER_ELEMENT);
The fix here is to pass in the field name indicating the TypedArray
view we want to write our data into instead of using the
view itself as the parameter.
[1] e427159553/src/preamble.js (L344)
Change-Id: I46cfb98f8bdf928b61690a5ced034a5961356398
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294516
Reviewed-by: Nathaniel Nifong <nifong@google.com>
This has become a weird target to optimize for. It's not hugely better
than SSE2, and AVX2 is common nowadays, so SSE4.1's sweet spot is small.
Change-Id: I31c62fb371a4902f084f78309e7a43955f1fc0c2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294320
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
989c790d4b..b10f4b94aa
2020-06-04 ynovikov@chromium.org doc: Chromium branch autorollers
2020-06-04 etienneb@chromium.org Add trace event to angle Program compilation API
2020-06-04 syoussefi@chromium.org Allow lines up to 72 characters in commit messages
2020-06-04 jmadill@chromium.org Roll third_party/vulkan_memory_allocator/ dac952eb0..431d6e572 (4 commits)
2020-06-04 jmadill@chromium.org Rename Platform.h to PlatformMethods.h.
2020-06-04 nguyenmh@google.com Allow line formats to pass description body's line length check
2020-06-04 spang@chromium.org Re-enable ShouldClearOpaqueFdRGBA8 on Pixel2 Vulkan
2020-06-04 xiaoxuan.liu@arm.com Vulkan: Cleanup unused variable in submitFrame()
2020-06-04 angle-autoroll@skia-public.iam.gserviceaccount.com Roll glslang from 78a3c915a1d7 to 232ba0d848d1 (1 revision)
2020-06-04 angle-autoroll@skia-public.iam.gserviceaccount.com Roll Vulkan-ValidationLayers from d43f5107400d to 4685ffce93ad (5 revisions)
2020-06-04 angle-autoroll@skia-public.iam.gserviceaccount.com Roll SwiftShader from 3ad285a60d82 to 0711869b5f20 (1 revision)
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 tdenniston@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
Bug: chromium:1064662
Tbr: tdenniston@google.com
Change-Id: I6403b64bac593b5fe3daa2acce8afdcd7af7c00a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294453
Reviewed-by: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Commit-Queue: skia-autoroll <skia-autoroll@skia-public.iam.gserviceaccount.com>
Mainly rearranging the code to perform all ICU iterations once
and cache the results for the next text layouts.
Change-Id: I514d04229d04778c1f2238064acccddf6b548c00
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294400
Commit-Queue: Ben Wagner <bungeman@google.com>
Commit-Queue: Julia Lavrova <jlavrova@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
This reverts commit 9c6f6bf031.
Reason for revert: likely hurt flutter perf
https://github.com/flutter/flutter/issues/58716
Original change's description:
> Add 2x2 bilinear downscale steps to GrSurfaceContext::rescale.
>
> Currently just uses 2x2 configuration but could be adapted easily
> to support other sample position layouts.
>
> Change-Id: If857bac5d1fab1b6ee04694e445d0f19431646f1
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/292079
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
TBR=bsalomon@google.com,michaelludwig@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I8c322c02e22622ab61b99e4a8c241b30a2a31b12
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294399
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
We are updating FPs to receive their input via a child FP where
possible, instead of relying on the input color.
Change-Id: Ib2bac2c7865d4285fbbf45c9de29bbd4d139d04b
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294396
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 4926b07217.
Reason for revert: fix wip
Original change's description:
> Revert "Improve scissor state tracking in GrRTC"
>
> This reverts commit 3b923a880b.
>
> Reason for revert: GrAppliedHardClip isn't tracking scissor state properly
>
> Original change's description:
> > Improve scissor state tracking in GrRTC
> >
> > At a low level, this changes GrScissorState from a rect+bool to a rect+size.
> > The scissor test is considered enablebd if the rect does not fill the
> > device bounds rect specified by the size. This has a number of benefits:
> >
> > 1. We can always access the scissor rect and know that it will be
> > restricted to the render target dimensions.
> > 2. It helps consolidate code that previously had to test the scissor rect
> > and render target bounds separately.
> > 3. The clear operations can now match the proper backing store dimensions
> > of the render target.
> > 4. It makes it easier to reason about scissors applying to the logical
> > dimensions of the render target vs. its backing store dimensions.
> >
> > Originally, I was going to have the extra scissor guards for the logical
> > dimensions be added in a separate CL (with the cleanup for
> > attemptQuadOptimization). However, it became difficult to ensure correct
> > behavior respecting the vulkan render pass bounds without applying this
> > new logic at the same time.
> >
> > So now, with this CL, GrAppliedClips are sized to the backing store
> > dimensions of the render target. GrOpsTasks also clip bounds to the
> > backing store dimensions instead of the logical dimensions (which seems
> > more correct since that's where the auto-clipping happens). Then when
> > we convert a GrClip to a GrAppliedClip, the GrRTC automatically enforces
> > the logical dimensions scissor if we have stencil settings (to ensure
> > the padded pixels don't get corrupted). It also may remove the scissor
> > if the draw was just a color buffer update.
> >
> > Change-Id: I75671c9cc921f4696b1dd5231e02486090aa4282
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/290654
> > Commit-Queue: Michael Ludwig <michaelludwig@google.com>
> > Reviewed-by: Brian Salomon <bsalomon@google.com>
>
> TBR=bsalomon@google.com,csmartdalton@google.com,michaelludwig@google.com
>
> Change-Id: Ie98d084158e3a537604ab0fecee69bde3e744d1b
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294340
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
TBR=bsalomon@google.com,csmartdalton@google.com,michaelludwig@google.com
# Not skipping CQ checks because this is a reland.
Change-Id: I2116e52146890ee4b7ea007f3c3d5c3e532e4bdd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294257
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
This reverts commit cc6349d390.
Reason for revert: Problems with MSAN
Original change's description:
> ICU optimization
>
> Mainly rearranging the code to perform all ICU iterations once
> and cache the results for the next text layouts.
>
> Change-Id: I2c2a502c705510eb169bf62efbfcc13b658591e3
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293336
> Commit-Queue: Julia Lavrova <jlavrova@google.com>
> Reviewed-by: Ben Wagner <bungeman@google.com>
TBR=bungeman@google.com,jlavrova@google.com
Change-Id: I7f7f759178c10349b4c879bafc68a7f8e1065b6a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294398
Reviewed-by: Julia Lavrova <jlavrova@google.com>
Commit-Queue: Julia Lavrova <jlavrova@google.com>
It's rare these days to find a CPU with AVX but not AVX2.
The memsets still make sense to leave, as we don't do anything
different there even with AVX2... all the instructions we want
are available with AVX.
Change-Id: I412f95c4548cfcc56c19b72348deceee83ba0610
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294319
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Mainly rearranging the code to perform all ICU iterations once
and cache the results for the next text layouts.
Change-Id: I2c2a502c705510eb169bf62efbfcc13b658591e3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293336
Commit-Queue: Julia Lavrova <jlavrova@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
This reverts commit 3b923a880b.
Reason for revert: GrAppliedHardClip isn't tracking scissor state properly
Original change's description:
> Improve scissor state tracking in GrRTC
>
> At a low level, this changes GrScissorState from a rect+bool to a rect+size.
> The scissor test is considered enablebd if the rect does not fill the
> device bounds rect specified by the size. This has a number of benefits:
>
> 1. We can always access the scissor rect and know that it will be
> restricted to the render target dimensions.
> 2. It helps consolidate code that previously had to test the scissor rect
> and render target bounds separately.
> 3. The clear operations can now match the proper backing store dimensions
> of the render target.
> 4. It makes it easier to reason about scissors applying to the logical
> dimensions of the render target vs. its backing store dimensions.
>
> Originally, I was going to have the extra scissor guards for the logical
> dimensions be added in a separate CL (with the cleanup for
> attemptQuadOptimization). However, it became difficult to ensure correct
> behavior respecting the vulkan render pass bounds without applying this
> new logic at the same time.
>
> So now, with this CL, GrAppliedClips are sized to the backing store
> dimensions of the render target. GrOpsTasks also clip bounds to the
> backing store dimensions instead of the logical dimensions (which seems
> more correct since that's where the auto-clipping happens). Then when
> we convert a GrClip to a GrAppliedClip, the GrRTC automatically enforces
> the logical dimensions scissor if we have stencil settings (to ensure
> the padded pixels don't get corrupted). It also may remove the scissor
> if the draw was just a color buffer update.
>
> Change-Id: I75671c9cc921f4696b1dd5231e02486090aa4282
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/290654
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com,csmartdalton@google.com,michaelludwig@google.com
Change-Id: Ie98d084158e3a537604ab0fecee69bde3e744d1b
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294340
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
At a low level, this changes GrScissorState from a rect+bool to a rect+size.
The scissor test is considered enablebd if the rect does not fill the
device bounds rect specified by the size. This has a number of benefits:
1. We can always access the scissor rect and know that it will be
restricted to the render target dimensions.
2. It helps consolidate code that previously had to test the scissor rect
and render target bounds separately.
3. The clear operations can now match the proper backing store dimensions
of the render target.
4. It makes it easier to reason about scissors applying to the logical
dimensions of the render target vs. its backing store dimensions.
Originally, I was going to have the extra scissor guards for the logical
dimensions be added in a separate CL (with the cleanup for
attemptQuadOptimization). However, it became difficult to ensure correct
behavior respecting the vulkan render pass bounds without applying this
new logic at the same time.
So now, with this CL, GrAppliedClips are sized to the backing store
dimensions of the render target. GrOpsTasks also clip bounds to the
backing store dimensions instead of the logical dimensions (which seems
more correct since that's where the auto-clipping happens). Then when
we convert a GrClip to a GrAppliedClip, the GrRTC automatically enforces
the logical dimensions scissor if we have stencil settings (to ensure
the padded pixels don't get corrupted). It also may remove the scissor
if the draw was just a color buffer update.
Change-Id: I75671c9cc921f4696b1dd5231e02486090aa4282
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/290654
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Apple's Clang has a bug rooted in these three conflicting constraints:
A) before we save a large amount of state to the stack, call
__chkstk_darwin() to um, do something to make that safe. It's
some sort of debug feature that involves marking various pages
as writable and unwritable to catch stack overflows.
B) before calling any function that doesn't use AVX, which includes
__chkstk_darwin(), call vzeroupper as a performance enhancement
C) we must save a large amount of state to the stack before it's
sound to call vzeroupper, so it can be restored after the
function call. Otherwise the upper 128 bits of all ymm registers
will be lost, zeroed.
There's no way to order A,B, and C to make them all happy.
Saving registers before zeroing them (C) is a correctness issue, so it's
got to take precedence. Zeroing the upper bits of ymm registers before
calling into code that's not ymm-aware (B) is a performance issue only,
and not actually even needed for __chkstk_darwin(). The whole
__chkstk_darwin() thing (A) is nice a safety feature we'll have to live
without.
The best fix would be to make it so that Clang doesn't issue vzeroupper
before __chkstk_darwin(), but we can't do that here. The next best
thing, since mucking with correctness isn't really viable, is to turn of
the stack checking.
Until now we've been using -O1 to reduce the stack usage down low enough
that the stack check calls are not generated. Maybe there's a known
safe single page, and the call's only needed when going above that?
Anyway, let's try focusing our workaround on -fstack-check itself.
Bug: skia:9709
Change-Id: Ie236101d2d464526b33e327db1f94574a7a86948
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294326
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Spin off the safer part of the SkOpts CL I had to revert.
Should we try that again, having this in place will make
it easier.
Change-Id: Ic34ce2f23c10466cf1036f9651a544d308f4edec
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294294
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
We are updating FPs to receive their input via a child FP where
possible, instead of relying on the input color.
This CL also adds a GM test for SwizzleOutput, since this did not appear
to be covered by existing unit tests.
Change-Id: I3d8176395bb42eab7ff471c9137597402b5b3a69
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293884
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
We are updating FPs to receive their input via a child FP where
possible, instead of relying on the input color.
Change-Id: Ic30224a45c6e4dba46c1cbaa27735013e8b3935f
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293937
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Move SkFontHost_mac.cpp into an 'optional' like the other fontmgr build
rules. This allows building with other fontmgrs on Mac and makes the lib
dependencies explicit. In the future this helps with splitting the out
the default factory.
Change-Id: Iecef9e428acb69f89b54afa00b4e779f5858c61a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294076
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Without this change, the generated code for an FP that includes a
fragment processor and a @setData block will erroneously include:
auto inputFP = _outer.inputFP;
(void)inputFP;
This is not a valid member name. (inputFP_index does exist.)
There does not appear to be any reason to generate this line.
Change-Id: I324878c0768959e56be33cfd8a85a2aa42794c7c
Bug: skia:10217
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/294039
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
These tests create a wgpu::Texture which outlives the wgpu::Device
which created it. This is problematic on Dawn, so just skip
the tests for now.
Bug: skia:10326
Change-Id: I29d5328e313baca19a2bbc2eedd80f2768cae77e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293936
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Mirrors what we have in Vulkan -- there's a cache of samplers in the
resource provider, and views are associated with their particular
texture resource.
Change-Id: I3751bce0817581a0044a7eea1b0c3ef9cc14d8e1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/293537
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
