The original code here used a std::recusive_mutex. Looking at the code
I don't see today how any of the code that grabs the mutex could call
back into a function that grabs the mutex again on the same thread.
Maybe this was more of an issue back when this was originally added?
Regardless today it should be safe to just use an SkMutex.
Change-Id: I81c8ea3bf5b2defe893dd9fff7a6a2eda10cace8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391256
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
There's no GPU dimension on these bots yet,
but we can use the CPU to constrain it just fine.
Bug: skia:11334
Change-Id: I2d0928ced7a9145a7160a40ea2a5bd2b1601a01b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390797
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Disabling it has caused regressions in some test slides.
Example: http://screen/5dpPVnJukBAcyKb
Change-Id: If302ffd39cd28215c557bbba444d9c8669dea1c8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391298
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>
This constructor takes a single argument and splats it diagonally across
an otherwise-zero matrix. These are also sometimes referred to as a
uniform-scale matrix.
Change-Id: I1ed8140f55f5cad4029015807b220d6475401daa
Bug: skia:11032
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390716
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
These changes cause massive ripple effects in the dehydrated code, so
it's been split out from http://review.skia.org/390716 to its own CL.
Change-Id: I5053ce3cf44ab5050c3a77ac764b4e9c8bfc00d5
Bug: skia:11032
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390896
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This will be used by the new stroke tessellator. All the other
tessellators should start chopping and chunking too. That will allow us
to quit cropping paths if we are afraid they might need more segments
than are supported.
Bug: chromium:1172543
Change-Id: I30f0ebb581f56cac099d8c05e0e181c4657c3db8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390096
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
We've found that the inliner only gives tangible gains in draw
performance on GLES devices. (See go/optimization-in-sksl-inliner)
On other devices, we can skip it and still get the same draw performance
regardless. A caps bit has been added to indicate a device that will
benefit from inlining, and the inliner is now disabled on platforms that
don't set this bit.
Change-Id: I61dfafd7e919deabf81529cea832bb11496410cc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390300
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This isn't something I expect to occur often. However, we have always
reported that array-ctors with constant arguments are compile-time
constants, and `compareConstant` is supposed to be implemented for all
compile-time constants.
Change-Id: I0bbfe2a3f78c29c14f69c3b6faca71067a0e45c8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391116
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:11334
Change-Id: I57343e80cc8b39644c5f189aeee0da229517a726
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390576
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
We don't directly support this today at all. In practice, though, simple
constant arrays are detected as equal in the constant-folding pass
because they hit the `x == x` self-equality check (using
`IsSameExpressionTree`).
This does not work for our inequality tests, though, so those do not
fold.
Change-Id: I6730a9a2d1da9ac613ee58889d651f3ff65b1d2d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391057
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>
Prevents introduction of a bug when modifying a type that is used
with one of these containers in such way that makes the use
unsafe.
Change-Id: I959e6f3bfbc9664a6b0ced636f5ae6cbe72eee1a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390676
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Add GrCPixmap, a GrPixmap but with const void* instead of void*. Share
impl via template base class GrPixmapBase.
Change-Id: I7dfdf24a73c1bc8557ff7b90f93a9399da2f3f75
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350022
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
We have another thread that may be adding pools back to the available
array the same time we are resetting it. Make sure we grab a mutex before
doing the reset.
Change-Id: I59c6acf7b654d326241f097fc995e17418d5c2e9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390480
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This adds the compute binding and dispatch methods to the command list.
For the moment we'll assume we're using root signature layouts that are
similar to the graphics setup -- this can be tweaked later. D3D has
separate bindings for graphics and compute in the same command list, so
the tracking for the active binding has been separated as well.
Change-Id: I5bec820d1fd17bb72a09366c354a7fa77325c6e8
Bug: skia:10446
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390516
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This specifically implements this in the vulkan backend to free up
command buffers/pools when the client is trying to free up as much gpu
memory as possible.
Change-Id: I1fecf55c80fe46cfc1d064f9dfbeb36b27c217d0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390479
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Simple refactor to make things reusable for the next non-hardware stroke
tessellator.
Change-Id: I0898b54a616e60f0475ac74cbd6f518e8696e0e5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390078
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
std::result_of is deprecated and removed in C++20, std::invoke_result is
in C++17 the library so not ready yet.
In the case of SkMakeArray there is no need for the full power of
std::invoke_result as SFINAE is not needed, nor is calling a method
supported. As a result a simple decltype is sufficient.
Bug: skia:11811
Change-Id: I6c07c380970cc2a991ab8bd457997320ba634e6c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390476
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Similar to existing SurfaceContextReadPixels but for writes. Tries all
combinations of src/dst color type and alpha type for write pixels.
Always reads back pixels for verification using the ImageInfo of the
tested surface context.
Bug: skia:8862
Bug: skia:11130
Bug: skia:11786
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348886
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Change-Id: I22b38a99c8b4e5b506cd0b94ab9cac3d347f6b2e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388236
Previously, the Compiler optimization overrides only supported force-
disabling a particular behavior; we assumed that the default state of
a Compiler was to allow all optimizations. This assumption is about to
be invalidated, as the Inliner will soon be off by default unless you
are using GLES.
The override flags are now a tri-state; optimization and inlining can
be set to "default", "on" or "off".
Change-Id: I5637693222ca1de74ca1073c24d86c8e7c5026f6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390136
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>
With the new drawing API in Ganesh it should be relatively
straightforward to chain FPs with different textures.
Bug: skia:11396
Change-Id: I57cde39a2c00c79ea8b9b907c0ee8a034f977a14
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/386616
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This reverts commit 8b6788079f.
Reason for revert: Fixed variable type for chrome & flutter
Basically, you can't use objc_externally_retained with `auto` type
variables because the compiler isn't smart enough to deduce the type
before checking that it's compatible with the attribute.
Original change's description:
> Revert "Use objc_externally_retained to cut down retain/release in Metal"
>
> This reverts commit 78381ee5e0.
>
> Reason for revert: breaking chromium and flutter mac builds
>
> Original change's description:
> > Use objc_externally_retained to cut down retain/release in Metal
> >
> > Until we get off of ARC, these extra retains and releases can really
> > add up in previous profiling of Metal, especially on iOS.
> >
> > Change-Id: Ib3aaa0b863c4469bdb6ad930395a170719c0293d
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381110
> > Commit-Queue: Adlai Holler <adlai@google.com>
> > Reviewed-by: Jim Van Verth <jvanverth@google.com>
> > Reviewed-by: Brian Salomon <bsalomon@google.com>
>
> TBR=jvanverth@google.com,bsalomon@google.com,adlai@google.com
>
> Change-Id: I36aff7698a70052a42b91a910b92fd512b3eecd4
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388717
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Change-Id: I5cad02fc9d8c7f21e22021858b2e6e9b038a3de6
Cq-Include-Trybots: luci.skia.skia.primary:Canary-Flutter
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389157
Commit-Queue: Adlai Holler <adlai@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
The main change here is just fusing the following three methods into recreateSKP:
deflateSKP
createCallbackContexts
createSKP
All the remaining changes are just API fallout from that.
Bug: skia:11728
Change-Id: Iae2ce65983ad56c8288bdc830e248394a0055bfb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389925
Reviewed-by: Adlai Holler <adlai@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Previously it aborted if at a downscale to 1 pixel the effective
sigma was greater than the max sigma.
For repeat/mirror the single col/row is the blurred result
For decal/clamp add a border of transparent/original border and then
recurse.
Bug: skia:11735
Bug: chromium:1174354
Change-Id: I3438f49fdc5167f6def61a49f9fc8eaef3912ec3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385279
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
This starts to move us in the right direction for DMSAA support. In
future CLs it will allow GrVkOpsRenderPass to depend/use a
GrVkFramebuffer object when setting things up instead of a GrVkRenderTarget.
This makes it easier to dynamically tell it to use an msaa color attachmet
or the normal non-msaa attachment from the render target.
A bonus to this is that we can remove the ugly addResources call on the
GrVkRenderTarget.
Bug: skia:11809
Change-Id: I7e5db64210830994ec507e6bbc4d3d1e1a9ed4e0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390036
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Variable and VarDeclarations cross-reference one another. They generally
get deleted around the same time, but this is not always the case. In
this CL we explicitly detach them from each other at destruction time
to avoid holding a stale pointer, removing the risk of accessing it
later. (Accessing null is still fatal, of course, but it's less
dangerous than using a recently-freed pointer, and easier to debug as
well.)
Modifying things in the symbol table requires a const_cast, but it's not
too risky to null out a pointer field on a conceptually-dead object.
A Variable without a VarDeclaration is inert.
Change-Id: Ie01244495a82a8007269522a561b2512c5f12384
Bug: oss-fuzz:32587
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390056
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This skp started causing
Perf-Debian10-EMCC-GCE-CPU-AVX2-wasm-Release-All-Puppeteer_RenderSKP
to timeout.
Change-Id: I248b4be8f2287ecb16969f71f985acd794bd8d11
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390298
Reviewed-by: Ravi Mistry <rmistry@google.com>
This is a reland of 2a88456c15
Original change's description:
> Don't reuse scratch textures when texsub image banned for FBO textures
>
> If writePixels uses a scratch surface it will likely bind it to a FBO.
> If this workaround is in play that texture then becomes "poisoned" as
> a temp surface for a future writePixels(). However, our caching
> mechanism will still find it and we'll incur an extra recursion in
> writePixels to allocate a new texture that has never been bound to a
> FBO. This pattern can repeat itself with increasing recursion depth
> until the cache budget has been reached and we start purging the
> "poisoned" textures.
>
> Disallowing reuse of scratch textures avoids this issue.
>
> Bug: skia:11204
> Change-Id: I8002a673900a9a34e889c92e816781d4b7f84688
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388696
> Reviewed-by: Robert Phillips <robertphillips@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:11204
Change-Id: I0585497e0d0d09dd68cbef050a0bbb8f9a85c16d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390058
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
I tested sksl_small, sksl_medium and sksl_large to check their ratio of
characters to AST nodes and the amount varied between 8:1 and 12:1. The
parser now reserves a pool of 1 AST node per 10 source characters before
beginning the AST translation. This gives a ~3% improvement in
sksl_large on my machine, not a huge win per se but still pretty
significant given the simplicity of the change.
Change-Id: I028c6c251504065cbbd36b6199b18728ff2414e5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/390057
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: chromium:1176969
Change-Id: Ie3853bffe1e004c4a930b83201d7e89f82d1df12
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385196
Reviewed-by: Stephen White <senorblanco@google.com>
Commit-Queue: Stephen White <senorblanco@google.com>
This allows me to separate the wrapper around the ID3D12PipelineState
from the other data we track with GrD3DPipelineState. In this way I can
create a GrD3DPipeline directly for compute experiments without worrying
about managing graphics-specific objects. In the long run, we'll
probably need GrD3DGraphicsPipelineState and GrD3DComputePipelineState.
Change-Id: I36ff90a93b6e53fae217aaca3f6e0e76d698aa57
Bug: skia:10446
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389896
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
We've had a rule that the max render target size can't be larger than
the max texture size. The fMaxTextureSizeOverride option put us in a
situation where that rule could have been broken.
Bug: skia:11396
Change-Id: I403d0fd6af8c294e8a72e379690581523b0ce819
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/387856
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This resetting occurs when merging ops tasks, when we decide to skip
an ops task because a later ops task in the list has a clear color.
You can't just delete the op chains, you have to get all the stuff
out of the chains first or else you'll hit an assert.
Bug: skia:10877
Change-Id: I5377f242a67ab65a656783a61a534be4756f6c21
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389959
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Commit-Queue: Adlai Holler <adlai@google.com>
Auto-Submit: Adlai Holler <adlai@google.com>
