DSL API changes mean that we no longer need to be so heavy-handed with
wrapping statements and expressions. Not only does this simplify the
code, it's a significant performance improvement.
Change-Id: I26ed9701db8012448b67e1ca16328d4f8f04aa03
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/449037
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This ReportErrors call is fairly heavy (we create a lot of expressions)
and no longer appears to be necessary given the error handling changes
elsewhere.
Change-Id: Ib9433e4346104e31e7577d6fe4b10208a277a9ff
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448936
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
When binding descriptor tables, their associated heaps need to be bound
as well. Previously we would bind those heaps when allocating from them.
However, if we re-use a descriptor table later, its heap may no longer
be bound. So we need to be sure to bind heaps for the current set.
To avoid unnecessary refs, rather than store a
sk_sp<GrD3DDescriptorTableManager::Heap> in each descriptor table, we
only store its ID3D12DescriptorHeap pointer. The Heap only needs to be
added to the command list once, when it is first used to allocate for
the current submit.
Bug: skia:12359
Change-Id: I70018368e4f08bf7757969b9e878b0ff42059486
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448836
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Change-Id: I8d0704a73c80db6729811dc4ec343c409e6a2fb3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448838
Auto-Submit: Greg Daniel <egdaniel@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Previously, we had various loops which would synthesize a `splat(0)`
instruction on every iteration of the loop. This was harmless and was
already deduplicated by SkVM internally, but in a long loop, this
deduplication process could occur many many times in a row. This hurts
performance needlessly. In particular, for-loops can run for thousands
of iterations in a row, so the savings here can be substantial.
This was inspired by investigation into oss-fuzz:37837.
Change-Id: I247d9ec397418a65219930e2f2b8cf977354bddd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448837
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>
We're being stricter in V2 wrt this. This CL is an experiment to see if this fallback is actually relied on.
Change-Id: Id19a4ef3e98ee41a94ed2817a06966b87e2ef522
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448267
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
We already had a test case here, but it wasn't actually in operation.
The test has been split into ES2 (square) and ES3 (non-square) halves,
returns the color like a proper runtime effect, and it's now running in
dm.
Also, Metal doesn't natively support matrixCompMult, so it injects a
helper function; I tweaked the helper so it no longer requires an extra
result variable.
Change-Id: Ie79242768966fcbe879ad73461d17b4fb8e55670
Bug: skia:12202
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448117
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We didn't have any tests which exercised the non-square matrix case
(because such a test requires ES3), so it was silently broken. It's
now fixed. The tests exposed a DIFFERENT Quadro P400 bug which will be
fixed separately.
Change-Id: Icf24acad5ea6f18aea3d8aa5a903e7bea41a5c23
Bug: skia:12443
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448379
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We are down to only one working device.
Replacement devices are being discussed.
Change-Id: I74ef94360daaf0e901ddf094b9141d525e52878c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448696
Commit-Queue: Joe Gregorio <jcgregorio@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
Bug: skia:12302
Change-Id: I7ff7bae388c5991f2c23c8945355fea55c42095a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447436
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This reverts commit 481a58dfe0.
Reason for revert: Fixing the build
Original change's description:
> Revert "Small changes for SkText that are not in experimental directory"
>
> This reverts commit 92f1bc0083.
>
> Reason for revert: Blocking Android roll.
>
> Original change's description:
> > Small changes for SkText that are not in experimental directory
> >
> > (also made utf 8<->16 conversion static on SkUnicode)
> >
> > Change-Id: Ie64d18ad21a35ec10bd0b350fb7887fb78a419d8
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448140
> > Reviewed-by: Mike Reed <reed@google.com>
> > Commit-Queue: Julia Lavrova <jlavrova@google.com>
>
> Change-Id: Ieb9eb0495dddfc0f9e9e74861b24efc0201fd520
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448656
> Auto-Submit: Brian Osman <brianosman@google.com>
> Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
> Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Change-Id: I1e5859c9dd943c701d4c4e648bdc3e44412eca54
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448676
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Julia Lavrova <jlavrova@google.com>
Flutter uses a minimum deployment version of 9.0, and we keep breaking
their roll with unguarded features. This will help catch those sooner.
Change-Id: Idd98b2ac985c36f5c793ff27b5a4b59014875ee5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448257
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Erik Rose <erikrose@google.com>
Bug: skia:12302
Change-Id: Ifc107ca2cf13c1daa59521b93fe4ad1d3c215258
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447297
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:11837
Change-Id: I42ad29e3b5765a81ef9366ebdf1e95739cd90b78
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445297
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Bug: skia:12402
Change-Id: I743724f66db8d7666d4d627d6945ce6bc3dc6bc3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448261
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This reverts commit 92f1bc0083.
Reason for revert: Blocking Android roll.
Original change's description:
> Small changes for SkText that are not in experimental directory
>
> (also made utf 8<->16 conversion static on SkUnicode)
>
> Change-Id: Ie64d18ad21a35ec10bd0b350fb7887fb78a419d8
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448140
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Julia Lavrova <jlavrova@google.com>
Change-Id: Ieb9eb0495dddfc0f9e9e74861b24efc0201fd520
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448656
Auto-Submit: Brian Osman <brianosman@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
This exposes a bug in the Metal code generator which will be resolved
in a followup CL.
Change-Id: If073835dbee474ea9a805eb92b42dc1fca2afbd0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448378
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
On the Intel Iris Pro 580 this is failing, and causing command list
issues.
Bug: skia:12359
Change-Id: I4de1d89eaa4668983cfa3290aef13bd1ef940187
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448376
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This reverts commit 10c9f36bdd.
Reason for revert:bad gms, maybe blocking chrome roll
Original change's description:
> New approach to GrProcessor uniforms.
>
> The important aspect is that it allows knowing the uniforms that will
> be used by a set of processors without having to create ProgramImpls.
>
> GrProcessor subclasses specify uniforms at creation time in a similar
> style to how GrGeometryProcessors already specify attributes. That is,
> they initialize a span of structs describing the uniform which may
> contain void uniforms that are skipped. Unlike attributes, the struct
> contains an offset into the processor where the data is stored.
>
> GrUniformAggregator is used to collect the uniforms from all processors
> that compose a draw and mangle their names. The ProgramImpl subclasses
> query the aggregator for their uniform names when emitting code.
>
> The old system for uniforms is left intact and only three processors,
> one GP, one FP, and one XP, are updated to use the new system.
>
> Some pieces that are missing before everything can be moved over:
> -support for uniforms not owned by GrProcessor (e.g. rt-adjust)
> -support for samplers
> -helpers for common patterns
> (e.g. GrGeometryProcessor::ProgramImpl::setupUniformColor(),
> and the various matrix helpers on ProgramImpl)
>
> Bug: skia:12182
>
> Change-Id: I21c1b7a8940eb9b8aad003f5a2569e43977a33d2
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/440841
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:12182
Change-Id: I6cc508900a599d27124f8ba48597593192d5d807
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448418
Auto-Submit: Brian Salomon <bsalomon@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
While browsing with bloaty-treemap, I found that these were not marked
as const. This pessimizes us a bit (e.g. the memory pages need to be
treated as writable)
This has the effect of moving about 16K from __data into __TEXT,__const.
Change-Id: I9816a8d8f2a92777cbde719db5129bb7f9bc2ea4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448263
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The important aspect is that it allows knowing the uniforms that will
be used by a set of processors without having to create ProgramImpls.
GrProcessor subclasses specify uniforms at creation time in a similar
style to how GrGeometryProcessors already specify attributes. That is,
they initialize a span of structs describing the uniform which may
contain void uniforms that are skipped. Unlike attributes, the struct
contains an offset into the processor where the data is stored.
GrUniformAggregator is used to collect the uniforms from all processors
that compose a draw and mangle their names. The ProgramImpl subclasses
query the aggregator for their uniform names when emitting code.
The old system for uniforms is left intact and only three processors,
one GP, one FP, and one XP, are updated to use the new system.
Some pieces that are missing before everything can be moved over:
-support for uniforms not owned by GrProcessor (e.g. rt-adjust)
-support for samplers
-helpers for common patterns
(e.g. GrGeometryProcessor::ProgramImpl::setupUniformColor(),
and the various matrix helpers on ProgramImpl)
Bug: skia:12182
Change-Id: I21c1b7a8940eb9b8aad003f5a2569e43977a33d2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/440841
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Fix: just check for maskfilter, and reject fast-cases if we have one.
New GM: points_maskfilter
before: draw red squares with no blurs, but red circles had blurs
after: both sets have the blur (i.e. the maskfilter)
Bug: skia:12435
Change-Id: I460908818d2edf8c6065db2fe99a9df0e7ffc543
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448266
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Reed <reed@google.com>
(also made utf 8<->16 conversion static on SkUnicode)
Change-Id: Ie64d18ad21a35ec10bd0b350fb7887fb78a419d8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448140
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Julia Lavrova <jlavrova@google.com>
Throughout SkSL we've begun using doubles as a convenient way to store
any SkSL value (int, float, bool) in a single type. This idea has now
been extended to literals. Rather than having three expression kinds for
integers, floats and boolean literals, we can have just one. These can
be accessed in a type-specific way (`floatValue`, `intValue`, and
`boolValue` return the expected type, or assert if it's not the
matching type), or in a type-agnostic way (`value` will return a double
and works on any type of Literal).
This allows us to remove a complex template trick (Literal<T> is gone),
removes two redundant Expression types, and and lets us reduce our code
size in ConstantFolder, FunctionCall, etc.
Most of the conversion process was pretty straightforward:
* `IntLiteral::Make` becomes `Literal::MakeInt`
* `x.is<IntLiteral>()` becomes `x.isIntLiteral()`
* `x.as<IntLiteral>.value()` becomes `x.as<Literal>.intValue()`
Change-Id: Ic328533611e4551669c7fc9d7f9c03e34699f3f6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447836
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
`removeprefix` was only added to Python in March 2020. It isn't
available on Python 3.8.2, which is the default macOS Python installed
version.
https://www.python.org/dev/peps/pep-0616/
Change-Id: I7b8db0dfc55f1448cc72f41ba9a6ad5f099d48ef
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448259
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:12086
Change-Id: I833a04d7dbc4624a85051aa8442a074520335b76
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448256
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
* When loading the MSAA texture from resolve, we set up the render
command encoder to be compatible with the next op. So we should just use
that command encoder rather than create one. This mainly helps with
clears as detecting compatibility in that case is difficult.
Bug: skia:12086
Change-Id: I859761ff970520c36629f1b8054f73e2fd29eb67
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447189
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Add a lerp that uses full 32-bit resolution for the multiply.
This gives the same answers as the floating point, and
demonstrates the error introduced by using the faster
16-bit multiply.
Change-Id: I90c46b350fae09f74d9c9cf2af7ef3d3d1d0f202
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448102
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Because the [-1, 1) lerp can handle a higher sampling rate in t,
decrease the t step to ensure everything is covered. The
limit needs to be trimmed by a 1/65536 to allow the older lerps
to run properly.
Remove the scaled test cases. They no longer make sense.
Change-Id: Ida841d65ebe794ba5d0ae67dfb145268764ec6a2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448101
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
The fuzzer found that the `DetectVarDeclarationWithoutScope` check was
placed too late in the function, and could be skipped over by for-loops
containing multiple variables. This was caught in ForStatement::Make,
which mirrors the Convert postconditions with matching assertions.
Change-Id: I6e9d97c7c9ca969aba65e601bbcd9fe676105838
Bug: oss-fuzz:38560
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/448116
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>
Using the extra bit improves the difference rate from .38% to .19%.
I think this uses the full resolution of all the values, and there
is no more accuracy by trying to get more bits in use. The maximum
difference is still 1.
This gives the same answer with the neon_vqrdmulhq_s16 and
ssse3_mm_mulhrs_epi16 instructions.
Change-Id: Ie801b68aead408c5e1845aaf94ea660121e414b9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447719
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit db38ad7b14.
Reason for revert: breaking g3 roll since it thinks the test case is "binary" not flagged as binary
Original change's description:
> Fixed DSL assertion error on source files containing nulls
>
> The assertion was there to make sure we weren't running off the end of
> the source, but naturally fails in the presence of legitimate embedded
> nulls.
>
> Change-Id: I3b80499e9b182c9ea046c479f35d7a965d548401
> Bug: oss-fuzz:38107
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447182
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
Bug: oss-fuzz:38107
Change-Id: I650d12d728b5d932bda79e81205b873d8b44771f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447936
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
We need to ensure all async reads are completed before we finish
encoding -- otherwise late reads will fire from the GrContext destructor
and crash because the encoder is no longer valid.
Change-Id: I0065561d2411ddd202897e625a187592e40ced87
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447184
Commit-Queue: Florin Malita <fmalita@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
No behavioral changes, just stylistic.
Change-Id: I95a54bb2ee5610272ecea4886163ba79ae2544a2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/447837
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
