Adding new tests in an upcoming CL. This sets the stage for that.
I also outlined the Register ctor so that the diff in the next CL is
more readable.
Bug: skia:10877
Change-Id: I604f1ce811cddf1d1105716f114e27ce12bf8249
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394998
Commit-Queue: Adlai Holler <adlai@google.com>
Auto-Submit: Adlai Holler <adlai@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
DSLVar is normally implicitly converted to DSLExpression wherever it is
used, but the naively template-ized nature of DSLFunction::operator()
meant that it was accepting DSLVar parameters directly. Since DSLVar is
non-copyable, this meant DSLVar couldn't be passed directly to a DSL
function.
The smarter templates in this change are able to pass DSLVar by
reference.
Change-Id: Id04531f909cefd29b377c46e37fc4525bb29953c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394161
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
GLSL (post-ES2) allows array comparison: http://screen/8gryPvb9T7gndyb
Unfortunately, because ES2 does not support array comparisons, we can't
add this test to the dm test suite.
Change-Id: I06b71683e49b2631669cff801dc647951a81a299
Bug: skia:11849
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394162
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: Icc710e414388e4026a5e9819a53b8dac8ee0a2d1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394896
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This is allowed in OpenGL ES2, and its absence in SkSL has been a pain
point for new users adopting Runtime Effects.
Change-Id: Id2ed78261a2cd2b14b49ad22cb74cdc9e0905f8a
Bug: skia:11368
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393418
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
No functionality changes, just making the naming consistent with our
other enums.
Change-Id: Ic9bc4a89f8373e4dc1060067a41468fb626e5fa1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394160
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This function will be used by the resource allocator to figure out
whether an allocation plan is feasible without committing to it.
Bug: skia:10877
Change-Id: I135b7b80d53d9c3541d2fa0313d91d14a1d54eb2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394156
Commit-Queue: Adlai Holler <adlai@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Auto-Submit: Adlai Holler <adlai@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
RSXForm is not stored directly in the glyph run. The Tx and
Ty portions are stored in the positions, and the scos, and ssin
portions in a span of SkVector.
This CL just adds the space for the extra information, but does
not utilize it yet.
Change-Id: I93e2140fb28a6638a57b6d811ed6dc823e8de6ae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394217
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Herb Derby <herb@google.com>
OpenGL ES2 allows structs to be compared: http://screen/6KnX4ZfkdLtqDWv
This already worked in GLSL, Metal and SkVM.
Change-Id: Iaf7029c0c1ea9d447348c8280a2788f0d36befad
Bug: skia:11846
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393598
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This allows it to access the resource cache as well as the provider,
and paves the way for memory budgeting inside the resource allocator.
Bug: skia:10877
Change-Id: I01e9c02e445494ac431b288f41006ae7360ff791
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/394116
Auto-Submit: Adlai Holler <adlai@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
As part of this we also create two fbos now for
multisampled_render_to_texture: one msaa and one not.
Bug: skia:11396
Change-Id: If519b52ec79de4b2abeac2c5658f766f1c7e8d9c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392746
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
We believe with advent of GrBackendSurfaceMutableState that this is
only needed for inline atlas uploads. Those call GrGpu directly.
Bug: 1191058
Change-Id: Iad851ebc9f5b7df21d16064a84caa9575da98634
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393082
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:11374
Change-Id: I63d605eabbe514a0469d00d8a671969874f3edd4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393081
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit 5a2de5e72f.
Reason for revert: Upon further investigation this still leaks
Original change's description:
> put an arena on GrSurfaceDrawContext
>
> This is part one of two CLs. In this CL, I put a
> sk_sp<GrArenas> on GrSurfaceFillContext where GrArenas wraps
> a SkArenaAlloc to add ref counting. Creating
> a GrOpsTask shares the GrArenas with the ops task. New plumbing
> was added to GR_DRAW_OP_TEST_DEFINE to allow a proper
> GrSurfaceDrawContext to be passed to GrAtlasTextOp's
> GR_DRAW_OP_TEST_DEFINE so the arena will have a proper lifetime.
>
> The second CL will work on replacing GrOpsTask's fAllocators
> system with the shared arena.
>
> Change-Id: Ife3be0ab265441cbffab360f2808f5eed86db8b3
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392936
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Herb Derby <herb@google.com>
TBR=bsalomon@google.com,herb@google.com
Change-Id: I9ca5c8b1e16b468003788cd3126eda1d40ff93ed
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393177
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This reverts commit f00d6a8efd.
Reason for revert: In the long run, it seems like this might leak.
Original change's description:
> remove fAllocators from GrOpsTask
>
> Change-Id: I5901f005c2758a92692e5cd70ba46a2b5ad797fd
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393116
> Commit-Queue: Herb Derby <herb@google.com>
> Reviewed-by: Adlai Holler <adlai@google.com>
> Reviewed-by: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=herb@google.com,robertphillips@google.com,csmartdalton@google.com,adlai@google.com
Change-Id: I895e47c999c6961fa14c356d6d5bde80dd86b63e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393176
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Also adds tests of non-uniform shader declarations. These are currently
allowed, but will be detected as an error in the next CL.
Bug: skia:11374
Change-Id: I3fee0a0c97ae590f7bc6952cb367f7e94436b891
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393080
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This shook out a long-standing bug; constant folding would treat a
matrix resize as if the cells not covered by the original matrix were
all zero. This is wrong; GLSL populates the unknown cells with an
identity matrix. We actually tested for the wrong behavior, so the tests
were updated to match the correct behavior, and an equivalent test was
added that does not constant-fold (to verify that our constant folder
matches reality).
Change-Id: I03df10ce646fbef0a36e9c1a841a7637182de122
Bug: skia:11032
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392916
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This is part one of two CLs. In this CL, I put a
sk_sp<GrArenas> on GrSurfaceFillContext where GrArenas wraps
a SkArenaAlloc to add ref counting. Creating
a GrOpsTask shares the GrArenas with the ops task. New plumbing
was added to GR_DRAW_OP_TEST_DEFINE to allow a proper
GrSurfaceDrawContext to be passed to GrAtlasTextOp's
GR_DRAW_OP_TEST_DEFINE so the arena will have a proper lifetime.
The second CL will work on replacing GrOpsTask's fAllocators
system with the shared arena.
Change-Id: Ife3be0ab265441cbffab360f2808f5eed86db8b3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392936
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Herb Derby <herb@google.com>
When growing a matrix, the new cells should contain the identity matrix.
`writeMatrixCopy` was populating every cell with zeros instead.
Also, `writeMatrixCopy` had some code for handling int matrices, but
we don't support this in SkSL anywhere else; replaced with an assert
that the component type is float.
Change-Id: Icb80ae711d5c4f1417b62168b203a661f52e8d7d
Bug: skia:11835
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/393037
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This is a reland of 8ce24b1d11
Original change's description:
> Make RGB_888x pixel operations work.
>
> Also remove vulkan-specific conversions in GrVkGpu and rely
> on higher level SurfaceContext to convert correctly.
>
> Bug: skia:8862
> Change-Id: Ib8b0541c8c5831148b7129c4bbed3f925660116e
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392378
> Reviewed-by: Robert Phillips <robertphillips@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:8862
Change-Id: I5d085e47398a4a2947228ee67f867552fa93123a
Cq-Include-Trybots: luci.skia.skia.primary:Test-Debian10-Clang-NUC5PPYH-GPU-IntelHD405-x86_64-Debug-All-Vulkan
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392842
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
The actual fix happened at prior CL http://review.skia.org/392197, which
reworked how vector-cast constructors function.
Change-Id: Ifb71ec913b349e65d38458dc615441e7a73efddc
Bug: oss-fuzz:32851
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392841
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>
As part of the new architecture I envision not compiling in stuff like
the GrClipStackClip when in NGA-mode. This ran/runs into trouble with
the unit tests.
Change-Id: I922dcf873c4bb1a3abfd2ce519f3dbf68e053770
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392839
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Making a VectorCast from a compile-time constant will perform the cast
at compile-time instead; previously, we did not apply this optimization.
This simplified a few test outputs in subtle ways. (In particular, the
SPIR-V codegen used to occasionally decorate OpConstantComposite of
constant numbers with RelaxedPrecision, and no longer appears to do
this. This should have no effect on results either way AFAICS.)
Because we don't return VectorCast constructors containing compile-time
constant values, we do not need to implement compareConstant for this
constructor; they only wrap non-compile-time-constant expressions.
Change-Id: I28c1f337f64d6f20fb86bc0f58e225af4bd7b26c
Bug: skia:11032
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392197
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Also remove vulkan-specific conversions in GrVkGpu and rely
on higher level SurfaceContext to convert correctly.
Bug: skia:8862
Change-Id: Ib8b0541c8c5831148b7129c4bbed3f925660116e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392378
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
No functional change yet, but fragmentProcessor is no longer allowed in
runtime effect SkSL (and errors about 'shader' variables will no longer
refer to 'fragmentProcessor').
Bug: skia:11813
Change-Id: I5161f9adbe79090ec31ed3faf7dcef0e290842b7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392440
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Allowed today, will soon be an error.
Bug: skia:11813
Change-Id: I5c13de7657fa85f13fa6d80e1d890225d8a3e868
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392439
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is a stopgap for dealing with OOM issues.
Bug: b/182959903
Change-Id: I25855b6fc1c9d8d270c1cb3fcbdc1c7f93623214
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391956
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Adlai Holler <adlai@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This reverts commit 92b35673c5.
Reason for revert: breaking the android roll (see ag/14076257)
Original change's description:
> Lift atlas clip FP creation out of GrClip::apply
>
> Atlas clips always had a potential point of failure: If the SDC's
> opsTask ever got closed between GrClip::apply and
> GrOpsTask::addDrawOp, their mask would have gotten sent to the wrong
> opsTask. It didn't _look_ like this could happen with the current
> code, but it could have also been inadvertently changed quite easily.
>
> This CL adds a "pathsForClipAtlas" array for GrClip::apply to fill out
> instead of creating FPs. The SDC then generates the actual clip atlas
> FPs once it knows exactly which opsTask they will belong in.
>
> Bug: chromium:928984
> Change-Id: I507ab13b2b5e8c3c3c1916d97611297dbbd8a522
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389926
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
TBR=robertphillips@google.com,csmartdalton@google.com,michaelludwig@google.com,adlai@google.com
Change-Id: I9597d822a9f31b7070aee691ddf3a52f4f424a24
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:928984
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/392339
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Derek Sollenberger <djsollen@google.com>
Atlas clips always had a potential point of failure: If the SDC's
opsTask ever got closed between GrClip::apply and
GrOpsTask::addDrawOp, their mask would have gotten sent to the wrong
opsTask. It didn't _look_ like this could happen with the current
code, but it could have also been inadvertently changed quite easily.
This CL adds a "pathsForClipAtlas" array for GrClip::apply to fill out
instead of creating FPs. The SDC then generates the actual clip atlas
FPs once it knows exactly which opsTask they will belong in.
Bug: chromium:928984
Change-Id: I507ab13b2b5e8c3c3c1916d97611297dbbd8a522
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389926
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Bug: skia:11803
Change-Id: I925f14be282b96355721986de6049090b35adf3d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/391856
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@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>
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>
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>
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>
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>
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
As soon as a single VarDeclaration is successfully created, its Variable
is added to the current symbol table. However, if a variable-declaration
line declared several variables in a row, we would stop if ANY of the
declarations contained an error and discard the entire statement, but
would continue processing the rest of the program. This left us in a
position where some Variables existed in the SymbolTable with valid,
reachable names, but their corresponding VarDeclaration statement had
been thrown away as erroneous. Since Variables point back to
VarDeclarations for their initialValues, this gave us a stale pointer.
Any future reference to that variable name which could trigger an
access to its initialValue would read from this dead pointer.
This CL fixes the conversion of VarDeclarations so that we no longer
throw away any VarDeclarations associated with a successfully-parsed
Variable.
Change-Id: If8ec3c160933e48a0e1f36414234b3a849d8978c
Bug: oss-fuzz:32587
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389636
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 is particularly relevant for "vector / scalar" divides, as these
end up generating an extra OpFDiv to take the reciprocal of the scalar.
This meant we were generating code to calculate "literal 1 / literal N"
at runtime whenever dividing vector V against scalar N. This fix moves
the reciprocal calculation to compile-time. And of course, it's a valid
strength reduction in all cases.
Change-Id: I0986057e516120fe872e70135f1ac365155d5377
Bug: skia:11792
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389156
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Due to missing type-checks, we would emit `OpVectorTimesScalar` when
multiplying an ivec with an int, or an `OpFDiv` when dividing an ivec
against an int.
Change-Id: Idc214dbe0ec208cb44f28b22e585584ac2ab7dae
Bug: skia:11267, skia:11788
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388742
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
We had constant-folding tests for vector-scalar arithmetic, but didn't
have an equivalent SkSL unit test for vector-scalar arithmetic that
actually needs to be computed at runtime.
This exposes two SPIR-V bugs: one was previously known, but the other
is a new discovery.
Change-Id: I28737128f20b445797c6c29872335d05f94cc95c
Bug: skia:11267, skia:11788
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388739
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Previously, these tests were never actually executed, only read during
code review. They are now properly tested for correctness whenever dm
is run. Non-ES2 compliant statements (do/while/switch) are unfortunately
excluded here, as they are not compatible with Runtime Effects yet.
Change-Id: I965c782baad6f8dd3961a400ae791fb2c1f844d3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389296
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
If sk_Caps.rewriteMatrixVectorMultiply is false, we generate the same
code as before.
If sk_Caps.rewriteMatrixVectorMultiply is true, and optimization is
enabled, medium-precision multiplication of `matrix * vector` is
rewritten in BinaryExpression as the sum of
(m[0]*v[0] + m[1]*v[1] + ... + m[N]*v[N]).
Change-Id: I9796158fd4203b5cd2278c0d0e9a2e8cda82b136
Bug: skia:11769
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388460
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
On ARM GPUs, the Vulkan driver does not honor relaxed precision when
evaluating SpvOpMatrixTimesVector. This leads to reduced performance
(compared to GLSL, where mediump matrices and floats do evaluate all
intermediate values at mediump).
This caps bit will be enabled on ARM GPUs and, in a followup CL, will
be used to toggle a workaround where `m*v` is rewritten as the sum of
(m[0]*v[0] + m[1]*v[1] + ... + m[N]*v[N]).
Change-Id: I310fa73639b6498552c9672e76860f2eded15d0a
Bug: skia:11769
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388459
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 7bba1f55e8.
Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
