Our SPIR-V output was missing many RelaxedPrecision decorations, which
was presumably impacting performance.
Change-Id: Iee32d4a42f37af167fe0e45f3db94c2142129695
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384178
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Bug: skia:11738
Change-Id: I1dd5e99830f70d72c292379a45c4e39a55588858
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/383706
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Change-Id: Ic2d1240ab785101365b0fd934562505fb5a3e599
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381816
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This CL will be used to test for potential performance regressions (or
improvements) that we might cause by disabling this optimization pass.
It will be reverted in ~1 day.
Change-Id: I26b7687c341eb6d81231406381c39869cfccf6d6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381259
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I7a7874e58bf53978afce8a41b26092406b6490ed
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380360
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This enables the ternary to be optimized away in code like:
const bool SHINY = true;
color = SHINY ? add_shine(x) : x; // to --> `color = add_shine(x);`
Without constant propagation.
Also, I added a unit test for ternary expression simplification; I
wasn't able to find an existing one.
When the optimization flag is disabled, this CL actually removes the
optimization of `true ? x : y` --> `x` entirely; previously, this
substitution would be made regardless of optimization settings.
Change-Id: I93a8b9d4027902d35f8a19cfd6417170b209d056
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379297
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Swizzle optimizations now occur at IR generation time. These
optimizations are redundant with the control-flow optimization phase so
they are mostly not visible in our test output, but they do affect DSL
test results. Interestingly, they do improve our test output slightly
as well, for various reasons (e.g. we do not fully optimize lvalues in
the control-flow pass).
Change-Id: I6ebe6d71a5c22d9823b5fa500e43078915cbfb45
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/372257
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This check now runs at function finalization time, before constant
propagation has occurred; this affected the "DeadIfStatement" test.
Our detection isn't smart enough to realize that a loop will run zero
times, so it treats `for` and `while` loops as always running at least
once. This isn't strictly correct, but it actually mirrors how the CFG
implementation works anyway. The only downside is that we would not flag
code like `for (i=0; i<0; ++i) { return x; }` as an error.
Change-Id: I5e43a6ee3a3993045559f0fb0646d36112543a94
Bug: skia:11377
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379056
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This should be legal, and we support this, but some versions of Android
do not: http://screen/3bkQewHF3xUMn5v There's no point in allowing
these shaders to exist; they can't compile on real-world clients, and
these vardecls are borderline meaningless (as the variables being
declared aren't reachable by any other statements).
Change-Id: Ie1351933c90caee9124eeab8983364ec030b2653
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379584
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 50b1b2b90d.
Reason for revert: ending experiment
Original change's description:
> Disable control-flow analysis in SkSL. (Performance experiment)
>
> This CL will be used to test for potential performance regressions (or
> improvements?) that we might incur by disabling this optimization pass.
>
> It will be reverted in ~1 day.
>
> Change-Id: I775cdb0c95df81fa25ebbd66e4ff01f64c660f68
> Bug: skia:11319
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378456
> Commit-Queue: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Ie385a82db237ff5651348d82b9651f8ba09375b9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379581
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This CL will be used to test for potential performance regressions (or
improvements?) that we might incur by disabling this optimization pass.
It will be reverted in ~1 day.
Change-Id: I775cdb0c95df81fa25ebbd66e4ff01f64c660f68
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378456
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This reverts commit e4da7b672f.
Reason for revert: breaks SkSLBench perf test
Original change's description:
> Migrate if-statement simplifyStatement logic to IfStatement::Make.
>
> This performs essentially the same simplifications as before, just at
> a different phase of compilation.
>
> Change-Id: Ia88df6857d4089962505cd1281798fda74fd0b02
> Bug: skia:11343, skia:11319
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376177
> Commit-Queue: John Stiles <johnstiles@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I0051188ffe69426904066eb60a932435efdc2af8
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11343
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379062
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This performs essentially the same simplifications as before, just at
a different phase of compilation.
Change-Id: Ia88df6857d4089962505cd1281798fda74fd0b02
Bug: skia:11343, skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376177
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This adds Analysis::IsConstantExpression, to determine if an expression
is a constant-expression. It now expands to cover 'const' local and
global variables, because we also enforce that the initializer on those
variables is - in turn - a constant expression.
This fixes 10837 - previously you could initialize a const variable with
a non-constant expression, and we'd emit GLSL that contained that same
pattern, which would fail to compile at the driver level. That should
not be possible any longer.
Bug: skia:10679
Bug: skia:10837
Change-Id: I517820ef4da57fff45768c0b04c55aebc18d3272
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/375856
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The expression `~123` was making a PrefixExpression of type $intLiteral.
It should be converted to type `int` when the ~ prefix is applied.
This change also changes the output from oss-fuzz:27614. Both programs
are essentially nonsense expressions with no real behavior, so this is
fine.
Change-Id: I586be149ce95136fabee72fdd3473814d54948cf
Bug: oss-fuzz:31410
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376620
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
These variables were later being eliminated by the dead-code-elimination
pass, so you can't see them directly in the final output, but removing
them affects the name mangling off all future symbols, so it causes an
enormous ripple effect in the diff. And of course, it's a waste of time
and memory to synthesize IRNodes just to destroy them later.
If we disable control-flow analysis, we lose the dead-code-elimination
pass entirely; this change is also beneficial for emitting better code
when optimizations are turned off.
Change-Id: I882b3be4f3fd99b77d99b6abe128f26bb9252c89
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/375776
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
The test inputs were removed at http://review.skia.org/360778
Change-Id: Ib2918f3f984cd80463bacb822ef510ee9feb1e77
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/373916
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The optimizer now properly recognizes all types of exits from a switch
statement. Break, continue and return are all potential exits and need
to be considered when determining the exit path from the switch.
Previously, dead code elimination was hiding the effects of this bug
from us, but it meant that an optimized switch had the potential to
generate lots of worthless IR nodes which then needed to be detected and
eliminated by the CFG. In particular, this affected the enum form of
blend, causing a catastrophic amount of extra work to be done.
Change-Id: If857e38cadfc016884624ea4db25a273ad3dce5b
Bug: skia:11352
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/372958
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
There are two forms. Swizzle::Make supports components XYZW only;
Swizzle::MakeWith01 also supports the 01 components, and restructures
the zeros and ones into a constructor (as IRGenerator::convertSwizzle
has historically done). This means that once we are past the initial
IR generation stage, and we know that the 01 components have been
eliminated, we can avoid the extra 01-handling logic and just call
Swizzle::Make directly. This isn't a huge deal but it means that call
sites like the inliner can avoid some extra work that will never happen.
Change-Id: I46690c3d6b07feb6327ee72e8f66f15592a35554
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/371398
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>
The CL at http://review.skia.org/366399 introduced a bug with
LValue::getPointer. Specifically, getPointer used to return zero when
no pointer is available. (This happens when the LValue is a swizzle.)
That CL changed the error code to -1. However, it did not fix up all
the call sites that checked the return value of getPointer().
This CL fixes up those call sites to use -1 consistently, and adds
TODOs in spots which do not check the result from getPointer() at all
(instead assuming it cannot fail). This will allow swizzled out-
parameters to work in SPIR-V as they did before. (Except in intrinsics,
where they seem to have been broken all along, but those are now marked
with a TODO at least.)
Note that we still do not fully emulate GLSL semantics for out
parameters, as out-parameters should only be copied back to the original
variable at the end of the function call to be fully GLSL compliant.
(This CL also replaces a tuple with a named struct for readability.)
Change-Id: I708dc7a69296a4244ba9ceb85c3e68d1f331bbc9
Bug: skia:11052
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368618
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Uses the pipeline-stage callback mechanism. It mangles the type name
(with a test to verify that this works), and then calls defineStruct
with the entire SkSL struct definition string.
Bug: skia:10939
Change-Id: If14cf1b11faaa80ad8d4086cdacf68532bac43fc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368809
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Disabled on Adreno 5xx/6xx as the tests do not pass on those GPUs:
http://screen/3Dkgs9syj37cjBV
Change-Id: Ib935d01e8f06dbfe7decd5cc4e52e0688b48be08
Bug: skia:11306, skia:11308
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368805
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit 4908a24d4b.
Reason for revert: test fails on Adreno 5xx/6xx, will land tests
separately and disable on Adreno
Original change's description:
> Revert "Add support for matrix == and != in Metal shaders."
>
> This reverts commit c501857188.
>
> Reason for revert: breaking many bots
>
> Original change's description:
> > Add support for matrix == and != in Metal shaders.
> >
> > We need to polyfill an operator== and != when these are first
> > encountered in the code.
> >
> > Change-Id: I539c838ee1871bcb0c4b66abb8a4a0f91146cd4f
> > Bug: skia:11306
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368496
> > Auto-Submit: John Stiles <johnstiles@google.com>
> > Reviewed-by: Brian Osman <brianosman@google.com>
>
> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
>
> Change-Id: Id583109a0d167c2c58a57644b14cd5f49d670737
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:11306
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368801
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
TBR=egdaniel@google.com,brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Bug: skia:11306
Change-Id: If7c628b8c7a2ce40d6c88599a7660ff91c4ac67a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368804
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
"Constant" is an address space qualifier and can't be applied to a
local variable. "Const" in GLSL (and hypothetically SkSL) is meant to
apply to a constant expression regardless of address space.
Our previous test was not finding any error because the optimizer was
eliminating the constant expressions entirely.
Change-Id: I6cfe8e2a621c79945b33e0166780d81e79890a1b
Bug: skia:11304
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368517
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit c501857188.
Reason for revert: breaking many bots
Original change's description:
> Add support for matrix == and != in Metal shaders.
>
> We need to polyfill an operator== and != when these are first
> encountered in the code.
>
> Change-Id: I539c838ee1871bcb0c4b66abb8a4a0f91146cd4f
> Bug: skia:11306
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368496
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Id583109a0d167c2c58a57644b14cd5f49d670737
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11306
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368801
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
We need to polyfill an operator== and != when these are first
encountered in the code.
Change-Id: I539c838ee1871bcb0c4b66abb8a4a0f91146cd4f
Bug: skia:11306
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368496
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The leftover tests in shared/ are not easily testable as Runtime
Effects; they do things that ES2 doesn't support or use a feature not
exposed directly by Runtime Effects.
Change-Id: I7ebe170cf713c4a0d2dbef333c1fcbac2410c67f
Bug: skia:11009
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/367059
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This takes away one of our gadgets for thwarting dead-code elimination
in unit tests, but it's the right thing to do. Comma expression left-
sides without side effects are clearly dead code.
Change-Id: Iaee490b4a742d06a0a0be94cddaa69a51543d8f5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/366719
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This fixes the OutParamsTricky test.
Change-Id: If59637bc946b71b141ae1d90cf1652bf80b163c4
Bug: skia:11269
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/366399
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
- MultipleAssignments
- NegatedVectorLiteral
- NumberCasts
- OutParams
- OutParamsTricky (disabled on GPU due to skia:11269)
Change-Id: I87dc9c5019931f3d2dc3aafbe1e02d0eee2e1a05
Bug: skia:11009, skia:11269
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/366400
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 reverts commit 38df4c8470.
Reason for revert: updated ArrayTypes test for ES2 compatibility
Original change's description:
> Revert "Improve support for arrays in Metal."
>
> This reverts commit dd904af566.
>
> Reason for revert: breaks ANGLE
>
> Original change's description:
> > Improve support for arrays in Metal.
> >
> > Arrays in Metal now use the `array<T, N>` type instead of the C-style
> > `T[N]` type. This gives them semantics much more in line with GLSL,
> > so they can be initialized and assigned like GLSL arrays.
> >
> > This allows the ArrayTypes and Assignment tests to pass, so they have
> > been added to our dm SkSL tests. (ArrayConstructors also passes, but
> > is not ES2-compliant so it is not enabled.)
> >
> > Change-Id: Id1028311963084befd0e044e11e223af6a064dda
> > Bug: skia:10761, skia:10760, skia:11022, skia:10939
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365699
> > Commit-Queue: John Stiles <johnstiles@google.com>
> > Auto-Submit: John Stiles <johnstiles@google.com>
> > Reviewed-by: Brian Osman <brianosman@google.com>
>
> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
>
> Change-Id: If6a18dea7d6a45fa7836e9129bf81c2e536f07e3
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:10761
> Bug: skia:10760
> Bug: skia:11022
> Bug: skia:10939
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365976
> Reviewed-by: John Stiles <johnstiles@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Bug: skia:10761
Bug: skia:10760
Bug: skia:11022
Bug: skia:10939
Change-Id: Ia1c4917f5d3c41162d282b3093814d861707ad30
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/366144
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit dd904af566.
Reason for revert: breaks ANGLE
Original change's description:
> Improve support for arrays in Metal.
>
> Arrays in Metal now use the `array<T, N>` type instead of the C-style
> `T[N]` type. This gives them semantics much more in line with GLSL,
> so they can be initialized and assigned like GLSL arrays.
>
> This allows the ArrayTypes and Assignment tests to pass, so they have
> been added to our dm SkSL tests. (ArrayConstructors also passes, but
> is not ES2-compliant so it is not enabled.)
>
> Change-Id: Id1028311963084befd0e044e11e223af6a064dda
> Bug: skia:10761, skia:10760, skia:11022, skia:10939
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365699
> Commit-Queue: John Stiles <johnstiles@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: If6a18dea7d6a45fa7836e9129bf81c2e536f07e3
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:10761
Bug: skia:10760
Bug: skia:11022
Bug: skia:10939
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365976
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Arrays in Metal now use the `array<T, N>` type instead of the C-style
`T[N]` type. This gives them semantics much more in line with GLSL,
so they can be initialized and assigned like GLSL arrays.
This allows the ArrayTypes and Assignment tests to pass, so they have
been added to our dm SkSL tests. (ArrayConstructors also passes, but
is not ES2-compliant so it is not enabled.)
Change-Id: Id1028311963084befd0e044e11e223af6a064dda
Bug: skia:10761, skia:10760, skia:11022, skia:10939
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365699
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We now catch this error at IR generation time; previously we'd send it
to the driver (where it would fail to compile).
Change-Id: I45890214ffa164be1c0f359320f942bc4dc479ca
Bug: skia:11265
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365697
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This uncovered a bug in Metal code generation of `matX *= matY` which is
now fixed. (It was emitting the helper function more than once.)
Change-Id: I0aeb0efe7ab5fbf5592a8ca6f4f5b50354d3d7f4
Bug: skia:11262
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/365489
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 works around a GLSL compilation bug on the Tecno Spark 3 Pro.
Change-Id: I516bd64745a8e99cccc87ee4bb2e1f5d5b26c130
Bug: skia:11255
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/364116
Commit-Queue: Robert Phillips <robertphillips@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Previously, our only test was invoking `sin(1)` which is a pretty
ineffective test. Now, we test args and return types for all the basic
scalars/vectors/matrices.
Change-Id: I7d335303eef8b9c9c6cfef2265a15bbd9bd73e0c
Bug: skia:11246
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/363943
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
These tests have updated to return green on success, or red on failure.
Some tests were modified slightly to conform to ES2 limitations, or
split into separate ES2 and ES3 parts.
Change-Id: Ib47aeca217aef33f3c4b5999d93afed5d42a1e62
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/363876
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
We already had this trick for scalar integers, this extends it to
integer vectors. As with prior work in this area, it would be better to
detect this case and produce an error, but now we at least produce
consistent and well-defined results (rather than undefined signed
integer overflow).
Bug: skia:10932
Bug: oss-fuzz:29494
Change-Id: I45526fe96b6ea42c0e88b9862f6961b316810321
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/363962
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
The test has been split into an ES2 version and ES3 version; the ES2
side omits unsupported integer ops like << >> & | ^ %.
Change-Id: Iba16d469a477809b17a823b1c68ae8937624c68e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/362616
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
SkVMCodeGenerator was lacking support for the comma operator entirely;
this has now been implemented.
Change-Id: I9350f54e6ee52764c620116e6dbfe4ca3e9cd47e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/363096
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Previously, a uniform not wrapped in an interface block would report a
SPIR-V error:
"Variables identified with the Uniform storage class are
used to access transparent buffer backed resources. Such variables must
be typed as OpTypeStruct, or an array of this type..."
Now, the SPIR-V code generator automatically detects such global
variables and synthesizes a struct named _UniformBuffer to hold them.
When these variables are accessed, an OpAccessChain instruction is added
to grab the variable out of the struct.
Change-Id: I5e852d4de01b866c291506cc8cf6eb547f097d66
Bug: skia:11225
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/360776
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously, structs that were defined as part of a variable declaration
would end up declared similarly in the generated code. Now, global
variable declarations that include a struct definition generate two
separate program elements.
Bug: skia:11228
Change-Id: Id7ddde6931fe07a250c2c9c46153879005535fb3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/361359
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This allows interface blocks in Metal to compile even if
`layout(binding=...)` is not specified. It will also be used in SPIR-V
in the followup CL, when an interface block is automatically synthesized
for top-level uniforms.
This CL also reorganizes the unit tests around uniforms a bit.
Change-Id: Ia898c536b454dda6f51677e232a8f6e6c3606022
Bug: skia:11225
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/360778
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This is a known deficiency of runtime effects, next step is to fix how
they manage function signatures to solve the problem.
Bug: skia:10939
Change-Id: Id934e0acdf774b03bd6edce78d7b2c077bdeae00
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/360603
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit 578f1acbe8.
Reason for revert: updated test to pass on Mac Intel 5100/6000
Original change's description:
> Revert "Add SkSL for-loop control flow test to dm."
>
> This reverts commit a0c266283a.
>
> Reason for revert: failing on Mac Intel
>
> Original change's description:
> > Add SkSL for-loop control flow test to dm.
> >
> > While loops and do-while loops remain untested in dm, as they are not
> > supported in ES2 (and therefore not available in Runtime Effects).
> >
> > Change-Id: I2f1bfccccd571cc4ced096bc18ebbb9ecc9f9b4a
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359556
> > 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>
>
> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
>
> Change-Id: I45335d16a695644eaeb8a535298c0efcc616c1ce
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359840
> Reviewed-by: John Stiles <johnstiles@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I2dc6e870393708a12286658001b723f25a6aec4a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359856
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This creates a helper function, _entrypoint, which invokes main() and
assigns its result into sk_FragColor. We also make sure to prevent
sk_FragColor from being dead-stripped from the code during IR
generation.
At present this is useful for allowing our SkSL test shaders to compile.
Change-Id: I2d7fab0e1959a77778ffdb18ca569e869bcaeece
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358525
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This lets us use descriptive names like `colorRed` and `colorGreen`
instead of `half4(1,0,0,1)` and `half4(0,1,0,1)`. It also lets us use
actual unknown values instead of synthesizing sorta-kinda-unknowns by
calling sqrt.
Change-Id: I61481c33b7ff42182955777b05cfa5fcc13e0efc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359567
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This allows uniforms to be specified without an explicit `layout(set=N)`
modifier. They will assume a default set value instead.
This turns out to fix a handful of tests in Metal/SPIR-V which were
written with GLSL in mind, or adapted from real generated GLSL code, and
didn't have layout information specified on their uniforms. It will also
make it easier to write SkSL tests using uniforms that can compile
either as a runtime effect or as plain Metal/SPIR-V code.
Change-Id: Id79ec06f278b913a45c09c2e6211195dc98b42c0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359838
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Previously, we would emit nothing at all, but that is not actually
valid if the Block is a child statement (e.g. the body of a loop).
Now we emit braces for empty blocks, even if the block was unscoped.
Change-Id: I456a8d7d306a3e59d85e39f80b9f15fe3347ea19
Bug: skia:11218
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359562
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
In most cases, this works properly and a `;` is emitted, but in one
particular case (int x, y;) we get nothing.
Change-Id: If88d92502f6a533284dd4e0f78daedaf1481ff3d
Bug: skia:11218
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359558
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>
Change-Id: I94094be7163a04bf48e86406230156a5433469b6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359140
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>
Change-Id: I924ac75b5f8a397f7af7a06925ef0c9deba5c509
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359141
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>
Change-Id: I8309940f8e40d0e84847ae272830896d010c39de
Bug: skia:11219
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359138
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 will allow us to load these inputs for unit testing in `dm`.
Change-Id: Id256ba7c30d3ec94b98048e47af44cf9efe580d5
Bug: skia:11009
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/357282
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Like _globals, it's not actually necessary to indirect through a
separate pointer at all. The output struct is now passed by reference
and the additional pointer variable is removed.
(Additionally, renamed _skGlobals back to _globals.)
Change-Id: Id089a20cb751cdaedc48462a52da78ee43783611
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/355632
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
We already had support for &&, ||, ^^ but somehow the common cases of
== and != were not implemented in the constant-folder.
This CL also updates the test to return a green/red color on success or
failure, instead of assigning arbitrary numbers into sk_FragColor that
don't mean anything. The long-term plan is to signal success or failure
of each test by color code; we can display these colors as swatches in a
GM slide for testing purposes.
Change-Id: I0810108b3c6b656a60cd8aa64ceefd765eff0157
Bug: skia:11112
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/355984
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
`globalStruct` is now named `_skGlobals` and is passed around directly
by reference, with no additional helper variable (`_globals`) at all.
Change-Id: Icc5566d2212afd14a4d43700e89f50bedcc8b45f
Bug: skia:11168
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/355717
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The Inliner likes to move function bodies around; after inlining, code
can inadvertently move upwards, above ProgramElements that the code
relies on. We work around this by always emitting functions last.
Change-Id: Ie5486cc3a79a478920342fb9f578d575486fb4cf
Bug: skia:11186
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354669
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The inliner discovered that when a binary expression is inlined, its
type is not cloned into the destination's SymbolTable. This meant that
when the inlined-from function was later dead-stripped, the type pointer
would become dangling. Did a quick pass over inlineExpression and
inlineStatement and ensured that types are always copied.
Also found that `copy_if_needed` was making a copy of eligible types
each time one was encountered, instead of making one copy and reusing
it. This is fixed as well.
Change-Id: Iee3259ab038dfb04034bf0110af1909ccffec3de
Bug: oss-fuzz:29444
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354219
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Unsized arrays are now allowed in exactly one place: On the declaration
of an interface block. This satisfies the one existing use-case, which
is the gl_in (sk_in) declaration for geometry shaders. There is no other
useful scenario, and most of our backends don't support them anyway.
Several spots were using less strict checks when attaching sizes to
arrays, allowing for zero or negative-sized arrays, so those are all
fixed now.
The existing tests that initialize arrays are still a problem, because
Metal doesn't support that (neither does GLES2). Also, ArrayConstructors
has gone from generating an error in the Vulkan backend, to invalid
SPIR-V.
Bug: skia:11013
Bug: skia:11127
Change-Id: Ib08dfe9aeec96bf605661665d6f166419d27e8bc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353817
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Change-Id: Icd8982d604881effee31cc1392e2717cb112d06d
Bug: skia:11172
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353629
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>
Previously, `writeVectorConstructor` did not consider boolean types at
all when converting scalars to a different type. Now, this code reuses
the existing logic from `castScalarTo(Float|SignedInt|UnsignedInt)`
which supports Booleans. Added `castScalarToBoolean` to cover going in
the opposite direction.
Change-Id: I5479ab181b9b721db7fbff0bdc01718ce8f9f9b9
Bug: skia:11171
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353625
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This revealed a gap in our SPIR-V scalar constructor support;
typecasting a number to bool would lead to an ABORT.
Change-Id: Idac6d7ba34adfd214ed3cad8139e22d7170456f0
Bug: skia:11172
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353628
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The test diffs look scary, but the only actual change is a minor
renumbering of IDs. The actual logic is the same.
Change-Id: I5ecc26c8581a4c01834932ff0291deba7d9e4618
Bug: skia:11171
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353622
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
These need to change because type coercion in SkSL is about to become
more strict in a followup CL; we are disallowing expressions that mix
ints and floats without a cast.
Change-Id: I0f6c3cba53fb67078f447345338262c153236c51
Bug: skia:11164
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353102
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Note that GLSL accepts these sorts of constructors natively, but Metal
and SPIR-V do not. In the generated IR we actually add a cast for
subexpressions where the type does not match. These casts can be seen in
the final output for both GLSL (where they are no-ops) and Metal/SPIR-V
(where they are essential).
This change exposed some missing SPIR-V functionality (vector casts do
not support bool types). This can be fixed up in a followup CL; these
casts were previously disallowed by SkSL entirely, so there won't be any
of them in existing code.
Change-Id: I54ae922e91b38bed032537496428747a081dc774
Bug: skia:11164, skia:11171
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353576
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
GLSL allows mixed types inside a vector constructor, but SkSL currently
doesn't handle it well; some cases don't compile, and others generate
bad code. This will be fixed in a followup CL.
Change-Id: Ia98b498f320b8fa91595404730f6cdc836615140
Bug: skia:11164
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353577
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>
I started unpacking the mechanics of type coercion, and realized that
the second half of the function was looking up the Symbol for a Type
based on its name (Types are already Symbols), converting that Symbol
back into a Type (we started with a Type anyway), wrapping that Type
in a TypeReference, then calling that TypeReference (which always
calls convertConstructor).
This CL cuts out the middle steps and simply calls convertConstructor
directly. A test was added to confirm that an earlier error encountered
on the CQ is no longer occurring.
Change-Id: I76aae455a301afe4e67ef989d9dfe11f47ed36ae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353105
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This CL also adds tests for vector*scalar and scalar*vector folding.
We currently do not constant-fold these, but support will be added in a
followup CL.
Change-Id: I68d7374ae15ab2f4d805a095803b645c92fb03d9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352237
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This optimization doesn't perceptibly improve the generated code; it
just replaces a binary expression with an equivalent unary one.
Change-Id: Ib6cd2732a22c26978665c57ee00d7b5e5d0a0aee
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352123
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This fixes SPIR-V code generation when encountering nested constructors
like `float3 v4 = float3(float2(1), 1.0);` as featured in our unit test
VectorConstructors.sksl.
Change-Id: I3a0c4b466b3cb17ba50bd264f899e59c55c768ed
Bug: skia:11141
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350032
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Many of our shaders generate the same vector constant dozens of times,
e.g. Gaussian blur uses float4(1) repeatedly. This change avoids
re-emitting redundant vector constants.
Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Previously, we had constant-value deduplication, based on the SkSL type
of the constant. However, we were still generating redundant constants,
because we would emit a separate constant for Float(n) and Half(n), or
Int(n) and Short(n), even though we generate the exact same instruction
for these constants. We now deduplicate based on the type's number-kind,
separating constant literals into three categories: floats, signed ints,
and unsigned ints. This better matches our type-handling in
getActualType.
Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously, the IR generator had code which could simplify conversion
constructors like `int(1.23)`. Separately, the optimizer's constant
propagation pass had its own separate implementation of these
simplifications as well.
This CL unifies the two implementations. Previously, the constant-
propagation pass version of the code only supported integer literals, so
this change also improves our code generation slightly.
Change-Id: I32c70a5f2aed210d03bef3166b1178a2d40cdabd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350024
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously `number(boolean)` casts were converted to a ternary during
IR generation, and `boolean(number)` casts caused an error.
Metal and GLSL should support this cast as written. SPIR-V needed a
little bit of logic to handle converting the boolean to a number via
OpSelect.
Change-Id: I0069781e2b5a26a25c8625ab41c2392342bfd10d
Bug: skia:11131
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349066
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This actually exposed a latent bug: we don't support bool(1.23) or
bool(1) casts, but these are valid in GLSL:
https://www.khronos.org/opengl/wiki/Data_Type_(GLSL)#Conversion_constructors
"to bool: A value equal to 0 or 0.0 becomes false; anything else is
true."
Change-Id: Ia929a09914ffc96f081d0402d7bb05b5428f8db6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349977
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Previously, the declaration didn't link back to function definition.
This makes the function appear to be undefined, which inhibits inlining
and also makes it difficult for us to validate the presence of a
definition for every called function.
Change-Id: I220ab502634cb3e1d337c23bac150af9aa6370b1
Bug: skia:10902
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349063
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 code was not using typeName() to emit its types, inadvertently
generating Metal code containing the `half` type.
We didn't have any unit tests which synthesized a matrix-construct
helper with half types, so Matrices.sksl was cloned into two separate
test files--MatricesFloat and MatricesHalf. These should be equivalent
except for float vs half types.
Change-Id: I19ecea994b8bc45594bb3f69e596896a3bcefe4d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348180
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
FrExp testing was moved to the intrinsic tests as part of
http://review.skia.org/341977, and the shared versions were removed from
sksl_tests.gni at that time.
Change-Id: Ife7f3622034d97a77b60d5a98c01f71630c161d6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348183
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
In Metal, matrix *= matrix is not natively supported and needs to be
injected via a helper function. This helper function now properly
converts `halfNxM` types to `floatNxM` types (as Metal does not support
half types). It also returns the result by reference instead of by
value to avoid an unnecessary copy.
Matrices.sksl now includes tests for operators += -= *=. Previously we
did not have any coverage for `matrix *= matrix` at all.
Change-Id: I7dfe468ced67eaf7c2405960e8c5efe6f2acf9e4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348178
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>
Previously, SKSL_INT was limited to an int32_t, so we couldn't
differentiate between -1 and 4294967295. We could paper over the
difference in some cases by relying on the expression's type, but this
was imperfect and left us unable to differentiate between an overflow
and valid results. SKSL_INT is now an int64_t; the code has been
updated to fix bugs that shook out as a result of the change.
This isn't a complete solution for overflow handling. There are still
lots of obvious places for improvement--e.g. constant folding can
easily overflow, and statements like `byte x = 1000;` are still
happily accepted.
Change-Id: I30d1f56b6f264543f3aa83046f43c2eb56d5fce4
Bug: skia:10932
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345173
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This fix is overly conservative in some situations (identity conversions
among vectors with the same component type), but fixes errors in two
existing unit test cases.
Bug: skia:11116
Change-Id: If852f8591fb26817528fdc37191c49129e17d6b3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347053
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This CL updates `compareConstant` to fail gracefully instead of
aborting if the passed-in types don't match. This lets us call
`compareConstant` without checking types first.
Change-Id: Id2acdbdf700e64bcb24825cdad2c0e000992e8cb
Bug: oss-fuzz:28904
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347038
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We need to rescan after optimizing away expressions that might exist
in the CFG/definition map, since we are rebuilding them from scratch and
not just stripping off excess parts from them.
Change-Id: I843a2ea3fc38428e7c0bd0e2bf7a7d41101345e3
Bug: oss-fuzz:28794
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344972
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
When we determine that a function only contains a single return
statement and it is at the top level (i.e. not inside any scopes),
there is no need to create a temporary variable and store the
result expression into a variable. Instead, we can directly replace
the function-call expression with the return-statement's expression.
Unlike my previous solution, this does not require variable
declarations to be rewritten. The no-scopes limitation makes it
slightly less effective in theory, but in practice we still get
almost all of the benefit. The no-scope limitation bites us on
structures like
@if (true) {
return x;
} else {
return y;
}
Which will optimize away the if, but leave the scope:
{
return x;
}
However, this is not a big deal; the biggest wins are single-line
helper functions like `guarded_divide` and `unpremul` which retain
the full benefit.
Change-Id: I7fbb725e65db021b9795c04c816819669815578f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
If we aren't wrapping the inlined function body in a loop, there's no
need to add a scopeless Block; we've already got one. This doesn't
affect the final output meaningfully--it just suppresses a newline--but
it's one fewer IRNode allocation.
Change-Id: Ib7b0014e908586d8acfcf6c23520873fad31d0b7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345163
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit e8e4aca955.
Reason for revert: can break ES2 for-loop rules
Original change's description:
> Declare all inlined variables at the topmost scope possible.
>
> By itself, this is uninteresting and even perhaps slightly
> counterproductive (as it separates vardecl from its initializer,
> increasing LOC). However, this enables a followup CL
> (http://review.skia.org/344665) which allows single-return functions to
> be inlined without the creation of a temporary variable at all. This
> applies to the majority of fragment processors in a typical Ganesh
> hierarchy. This change will greatly reduce the number of inliner-created
> temporary copies when compiling a typical tree of FPs.
>
> Change-Id: I03423a13cf35050637dabace4a32973a08a4ed0a
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344764
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Ica01d6906bcb9cef1f49d22dda714fc9cbfa3885
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345121
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 345d72124d.
Reason for revert: can break ES2 for-loop rules
Original change's description:
> Eliminate inliner temporary variables for functions with a single exit.
>
> When we determine that a function only contains a single return
> statement, there is no need to create a temporary variable and store the
> result expression into a variable. Instead, we can directly replace the
> function-call expression with the return-statement's expression.
>
> This dramatically simplifies the final optimized output from chains of
> very simple inlined functions, which is a very common pattern for trees
> of Skia fragment processors.
>
> Change-Id: I6789064a321daf43db2e1cef4915f25ed74d6131
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344665
> Commit-Queue: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I60845f22159605a06047b030e2686a769121a35a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345120
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
When we determine that a function only contains a single return
statement, there is no need to create a temporary variable and store the
result expression into a variable. Instead, we can directly replace the
function-call expression with the return-statement's expression.
This dramatically simplifies the final optimized output from chains of
very simple inlined functions, which is a very common pattern for trees
of Skia fragment processors.
Change-Id: I6789064a321daf43db2e1cef4915f25ed74d6131
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344665
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
By itself, this is uninteresting and even perhaps slightly
counterproductive (as it separates vardecl from its initializer,
increasing LOC). However, this enables a followup CL
(http://review.skia.org/344665) which allows single-return functions to
be inlined without the creation of a temporary variable at all. This
applies to the majority of fragment processors in a typical Ganesh
hierarchy. This change will greatly reduce the number of inliner-created
temporary copies when compiling a typical tree of FPs.
Change-Id: I03423a13cf35050637dabace4a32973a08a4ed0a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344764
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
The additional scopes were harmless, but didn't really add any value.
Originally they were used to tightly scope inlined variables, but we now
mangle inlined variable names.
Change-Id: I7b35e737598036c0b6d3d9f71cbcd4a53d609ce9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344757
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Change-Id: I1a96060b2e52cddb50948a48520aab30bd097bbd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/343577
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Change-Id: Id38a3d04fcb1904a4c666d92087b4fe14bd03a27
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/343110
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This does not give us 100% coverage of intrinsics yet, but it is a
pretty good start.
Change-Id: I97d49324db1afd9f2975c2eeafbacdead710d4aa
Bug: skia:11054
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341977
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
We now insert helper functions which defer the assignment of out-
parameters back into their original variables to the end of the
function call. This allows us to match the semantics listed the GLSL
spec in section 6.1.1:
"All arguments are evaluated at call time, exactly once, in order, from
left to right. [...] Evaluation of an out parameter results in an
l-value that is used to copy out a value when the function returns.
Evaluation of an inout parameter results in both a value and an l-value;
the value is copied to the formal parameter at call time and the lvalue
is used to copy out a value when the function returns."
This technique also allows us to support swizzled out-parameters in
Metal, by reading the swizzle into a temp variable, calling the original
function, and then re-assigning the result back into the original
swizzle expression.
At present, we don't deduplicate these helper functions, so in theory
there could be a fair amount of redundant code generated if a function
with out parameters is called many times in a row. The cost of properly
deduplicating them is probably larger than the benefit in the 99% case.
Change-Id: Iefc922ac9e2b24ef2ff1e9dacb17a735a75ec8ea
Bug: skia:10855, skia:11052
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341162
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: Iac8096f6c225258b430858bad90199ec00b93b6c
Bug: skia:10998
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/342304
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>
Also fixes some additional style mishaps in class method names.
Change-Id: I49e7ac1aa91d84fef5fbc636552f040a2cb58c78
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341466
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I2c958b7aca972b7eec07e10d6c8af95fa53e761a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/342117
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Our optimizer ignores index expressions, but has a few simplifications
that it can perform on swizzles. (Added extra code to SwizzleByIndex
which demonstrates this.)
Change-Id: If3c85a0456d98749008d796e422944b602ee6933
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341460
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Short/ushort types are valid as-is and don't need to be coerced to int.
Change-Id: I41d6a537094e0c3f968e47926f541e0f6a3f92b4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341459
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>
Change-Id: Ib9117dbd1bcd2c3581fba02416d9eabda1dfc6dd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341458
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
We will need to emit a helper function to work around this case, as
GLSL supports swizzled out params, but Metal does not. In this CL, we
do not yet synthesize the helper function, but we annotate the code with
a comment indicating affected calls. (Of course, this will be replaced
with a helper function in a followup CL)
Even detecting a swizzle is actually an interesting problem, because
index expressions are sometimes actually swizzles, depending on the type
of the base expression. Also, the index or swizzle might be nested in
several other valid assignable expressions.
Change-Id: I8c74f9a7daec08eff1f32387f8b6b96851c1bd6e
Bug: skia:10855
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341057
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Pointers require decorating the variable with a * to read back the
value, which the code generator did not properly handle. There was a
special case to add the * but it only supported assignment into the
variable, not reading back. References require no special decoration.
This change fixes compile errors in Functions.sksl with the "bar"
function. (This test marks `x` as an inout but never actually mutates
it.) It also allows us to remove a special-case workaround for `frexp`,
an intrinsic function which uses a reference for its out-parameter.
Additionally, this CL adds a non-inlining copy of "OutParams.sksl" to
the Metal test directory, as most of our tests which use out-parameters
end up inlining all the code, which hides these sorts of bugs.
Change-Id: I31c4db04f6b512b4cd4fe65b3347b82bdbf039cd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/341000
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Previously, we would emit an invalid [[buffer(-1)]] annotation on the
block, causing the Metal compilation to fail.
Change-Id: I68b2439c05db3163686e84c5dcc9a5c43870ff67
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/340761
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
It's not legal to use identifiers like "int" or "sampler" to name your
variables (or enums, or structs, etc.). SkSL will now report this as an
error instead of relying on the driver to catch this.
(Note that in some contexts, it might be legal by the spec to reuse a
name that you introduced yourself, depending on the scope. In practice,
this confuses Apple GLSL, so we shouldn't support it anyway.)
This caught several existing places in our code where we used the name
"sampler." These were never exposed to the driver (they were intrinsics
that we would replace during compilation) so they were harmless before.
Change-Id: Ia6dcfca8c500d02e1eb5f9427bed8727e114dfc2
Bug: skia:11036
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/340758
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
GLSL only allows one-dimensional arrays. This CL lowers SkSL's array
dimensionality limit from eight to one, and fixes all the tests that
this breaks. The rest of the code still technically supports
arbitrarily-deep array dimensionality; there are many opportunities for
code cleanup and simplification in followup CLs.
Change-Id: I0fc31e4626649ec69d40c5f5597b3924de298df0
Bug: skia:11026
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/340339
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This is illegal in older versions of GLSL and in Metal. We now fail at
SkSL compilation time and properly report the error.
Change-Id: I6ddaeabff5386a1ed6ca3eb8703a6035476ec77a
Bug: skia:11021
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339298
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This CL fixes cases where array dimensions could be placed on the type
instead of the variable (`float[2] x` instead of `float x[2]`). It also
reports errors in cases where arrays aren't syntactically valid in
Metal, rather than emitting unusable Metal code. (Some of these cases
are actually invalid GLSL as well! But those fixes are coming in
followup CLs.)
Change-Id: I22279127c8a9aa2f22bf5ea3d225e563c2e254f2
Bug: skia:10926, skia:10760, skia:10761
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/340137
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The proper approach for creating multi-dimensional array types is
complicated, so I added a function in SymbolTable which does it the
right way (addArrayDimensions). I found all the places in SkSL which
created arrays from base types and size arrays, and refactored them to
call addArrayDimensions instead of doing it manually.
I believe that this approach fixes a bunch of minor issues with multi-
dimensional array types; some are visible in the current codegen output,
and others are latent bugs. e.g. in some instances, a Variable's type()
was silently holding flipped array dimensions, but this never led to
a visible bug because we ended up using the VarDeclaration's baseType()
plus sizes() everywhere that the type was used. (In particular, this
caused debugging headaches in http://review.skia.org/340137 where I'd
use a Variable's type and suddenly its array dimensions would be wrong.)
Change-Id: Idd6a86aa5d1dce8918d02a53bcc2f7d7886e3ac5
Bug: skia:11016, skia:10924
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339860
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reintroduces the flipped-array-dimensions bug in skia:10924. It
will be fixed in followup CLs.
Change-Id: I24ec687209b397f5fd0cf44194d0e21fe30dc32c
Bug: skia:10924
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339797
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The Metal return type from main() diverges from the SkSL source, so we
patch it in the Metal code generator. This CL improves the patching
process in multiple ways:
- A `return` statement from a fragment processor main() is rewritten to:
return *_out;
- A `return` statement from a vertex processor main() is rewritten to:
return (_out->sk_Position.y = -_out->sk_Position.y, *_out);
- We avoid emitting a duplicate `return *_out;` statement if we can
determine that main() already ends in a return statement. This is
harmless either way so it doesn't necessarily catch everything. (e.g.
it doesn't detect an if/else which returns at the end of both blocks.)
Also added a unit test which returns from the middle of a vertex shader,
since we didn't test this anywhere and we need to verify that
sk_Position.y will be negated. (This didn't work properly before.)
Change-Id: I14cf18375894fc712fa6c6466df3888ebaeba7c8
Bug: skia:10903
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339636
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously, this would generate invalid code such as `[[user(locn-1)]]`.
We now generate a more-useful error at SkSL compilation time.
Change-Id: Ifbe335ec6d4abcbdfe89b892ba51063c94d22b11
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339397
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Just filling in a gap in our tests. The output is a little strange as it
exposes a missed opportunity to constant-fold array accesses, but it
seems fine otherwise.
Change-Id: I6df13e0f9a49455015ceb47d7802bb5e1bbdaa1a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339217
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>
Constructors such as `float[2](0, 0)` add a type to the symbol table;
this type needs to be copied into the new symbol table if the
constructor is cloned by the inliner.
Change-Id: Ifa8d2dec87103c6223ce493e2201a904c14c2137
Bug: oss-fuzz:28050
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339168
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Previously, GLSL and Metal code generators would emit a struct wherever
the type was first used in the code, regardless of where it was
originally defined or what scope the type needs to live in. This CL adds
a ProgramElement for struct definitions, so that structs will now appear
at the top-level as they were originally defined. In the case of Metal,
some special handling is also needed to handle the Globals struct
properly.
Not yet fully supported:
- No special handling for structs declared inside functions yet
- No support for structs in separate scopes with overlapping names
The severity of the remaining issues depends mostly on whether we want
to support structs inside functions in Runtime Effects.
Change-Id: Ia95d4529506cb3fa6da63f5cb548199a93e1c0c5
Bug: skia:10922, skia:10923, skia:10925, skia:10926
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338600
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This test verifies that dead-stripping works on both built-in and user
functions, if their function call is optimized away.
Change-Id: I3125a34640c69de43c383343cd00d97e5a32ac60
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338836
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Enums are an SkSL-only concept--when we output code, we emit plain
IntLiterals--so the fix is simply to ignore the Enum program element
when we encounter it. This is what GLSLCodeGen does as well.
Also added a unit test to confirm that enums work normally, and that
enums are subject to optimization and static-comparison checks just as
ints would be.
Change-Id: Ic4f8da7a27983add9eb41b936d46f6638d22bd4b
Bug: skia:11003
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338800
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
^^ is not an operator in Metal. != can be used for the same purpose.
Change-Id: If75b000076ebe0aa81d0ab354a8ae33e6ed52101
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/339156
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>
Change-Id: Ifc1f0921d983ee09d7bc2632aeca41689f1bf0c3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338603
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>
GLSL does not allow most binary operations on bvec types; we can now
detect these and properly flag them as errors.
Note that `determine_binary_type` was also refactored. It originally
started with an enormous omni-switch over every possible Token type,
used to set various bools describing the type of binary expression at
hand. Instead of one big switch, this has been refactored into several
small switches in standalone functions that simply switch on the op and
immediately return true or false. Conceptually this seems like more
work (checking the op multiple times), but these tiny switches actually
boil down to little branchless shift-and-mask functions, so in practice
they should be quite efficient compared to the original omni-switch.
Change-Id: I81b473d98c65da1edd136f35fc8f656261f8930d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338346
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This is very unlikely to occur in real-world code, as it's somewhat
nonsense to use the comma operator in this way. However, it's better to
fail cleanly than to assert.
Change-Id: I76481cd8a993cb1a798ee16956400a512efd4c15
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337636
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Fiddled with the logic a bit so that when we're in unit test mode, the
output still includes all of the SPIR-V (as well as the validation error
message), so that tracking them down is easier.
Bug: skia:10694
Change-Id: I15e7777af3d268a5952765dbe5d63612cad0ac07
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338320
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Fix code generation for Metal and Vulkan with geometric
intrinsics that have scalar versions in GLSL/SkSL, but no
native support in MSL/SPIR-V.
Change-Id: Id4538a00172e0d233ad9d5ed8d33db6436b83208
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/338276
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Previously, we'd gauge suitability for inlining by counting the nodes in
a function; past a certain limit, the function was considered "too big."
Now, we also incorporate the number of times that function is called.
So if a function is called three times, and its size is 20 nodes, it
would be considered to have an inlining cost of 60 (3 * 20) instead of
20.
This should tamp down the aggressive nature of the inliner in cases like
gaussian convolution or complicated blends, and will hopefully satisfy
Pinpoint.
No change visible in Nanobench (which doesn't test any of these sorts of
patterns, but certainly inlines things): http://screen/AwD5hkgkEfjVx4g
Change-Id: Ie5e32898245ac854adb9ddd52d87001df6a67125
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337676
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
- float4(float2(1, 2), 3, 4) --> float4(1, 2, 3, 4)
- half3(z, half2(fn(x), y*2)) --> half3(z, fn(x), y*2)
Single-argument constructors will be ignored by this optimization; these
might be casts or splats.
This had an unexpected side benefit of simplifying some Metal output,
as we need to output fewer Metal matrix construction helper functions
when matrices use more simple scalars for construction.
Change-Id: I0a161db060c107e35247901619291bf83801cb11
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337400
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Previously, every output was labeled ".asm.frag" regardless of the
actual type.
Change-Id: Icf3a56bb04d88cc0443f12c2dfb99c66ee00dff0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337717
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>
Change-Id: I1be21b428939d17bbf3a9347a64db56c7cd69eb4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337638
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>
These are not actually supported operators in GLSL, Metal or SPIR-V and
we don't emulate them. Their absence was causing SPIR-V to fail the
Operators.sksl test.
Change-Id: Ia6933788392aea48836b7be19e32b9969805f254
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337185
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
We now have SPIR-V golden outputs for `blend` and `shared` tests.
This exposes a handful of SPIR-V limitations for us to address.
Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac
Bug: skia:10694
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182
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 allows swizzle removal to apply in more cases; in particular, we
can now optimize away extra swizzles caused by zero/one swizzle-
components quite effectively.
The "trivial expression" code was lifted from the inliner. Some subtle
changes in trivial-expression determination affect the inliner's results
in boring, non-meaningful ways. In particular, multi-argument
constructors containing all-constant values are now considered trivial,
whereas previously only single-argument constructors made the trivial-
ness cut. This allows the inliner to propagate some values that it
wouldn't have before.
Change-Id: I9a009b6803d9ac9595d65538252ba81c2b7166a7
Bug: skia:10954
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/336156
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
When values from the same argument are used consecutively by the outer
swizzle, they can be merged in the inner swizzle. Merging isn't always
possible, of course, but it will be used where it can be:
`half4(1, colRGB).yzwx` --> `half4(colRGB.xyz, 1)`
`half4(1, colRGB).yxzw` --> `half4(colRGB.x, 1, colRGB.yz)`
Change-Id: Id164b046bc15022ded331c06d722f1ae3605a3bd
Bug: skia:10954
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335872
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>
This will reorder constructors with swizzles applied, such as
`half4(1, 2, 3, 4).xxyz` --> `half4(1, 1, 2, 3)`
`half4(1, colRGB).yzwx` --> `half4(colRGB.x, colRGB.y, colRGB.z, 1)`
Note that, depending on the swizzle components, some elements of the
constructor may be duplicated and others may be eliminated. The
optimizer makes sure to leave the swizzle alone if it would duplicate
anything non-trivial, or if it would eliminate anything with a side
effect.
Change-Id: I470fda217ae8cf5828406b89a5696ca6aebf608d
Bug: skia:10954
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335860
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This was found at https://oss-fuzz.com/testcase-detail/5155684475469824
but the associated oss-fuzz issue ID appears to be misdirected (it's
showing oss-fuzz:24498, an unrelated issue).
PrefixExpressions can return true for `isCompileTimeConstant` but did
not implement `compareConstant`; the fuzzer discovered this. Because
compile-time constants can only be compared if they are of the same
kind, this means that `compareConstant` is actually comparing a pair of
expressions that are both negated. These negations will just cancel
out, so `compareConstant` on a pair of PrefixExpressions can just call
`compareConstant` on the inner operand of each expression.
Change-Id: I7793e25314e6c8a74278b73299d310794baf71f4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335870
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>
The fuzzer managed to create a test case which temporarily evaluates to
expression `half2(half(0.2)) + 2` as it is optimized. This requires a
bunch of temporary nonsense math as the IR Generator is attempting to
simplify as it goes; various attempts to remove terms from the fuzzer
test-case would cause it to stop reproducing the error.
Constructor::getVecComponent assumed that any constructor with a single
scalar argument would always implement `getConstantFloat` and
`getConstantInt`; however, constructors themselves did not actually
implement these methods. This meant that nesting a scalar constructor
inside a non-scalar constructor would abort when it tried to deduce the
value inside the inner constructor.
This has been fixed by implementing `getConstantFloat` and
`getConstantInt` for Constructors. These methods will assert if the
constructor has more than one argument or is a non-scalar type. This
should allow any number of nested constructors, e.g.
`half4(half(half(half(1))))` should recursively evaluate properly,
should we somehow generate this as an intermediate expression.
Change-Id: Iaee4284cba03974443cd7b5dccfd7909c1a5f3a6
Bug: oss-fuzz:27614
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335868
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>
The optimizer can now turn the expression `half4(1).xyz` into
`half3(1)`, or `half4(1).w` into `1`. This is actually a somewhat common
case when inlining chains of fragment processors, as inputs are often
overridden to `half4(1)` or `half4(0)`. This optimization also applies
to more complex cases, e.g.:
`half2(anyFunc(sqrt(2))).yxyx` --> `half4(anyFunc(sqrt(2)))`
Since the interior of the constructor is always evaluated once in either
case, it does not actually matter what the constructor contains.
Change-Id: I8d5f358502eaa8e35d4968e74fbd6b0ce2ab6365
Bug: skia:10954
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335818
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously, the Type's fOffset was set to -1 during parsing, so any
errors related to the Type would be reported on line 1.
Change-Id: I9834f733bc763c5946b3ff81d8aef4807cdc13d1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/335584
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>
Metal-specific tests are pretty thin on the ground here, and some of
the remaining tests no longer added value as they were already covered
pretty well by existing tests in Shared. The majority of remaining tests
were specific to Metal's lack of flexible matrix casting (and SkSL's
ability to paper over this with helper functions).
Change-Id: I7b3c445268b95320e7f46ec88d793c315d43ee8a
Bug: skia:10694
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/334956
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This addresses a sanitizer issue discovered in
https://oss-fuzz.com/testcase-detail/4908118777266176 (it has not been
assigned an oss-fuzz bug number yet; coming soon)
This puts an upper bound on struct nesting, again to prevent memory-
layout and other recursive type-handling code from overflowing the
stack. Coincidentally, while researching GLSL behavior around this bug,
I learned that WebGL has a similar limitation but caps nested structs to
4 deep. (I could not find any documented GLSL upper bound.)
Note that both the GLSL and Metal outputs for StructMaxDepth are badly
malformed. (Structs cannot be embedded within another struct in GLSL;
structs SA7 and below are never declared in GLSL; the array list for SA7
is backwards in GLSL; Metal is missing structs SA1 through SA8; Metal
puts the array list on the type instead of the variable name.)
These issues will be addressed in separate CLs.
Change-Id: I0f1059b6faa400cd0647dd7010ec839f73779a36
Bug: skia:10922, skia:10923, skia:10925, skia:10926
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/333316
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This addresses a sanitizer issue discovered in
https://oss-fuzz.com/testcase-detail/4908118777266176 (it has not been
assigned an oss-fuzz bug number yet; coming soon)
We need to set some sort of limit here to avoid stack overflow. Eight
array dimensions seems like more than enough for any sort of code that
we might realistically need, but the limit is definitely flexible if we
wanted to increase it. (The fuzzer needed to generate a several-
hundred-dimensional array before encountering a crash.)
Change-Id: I3630ab40e47cc58a2280ba200b485e1958371fdc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/333160
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
A GLSL function like:
void fn(int x[1][2][3]) {...}
Will emit SkSL with the array dimensions in reverse order:
void fn(int x[3][2][1]) {...}
Trying to invoke the function will fail because it expects a reverse-
dimensioned array.
Change-Id: I24431aabd2f6111b5493f63f0a85f9c78514d522
Bug: skia:10924
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/333317
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This implements constant folding optimizations on int vectors
(== != + - * /) that were previously only supported on float vectors.
Bug: skia:10908
Change-Id: Ibf61ab43eb7ae2ce8e99cce21cc55777359817e5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/332424
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This test is meant to demonstrate that constant folding for int and
float vectors is not on equal footing. Float vectors currently generate
better-optimized output.
Change-Id: Ib4822c7b594e9bc4eb4fb9cfe6ab46f7f76268d6
Bug: skia:10908
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/332423
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This CL improves on the previous fix for oss-fuzz:26789 by actually
propagating the negation from the PrefixExpression inside the
constructor, which unblocks further optimizations.
Interestingly, this fix also exposes a further missing optimization--we
optimize away comparisons of constant-vectors for floats, but fail to
do the same for ints.
Change-Id: I9d4cb92b10452a74db96ff264322cdc8a8f2a41f
Bug: oss-fuzz:26830, oss-fuzz:26789, skia:10908
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/332263
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This CL solves the fuzzer crash. Constant propagation of the negative
sign into the vector will be investigated in a followup CL.
This CL also adds a few cleanups into IRGenerator::constantFold.
Change-Id: If73a4fe2a5777265e7d43cc4f482653a38cb59af
Bug: oss-fuzz:26830, oss-fuzz:26789
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/332261
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Change-Id: Ibc1a8d3ebbf62cc55d013f7d9146f6b155d11da2
Bug: oss-fuzz:26830, oss-fuzz:26789
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/332377
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>
Previously, when a prototype was parsed, this added a function
declaration to the symbol table, but the prototype itself was not
re-emitted during code generation. This meant that the final code might
not be valid, since the absence of prototypes meant that the code might
attempt to invoke a function before its declaration. Now, prototypes are
stored in the ProgramElement list and re-emitted during code generation
for GLSL/Metal/CPP. (SPIR-V doesn't name its functions at all.)
Change-Id: I76446c796000eb0b56f964d82457122182c28b87
Bug: skia:10872
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/331136
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
When eliminating a CFG node, we now flag its exit nodes; if our
optimization pass reaches one of those flagged nodes, we stop the
current optimization process in its tracks and initiate a rescan.
We do NOT recursively mark the exits of the exit nodes, so this fix is
reliant on the CFG being ordered in a non-chaotic fashion, but in
practice this seems to be sufficient for the CFGs we generate today.
Change-Id: I892805361c5f4297e02146f37a759dfda83f5488
Bug: oss-fuzz:26942
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/331597
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: Ief57d9c102b3c7658738920cdf54ccd4d21c5c5e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/331656
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>
Change-Id: I19a9564ac4d52b709b8fdd757b99222372c626f4
Bug: oss-fuzz:26942
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/331598
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>
- Prototypes for never-declared functions
- Prototype before use
- Prototype after use
- A variety of inputs and outputs on the prototyped functions.
- Calling declared-but-undefined functions
Currently, the prototypes are not actually emitted in the generated GLSL
or Metal output at all. This CL is demonstrates our baseline before
proper prototype support is added.
Change-Id: I6112e0a89ab9bbecefccaca9fba985bb8011fff1
Bug: skia:10872
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/331376
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This improves the test output for Metal. Previously, the Metal output
was just an error message, since 1D textures were unsupported. Now we
have a valid golden output for the 2D case in Metal. (1D is still
unsupported and is likely to remain unsupported; Skia currently has no
use case for 1D textures.)
Change-Id: I91977712030f08e371cc6bfb2afa578940ca00b7
Bug: skia:10797
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/330940
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
(This CL also adds modulo to the IntFolding shared test, since this was
absent from the test. It's implemented and working properly already.)
Change-Id: I24a947ab38754bff2624cd5b58cf7a39553ca888
Bug: skia:10870
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/330596
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>
In cases where multiple variables were declared on a single line, it is
legal for variable initialization-expressions to reference variables
declared earlier in the var-decl statement. It is NOT legal for the
inliner to move those references up to the previous statement, where the
variable doesn't exist yet.
This is mitigated by disabling the IRGenerator inliner for var-decls
past the first one in a var-decls statement. (The optimizer will still
pass over this code later and is able to inline it correctly, if it is
worth doing.)
Change-Id: I7a0d45eab20e30ed9f6b2f5c1251b6e0d8eeaea3
Bug: oss-fuzz:26167
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/329357
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
PLS and discard-only shaders are the only time this has an impact,
and it doesn't seem like a problem to have the declaration?
Removes one use of variable reference counts, which are going to be
refactored.
Change-Id: Idb8d06087eed56070252ee02dcf907bf0d24c5a1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/328796
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Our Metal codegen assumes that out params are pointers, but Metal's
built-in frexp actually takes a reference for the exponent, not a
pointer. We now add in a helper function to translate.
Change-Id: I24686347d07151dd99a1ff1c43aff2b35c3181e5
Bug: skia:10762
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/328387
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Declaring max_vertices before invocations fails to adjust max_vertices
when invocation support is not present. (It should be 4, not 2 in this
case).
Bug: skia:10827
Change-Id: Ief7af97eabf5414ea8363808fc1ad2e9c480fe10
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/325664
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This reverts commit 941fc7174f.
Reason for revert: performance now seems to be roughly equal or better
(~1%) over several trials.
Nanobench: http://screen/A8e8sojaXBgbMgF
Original change's description:
> Revert "Remove inliner from IR generation stage."
>
> This reverts commit 21d7778cb5.
>
> Reason for revert: Pinpoint absolutely hates this change
>
> Original change's description:
> > Remove inliner from IR generation stage.
> >
> > There is no need to inline code during IR generation, as the optimizer
> > can now handle this.
> >
> > Change-Id: If272bfb98e945a75ec91fb4aa026e5631ac51b5b
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/315971
> > 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>
>
> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
>
> Change-Id: I62c235415bcdc92a088e2a7f9c3d7dbf7e1bf669
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/317976
> Reviewed-by: John Stiles <johnstiles@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I6189806c678283188f4b67ee61e5886f88c2d6fc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/324891
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This is a reland of 6bbf026b54
Original change's description:
> Add sk_Caps.builtinDeterminantSupport and use it in cross().
>
> This CL partially relands http://review.skia.org/321790.
>
> Change-Id: I26a1aefda8a01167783e6e7fa15a51aa35ee5d82
> Bug: skia:10819, skia:10810
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/323784
> Reviewed-by: Chris Dalton <csmartdalton@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
Bug: skia:10819
Bug: skia:10810
Change-Id: I7731f93db07bc917707cbbe1daca2e5ce0f763d7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/324620
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 6bbf026b54.
Reason for revert: Breaking Metal bot.
Original change's description:
> Add sk_Caps.builtinDeterminantSupport and use it in cross().
>
> This CL partially relands http://review.skia.org/321790.
>
> Change-Id: I26a1aefda8a01167783e6e7fa15a51aa35ee5d82
> Bug: skia:10819, skia:10810
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/323784
> Reviewed-by: Chris Dalton <csmartdalton@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
TBR=csmartdalton@google.com,johnstiles@google.com
Change-Id: I4a6c1a63dc38682dd965f78f0c1da98f35b6dbad
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:10819
Bug: skia:10810
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/324264
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This removes VarDeclarationsStatement entirely. VarDeclaration instances
appear directly as statements in Programs. SkSL that declares multiple
variables in a single declaration is transformed to represent that as a
series of VarDeclaration statements.
Similarly, global variable declarations are represented by
GlobalVarDeclaration program elements, one per variable.
Bug: skia:10806
Change-Id: Idd8a2d971a8217733ed57f0dd2249d62f2f0e9c5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/323102
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
We were letting this get further, then asserting.
Bug: skia:10797
Change-Id: Iff6fe43aa32450b5a517c94773031d593f1f62a2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/321794
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
For instance, `foo[0].x` is now considered trivial to inline. It
combines two trivial cases: array-indexing by an int literal, and a
swizzle.
Change-Id: Ibb3ca1f324bbee0e9b3556e66644923fc9e0cf45
Bug: skia:10786
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320768
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The following types of expression are hoisted directly into the
inlined code:
- Struct field access: `myStruct.myField`
- Swizzles: `myVector.xzy`
- Simple array indexes: `myArray[0]`
This significantly reduces the number of temporary variables generated
by the inliner.
Change-Id: Ifed226ecc87b096ec1e38752c0c38ae32bd31578
Bug: skia:10737, skia:10786
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319919
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The right side of assignments will collapse redundant swizzles, but the
left side does not. Code like this can actually cause the ByteCode-
Generator to assert if it is run (it asserts in `swizzle_is_simple`
during code generation).
Change-Id: I891912fe0b5de2670dfa95f6702a86d5c42bb2ec
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320296
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 was adapted from a test in SkSLInterpreterOutParams and presents a
challenging double swizzle.
Change-Id: Icb7b3bbb18d4b3cfa0c26acb524c08812ba88096
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319920
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 a reland of 435b482638
inlineStatement now takes a `const Expression* resultExpr` instead of
`const Expression& resultExpr` because resultExpr will be null for a
void function.
Original change's description:
> Support out parameters that use a swizzle.
>
> This CL also removes the `VariableExpression` class that was briefly
> added in a prior CL. This class was intended to support cloning an
> expression while changing the refKind of a VariableReference inside of
> the expression, but it added state and complexity. In this CL, rather
> than track this via extra state, the inliner just recurses into the
> expression as needed to find its VariableReference. Since most relevant
> expressions are just a VariableReference anyway, this is inexpensive.
>
> Change-Id: Id4d926b7d7520b5e6ce455446c05a6d59ef62a84
> Bug: skia:10756
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319917
> Commit-Queue: John Stiles <johnstiles@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
Bug: skia:10756
Change-Id: I35f76c21eccf0ba2ab47e4313e131f7aa26980fa
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320223
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 435b482638.
Reason for revert: ASAN/UBSAN unhappy.
Original change's description:
> Support out parameters that use a swizzle.
>
> This CL also removes the `VariableExpression` class that was briefly
> added in a prior CL. This class was intended to support cloning an
> expression while changing the refKind of a VariableReference inside of
> the expression, but it added state and complexity. In this CL, rather
> than track this via extra state, the inliner just recurses into the
> expression as needed to find its VariableReference. Since most relevant
> expressions are just a VariableReference anyway, this is inexpensive.
>
> Change-Id: Id4d926b7d7520b5e6ce455446c05a6d59ef62a84
> Bug: skia:10756
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319917
> Commit-Queue: John Stiles <johnstiles@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Ibdda47607f9e6e7f3a7459915067cf5e20919993
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:10756
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/320220
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This CL also removes the `VariableExpression` class that was briefly
added in a prior CL. This class was intended to support cloning an
expression while changing the refKind of a VariableReference inside of
the expression, but it added state and complexity. In this CL, rather
than track this via extra state, the inliner just recurses into the
expression as needed to find its VariableReference. Since most relevant
expressions are just a VariableReference anyway, this is inexpensive.
Change-Id: Id4d926b7d7520b5e6ce455446c05a6d59ef62a84
Bug: skia:10756
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319917
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Swizzles in combination with out params are currently broken in SkSL.
They are fixed in the followup CL at http://review.skia.org/319917
Change-Id: I22d9436a15631e6ee2acf9fc312a8634cf3b5407
Bug: skia:10756
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319918
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>
GLSL does not support assigning to ternaries, and will fail to compile
and/or generate non-functional shaders if we pass in a shader that tries
to assign into a ternary expression.
If SkSL is able to completely eliminate the ternary (e.g. if it boils
down to a simple `true ? x : y` or `false ? x : y`), SkSL can strip out
the ternary entirely and generate valid GLSL. This case is harmless and
so it is still allowed.
Change-Id: I960f119fb9934f998697634e6c4e519cd77d3780
Bug: skia:10767
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319679
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: I452e52a87d89cefb5c21a0d9d57e9771f3038d73
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319783
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>
This was unused and did not work on non-GLSL backends.
Change-Id: I6bd314d43cfefa64871b5c0e964b5ae52e494164
Bug: skia:10757
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319778
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Chris Dalton <csmartdalton@google.com>
These tests will also be used for Metal and SPIR-V testing. A small
handful of GLSL-specific stragglers (#version-specific or type-precision
related) will remain in /glsl/.
Change-Id: I7f2b2bd92825c327922c8ce74e438d2daa440dff
Bug: skia:10649
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/319408
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
