When optimization is disabled, I did not expect to see arithmetic like
`color * ONE` disappear (even though it's just a constant fold):
const vec4 ONE = half4(1);
vec4 x = color * ONE;
Change-Id: Id24a632712591c73a904b90e5efdba092d846c0e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382477
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
As you might expect, a function tagged with `noinline` will never be
considered as a candidate for inlining.
Change-Id: Ia098f8974e6de251d78bb2a76cd71db8a86bc19c
Bug: skia:11362
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382337
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@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>
It turned out that everywhere we were using or testing DSL code either
directly or indirectly imported big chunks of the SkSL library. These
imports turned out to be necessary; code written using just DSL.h would
fail with various template instantiation errors.
Change-Id: Iae72d15b0d6ef14614ac1a4ff08c36bc1876cd4d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381638
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
These were accidentally omitted from the supported operator list.
Change-Id: Idecd17adb8b3f5043e36328c65ca12be33e990f4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381637
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Expressions like `value == value` or `color.a != color.a` can be
replaced by `true` or `false` on sight. The GLSL spec makes it clear
that checking for NaN is optional:
4.7.1 Range and Precision
"... NaNs are not required to be generated. Support for signaling NaNs
is not required and exceptions are never raised. Operations and built-in
functions that operate on a NaN are not required to return a NaN as the
result."
Change-Id: I2ad9f2dc505b638ea2904bef41b7a79a2b329551
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381262
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@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>
This can cause unreachable-code warnings when control-flow analysis is
enabled, and CFA has a matching optimization pass, so we only do this
when CFA is turned off.
Change-Id: I6c90bbcc0a9de2478a22dd9054567dca1b13b84a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381258
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>
Reviewed-by: Brian Osman <brianosman@google.com>
Use placement-new to copy AST data into fBytes; this gives us a "real
object" at that memory location which we can later access via
reinterpret_cast. In practical terms, this lets us return a reference
that points inside of fBytes without causing UB. This saves us from
needing to copy the data back out every time it is accessed.
For large objects, this can be a substantial savings. For sufficiently
small objects (<= 8 bytes) the data is copied back out regardless, since
a single load is more than fast enough.
Change-Id: I3ad246acda4feb7e7aa1987d966abd607983e205
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380996
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This lets us see at a glance what each optimization pass contributes to
our final compiled SkSL output.
Change-Id: I52c56c92c408eee34045c5e6f60298cf9548ff5d
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381257
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: 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>
Expressions like `x * 1`, `x *= 1`, `x + 0`, `x * 0`, or `0 / x` don't
actually do anything, and can be simplified to just `x` or `0`. (The
zero case must also check that `x` doesn't have side effects, because
`0 * myFunction()` still needs to call `myFunction`.)
`0 - x` is also detected and rewritten as `-x`.
`0 / 0` is left as-is.
This logic works for scalars and vectors; matrices are left as-is.
A similar optimization also occurs in the constant-propagation pass, so
we see almost no diffs in the tests. If control-flow analysis is turned
off, we do see some improvements. (I didn't reuse the existing code at
all, since it was designed around rewriting the CFG tree, but the
concept was identical.)
Change-Id: Ia99cd81f1d4cd3dafaa43ccac6a2261e3257a185
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380356
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This gives Viewer a way to dynamically toggle CFA off and on, so we can
immediately see the impact on shader optimization.
Change-Id: I912df70de126a74a1f0d11c511b2acc64f1e9f28
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380456
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Declare
This solves several issues caused by the lack of ordering guarantees in
C++; it was possible for the SkSL backend to look for the value of a
variable before its Declare() call gets processed. Moving the initial
value out of Declare should fix this whole class of problems.
Change-Id: I428fe230f1c312a0128c1f00c2a36cb95f4590a6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380358
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
- Values from constant variables are folded in when it helps, e.g.:
const bool SHINY = true;
const float SHININESS = 2;
if (!SHINY) { // <-- optimizes directly to `false`
param = -SHININESS; // <-- optimizes to `-2`
- Doubled-up logical-not and negation are stripped:
y = -(-x); // <-- optimizes to `y = x;`
b = !!a; // <-- optimizes to `b = a;`
Removal of doubled-up negation and logical-not was actually never
implemented in the constant-propagation phase; I just noticed it while
I was here and thinking about it.
Change-Id: Ie28bb9b5af91376f03d926e26e37f4a131bbf550
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379298
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@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 performs the same simplifications as the control-flow phase, but at
IR generation time. There's no visible difference in the tests (besides
DSL) because these aren't new optimizations; they're just happening
at a different phase of compilation.
Change-Id: I26d241167b0e690b23f8f4370339714783c8d6fd
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/371482
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 check is redundant; we now determine this at function finalization
time (http://review.skia.org/379056) and error out before this check
can be reached.
Change-Id: I1706adb90435b430883d7d1ab44cb5dc98fccd8c
Bug: skia:11377
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379836
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This was removed in http://review.skia.org/377842 to allow compilation
without control-flow analysis. We now support const bools in @if
statements (as of http://review.skia.org/379063) and can relax this
restriction a bit.
Change-Id: If3e8018c599d765075568a90af94fb4e4cb25118
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379797
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@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>
In addition to the unsurprising changes to eliminate references to
src/, we also had to tighten up some C++17-isms as they are not
permitted in public headers.
Change-Id: Ie5005a33d7a135e69fb66beca5e7a5f960dbd453
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378496
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>
The original dispatched all the SwitchCase statements from within the
Switch handler. This splits it apart, so that individual SwitchCases can
be handled in a ProgramVisitor subclass' `visitStatement`. No existing
Analysis passes look for SwitchCase statements (they weren't sent!) so
this is currently a no-op change. This functionality will be used in a
followup CL.
Change-Id: I4c18cf4f67130df7e37a4d78a7d1d584764d75ff
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378780
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Covers some common geometry processors, texture effect, etc.
This also rearranges how fp keys are arranged in the overall key. We no
longer include the key size as part of the key - this made no sense.
Instead, we explicitly include the number of children. We also put all
data for one fp before any children, so the tree can be reconstructed
more-or-less top-down.
Finally, added an "addBool" helper that reads nicer than addBits(1)
everywhere.
Bug: skia:11372
Change-Id: I4e35257fb5923d88fe6d7522109a0b3f4c4017d4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379059
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@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>
All layout(key) fields include the field name meta-data, and use as few
bits as possible.
Bug: skia:11372
Change-Id: Ie12b3e0d01148457e5ea078cbf7d0a4bff35302e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378596
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This gives a nicer interface (name + SKSL_INT) to getting all the values
of an enum (rather than iterating over the symbol table). Will be used
in a follow-up CL.
Change-Id: I1c104623bd06f770c57b88ca9996d3e8c87a3ca7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379057
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: 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>
Rather than have the inliner own this responsibility, the function
finalizer now detects if a function is supposed to return a value but
never actually does. This will allow us to detect this error case even
if the inliner is disabled. The inliner should no longer encounter
functions that claim to return a value but don't, so it will now assert
if one is encountered. (The inliner still has the logic to handle this
case gracefully, just in case.)
The check is currently very simple and doesn't analyze the structure of
the function, so it won't report cases where some paths return a value
and others don't, e.g. this will pass the test:
int func() { if (something()) return 123; }
(This is good enough to resolve the inliner issue, though, as it only
occurred in functions with no value-returns at all.)
Change-Id: I21f13daffe66c8f2e72932b320ee268ba9207bfa
Bug: oss-fuzz:31469, oss-fuzz:31525, skia:11377
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377196
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
SkSL IRNodes now have two different factory functions:
XxxxxStatement::Convert -- this is the equivalent of
IRGenerator::convertXxxxx. Type coercion is fully supported. Errors
are reported via ErrorReporter. Failure is handled gracefully by
returning null.
XxxxxStatement::Make -- this is treated more like `new XxxxxStatement`
but with optimizations applied. Types must already be correct.
If the passed-in arguments don't make sense, errors are reported
via SkASSERT. `Make` is always expected to return an IRNode--
either exactly what was passed in, or an equivalent simpler form.
(In some rare exceptions, supporting Make was too difficult and is
not handled in this CL. Constructor was the biggest edge case; it
is extremely complicated [skia:11032], and supporting Make would require
duplicating most of its logic. It was too much copy-and-paste for me.)
Rationale for this change:
Originally, when the Inliner or Rehydrator would create an IRNode, it
would just allocate it via `new` or `make_unique` and take it as-is.
No error-checking, no optimization.
I decided to allow those callers to benefit from simple optimizations.
This meant porting logic from the IRGenerator (e.g. `convertIf`, which
removes empty branches) into widely available factory functions.
However, the Inliner and Rehydrator don't need the type coercion, and
we'd rather have them assert/succeed instead of getting a null and
crashing (in the unexpected case that recreating an IRNode causes an
error to be detected).
Change-Id: I6949328e904258a627106524ed4134b524e012b6
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377843
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This can be used in places which simplify an expression at IR generation
time. e.g. `const bool x = true; @if (x) {...}` could be evaluated at
compile time by calling GetConstantValueForVariable before checking for
a BoolLiteral and flattening the if. It is also used at the top of
ConstantFolder::Simplify to reduce either side of the binary
expression being simplified, in case it's a const variable.
We don't use a lot of const variables in our tests, but this does
improve the code generation in ConstVariableComparison.sksl
when constant-propagation is disabled: http://screen/gnWPhQG8Jc5cER9
Change-Id: I26309769cd16a6833b74b11a115b87c3dc312514
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378017
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This will allow @if statements to work properly with layout(key) values
even with control-flow analysis disabled.
Change-Id: Ib76b8c5553621815decda7fb0eefccb2c82df02b
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378416
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I34353769f4c65bdfa7205d460671a78f3463a5f2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378317
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:11356
Change-Id: I16322e6396dc7e7c8c50ba1d39e07311cf3bd346
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376116
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Conceptually, ForLoopIsValidForES2 is meant to prevent the creation of
an invalid ForStatement, so ideally, we shouldn't need to actually
allocate the ForStatement to check its correctness. This change will
be more meaningful in a followup CL where correctness tests are factored
away from object creation more strongly.
Change-Id: I0d924f741d86f4e157543d801abf4c45d0849dd8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377838
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Interestingly, this improves our codegen even with the optimizer fully
enabled, as apparently statement chains like:
`x = true; x = x; x = x;`
were getting transformed by constant-propagation into:
`x = true; x = true; x = true;`
making them no longer candidates for self-assignment elimination.
Change-Id: I6d94a809e94b01a00fd92459fcbce898b3cbbb11
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377100
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Interestingly, this reduces the size of our dehydated data
significantly, because we were storing the result type of the binary
expression explicitly. Now the result type is deduced automatically from
the left and right types by the Make call.
Change-Id: Ic6187a5c36774f5a4ed2ec579e9cd13b331832b5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377099
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
In http://review.skia.org/375776, an optimization was added to the
Inliner, causing it to skip generation of unnecessary temporary
variables. The fuzzer immediately discovered a flaw in this logic: the
"unnecessary" variable was actually used in the rare case that a
function failed to actually return a value. The inliner didn't detect
this case. Of course, this isn't a valid program either, so now we
report the error and cleanly fail.
Change-Id: I1f201cfd33f45cace3be93765a4e214e43a46e69
Bug: oss-fuzz:31469, oss-fuzz:31525
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377101
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This will be used in the Make functions for BinaryExpression,
TernaryExpression and Function.
Change-Id: Ifde9ca83ab415a00c4712eaa43fd93c695b915cb
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377097
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
The binary and ternary operators need this functionality, and it doesn't
actually have anything specific to do with IRGenerator.
Change-Id: I89a3b0da335ff1fd0b029730934dc1f58c93dd85
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377096
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
We have no optimizations on postfix-expressions at all so this is a very
straightforward cut-and-paste.
Change-Id: I73b3bc73ad833e668306301e6d6859e44230393a
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376846
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Surprisingly, this actually improved our error detection slightly.
The expression `- -half4(0)` can now be simplified to `half4(0)` at
IR generation time, which allows the constant-folder to detect a
constant zero (and from there, a division by constant zero).
Change-Id: I8c4f6ab522efab5bf98913f9c6a1487b7af39a99
Bug: skia:11342, skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376842
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Fundamentally this function wasn't a generic "set the refKind on an
expression" at all; it was intended specifically as "this expression
must be IsAssignable and contain a VariableReference." Also, it needs
nothing at all from the IRGenerator. Moving it out is important because
it will be used by Make functions for things like VariableReference.
Change-Id: I52d613989cd5fb0feb2e98ae454943cdacb9404b
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376841
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>
For now, just use this to prevent *any* layout qualifiers from appearing
on functions, or their parameters.
Bug: skia:11301
Change-Id: I05d8118c7121048c6ef49695a54e3714a8f8687e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376796
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
