Evaluating either kind of expression now works like all other
expressions - evaluate the inner part, then work with the resulting
values. Added unit tests for both of these that previously failed.
With this change, writeVariableExpression is only used for
VariableReference expressions, so adjust that, too.
Reland now safe, after fix to Value::operator[]
This reverts commit 1ea6d6051e.
Bug: skia:11178
Cq-Include-Trybots: luci.skia.skia.primary:Test-Debian10-GCC-GCE-CPU-AVX2-x86-Debug-All-Docker
Change-Id: I14782fcdfef33a47a46334447c5847976721b21f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359564
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
An upcoming change to SkVMGenerator triggers a gcc-8 x86-only bug. This
change prevents that, by limiting the conversions back and forth from
size_t to int (particularly here, where the original value is also
stored using a bitfield).
Change-Id: Ie7f8f3322dd387b0627ab2b2325ce7d82ce6be39
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/359563
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
It's very easy to add intrinsics to the VM generator this way, (and
there are others that can't be written in SkSL itself), so this makes
everything consistent.
Bug: skia:10913
Change-Id: I1fcd0047136459f1f6aee2cf4fd87e875edb7e3d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348879
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@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>
Change-Id: I4e771083e90f3c60b61f7ce7c8e6697e7bf7c7e1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358518
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
In GLSL and SkSL, control statements don't require explicit braces
around single-statement children. That is, the `match = true` child
statement here doesn't need to be braced.
if (condition) match = true;
Because there are no braces, we never create a Block or a dedicated
SymbolTable here. This is normally not a problem, but the fuzzer
discovered that it can dump things into the symbol table inside a child
statement:
if (condition) int newSymbol;
This becomes problematic because the symbol name now outlives its block.
This means `newSymbol` can be referred to later, which should be illegal
(and can cause the optimizer to blow up since the structure is bogus).
There doesn't seem to be any reason to allow this code to compile; the
user can add an explicit scope here to make it reasonable, and it's
(almost) meaningless to declare a symbol that's instantly going to fall
out of scope. This code is now rejected with an error message.
Change-Id: I44778e5b59652d345b10eecd4c88efbf7d86a5e0
Bug: oss-fuzz:29849
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358960
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The previous change caused varDeclarations() to sometimes return an
expression-statement. This only made sense in the context of being
called from Parser::statement(). Other places which called
varDeclarations() expect vardecls and nothing else.
Change-Id: I562657cadfa20dcd77b527f2dc43dca0c6bf389f
Bug: oss-fuzz:29845
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358528
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit f619079545.
Reason for revert: Some bots unhappy.
Original change's description:
> Fix field access and indexing of complex expressions
>
> Evaluating either kind of expression now works like all other
> expressions - evaluate the inner part, then work with the resulting
> values. Added unit tests for both of these that previously failed.
>
> With this change, writeVariableExpression is only used for
> VariableReference expressions, so adjust that, too.
>
> Bug: skia:11178
> Change-Id: Ia595be473b55f4bb03ec25897f9929835177257c
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358529
> Reviewed-by: Mike Klein <mtklein@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=mtklein@google.com,brianosman@google.com,johnstiles@google.com
Change-Id: I56776139f9164b24b35a93307774e9b12c50054e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11178
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358959
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Change-Id: I59821d748768d74b251d9339787f193c23e3d4dc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358526
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Evaluating either kind of expression now works like all other
expressions - evaluate the inner part, then work with the resulting
values. Added unit tests for both of these that previously failed.
With this change, writeVariableExpression is only used for
VariableReference expressions, so adjust that, too.
Bug: skia:11178
Change-Id: Ia595be473b55f4bb03ec25897f9929835177257c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358529
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This attempt to retain a std::move'd value was inadvertently left in
place in a previous CL. It's harmless, but should have been removed.
Change-Id: I485e0f14d921fe6c87a62520c7478a28e8ce28d8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358517
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This allows us to write SkSL shaders which are valid both for use as
Runtime Effect, and for compilation with skslc targeting Metal.
Change-Id: I74e125d81865d4092e657a7d9948d2e72054bda5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/357777
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>
(and Ternary for good measure)
Change-Id: I4afa121d54ab9ba8d0814693ce53da7cb73ef340
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353626
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Added asserts that verify we don't try to emit the same struct or array
with two different memory layout rules. Some code paths were failing to
inspect the associated variable, leading to incorrect errors about the
attached offsets of members.
Added a test case that triggered that error, and also triggers the new
asserts.
Then, fixed the underlying cause: writing out the struct definition as a
side effect of accessing a member in getLValue().
Bug: skia:11205
Change-Id: I6e5fb76ea918ec9ff10425f2d519ddbc54404b27
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/357436
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This allows us to roll the Parser back to an earlier state if we need
to do so. This includes:
- rewinding the lexer
- restoring the previous Pushback node
- backing out AST nodes
- backing out errors
This functionality is used to back out of parsing a vardecl if we
discover mid-stream that it is actually an expression statement that
coincidentally starts with the name of a type.
Change-Id: Ia5feb45019693931c1e6870e3ff7a5398924c863
Bug: skia:11198
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356997
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The lexer can be reset to an earlier state simply by overwriting its
offset to a previous position.
Change-Id: I571c7981dbec3c43a894fe599f2e2a167bc380df
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356976
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This should not cause any functional changes, and is just a prerequisite
for upcoming DSL work.
Change-Id: Iea165d3b7ede39ccc9cf5f5d78f623bc883b391e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356816
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Change-Id: Ib150e6d6d3de34a85ce8051eea843ab3b2d7ab75
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356921
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This was the last remaining user of ByteCode. The skvm solution
is faster, and lets us delete the ByteCode system.
Testing on 15 instances of sinusoidal_emitter (90k particles):
- ByteCode ~9 ms
- ByteCode (older, optimized): ~5.5 ms
- skvm ~2.1 ms
Change-Id: Ia2e5c9ab2d36c97e59af28a6f989bf212889e439
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356919
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@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>
The SPIR-V code generator would sometimes invoke helper functions in an
unsequenced way, causing the output to vary across different compilers
or optimization settings. This CL explicitly sequences these operations,
which should make the output consistent across platforms.
Change-Id: Iaaf75b80e7495768d73dd6afa5b6d03f9cf3f262
Bug: skia:11175
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356844
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We use our own custom lexer now.
Change-Id: I6a2a408094ba37c2eef7092f5a5caa107938613b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356476
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The previous implementation assumed that SkSL expressions do not have
side effects and so treated either side of a Boolean expression as
short-circuitable. That is, `foo() && false` and `false && foo()` would
both be optimized to `false`, eliminating the `foo()` call.
We now check for side effects first. An expression like `expr && false`
can only be optimized to `false` if `expr` has no side effects. (If
`expr` does have side effects, the expression is left as-is.)
Change-Id: I473cf026a8afe35d6a8d9518498f2b26d8996e60
Bug: skia:11162
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356357
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>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Additionally, restructure the unit test to return a color (green for
pass, red for fail).
Change-Id: Ib1bb6bd8771c72cc751d8d2c65cc14a693166d4c
Bug: skia:11112
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356301
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: I81c9a6ddcd65d40ebced18c7cfb1cad51066ba7b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356297
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
external callers
This should not result in any functional changes.
Change-Id: I5ef0301cf63bee561871c0110fda2692849b53d8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356102
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
No functionality changes, but the new variants of the methods make them
more accessible from upcoming DSL code.
Change-Id: Iaa9f1fab31cb7db00007b00d7d3b88ff5b9696b6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356103
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This exposed a preexisting error: GrGrSLTypesAreSupported.fp used non-
square matrices, which are not actually present in GrSLType. The test
has been updated to use square matrices.
Change-Id: Ib51141cc14a0c3fcd1c3c3abf378f190d457b95f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356077
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@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>
This reverts commit b7e836cee9.
Change-Id: I3c39a928ba4a9a2863b616f2a500975294b03860
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/355980
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>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit ebf569004f.
Reason for revert: std::clamp is c++17
Original change's description:
> Support indexing by loop variables in SkVMGenerator
>
> Bug: skia:11096
> Change-Id: I25a91bacf1c3455ac67422fb0e59b9b152c2054a
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354667
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,brianosman@google.com,johnstiles@google.com
Change-Id: I0590cf7fe626fb59be3381b5e8eb66a9a2a9e8cb
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11096
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/356056
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Bug: skia:11096
Change-Id: I25a91bacf1c3455ac67422fb0e59b9b152c2054a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354667
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@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>
- Remove unexplained special case where we avoided coercing uints to int
- Replaced the set<int> of case values with SkTHashSet<SKSL_INT>. Should
be faster and more type-correct.
Change-Id: I3286bd50253cc7a1ff6a550dc429bdfebbb8d8c5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354670
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 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>
This enforces an even stricter version of the rules from GLSL ES 1.0
Appendix A, Section 5. Essentially, indices (to arrays, vectors,
matrices) must be made of literals, loop indices, and expressions made
of those two.
Bug: skia:10837
Bug: skia:11096
Change-Id: I437a5ed64da58e24d5991ddbde68859f5214e98b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354665
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
As far as I know, there shouldn't be a way to introduce a struct or enum
other than at global scope; the keywords are not accepted inside a
function body. In fact, I wasn't able to find a way to exercise these
code paths in practice. But we now have concrete assurance that any
possible type can be cloned into a symbol table safely; all Types are
either built-in (available everywhere by design) or are clonable.
Change-Id: I4b006b6cab995b3e598b683736ab9689828629c9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354664
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The inliner needs to clone Types from one SymbolTable to another when
cloning blocks of code. However, it seems like a poor division of
responsibility for the inliner to need to know how to clone every Type
correctly. It makes more sense for this logic to exist within the Type
class itself.
Change-Id: I4a383d5e22d5084eb35992a0b5c24865d2030282
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354662
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>
Change-Id: I62755dfcf5c8bb5bd7e0cc742e675aacf4afbb78
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354658
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This CL limits the availability of functions like "MakeLiteralType" and
"MakeScalarType" to the BuiltinTypes class. This allows us to know,
definitively, that whenever we encounter these types, they are part of
the BuiltinTypes class and available anywhere in the program without
needing to be cloned.
The remaining MakeXxxxxxTypes in SkSL::Type are constructed dynamically
and injected into child symbol tables during parse/IR generation. These
types must be treated with special care when being moved or cloned:
- Arrays (MakeArrayType)
- Structs (MakeStructType)
- Enums (MakeEnumType)
Change-Id: I5490d6739c2a5ffdd54195f5a0b9b5be05d07953
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354878
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
No code or functionality changes.
Change-Id: I24317767570ae9ebbfea56f873d98709fb22d8b9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354876
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 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>
Change-Id: I93d580e67cb66d388bd66b280ff229cb79e5b154
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/354218
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@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: Ia3ac338bef376aa1649569b9ebd3f7feb23ffd52
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353936
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This better matches GLSL's type coercion behavior.
Change-Id: I73fcfd8a9e57fd4cdb1692074d73ebd8fb788ac2
Bug: skia:11164
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353712
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Literals are still flexible; we still allow `1` to coerce to float.
However, we no longer accept code like `int x = sqrt(2);` or
`int x = 0; float y = x;` without an explicit cast.
Change-Id: Ieb294a4877447e2336252f876e8bc489d1e4a59a
Bug: skia:11164
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353417
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: 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>
Bug: skia:10680
Change-Id: Ic77f7355866363ef476a93d8da180cf53207fa6d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353707
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
It's not sound to pass undefined (skvm::NA) values into select(),
but this is working today because the F32a argument is 'fixing' it.
The first time through this snippet updating fReturn value,
int i = 0;
for (skvm::Val& slot : currentFunction().fReturnValue) {
slot = select(returnsHere, f32(val[i]), f32(slot)).id;
i++;
}
the call to f32(slot) creates an F32{builder, NA}. We pass that to
select() and that argument's F32a(F32) constructor, resulting in
F32a{builder, NA, 0.0f}. Then when we need that as an F32, we resolve
it as splat(0.0f) because the F32a's id field is NA.
In short, best to remove F32a. :)
Added some SkASSERTs that would have caught this.
Change-Id: I67324cf20ad39ca555e69b9c407f379d14046043
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353838
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
I'm looking to phase out I32a/F32a, and rewriting this expression to
avoid select() makes it easier, making these types unused except in
SkVM.{h,cpp}.
There's no particular reason beyond making that refactor easier to do
this: SkVM can convert select(cond, splat(1), splat(0)) into cond & 1
itself, and once I'm done with removing I32a/F32a, if we prefer select
we should be able to rewrite this back as
dst[i] = skvm::select(i32(src[i]), 1, 0);
Change-Id: I562a112e54fdc2578802db02f6754c64a12798cc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353837
Commit-Queue: Mike Klein <mtklein@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>
Change-Id: Ibb9dca7b8e04485924f6a7d8dd4b17df90d5f320
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353779
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>
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>
Such loops must be unrollable, so that's what we do.
Bug: skia:11094
Change-Id: I1b34917b6f2d015ae7867415d0120a5df0ffd618
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353619
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Bug: skia:11094
Change-Id: I68a08e79d29579901b74daca3c22f5112fbb3c8c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353356
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@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>
Change-Id: I9859081a14b110731f943e09fdd94dc10e0c9dfc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353580
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This adds basic operator overloading (everything but array indexing) and
related tests to the SkSL DSL.
Change-Id: Ic9fdc9a02a5496e2706d18fb435d838b4ee53ad7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353103
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Previously, we blindly trusted that the scalars inside a compile-time-
constant vector Constructor would be of matching type. (e.g. the scalars
inside a half3 constructor would always be of type half) This *should*
be true if we've done all our type-checking homework correctly, but we
would ABORT if anything went wrong. Now, we are less trusting and will
actually verify types and do the right thing if the scalar inside turns
out to be an int somehow. In practice, there's no real downside to
doing the conservative type-checks.
Change-Id: If0a255eb96a0ccfec7d0a261caad542cae93d078
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353556
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>
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>
The new signature is more amenable to being called from outside code,
in particular the DSL.
Change-Id: I4c7650b71f02471f27e348b0c44abe8511691db1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353038
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Previously, the only way to detect a literal Type was to explicitly
check against the values fFloatLiteral/fIntLiteral or compare the name
against $floatLiteral or $intLiteral.
Now, you can call type.isLiteral() to see if the type corresponds to a
literal, or call type.scalarTypeForLiteral() to promote to the
equivalent non-literal type (fFloat/fInt). Calling
type.scalarTypeForLiteral() on a non-literal type is safe and will just
return the input type unscathed.
Change-Id: Ib939dda9bcd43b3ef4159211c5349d71fc857b95
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353037
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This will allow errors to be reported outside of the IRGenerator more
easily (without passing around an ErrorReporter object).
Change-Id: I4bcb59fcd526599fa593fcb3b1de0a5ae64ab901
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352737
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This is the first test that used uniform data, so fix up how uniforms
work in the generic SkSL-to-SkVM function.
Bug: skia:11094
Bug: skia:11096
Change-Id: Ie391c1a6b8b68f0f4f014d7e767d7b5101341fab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352739
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Change-Id: I96b547de4fe4b73096fb26d0ef21a4e7555ca06a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352238
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:11127
Change-Id: I3a7fda8bb62f9cf9b6c83441703f537e75461d07
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352509
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Previously, these were in SkSL::Context directly. This change doesn't
remove them from the context entirely, but it gives them a dedicated
subclass and firewalls them off from the rest of the context.
Change-Id: I0c344bf7436a11b8494a5fe7542d0a4ef1ece964
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352502
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This doesn't change any logic, just makes the IR generator a few
hundred lines shorter.
Change-Id: I92010191ee9283c33499c819d65fc85913f25824
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352121
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@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>
Swizzle::constantPropagate is a limited form of this optimization:
https://osscs.corp.google.com/skia/skia/+/master:src/sksl/SkSLCompiler.cpp;l=1159
It predates this optimization code, though. It also flattened out an
unnecessary constructor that would have otherwise required an extra pass
to eliminate.
Updated the optimization code at L1159 to simplify away the unnecessary
constructor where possible, and then removed Swizzle::constantPropagate.
This has no effect on generated code.
Change-Id: I0f43d5c51761965230c853f309a6ef068f9aef77
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352120
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This tripped us up at least once recently.
Change-Id: I902759929e359ce639557bb766b0c6593badc07b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352093
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Bug: skia:10680
Change-Id: I8697bdc157d250f3c390c7f49074318aa8c7bdab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351918
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Now that BoolLiterals can be constructed with a Type*, there's no need
to pass the context through.
Change-Id: Ic2003aa03f5c2428e73e81c95eb12c862700554a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352025
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
We were dropping the modifiers when re-emitting the function
declarations. We rarely noticed this, thanks to aggressive inlining in
our existing tests. Upcoming changes to the test suite (to disable
inlining) exposed this bug.
Change-Id: I773d5ad6f7694d4a491be525c05b730794216e43
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352039
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>
This makes almost all existing code read more clearly.
Change-Id: I314331d9aa2ecb4664557efc7972e1a820afc250
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352085
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: Ia49b9dceccb83f8316698ba13d0920a988335010
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/352040
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: I0bbda6a41391fc2a11dc812be5e9c0c0d14c4d75
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351921
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
The CFG/definition map are no longer valid after replacing an expression
entirely. Swizzle-of-swizzle optimization was another case where the
optimizer would replace an expression wholesale, but failed to set the
needs-rescan flag.
Change-Id: Ida0363d738cd1d3ac2a48c824aa04065a7ca16b7
Bug: oss-fuzz:29085
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351776
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Change-Id: I4a130b0c776888587122a7ae25e1b32b10011d55
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351499
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
The code at this point doesn't do anything useful, but establishes some
of the basic types and patterns.
Change-Id: I580a9e75ffa3162879893450fb7d1f0905a10687
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350697
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Change-Id: I75f907ca673ee67f5d623b032128b97833070a0b
Bug: skia:10931
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351504
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>
All errors need to be caught earlier, so turn any remaining TODO items
into asserts.
Bug: skia:11127
Change-Id: I6731f947233522df6397b3444c26d5ebc417c431
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351506
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This will flatten out expressions such as `!false` or `!true`. We
already had a similar fix-up at IR generation time which handled simple
cases, but this will catch more complicated ones like `!sk_Caps.xxxxx`
(since caps bits are only flattened out at constant propagation time).
Change-Id: I04282809d9a784266a64dbcafd097f3b0662806c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351497
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 not actually necessary now that constantPropagate can fully
flatten out unary negation into its constant operands. The compilation
results don't change at all.
Change-Id: I7ab55bd3720413609d799dd866e1703973cb2626
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/351202
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, we used a union of (int64, float) to store the cached value.
However, the union-based code turned out to be more complex than just
type-punning the float's bits to an int via memcpy (which we needed to
do anyway). The union-based approach was also likely to be UB by the
letter of the law--we were creating float keys by storing into fFloat,
then checking the map by comparing equality on fInt.
Change-Id: I62c7ff4b5fab8bb1be8836c23f746ef254053b6f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350957
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>
Like getIVecComponent or getFVecComponent, this retrieves the n'th
element of a Boolean compile-time constant vector. This will be used in
followup CLs.
Change-Id: Ib41c9c89cb773251e4c0d6cdcaea0437d8074e48
Bug: skia:11141
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350918
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This will be used in a followup CL to implement getBVecComponent. At
present it's not called.
Change-Id: Idd6f18314d0835af3946ea7458e6650384f505ea
Bug: skia:11141
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350703
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Previously ExternalValues were flexible, and could be used as raw values
(with the ability to chain access via dot notation), or they could be
callable. The only non-test use-case has been for functions (in
particles) for a long time. With the push towards SkVM, limiting
ourselves to this interface simplifies things: external functions are
basically custom intrinsics (and with the SkVM backend, they'll just get
access to the builder, and be able to do any math, as well as
loads/stores, etc).
By narrowing the feature set, we can rename everything to reflect that,
and it's overall clearer (the SkSL types now mirror FunctionReference
and FunctionCall directly, particularly in how they're handled by the
CFG and inliner).
Change-Id: Ib5dd34158ff85aae6c297408a92ace5485a08190
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350704
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Continue to test everything with the ByteCode interpreter, and run most
tests with the new SkSL-to-SkVM utilities, as well. A few tests rely on
features that aren't yet implemented (function calls, looping), and some
of the bespoke tests (that don't use the test() helpers) use even more
exotic features that need to be implemented or disallowed in the IR
generator. This is getting us closer to not needing ByteCode at all,
though.
Refactored a bunch of the helper code to reduce copy-paste among the
many different 'test' functions.
Change-Id: I138d4a24266f2d862742245c5ee895d86c01018e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350560
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@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>
This CL adds additional variants of some IRGenerator functions to enable
them to be more easily called from upcoming DSL code. This should not
change functionality (other than a slightly different error message in
one case).
Change-Id: Ic727c194fca5111d4283007f4ec23653598a853b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350017
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
IntLiteral and FloatLiteral each had a pair of public constructors: one
which takes a Type (what type of Int/Float is this?), and another
constructor which takes a Context (assume the generic Float/Int type).
BoolLiteral had a public Context-based constructor, but its Type-based
constructor was private. The Type-based constructor is now public to
match its siblings. This allows us to create Booleans without needing
a Context&, if we have an fBool_Type, and fixes make_unique (which
doesn't like private constructors).
Change-Id: Iffb4980947a73b3a89aa1452cc90dce63620bd67
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350636
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: Ia4a1c38161046b94dc56a1a76704766f1e14aab7
Bug: skia:11131
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350019
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@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>
Fixes terminology inside the generator to correctly use 'arguments'
rather than 'parameters'. Retains the caller-supplied argument span, so
that generateCode can push changes to 'out' and 'inout' parameters back.
Adds a useful helper function for getting a named function from a
Program.
All of this is used more extensively in upcoming CLs that migrate
interpreter unit tests (and particles) to SkVM.
Change-Id: I9d9aef4b7c4daf93d822e22f813162f4ed2385be
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350023
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The code didn't take into account that x and y might be different types.
(This bug was not actually harmful; type coercion allowed the code to
compile even with the wrong type. The float would be silently splatted
into a vec and the rest of the code would work as-is.)
Change-Id: Ib76bc733f76304e451ef9197421b4bc22e29e49c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348888
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: Ieb7698d357c9be05ca1f17de84215add54553f84
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349065
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I2c39df532803d827d7cad876021f2ead81145f1d
Bug: skia:10902
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349064
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>
Previously, we did very little to distinguish between a built-in
intrinsic and a user-defined function whose name matches an intrinsic.
This could lead to all sorts of surprising outcomes, as our intrinsic-
rewriting code is able to make assumptions that might not hold true for
arbitrary user-defined functions.
Change-Id: I4180e0c5becdeb6a0a162534eaecfc90dda3392c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349062
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, we had three separate implementations of intrinsics:
- "Special" intrinsics (the common case)
- "Metal" intrinsics (the GLSL bvec comparison functions)
- A hardcoded list of if-then workarounds in `writeFunctionCall`
There doesn't seem to be any particular rhyme or reason for why things
were organized in this way. These all been merged into the "Special"
intrinsic handling.
Change-Id: I10e4eed00a2dd2c7cb1f79a4b96460fc43a68ebb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/349059
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: I6c0a6192a78ce60be60a71ed75350ca1bc256d57
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348890
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Change-Id: I954af70f545a2258babd82af0d43d509201fdc59
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348889
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:10913
Change-Id: I430e5eb3fecb0f15775db03699819194d44271b6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347958
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@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>
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>
4x4 was dividing a matrix by a scalar - this isn't allowed, multiply by
the scalar's inverse instead.
The types in the signature were derived from type.name(), which wasn't
applying the half->float re-mapping.
Finally, use raw strings so the resulting shader code isn't all crammed
on one line.
Bug: skia:10913
Change-Id: Ie28373fc138445b8c195dbd37687e4ad4504e918
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/348177
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@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>
Bug: skia:10913
Change-Id: I59f5b0fb2d015f8543b4038c2c5b18ce24c194a8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347956
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Previously, some types of overflow were detected, but most would assert
or silently generate invalid code. Now, the parser will properly report
an error if it encounters any integer that exceeds UINT_MAX or any float
that exceeds FLT_MAX.
This fixes test OverflowUintLiteral.sksl. Added a test for floats as
well, OverflowFloatLiteral.sksl.
OverflowIntLiteral.sksl does not fail yet, because its values are larger
than INT_MAX, not UINT_MAX. These are legal from the perspective of the
parser. This must be caught later at IR generation time.
Change-Id: Ia5a904d01427cdc9f2ab5f4174154418737835e6
Bug: skia:10932
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347176
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
These are part of the GLSL ES 1.0 spec.
Bug: skia:10913
Change-Id: I61bfcadc92209fb3a6aa07b3d6ad579460ad2ecd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347049
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This was only used for nullable fragment processors, which are gone.
Change-Id: I1ea805c683995367a7525b787c9113ae6d2d0ae0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347051
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@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 feature had devolved to just an assert, and one that isn't really
necessary - all of Ganesh is built to handle any child processor being
null. The next step is to remove nullable types entirely -- a large
amount of code.
Change-Id: I612a5867f8690400b405aa1f5c929e76cf5918fd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347050
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Salomon <bsalomon@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>
'int' is the only integral type that exists in GLSL ES 1.0 (and it's not
really guaranteed to be an integer). This enforces the same restriction
on runtime effects - no unsigned integers, and no short or byte types.
Bug: skia:11093
Change-Id: I938f1e0e125dc8347507f428b46b51c66033c752
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347046
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
More GLSL-isms that weren't used, and weren't really ready for use (most
were defined as "Other", not "Sampler"). If/when we need these, it's
easy to add them back. In the meantime, we should have a simpler system
for reserving keywords that doesn't pollute the type system.
Bug: skia:11115
Change-Id: I436c1e4de6e6b92ff14fc99ed1d47e0c5d1e3aff
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347045
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Missed these when I moved the float versions earlier. Non-square
matrices don't exist in our minimum spec (GLSL ES 1.0).
Bug: skia:11093
Change-Id: I09b3ab71199bc70d9b54302c14b93bc3f3dec2d0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347042
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
These are GLSL-isms that weren't really implemented - each one was a
"generic" type that only resolved to a single underlying type. We've
got along just fine without them for years, so update our sample()
declarations to take the actual underlying type. (Note that we had
worked around this by declaring an integer version of sample where
necessary, so we can presumably keep doing that in the future).
Change-Id: I4c46a2fa0c1f19e6278298c8005a2760329e7abf
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347040
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Nullable fragment processors still exist, but they're handled
transparently by sample() within C++, so there's no need for .fp files
to ever do these tests manually.
Change-Id: Idf2bc58505207560553066c0126a2a036c5d970b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347039
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
These were unused, and I don't think they were correctly defined, given
the inconsistency in naming and construction. If we do end up needing
something like this (a variant type for returns from sample(), we can
re-add them correctly).
Bug: skia:8863
Change-Id: I32a9487ba0e247f67696947babbfd6b64f0a047c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/347037
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: I2d396907387de4a5f3407d81efb9d2cd80e430d1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346265
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Opaque types can no longer be copied via assignment or construction, and
various restrictions originally applied to the "fragmentProcessor" type
have been extended to cover opaque types in general.
Change-Id: I55ab7aefd1e6ef277e56a9408b430e1de5ba12ca
Bug: skia:11027
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346264
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 intrinsic was previously lacking a unit test, and wasn't actually
implemented in Metal or SPIR-V. Fortunately it's trivial to add.
Change-Id: I68bbdc58376b579c7f3f0ae5f49323b389c2b8c4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346263
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 slightly simpler than having three separate overloads, and
provides the same results.
Change-Id: Icc7a749fd642f6d6a9e69b769494c566569ea8f4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346262
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>
Updated Pair type in SkTHashMap to derive from std::pair to fix C++14
issues with structured bindings.
Original change's description:
> Add support for range-based for loops to SkTHashSet/Map.
>
> This allows loops over SkTHashes to break in the middle, and also
> removes the need to use lambda captures to bring variables inside the
> loop's scope.
>
> Change-Id: Ief55d776b2c57a44b24cfe1c94493a5d514791c8
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346496
> Reviewed-by: Mike Klein <mtklein@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>
Change-Id: I2ac5b2c59e70ed0ec3b42b32e7994d6bcdf56b40
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346502
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
I checked that this prevents diffs on my M1 Mac.
Without this we'd see FMAs.
Change-Id: I24fa2e301cde556de556332b472399b97e6a08fe
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346676
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Previously, a return statement inside a scoped Block would always result
in the return expression being assigned to a temporary variable instead
of replacing the function-call-expression directly. This was done
because there might be variables inside the Block; these would have
fallen out of scope when the expression is migrated to the call site,
resulting in an invalid expression.
We aren't actually examining the return expression so we don't know if
it uses variables from an inner scope at all. (Inspecting the return
expression for variable usage is certainly possible! But it's a fair
amount of code and complexity for a small payoff.)
However, we can very easily get most of the benefit here without paying
for the complexity. In this CL we now look for variable declarations
inside of scoped Blocks. If the code doesn't add any vardecls into
scoped Blocks, there's no risk of scope problems, and we don't need to
use a temp-var to store our return expressions. If any vardecls are
added, we go back to using a temp-var as before.
Change-Id: I4c81400dad2f33db06a1c18eb671ba2140232006
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346499
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
These don't exist in our minimum spec (GLSL ES 1.0)
Bug: skia:11093
Change-Id: Ia2d871199fff2a98dcd517c1eebe46decb0c2dfb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/346657
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
