Commit Graph

431 Commits

Author SHA1 Message Date
John Stiles
2a6c41571b Fix Metal codegen error with structs containing compound types.
While working on an unrelated test, I accidentally triggered a bug in
Metal code generation. Our struct-equality helper functions did not
properly handle vector fields. Wrapping each comparison in `all(...)`
fixes the problem. (all() on a scalar is allowed and does nothing.)

Our struct comparison tests now include a vector and a matrix.

Change-Id: I59061ae9c3c3ab2c2dbdcb5257bc23e2257152af
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470399
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-12 15:54:41 +00:00
John Stiles
1991780081 Update LoopFloat/LoopInt tests to reduce hoisting.
Previously, none of our `runtime` tests relied on the input coordinate
in any way, so all of the logic was hoisted above the main loop in every
test. This CL adds an artificial reliance on the input coordinate so
that we have at least some SkVM tests with real code in the main loop.
This lets us see debug trace instructions interleaved with real code.

The input coordinate is clamped against a known uniform value
(`colorGreen` always contains 0101) so that the final test output
remains consistent in practice.

Additionally, I noticed that this test was only enabled in ES3, but
it doesn't seem to have anything ES3-specific in it, so it's now
enabled across the board.

Change-Id: Ie82f40b1060edb6071e300040ac59fb7d27094b0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470397
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2021-11-12 13:45:59 +00:00
John Stiles
05c41162fb Allow inlining of functions with unassigned out-params.
Our inliner would ignore any functions with `inout` parameters, because
inlining them properly was more complex than just leaving the function
call. However, real-world code can sometimes contain helper functions
that have `inout` params that are never used at all (e.g. an uber-shader
with some features turned off).

We now read the ProgramUsage and check to see whether or not the
`inout`-qualified parameter is actually modified. If it's never changed,
the function now remains a candidate for inlining.

Change-Id: I92e494f94cc070801cb9aa28bd13faa689b806b6
Bug: skia:12636
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470299
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-11 19:33:06 +00:00
John Stiles
6ee5d9e3c9 Improve index-folding of arrays and matrices.
Yesterday's implementation was close but I realized later that it wasn't
quite ideal.
- Array index-folding was gated on `isCompileTimeConstant`, which is too
  strict. The real limitation is `hasSideEffects`. If an array contains
  a side-effecting expression, we should leave it alone. Otherwise it
  is safe to pluck out an element from the array and toss the rest.
- Matrix index-folding was gated on `getConstantSubexpression` for the
  extracted elements, but did not check the other elements at all. This
  was too lenient; we now only proceed to the folding step if
  `hasSideEffects` returns false.

I added some tests to verify the final behavior and also discovered a
small related issue. Diagonal matrices were not substituting literals
in for constant-values, which inhibited folding as well and would break
constant-expression evaluation. This is now fixed.

Change-Id: Idda32fd8643c1f32ba21475251cd4d4dd7cea94c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470396
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-11 18:59:04 +00:00
John Stiles
74410023ac Add tests for inlining functions with out/inout params.
Functions which don't write to their out params should be safe to
inline, but we currently don't recognize this.

Change-Id: I753e48067c7be4473675ef6c95e61af17dc5ae41
Bug: skia:12636
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470298
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-11-11 18:46:21 +00:00
John Stiles
107ea94139 Add support for clamp($genUType, ...) in SkSL.
We never used it internally, but the shaders used by Filament rely on
it. It doesn't exist in ES2 so this doesn't affect Runtime Effects.

Change-Id: Idb2afb15ff160b950ad02101bf6381a5d5c56468
Bug: skia:12635, skia:11209
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470156
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-11 18:43:42 +00:00
John Stiles
6fae052362 Implement constant folding for index expressions into matrices.
Indexing into a constant matrix is a constant expression, so we are
obligated to support it for ES2 compatibility.

Change-Id: Ibe1e5bac39d9a88ce0222997a38e8b6952fdb336
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469819
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-10 21:38:56 +00:00
John Stiles
d92f9c2f8d Add tests for matrix constant-expressions.
We should support constant-expressions involving matrices (GLSL ES2
does, WebGL does). We currently don't. We do properly report out-of-
range indexing, but we don't optimize away valid matrix index
expressions or allow matrices to be indexed in a constant-expression
context.

Change-Id: If58aa4c5f15abef421a412957072f3617b4176df
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469818
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-10 20:59:36 +00:00
John Stiles
3af95eb87a Fix test when optimizations are off.
When optimizations are disabled, the compiler flags this function as
exiting without returning a value (since it doesn't convert the ifs from
maybe-taken into definitely-taken).

Change-Id: I226d0f6ba8cc664aecf1c5afaaf03e92038185ba
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469821
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>
2021-11-10 20:17:30 +00:00
John Stiles
76c1ff1566 Optimize indexing into an array with a constant-expression.
Previously, SkSL was unable to resolve the constant expression `x[y]`
for a constant-array `x` and a constant-integer-scalar `y`. Now, if `x`
and `y` are known, we can replace `x[y]` with the indexed array element.

Note that we need to be careful here, as it's not a valid optimization
to eliminate array elements that have side effects. We preserve side-
effecting expressions using the comma operator.

Change-Id: I5721337eb42b48c0b05f919c1cadfae19dd3b84f
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469839
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>
2021-11-10 19:43:58 +00:00
John Stiles
0b84159e3b Improve array-indexing tests.
Previously, we didn't have tests which leveraged constant-evaluation of
array indexing (because we didn't support it), and our test files
commingled constant-indexing into vectors with constant-indexing into
arrays.

The test files now separate vector- and array-handling into separate
tests, and a ton of new cases have been added to ArrayFolding. The
ArrayFolding tests now require constant-evaluation of array indexing,
so they fail in this CL, but will be fixed in the followup CL.

Change-Id: I3b663e743d97d6db80627bc9b7808f88c99917a7
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469528
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-10 18:31:28 +00:00
John Stiles
6e6ae1b762 Fix inlined out-of-range vector access.
Previously, this code assumed that IndexExpression::Convert had done
range checking and that it was safe to access the base expression at
the passed-in index. The inliner violates this assumption, because it
can replace unknowns (where out-of-range access is undefined but non-
fatal) with knowns (where out-of-range access is forbidden).

We now do range-checking inside IndexExpression::Make and report the
error cleanly, instead of asserting inside of Swizzle::Make due to an
invalid component index.

Change-Id: If0f31b1f694bcc2a875d124f70be311d6634c77b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469535
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-10 14:44:38 +00:00
John Stiles
32385b7070 Report incomplete expression-statements as errors.
Previously, a dangling type or function reference would be eliminated
silently with optimizations on, or would assert when optimizations were
off.

Change-Id: Ib2e273b6f069724e8872c9cb97351b647b875a62
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469525
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-09 22:10:18 +00:00
John Stiles
ee525493ea Add test for incomplete expressions.
The ExpressionStatement currently eliminates dangling references without
reporting them as an error. This happens due to optimization; these
expressions (being meaningless) have no side effects, and so the
optimizer replaces them with Nop. When the optimizer is off, these
programs trigger an assert:

https://osscs.corp.google.com/skia/skia/+/main:src/sksl/SkSLAnalysis.cpp;l=582;drc=e7a953524787e3bd0c437ec52de4e40986689825

A followup CL will fix ExpressionStatements so that they report
incomplete expressions as an error.

Change-Id: Ica49166032e670749fc1b4e7a869fbab03364d4f
Bug: skia:12472
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469524
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-09 22:09:49 +00:00
John Stiles
70ae43148d Implement trace_var opcode.
This writes an entry to the trace buffer every time a slot value is
changed.

Change-Id: Iac3912be71ad654f70a7158e306e0643086c6cb0
Bug: skia:12614
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/469179
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-09 15:24:00 +00:00
John Stiles
15384b1195 Add a trace_line opcode to SkVM.
This will be used to populate a trace buffer for the SkSL debugger.
See http://go/sksl-tracing for details and rationale.

Change-Id: I4c218c65ff01c339cf460e97e41566860a694720
Bug: skia:12614
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/468436
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-11-05 19:55:39 +00:00
John Stiles
afa75bc522 Add a SkVMDebugTrace flag to ProgramSettings.
When this is enabled, the SkVM code generator will emit trace
instructions for debugging purposes. (The trace instructions are a work
in progress, so at present, the flag doesn't do anything meaningful.)

Change-Id: Ia7d66840d915b1a7e531a3069e641c840bb9c0eb
Bug: skia:12614
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/467764
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-11-05 13:28:40 +00:00
John Stiles
e7a9535247 Enforce basic limits on global size in SkSL.
Much like http://review.skia.org/467759, this CL defensively guards
against programs which consume more space than is reasonable. Globals
exist outside of functions, so they wouldn't be caught by the stack size
checks.

Change-Id: I035f27d57bc329508820a729a1e367ecaadfe156
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/467760
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-11-04 18:34:19 +00:00
John Stiles
7cde28909f Enforce basic limits on function stack size in SkSL.
Functions that declare variables totaling more than 100,000 slots will
now generate an error.

This is only a partial mitigation to the problem, as a sophisticated
attack could still chain/nest multiple functions together to consume
extremely large amounts of stack. However, this mitigation is still more
sophisticated than our peers; both WebGL and glslang are susceptible to
similar problems, and in the general case (ES3+ with full flow control)
it's intractable.

Change-Id: I153c75267c017a23f59fe9e59f6e391197ee6101
Bug: oss-fuzz:40304, oss-fuzz:40694
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/467759
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-04 18:17:44 +00:00
John Stiles
390edeb88d Fix fuzzer-discovered error with no-op arithmetic.
The fuzzer triggered this error in a strange way that involves parsing a
TK_INVALID token. The fuzzer's original input used \xFF bytes in the
shader text to do this. I replaced these with the ` character since it
behaved the same, but allows our test inputs to remain basic ASCII.

The root problem is that `cast_expression`, part of no-op arithmetic
simplification, can now fail because expressions like `int(4000000000)`
no longer get past Constructor::Convert. Previously we had assumed
`cast_expression` could never return null; now we check its result for
null before using it.

Change-Id: I7335395bab0daf1f788b0c7c154904b2372ae13f
Bug: oss-fuzz:40660
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/467316
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>
2021-11-03 14:57:48 +00:00
John Stiles
293bb46b2d Enable more ES3-specific SkSL tests.
Updated ReturnsValueOnEveryPathES3 to remove overlap with the ES2 tests,
and fixed some broken cases. Disabled the ReturnValueOnEveryPathES3 test
on Intel + Windows because switch statements on Intel + Windows are
pretty broken.

Change-Id: Id93e8af1ef7bf11fd74ef12a464c77d56cc032a0
Bug: skia:11209, skia:12465
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/467078
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-03 14:02:05 +00:00
John Stiles
aa369d2b8e Fix error with inlined literals overflowing their types.
It's possible to write code containing errors that are only apparent
once the inliner runs. For instance, a function which takes a short and
returns its negative it is valid for most inputs, but undefined for
-32768 (because +32768 does not fit in a short). A function which takes
floats and casts them to ints is valid for many inputs, but not valid if
you pass in 5 billion.

This CL restructures our out-of-range integer error detection to report
errors cleanly in these cases instead of asserting. It also refactors
the range checking code to be usable in situations where we don't yet
have a Literal expression.

Change-Id: I98f0be63bf9afbbf1ab90233fa86d380cfae42b4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/466439
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-02 14:26:44 +00:00
John Stiles
c7c49f5656 Mark symbols starting with gl_ as reserved words.
Change-Id: I01d82447658c7acc5fe9eb230eb7020b49fa6c4f
Bug: skia:12498
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/466447
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-11-01 22:43:44 +00:00
John Stiles
81c86e8608 Fix fuzzer-discovered assertion with nonsense array sizes.
Change-Id: I7512491f55c10118f0ab058500f6ce9b5b8545cd
Bug: oss-fuzz:40557
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/466296
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>
2021-11-01 15:26:14 +00:00
John Stiles
9b9805959d Reland "Add support for half-precision types in Metal."
This reverts commit 9d24b02c2f.

Reason for revert: needs premul/unpremul conversion fix (http://review.skia.org/465798)

Original change's description:
> Revert "Add support for half-precision types in Metal."
>
> This reverts commit d90e09b1ae.
>
> Reason for revert: MacMini failing CompressedBackendAllocationTest
>
> Original change's description:
> > Add support for half-precision types in Metal.
> >
> > This will hopefully improve performance on lower-end GPUs.
> >
> > Change-Id: I9c2ee6dc31acd08bec0bfb5f59edc3cf90163f9e
> > Bug: skia:12339
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465078
> > Auto-Submit: John Stiles <johnstiles@google.com>
> > Reviewed-by: Brian Osman <brianosman@google.com>
> > Commit-Queue: John Stiles <johnstiles@google.com>
>
> Bug: skia:12339
> Change-Id: Ic5aa4bef454ca67f5ce26c600444d9565e0158cb
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465796
> Auto-Submit: Brian Osman <brianosman@google.com>
> Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
> Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>

Bug: skia:12339
Change-Id: I53a8a6fef299da15d206d884ba7029820ffcff43
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465799
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-10-30 14:33:52 +00:00
Brian Osman
9d24b02c2f Revert "Add support for half-precision types in Metal."
This reverts commit d90e09b1ae.

Reason for revert: MacMini failing CompressedBackendAllocationTest

Original change's description:
> Add support for half-precision types in Metal.
>
> This will hopefully improve performance on lower-end GPUs.
>
> Change-Id: I9c2ee6dc31acd08bec0bfb5f59edc3cf90163f9e
> Bug: skia:12339
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465078
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: John Stiles <johnstiles@google.com>

Bug: skia:12339
Change-Id: Ic5aa4bef454ca67f5ce26c600444d9565e0158cb
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465796
Auto-Submit: Brian Osman <brianosman@google.com>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
2021-10-29 23:04:09 +00:00
John Stiles
d90e09b1ae Add support for half-precision types in Metal.
This will hopefully improve performance on lower-end GPUs.

Change-Id: I9c2ee6dc31acd08bec0bfb5f59edc3cf90163f9e
Bug: skia:12339
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465078
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
2021-10-29 19:47:02 +00:00
John Stiles
ba9d5362e4 Detect and properly handle swizzles of out-of-range literal values.
Change-Id: Ic30c48dce0cb0072f07defcdb0b9e60b94f50818
Bug: oss-fuzz:40479
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465392
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>
2021-10-29 19:39:33 +00:00
John Stiles
378e4aecfe Report int-literal overflow via cast-from-float.
The fuzzer discovered that SkSL could create an out-of-range int literal
by casting from a floating point literal. We were only doing range
checks when the starting literal was an integer. Since we now assert
when an out-of-range int literal is created (as of
http://review.skia.org/464124), the fuzzer can detect this error.

Change-Id: Ie66f60ddbe7b4fbe5b648c17292c59a4ba079716
Bug: oss-fuzz:40456
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/465385
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>
2021-10-29 14:55:30 +00:00
John Stiles
4363cdb5ea Fix for fuzzer-discovered error with bitwise-not.
A recent CL (http://review.skia.org/464121) made it an error to coerce a
literal value to a type that cannot hold the value. The fuzzer found a
case where we assumed type-coercion of a literal would always succeed,
and failed to null-check the result. We now null-check the result.

Change-Id: Id97c6016e56c20ef724028f71bbf4688dde3c064
Bug: oss-fuzz:40428
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464919
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-10-28 16:44:09 +00:00
John Stiles
f2d016f12e Fix for fuzzer-discovered error with negation.
Yesterday's negation-related changes (http://review.skia.org/464123)
exposed a flaw that the fuzzer was able to exploit. We were previously
able to assume that `simplify_negation` would always return a non-null
expression; in some cases, that is no longer true.

Change-Id: Ia585232b0e35fafe0c642384a59ef94ce743ffd5
Bug: oss-fuzz:40427
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464916
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>
2021-10-28 16:14:42 +00:00
John Stiles
aa68a6b3a5 Reject scalar casts of literals to types that can't hold them.
With this change, we no longer have any SkSL tests which are able to
make a Literal integer that overflows its type. Literal::MakeInt now
asserts that its value is within bounds. I look forward to the fuzzer's
inevitable attempts to trigger these assertions.

Change-Id: I7b15e862caaf65984d33f5d72d2c1de816d1d292
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464124
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-10-28 14:05:14 +00:00
John Stiles
1fa2c28ee1 Fix dFdy test.
This was mistakenly using dFdx in some portions (copy-paste error).

Change-Id: Ifb159b3c44185d9166c10725b24002a28a0895b2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464381
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>
2021-10-27 21:25:32 +00:00
John Stiles
2bdb65dcf2 Refuse to negate a literal if it would over/underflow its type.
Previously, we would create a Literal with the negated value even if it
was outside the type's minimum/maximum values. Error reporting would
happen elsewhere, if at all (e.g. during assignment or coercion).

Change-Id: I020a93daf2b0f5741fb805a58a690489d7578dab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464123
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-10-27 17:54:34 +00:00
John Stiles
94b1f0dfa1 Add overflow protection to vector-arithmetic folding.
This was causing errors in UBSAN when compiling some of our existing
SkSL tests.

Change-Id: I66f22607094df77d47ff70948a139c77feae8624
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/464118
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-10-27 16:49:02 +00:00
Brian Osman
2d1207acb9 Remove the "in blend modes randomly fail for all zero vec" workaround
At this point, it seems like this was a mis-diagnosis of the underlying
issue around dual-source blending (and its interaction with other blend
state).

Change-Id: I11af0c9b70c32e14c353848db3d6adbfe5f08225
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/462176
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
2021-10-21 16:54:51 +00:00
John Stiles
9f43ceefa3 Allow vector operator~ in SkSL.
This is supported in GLSL ES3. (Strangely, vector operator! isn't.)
Previously, this was flagged as an error: http://review.skia.org/459885

Change-Id: I2c4299159fff58fefe8bd131c8d317cd82974a62
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459886
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>
2021-10-18 15:03:33 +00:00
John Stiles
319d75d6f6 Add error test for unary-negating and unary-plussing arrays.
At present, we only detect four errors here. We should detect six.

Change-Id: I226854ab930a273695c42cf2f7bdb1d5cd97e50b
Bug: oss-fuzz:39998
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459882
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>
2021-10-18 14:30:36 +00:00
John Stiles
9e952b379b Allow const int variables to be used for array sizes.
See http://review.skia.org/460037 for an example of the existing
behavior. Const variables are constant-expressions and should be allowed
here.

Change-Id: I41383d79668785f270b7825485e9f6fa56c553c1
Bug: skia:12549
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/460036
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>
2021-10-18 14:23:33 +00:00
John Stiles
409bb0195f Fix up the last of the ES2 conformance tests.
Four of the tests relied on approximate floating values to pass.
One of the tests is legitimately not passing in SkSL yet (we don't
reject keywords beginning with gl_, and it breaks existing code if we
do).

Change-Id: I2d5b3896787689dad16bd7b2805c0aa6a4c45603
Bug: skia:12498, skia:12484
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459879
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2021-10-15 19:32:48 +00:00
John Stiles
c8a96076b1 Added four more reserved words to SkSL.
We now detect attribute, varying, precision and invariant as reserved.

Change-Id: I8c90655a70b1bad31bf6143c3fdcb2ce582320b1
Bug: skia:12484
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459479
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2021-10-14 16:29:32 +00:00
John Stiles
f7d2673643 Fix up samplerCube/textureCube for ES2 conformance.
`samplerCube` is a type which we don't support at all. It has been added
to the reserved-word list.

`textureCube` was in our list of built-in types, but was not actually
used in any way; it wasn't actually added to the root or private symbol
tables, and was totally unreferenced by the code. It's been deleted.

Change-Id: I4f79ce5d40ac6ebdb2a7067fa60cc79e316b01b6
Bug: skia:12484
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459123
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2021-10-14 15:21:31 +00:00
John Stiles
ee36412765 Add ES2 conformance test harness to dm.
This CL adds a few more exceptions to our ES2 test import, and adds the
dm code which actually runs the tests.

Change-Id: If6691dd35931f4f10262d3a1eff020c2c347ca59
Bug: skia:12484
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459124
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2021-10-14 14:50:06 +00:00
John Stiles
21fe518fbb Revert "Disallow matrix ctors which overflow a column."
This reverts commit eb68973c2f.

Reason for revert: ES2 conformance test checks this

Original change's description:
> Disallow matrix ctors which overflow a column.
>
> The GLSL spec allows matrix constructors containing vectors that would
> split between multiple columns of the matrix. However, in practice, this
> does not actually work well on a lot of GPUs!
>
> - "cast not allowed", "internal error":
> 	Tegra 3
> 	Quadro P400
> 	GTX 660
> 	GTX 960
> - Compiles, but generates wrong result:
> 	RadeonR9M470X
> 	RadeonHD7770
>
> Since this isn't a pattern we expect to see in user code, we now report
> it as an error at compile time. mat2(vec4) is treated as an exceptional
> case and still allowed.
>
> Change-Id: Id6925984a2d1ec948aec4defcc790a197a96cf86
> Bug: skia:12443
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/449518
> Commit-Queue: John Stiles <johnstiles@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>

Bug: skia:12443
Change-Id: I5a32744c88b9b830ad657488824c8c7dd0b0a652
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458056
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
2021-10-14 01:30:08 +00:00
John Stiles
64c907c052 Mark private types as invalid in the public symbol table.
Previously, in public code, private types didn't exist anywhere in the
symbol table chain, and those names were free for the taking. Now, we
register them as invalid types in the public symbol table. This prevents
them from being used as variable names, and gives a more explicit error
if you try to use them as a type.

Change-Id: I9a943bf923639b72cbf36b1acf4b4fbe70982786
Bug: skia:12538
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/459119
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>
2021-10-13 20:41:09 +00:00
John Stiles
abb6814cc7 Add support for 0XABCD style hex literals.
To my great surprise, a capital X is allowed in hex literals. In fact,
this is allowed in both GLSL and C. The ES2 conformance suite tests
this, so now SkSL supports it as well.

Change-Id: If795c6033b301420669f002530ee1d14fec29f96
Bug: skia:12533
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458723
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-10-13 12:41:54 +00:00
John Stiles
5420cbcf65 Match GLSL scoping rules more closely in SkSL.
GLSL treats builtin types and user-defined types differently; `int` and
`float` are keywords and cannot be used to name variables. However, it's
fine for a user type like `struct xyz` to be hidden by a variable
`int xyz` or even `xyz xyz` (i.e., a variable of type `struct xyz` named
`xyz`).

We now honor that distinction and include tests for it. This will fix
several ES2 conformance tests (local_struct_variable_hides_struct_type,
local_int_variable_hides_struct_type, etc.).

Change-Id: I7a45c70707087f9f355ce5b06b032fed16683f3e
Bug: skia:12527
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458721
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-10-12 21:53:28 +00:00
John Stiles
980169a5b5 Allow non-trivial constant expressions for array sizes.
This fixes GLSL ES2 conformance test `array`.

Change-Id: I6ebee9253e1e8c394d9ddb6899e3a0940b7a38ef
Bug: skia:12495
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458718
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
2021-10-12 21:04:33 +00:00
John Stiles
fedb3957a9 Add exception list for GLSL tests which fail under SkSL.
SkSL intentionally differs from GLSL in some edge case behaviors.
For instance, we intentionally disallow functions that can exit without
returning a value, and reject constructors that shrink the size of a
vector (swizzles can do this in a more intentional way). In these cases,
we update the test's expected outcome from "pass" to "fail."

Change-Id: I671d6eb7d9ae06caa2895c3310356a399b36b2bf
Bug: skia:12484
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458596
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
2021-10-12 21:01:26 +00:00
John Stiles
116d2e0e48 Fix ES2 conformance test 'in_vs_no_in'.
SkSL treated these two functions as distinct, even though they are not:
    void func(in float x);
    void func(float x);

The `in` modifier on a function parameter is the default state, making
these two prototypes functionally identical. We now strip off an `in`
modifier on a function definition. This gives us three potential states
for each param: nothing (meaning `in`), `out`, and `inout`.

Change-Id: Id2acb53ecaca98f86a7f6a83e0b9a375f9abe2b8
Bug: skia:12525
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/458257
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>
2021-10-11 22:27:37 +00:00