Transforms are meant to be reusable. Those that hold state cannot be used concurrently.
State leakage between runs is dangerous.
To fix this:
* Add transform::Data - A new base class for extra information emitted by transforms.
* Add transform::DataMap - A container of Data, keyed by type.
* Add a transform::DataMap field to Transform::Output.
* Have FirstIndexOffset emit a FirstIndexOffset::Data.
* Deprecate the getters on the transform.
Mutability of the transform config is also dangerous as setters can be called while a transform is actively running on another thread.
To fix:
* Expose a VertexPulling::Config structure and add a constructor that accepts this.
* Deprecate the setters on VertexPulling.
Also deprecate Transform::Output::diagnostics.
Put all the transform diagnostics into the returned Program. Reduces error handling of the client.
Change-Id: Ibd228dc2fbf004ede4720e2d6019c024bc5934d1
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42264
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
The transform::Manager no longer does anything special. It is now a container of transforms that acts like a single transform.
Change-Id: I5e51e250cfa610b651445b7cd5efd29811c56d2e
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42261
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
postfix_expr() handles `[]` array accessors and call expressions `()`.
Have postfix_expr() use sync to parse these:
* It will use the end bracket token to attempt to resynchronize the parser on error
* It also considers maximum parser recursion depth, avoiding stack overflows
Fixed: chromium:1180573
Change-Id: I8c1c62c68e24a564e0e4e7d0de9f5a3fa7032369
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42222
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Have the TD resolve swizzles down to indices, erroring out if they're not valid.
Resolving these at TD time removes swizzle parsing in the HLSL writer, and is generally useful information.
If we don't sanitize in the TD, we can end up trying to construct a resulting vector of an invalid size (> 4) triggering an assert in the type::Vector constructor.
Fixed: chromium:1180634
Bug: tint:79
Change-Id: If1282c933d65eb02d26a8dc7e190f27801ef9dc5
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42221
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Change-Id: Ia6c21dfc33445ba828b2f244d8a3479fb3328805
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42263
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Add test.
Change-Id: Ide47df8b3a946818ed9a8ffd62411284f568609b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42262
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
IdentifierExpressions that resolve to functions and intrinsics were not being assigned a semantic::Expression as there is no function / intrinsic type to give them. This lead to NPEs in the writers when these identifiers were reached and TypeOf() is called.
Adding a new tint::type::Callable is an option, but until functions become a type in the language, this seems like a very large and debatable change.
Attempting to detect IdentifierExpressions with no semantic node in the validator is another option, but this is clunky as it has to detect incomplete semantic info from the TD, and cannot identify whether this actually resolved to a function or an intrinsic.
Instead we now error in the TD if encounter an IdentifierExpression that resolves to a function or intrinsic which is not called (ident is not followed with parenthesis).
Fixed: chromium:1180544
Fixed: chromium:1180814
Change-Id: I121dd194356419f94b09c7ee1ed544a350a114b3
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42220
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
WGSL recently removed this concept, since it didn't exist in WebGPU,
so excising it from the code.
BUG=tint:515
Change-Id: Ibbca6bd643fd96c2fb10bd33f471c9e9e58de535
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42320
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
If the parser hits a `maximum parser recursive depth reached` situation, then we need to try and resynchronize the parser.
If we fail to do this, then the synchronized_ flag may remain true, and the parser will believe progress is still being made.
In this situation the parser may try to reparse the same token, forever.
By calling sync_to() we either find the end of the block, and forward progress can be made, or synchronized_ is set to false, and the parser can error out cleanly.
Add test case from fuzzer report.
Fixed: chromium:1180128
Change-Id: I893077677fd3dfbd4b9b400cd32db842b06db500
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42029
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
An unbuilt program doesn't have an ast::Module, so Program::AST() will just explode.
* Have a Program default to false for IsValid()
* Initialize the ast_ field to nullptr. This was previously floating.
* Return from CommonFuzzer::Run() earlier if the InputFormat is not recognised.
Fixed: chromium:1180130
Fixed: chromium:1180157
Change-Id: I9b67daa10746f386f44919a7b9ac5c171092d6e5
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42028
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: David Neto <dneto@google.com>
HLSL validation will be disabled by default, and is now opt-in with
--validate-hlsl and/or with a --dxc-path
Change-Id: Ia98d7b1fbba50168bbe85c7982b73598100016d7
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42024
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Change-Id: I4e3b22e64554427f341a1bb260d72aa202c6b53c
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42027
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Add Source::FileContent to hold the file source content and per-line data.
Have Source hold an optional pointer to a FileContent, and add a file_path field.
This allows us to kill the `FreeInternalCompilerErrors()` filth as we're now able to construct Sources that hold a file path without file content.
Change-Id: I03556795d7d4161c3d34cef32cb685c45ad04a3d
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42026
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
Some comments from https://dawn-review.googlesource.com/c/tint/+/42060
were not addressed before it landed, this is a follow-up to address
them.
BUG=tint:489
Change-Id: Ieafe143e867ce0f5edf6cf2258db3ed12ffc13fb
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42080
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Chromium has its own test main() entrypoint.
To ensure that Chromium doesn't panic about memory leaks with the tests that exercise the ICE cases, we have to explicitly call the FreeInternalCompilerErrors() functions in these tests (at least until I can add this to end of Chromium's test main() function)
Change-Id: I2ea5109fcdb5f68f56a19709a1ec35ed72c0f760
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42025
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
BUG=tint:489
Change-Id: I03a15fdd156bf0c85b1afebb6e0e26fd44dff51f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42060
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Use a type alias for the most common and lengthy return type.
Drop the `const` from the return type - it is (or at least should be) meaningless.
Fix `auto& referenced_variables =` local variable declaration in inspector.cc.
You cannot take a reference to a non-ref return value of a function.
I have no idea why the compiler wasn't erroring about this.
Change-Id: Icc0b4a9bf4b4c9d03b44670eda27214ef04ad41c
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42021
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Lessens the friction of using these macros.
Also allows you to add a message to TINT_UNREACHABLE(), which you couldn't do before.
Change-Id: Ida4d63ec96e1d99af71503e8b80d7a5a712e6a47
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/42020
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Looks for DXC on PATH.
If found, it will invoke DXC with the shader program generated by the writer to verify the shader can compile.
Change-Id: Iad8b4021bac16d01214b500ddb89b5f743927ce9
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41943
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Command is a helper used by tests for executing a process with a number of arguments and an optional stdin string, and then collecting and returning the process's stdout and stderr output as strings.
Will be used to invoke HLSL and MSL shader compilers to verify our test generated code actually compiles.
Change-Id: I5cd4ca63af9aaa29be7448bb4fa8422e6d42a8ce
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41942
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
TmpFile constructs a temporary file that can be written to, and is automatically deleted on destruction.
Will be used to create a temporary source file for verifying generated HLSL and MSL against their shader compilers.
Change-Id: Ieaa6f257b93f4f2193dafe6297603816f6964928
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41941
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Remove the Parser classes from the wgsl and spirv namespaces.
These have been replaced with a Parse() method.
Remove the TypeDeterminer::Run() method, this was not called by tint and
the TypeDeterminer is now non-public API.
Change-Id: I5ddb82768da04398ab3958d1647be44f9fe30c21
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41840
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Appends an error message with the tint compiler source location to the
provided diagnositic list, and then calls the global error handler if
one is set.
Tests and the sample app now register an error handler to print the
diagnostic list to stderr and abort when NDEBUG is not defined.
All uses of assert(false) have been fixed up to use these macros.
Change-Id: I2f63e51ed86ac23883301d280070bd1a357c6cb2
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41620
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
When copying or moving a ProgramBuilder, we always want the ProgramBuilder::ty to point back to the owner.
We were previously std::move()'ing the ty field, which is not correct - this will result in the TypesBuilder pointing to the wrong ProgramBuilder.
I'm not sure why we've not seen any issues with this using clang, but running under MSVC immediately highlighted this brokenness.
Change-Id: I4293bb00ac4fbdfa66d12b1504a7bd060e014cd6
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41861
Commit-Queue: Ben Clayton <bclayton@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Most recursive control flow passes through Sync().
Error out if the Sync() function is recursively called too many times.
This replaces the more specific kMaxConstExprDepth, which also passes
through Sync().
Fixed: chromium:1178436
Change-Id: I64a05f9f6a4fe6d2b53a3ca75642b30e98c7a35f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41724
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
The CloneContext was previously dealing with pointers to CastableBase, which has no guarantees that the object was actually cloneable.
Add a Cloneable base class that CloneContext can use instead.
Improves readability and produces cleaner compiler errors if you try to clone a non-cloneable object.
Change-Id: I4352fc5dab3da434e4ab160a54c4c82d50e427b4
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41722
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
A large chunk of validation is now handled by the IntrinsicTable. Remove this.
Also fail validation and propagate diagnostics from the Program to the validator if attempting to validate a broken program. This should prevent undefined behaviour if the user forgets to check the program.IsValid() after parsing.
Change-Id: I2972e8ce296d6d6fca318cee48bc6929e5ed52db
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41549
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: David Neto <dneto@google.com>
Semantic info is no longer part of the ast, so it is now odd to mention semantic info on a clone method for the AST.
Improve the documentation around cloning on the Program methods and the CloneContext.
Change-Id: Ib1cf255acfd994521aaa5add2789e5117db6b072
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41548
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: David Neto <dneto@google.com>
Storage classes are unused for constants.
Also trim extra arguments to these variable constructor functions that are already defaulted to the same value.
Change-Id: Ia0b44364de000bfb8799de15827ce08fcce5f5be
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41541
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: David Neto <dneto@google.com>
Move the storage parameter after type.
Const() has no use for storage classes, so this parameter will be removed in the next change.
This reordering keeps Var() and Const() parameter types identical for the first two non-optional fields
Change-Id: I66669d19fa2175c4f10f615941e69efcab4c23e1
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41540
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Move these into a separate const variable declaration statement just above the before the use of the array initializer.
HLSL does not allow array initializers as part of a sub-expression
Fixed: tint:406
Change-Id: I98f93f2eca172865691831011c852de5e57c5ad6
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41484
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: James Price <jrprice@google.com>
Inserts objects before others when cloning
Change-Id: Ibf247abae3aeb3d351048f1182db2a2b42b2c677
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41547
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
And assert that the cast succeeded.
There is a danger with Replace() or ReplaceAll(), where you can end up replacing a node with another node of an incompatible type for some reference of that object. Previously this would silently cast to the incorrect type, and Bad Things would happen. Now we will assert in this situation.
I have not observed this issue happening (all current uses of Replace() and ReplaceAll() are believed to be safe). This is just an edge case I've spotted and wanted to add some safety belts for.
Change-Id: Icf4a4fe76f7bc14bcc6b274de68f7d0b3d85d71f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41546
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Reviewed-by: David Neto <dneto@google.com>
Use a sanitizing transform to convert scalar `sample_mask_{in,out}`
variables to single element arrays.
Add the `SampleRateShading` capability if the `sample_index` builtin
is used.
Bug: tint:372
Change-Id: Id7280e3ddb21e0a098d83587d123c97e3c34fa1b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41662
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: James Price <jrprice@google.com>
Returns a list of ast::IdentifierExpression* nodes that reference the
variable.
Change-Id: I36f475c6ddf5482f9ae9b432190405625f379f0d
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41661
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
inspector_test.cc: Remove unused method
test_helper.h: Fix comment type on member
Change-Id: I53b1392fa1fa8154efdcd471d1df1bbf27e04dc9
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41728
Commit-Queue: Ben Clayton <bclayton@google.com>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Add Stmt() accessor on all semantic::Expressions so the owning statement can be retrieved.
Change-Id: I5d584335a6d137fdeab0b8d74a161fcae9b46080
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41545
Reviewed-by: James Price <jrprice@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Auto-Submit: Ben Clayton <bclayton@google.com>
Other builtins use WGSL terms instead of SPIR-V terms too, and the
WGSL writer is relying on the output of `operator<<(Builtin)`, which
just stringifies the name of the enum. This also matches the
equivalent `semantic::Usage::kSampleIndex` enum.
Added test coverage for WGSL builtin generation.
Bug: tint:372
Change-Id: I8077d22c4a5ddf67b1ad07e7365453db74db8e7d
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41660
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Auto-Submit: James Price <jrprice@google.com>
BUG=tint:489
Change-Id: I28e4b0e568aea463971e9d2f5a04f04e942e2564
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41600
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
This is the first step in being able to read code generated
by Clspv.
Actively ignore the instructions instead of applying stripping
transform before hand. That way we have a chance at properly counting
instructions, which helps produce better diagnostics.
Bug: tint:3
Change-Id: I82bde88897485380d70dc8b287c3843eae5489b6
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41641
Auto-Submit: David Neto <dneto@google.com>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
These files were added in dfd1714, but only added to CMakeLists.txt.
Change-Id: I775e8fef68b2f4900a6c562533512f0eda01b795
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41700
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
In C++ argument evaluation order is undefined. MSVC and Clang evaluate these in different orders, leading to hilarity when writing tests that expect a deterministic ordering.
Pull out all the argument expressions to create() in the clone functions so a cloned program is deterministic in its ordering between compilers.
Change-Id: I8e2de31398960c480ce7ee1dfaac4f67652d2dbc
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41544
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Auto-Submit: Ben Clayton <bclayton@google.com>
Change 41302 correctly fixed up Module::Clone(), but this wasn't actually called by the CloneContext, as Module::Clone() returns a new Module, where as the CloneContext needs to clone into an existing Module.
Refactor the code so that this duplicated logic is moved into a single Module::Copy() method.
Fixed: 1177275
Change-Id: Ia8c45ef05e03b2891b5785ee6f425dd01cb989c6
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/41542
Reviewed-by: James Price <jrprice@google.com>
Reviewed-by: dan sinclair <dsinclair@chromium.org>
Commit-Queue: dan sinclair <dsinclair@chromium.org>
Ben Clayton
Ryan Harrison
James Price
Vasyl Teliman
David Neto
dan sinclair