Generate an OpSpecConstantComposite instruction decorated with the
WorkgroupSize builtin. Only support a single stage with an overridable
workgroup size.
Bug: tint:713
Change-Id: I139123c0af8326fcbd796cb2fc9d223882206e19
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51263
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Change the type of the values in an ast::WorkgroupDecoration to be
ast::Expression nodes, so that they can represent both
ast::ScalarExpression (literal) and ast::IdentifierExpression
(module-scope constant).
The Resolver processes these nodes to produce a uint32_t for the
default value on each dimension, and captures a reference to the
module-scope constant if it is overridable (which will soon be used by
the inspector and backends).
The WGSL parser now uses `primary_expression` to parse arguments to
workgroup_size.
Also added some WorkgroupSize() helpers to ProgramBuilder.
Bug: tint:713
Change-Id: I44b7b0021b925c84f25f65e26dc7da6b19ede508
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51262
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This reverts commit 594075a2f0.
Reason for revert: This is not a complete solution. It does not cover assignment of variables of the same struct type, but of different access types. For example:
```
[[block]] struct S{ i : 32; };
var<storage> gr : [[access(read)]] S;
var<storage> gw : [[access(write)]] S;
fn f() {
var a : S;
a = gr;
gw = a;
}
```
With my CL, we currently generate invalid assignments because the new types of 'S' for 'gr' and 'gw' are now all different types. We would need to generate functions to assign between them.
In the end, we will drop this strategy, and instead, Dawn will be modified to not rely on names.
Original change's description:
> Add DuplicateStorageStruct transform
>
> This transform is currently required by Dawn for it's GLSL backend so
> that SPIRV-Cross does not end up renaming the structures internally,
> which it does to fix aliasing problems. We can remove this transform
> if/once Dawn using binding numbers rather than names for uniform/storage
> buffer data.
>
> Bug: tint:386
> Bug: tint:808
>
> CloneContext: allow insert before/after of a node marked for removal
> Change-Id: I3ff3b37bca62db07d5c759250dd4777e279b7a4f
> Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51403
> Commit-Queue: Ben Clayton <bclayton@google.com>
> Kokoro: Kokoro <noreply+kokoro@google.com>
> Reviewed-by: Ben Clayton <bclayton@google.com>
TBR=bclayton@google.com,amaiorano@google.com,noreply+kokoro@google.com
Change-Id: I695347b0ffe8be23b9dc44885af378be56512406
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: tint:386
Bug: tint:808
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51500
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
This transform is currently required by Dawn for it's GLSL backend so
that SPIRV-Cross does not end up renaming the structures internally,
which it does to fix aliasing problems. We can remove this transform
if/once Dawn using binding numbers rather than names for uniform/storage
buffer data.
Bug: tint:386
Bug: tint:808
CloneContext: allow insert before/after of a node marked for removal
Change-Id: I3ff3b37bca62db07d5c759250dd4777e279b7a4f
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51403
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This change implements pointers and references as described by the WGSL
specification change in https://github.com/gpuweb/gpuweb/pull/1569.
reader/spirv:
* Now emits address-of `&expr` and indirection `*expr` operators as
needed.
* As an identifier may now resolve to a pointer or reference type
depending on whether the declaration is a `var`, `let` or
parameter, `Function::identifier_values_` has been changed from
an ID set to an ID -> Type* map.
resolver:
* Now correctly resolves all expressions to either a value type,
reference type or pointer type.
* Validates pointer / reference rules on assignment, `var` and `let`
construction, and usage.
* Handles the address-of and indirection operators.
* No longer does any implicit loads of pointer types.
* Storage class validation is still TODO (crbug.com/tint/809)
writer/spirv:
* Correctly handles variables and expressions of pointer and
reference types, emitting OpLoads where necessary.
test:
* Lots of new test cases
Fixed: tint:727
Change-Id: I77d3281590e35e5a3122f5b74cdeb71a6fe51f74
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50740
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Rename:
* type() to StoreType()
* storage_class() to StorageClass()
Move away from snake_case methods in the semantic namespace.
Try to avoid generic 'type()' method names.
Also add an assertion to detect doubly nested references (these are
invalid).
Bug: tint:727
Change-Id: I975a3f1e5fbed7947cc2fc156fee892b282c63de
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51220
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Once the resolver correctly handles address-of and indirections,
the logic around these will be different enough that a parameterized
test no longer makes sense.
Bug: tint:727
Change-Id: I79c138b12bd6f49d9dbee02c62e594a9b1b4a2db
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51185
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: David Neto <dneto@google.com>
Semantic information about block statements the resolver would
temporarily create while resolving is now exposed in a
sem::BlockStatement class.
In the process, semantic information about statements in general is
overhauled so that a statement has a reference to its parent
statement, regardless of whether this is a block.
Bug: tint:799
Bug: tint:800
Change-Id: I8771511c5274ea74741b8c86f0f55cbc39810888
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50904
Commit-Queue: Alastair Donaldson <allydonaldson@googlemail.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
In preparation for implementing
https://github.com/gpuweb/gpuweb/issues/1604, this change removes the
sem::AccessControl node. Instead, the ast::AccessControl::Access enum is
now on the sem::StorageTexture class, as well as on sem::Variable. For
sem::Variable, the field is set when the variable's type is either a
storage buffer or a storage texture.
Bug: tint:802
Change-Id: Id479af36b401d067b015027923f4e715f5f69f25
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/51020
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
The backends now use sem::Variable::ConstantId() instead, since the
AST does not have the correct ID when the parameterless version of the
override attribute is used.
Change-Id: I816295c8b2e4baa1671c9d042920bf001b1b4336
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50845
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Nothing currently generates these.
Resolver currently doesn't know how to handle these.
Backends currently stubbed enough to build without warnings.
Bug: tint:727
Change-Id: I4b9863ae098b903b51a63c36c10bc6e390efbbb3
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50746
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: James Price <jrprice@google.com>
There's now one remainin use of the semantic info in the WGSL writer - the transformation of [[offset]] into padded fields.
This does not belong in the WGSL writer, but instead part of the transform::Wgsl sanitizer.
Bug: tint:798
Change-Id: I95ba11f022c41150cc12de84a4085cd7d42019fe
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50822
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: James Price <jrprice@google.com>
To declare a local variable, we write `var name : type`, not `var<function> name : type`.
This change fixes all the places where we were feeding StorageClass::kFunction into variable declarations.
Note that the resolved, semantic variable correctly infers the `kFunction` StorageClass.
Change-Id: I6221fabae1de0435044f29b9a91808421d5cace6
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50821
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Commit-Queue: David Neto <dneto@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Add a new parameter to BindingRemapper::Remappings that allows resulting binding points to collide.
When enabled, the output of the transform contains two or more module-scoped variables with the same binding point, used by the same entry point, then these variables will be decorated with an internal decoration to disable validation for the collision.
This is to work around collisions generated for the HLSL backend where the variables actually exist in different register classes, which is permitted by D3D12.
The transform will only generate these decorations if it needs to.
Fixed: tint:797
Change-Id: Id8a87523801bd0cd0dd54227ebabd4299bc20c27
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50742
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
An [[internal]] decoration that specifically disables certain validation checks.
Begin with a single kFunctionHasNoBody mode.
Migrate the Resolver to using this instead of allowing any InternalDecoration to disable the checks for no-body.
Bug: tint:797
Change-Id: I213b9a6844a456775ede06d60e456d9f77a449d0
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50741
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Variable() is called for globals, locals and parameters. Much of the logic is the same.
Move all the common logic down into Variable(). This:
* Removes some yucky default parameters
* Adds type validation that was missing for globals (broken tests fixed)
* Gives me a single place to implement the Reference type wrapping
Bug: tint:727
Change-Id: I70f4a3603d7fa781da938508aa2a1bc80ec15d77
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/50580
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
