Use the local (not global) invocation id to decide which invocation should do the initialization, so that every workgroup gets initialized, not just the first.
Fixes#1745: Support out-of-order module scope declarations in WGSL
Fixes#1044: Forbid local variable shadowing in WGSL
Fixes#2076: [wgsl-in] no error for duplicated type definition
Fixes#2071: Global item does not support 'const'
Fixes#2105: [wgsl-in] Type aliases for a vecN<T> doesn't work when constructing vec from a single argument
Fixes#1775: Referencing a function without a return type yields an unknown identifier error.
Fixes#2089: Error span reported on the declaration of a variable instead of its use
Fixes#1996: [wgsl-in] Confusing error: "expected unsigned/signed integer literal, found '1'"
Separate parsing from lowering by generating an AST, which desugars as
much as possible down to something like Naga IR. The AST is then used
to resolve identifiers while lowering to Naga IR.
Co-authored-by: Teodor Tanasoaia <28601907+teoxoy@users.noreply.github.com>
Co-authored-by: Jim Blandy <jimb@red-bean.com>
* hlsl-out: don't output interpolation modifier if it's the default (linear/`Interpolation::Perspective`)
* Remove linear interpolation modifiers from HLSL output tests
Replace uses of `call_unique` with uses of `call` and `call_or`, which becomes
public. It's not clear when `call_unique` is correct to use, and avoiding a few
numeric suffixes here and there isn't worth it.
Replace `Module::apply_common_default_interpolation` with a simpler function
that handles a single `Binding` at a time. In exchange for the simplicity, the
function must be called at each point function arguments, function results, and
struct members are prepared. (Any missed spots will be caught by the verifier.)
This approach no longer requires mutating types in the arena, a prerequisite for
properly handling type identity.
Applying defaults to struct members when the struct declaration is parsed does
have a disadvantage, compared to the old whole-module pass: at struct parse
time, we don't yet know which pipeline stages the struct will be used in. The
best we can do is apply defaults to anything with a `Location` binding. This
causes needless qualifiers to appear in some output. However, it seems that our
back end languages all tolerate such qualifiers.
* [hlsl-out] flatten the entry point inputs
Previously, the logic was re-ordering the inputs according to the binding.
This breaks if one of the inputs is a struct. With this change, the struct
fields are also flattened into the fake entry point struct. We also
construct the original arguments at the beginning of the function.
* hlsl-out: completely separate the flattened IO structs from the original IR structs
Previously, we had heuristics to detect if a particular struct needs the fields
to be re-ordered. We'd re-order interface structs without layout, and the detection
was very fragile and easily wrong.
The new logic is spawning separate struct types if we need any re-ordering to happen.
It's solid, there are no heuristics.