Enum naga::TypeInner

pub enum TypeInner {
    Scalar {
        kind: ScalarKind,
        width: Bytes,
    },
    Vector {
        size: VectorSize,
        kind: ScalarKind,
        width: Bytes,
    },
    Matrix {
        columns: VectorSize,
        rows: VectorSize,
        width: Bytes,
    },
    Atomic {
        kind: ScalarKind,
        width: Bytes,
    },
    Pointer {
        base: Handle<Type>,
        space: AddressSpace,
    },
    ValuePointer {
        size: Option<VectorSize>,
        kind: ScalarKind,
        width: Bytes,
        space: AddressSpace,
    },
    Array {
        base: Handle<Type>,
        size: ArraySize,
        stride: u32,
    },
    Struct {
        members: Vec<StructMember>,
        span: u32,
    },
    Image {
        dim: ImageDimension,
        arrayed: bool,
        class: ImageClass,
    },
    Sampler {
        comparison: bool,
    },
    AccelerationStructure,
    RayQuery,
    BindingArray {
        base: Handle<Type>,
        size: ArraySize,
    },
}

Enum with additional information, depending on the kind of type.

Variants

Scalar

Fields

kind: ScalarKind

width: Bytes

Number of integral or floating-point kind.

Vector

Fields

size: VectorSize

kind: ScalarKind

width: Bytes

Vector of numbers.

Matrix

Fields

columns: VectorSize

rows: VectorSize

width: Bytes

Matrix of floats.

Atomic

Fields

kind: ScalarKind

width: Bytes

Atomic scalar.

Pointer

Fields

base: Handle<Type>

space: AddressSpace

Pointer to another type.

Pointers to scalars and vectors should be treated as equivalent to ValuePointer types. Use the TypeInner::equivalent method to compare types in a way that treats pointers correctly.

Pointers to non-SIZED types

The base type of a pointer may be a non-SIZED type like a dynamically-sized Array, or a Struct whose last member is a dynamically sized array. Such pointers occur as the types of GlobalVariable or AccessIndex expressions referring to dynamically-sized arrays.

However, among pointers to non-SIZED types, only pointers to Structs are DATA. Pointers to dynamically sized Arrays cannot be passed as arguments, stored in variables, or held in arrays or structures. Their only use is as the types of AccessIndex expressions.

ValuePointer

Fields

size: Option<VectorSize>

kind: ScalarKind

width: Bytes

space: AddressSpace

Pointer to a scalar or vector.

A ValuePointer type is equivalent to a Pointer whose base is a Scalar or Vector type. This is for use in TypeResolution::Value variants; see the documentation for TypeResolution for details.

Use the TypeInner::equivalent method to compare types that could be pointers, to ensure that Pointer and ValuePointer types are recognized as equivalent.

Array

Fields

base: Handle<Type>

size: ArraySize

stride: u32

Homogenous list of elements.

The base type must be a SIZED, DATA type.

Dynamically sized arrays

An Array is SIZED unless its size is Dynamic. Dynamically-sized arrays may only appear in a few situations:

• They may appear as the type of a GlobalVariable, or as the last member of a Struct.

• They may appear as the base type of a Pointer. An AccessIndex expression referring to a struct’s final unsized array member would have such a pointer type. However, such pointer types may only appear as the types of such intermediate expressions. They are not DATA, and cannot be stored in variables, held in arrays or structs, or passed as parameters.

Struct

Fields

members: Vec<StructMember>

span: u32

User-defined structure.

There must always be at least one member.

A Struct type is DATA, and the types of its members must be DATA as well.

Member types must be SIZED, except for the final member of a struct, which may be a dynamically sized Array. The Struct type itself is SIZED when all its members are SIZED.

Image

Fields

dim: ImageDimension

arrayed: bool

class: ImageClass

Possibly multidimensional array of texels.

Sampler

Fields

comparison: bool

Can be used to sample values from images.

AccelerationStructure

Opaque object representing an acceleration structure of geometry.

RayQuery

Locally used handle for ray queries.

BindingArray

Fields

base: Handle<Type>

size: ArraySize

Array of bindings.

A BindingArray represents an array where each element draws its value from a separate bound resource. The array’s element type base may be Image, Sampler, or any type that would be permitted for a global in the Uniform or Storage address spaces. Only global variables may be binding arrays; on the host side, their values are provided by TextureViewArray, SamplerArray, or BufferArray bindings.

Since each element comes from a distinct resource, a binding array of images could have images of varying sizes (but not varying dimensions; they must all have the same Image type). Or, a binding array of buffers could have elements that are dynamically sized arrays, each with a different length.

Binding arrays are in the same address spaces as their underlying type. As such, referring to an array of images produces an Image value directly (as opposed to a pointer). The only operation permitted on BindingArray values is indexing, which works transparently: indexing a binding array of samplers yields a Sampler, indexing a pointer to the binding array of storage buffers produces a pointer to the storage struct.

Unlike textures and samplers, binding arrays are not ARGUMENT, so they cannot be passed as arguments to functions.

Naga’s WGSL front end supports binding arrays with the type syntax binding_array<T, N>.

Implementations

impl TypeInner

pub fn indexable_length( &self, module: &Module ) -> Result<IndexableLength, IndexableLengthError>

Return the length of a subscriptable type.

The self parameter should be a handle to a vector, matrix, or array type, a pointer to one of those, or a value pointer. Arrays may be fixed-size, dynamically sized, or sized by a specializable constant. This function does not handle struct member references, as with AccessIndex.

The value returned is appropriate for bounds checks on subscripting.

Return an error if self does not describe a subscriptable type at all.

impl TypeInner

pub const fn scalar_kind(&self) -> Option<ScalarKind>

pub const fn scalar_width(&self) -> Option<u8>

pub const fn pointer_space(&self) -> Option<AddressSpace>

pub fn is_atomic_pointer(&self, types: &UniqueArena<Type>) -> bool

pub fn size(&self, _gctx: GlobalCtx<'_>) -> u32

Get the size of this type.

pub fn canonical_form(&self, types: &UniqueArena<Type>) -> Option<TypeInner>

Return the canonical form of self, or None if it’s already in canonical form.

Certain types have multiple representations in TypeInner. This function converts all forms of equivalent types to a single representative of their class, so that simply applying Eq to the result indicates whether the types are equivalent, as far as Naga IR is concerned.

pub fn equivalent(&self, rhs: &TypeInner, types: &UniqueArena<Type>) -> bool

Compare self and rhs as types.

This is mostly the same as <TypeInner as Eq>::eq, but it treats ValuePointer and Pointer types as equivalent.

When you know that one side of the comparison is never a pointer, it’s fine to not bother with canonicalization, and just compare TypeInner values with ==.

pub fn is_dynamically_sized(&self, types: &UniqueArena<Type>) -> bool

pub fn components(&self) -> Option<u32>

pub fn component_type(&self, index: usize) -> Option<TypeResolution>

Trait Implementations

impl Clone for TypeInner

fn clone(&self) -> TypeInner

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TypeInner

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Hash for TypeInner

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for TypeInner

fn eq(&self, other: &TypeInner) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

This method tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Eq for TypeInner

impl StructuralEq for TypeInner

impl StructuralPartialEq for TypeInner