naga/back/hlsl/

mod.rs

/*!
Backend for [HLSL][hlsl] (High-Level Shading Language).

# Supported shader model versions:
- 5.0
- 5.1
- 6.0

# Layout of values in `uniform` buffers

WGSL's ["Internal Layout of Values"][ilov] rules specify how each WGSL
type should be stored in `uniform` and `storage` buffers. The HLSL we
generate must access values in that form, even when it is not what
HLSL would use normally.

Matching the WGSL memory layout is a concern only for `uniform`
variables. WGSL `storage` buffers are translated as HLSL
`ByteAddressBuffers`, for which we generate `Load` and `Store` method
calls with explicit byte offsets. WGSL pipeline inputs must be scalars
or vectors; they cannot be matrices, which is where the interesting
problems arise. However, when an affected type appears in a struct
definition, the transformations described here are applied without
consideration of where the struct is used.

Access to storage buffers is implemented in `storage.rs`. Access to
uniform buffers is implemented where applicable in `writer.rs`.

## Row- and column-major ordering for matrices

WGSL specifies that matrices in uniform buffers are stored in
column-major order. This matches HLSL's default, so one might expect
things to be straightforward. Unfortunately, WGSL and HLSL disagree on
what indexing a matrix means: in WGSL, `m[i]` retrieves the `i`'th
*column* of `m`, whereas in HLSL it retrieves the `i`'th *row*. We
want to avoid translating `m[i]` into some complicated reassembly of a
vector from individually fetched components, so this is a problem.

However, with a bit of trickery, it is possible to use HLSL's `m[i]`
as the translation of WGSL's `m[i]`:

- We declare all matrices in uniform buffers in HLSL with the
  `row_major` qualifier, and transpose the row and column counts: a
  WGSL `mat3x4<f32>`, say, becomes an HLSL `row_major float3x4`. (Note
  that WGSL and HLSL type names put the row and column in reverse
  order.) Since the HLSL type is the transpose of how WebGPU directs
  the user to store the data, HLSL will load all matrices transposed.

- Since matrices are transposed, an HLSL indexing expression retrieves
  the "columns" of the intended WGSL value, as desired.

- For vector-matrix multiplication, since `mul(transpose(m), v)` is
  equivalent to `mul(v, m)` (note the reversal of the arguments), and
  `mul(v, transpose(m))` is equivalent to `mul(m, v)`, we can
  translate WGSL `m * v` and `v * m` to HLSL by simply reversing the
  arguments to `mul`.

## Padding in two-row matrices

An HLSL `row_major floatKx2` matrix has padding between its rows that
the WGSL `matKx2<f32>` matrix it represents does not. HLSL stores all
matrix rows [aligned on 16-byte boundaries][16bb], whereas WGSL says
that the columns of a `matKx2<f32>` need only be [aligned as required
for `vec2<f32>`][ilov], which is [eight-byte alignment][8bb].

To compensate for this, any time a `matKx2<f32>` appears in a WGSL
`uniform` value or as part of a struct/array, we actually emit `K`
separate `float2` members, and assemble/disassemble the matrix from its
columns (in WGSL; rows in HLSL) upon load and store.

For example, the following WGSL struct type:

```ignore
struct Baz {
        m: mat3x2<f32>,
}
```

is rendered as the HLSL struct type:

```ignore
struct Baz {
    float2 m_0; float2 m_1; float2 m_2;
};
```

The `wrapped_struct_matrix` functions in `help.rs` generate HLSL
helper functions to access such members, converting between the stored
form and the HLSL matrix types appropriately. For example, for reading
the member `m` of the `Baz` struct above, we emit:

```ignore
float3x2 GetMatmOnBaz(Baz obj) {
    return float3x2(obj.m_0, obj.m_1, obj.m_2);
}
```

We also emit an analogous `Set` function, as well as functions for
accessing individual columns by dynamic index.

## Sampler Handling

Due to limitations in how sampler heaps work in D3D12, we need to access samplers
through a layer of indirection. Instead of directly binding samplers, we bind the entire
sampler heap as both a standard and a comparison sampler heap. We then use a sampler
index buffer for each bind group. This buffer is accessed in the shader to get the actual
sampler index within the heap. See the wgpu_hal dx12 backend documentation for more
information.

[hlsl]: https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/dx-graphics-hlsl
[ilov]: https://gpuweb.github.io/gpuweb/wgsl/#internal-value-layout
[16bb]: https://github.com/microsoft/DirectXShaderCompiler/wiki/Buffer-Packing#constant-buffer-packing
[8bb]: https://gpuweb.github.io/gpuweb/wgsl/#alignment-and-size
*/

mod conv;
mod help;
mod keywords;
mod ray;
mod storage;
mod writer;

use alloc::{string::String, vec::Vec};
use core::fmt::Error as FmtError;

use thiserror::Error;

use crate::{back, ir, proc};

#[derive(Copy, Clone, Debug, Default, PartialEq, Eq, Hash)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
pub struct BindTarget {
    pub space: u8,
    /// For regular bindings this is the register number.
    ///
    /// For sampler bindings, this is the index to use into the bind group's sampler index buffer.
    pub register: u32,
    /// If the binding is an unsized binding array, this overrides the size.
    pub binding_array_size: Option<u32>,
    /// This is the index in the buffer at [`Options::dynamic_storage_buffer_offsets_targets`].
    pub dynamic_storage_buffer_offsets_index: Option<u32>,
    /// This is a hint that we need to restrict indexing of vectors, matrices and arrays.
    ///
    /// If [`Options::restrict_indexing`] is also `true`, we will restrict indexing.
    #[cfg_attr(any(feature = "serialize", feature = "deserialize"), serde(default))]
    pub restrict_indexing: bool,
}

#[derive(Clone, Debug, Default, PartialEq, Eq, Hash)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
/// BindTarget for dynamic storage buffer offsets
pub struct OffsetsBindTarget {
    pub space: u8,
    pub register: u32,
    pub size: u32,
}

#[cfg(any(feature = "serialize", feature = "deserialize"))]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
struct BindingMapSerialization {
    resource_binding: crate::ResourceBinding,
    bind_target: BindTarget,
}

#[cfg(feature = "deserialize")]
fn deserialize_binding_map<'de, D>(deserializer: D) -> Result<BindingMap, D::Error>
where
    D: serde::Deserializer<'de>,
{
    use serde::Deserialize;

    let vec = Vec::<BindingMapSerialization>::deserialize(deserializer)?;
    let mut map = BindingMap::default();
    for item in vec {
        map.insert(item.resource_binding, item.bind_target);
    }
    Ok(map)
}

// Using `BTreeMap` instead of `HashMap` so that we can hash itself.
pub type BindingMap = alloc::collections::BTreeMap<crate::ResourceBinding, BindTarget>;

/// A HLSL shader model version.
#[allow(non_snake_case, non_camel_case_types)]
#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq, PartialOrd)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
pub enum ShaderModel {
    V5_0,
    V5_1,
    V6_0,
    V6_1,
    V6_2,
    V6_3,
    V6_4,
    V6_5,
    V6_6,
    V6_7,
}

impl ShaderModel {
    pub const fn to_str(self) -> &'static str {
        match self {
            Self::V5_0 => "5_0",
            Self::V5_1 => "5_1",
            Self::V6_0 => "6_0",
            Self::V6_1 => "6_1",
            Self::V6_2 => "6_2",
            Self::V6_3 => "6_3",
            Self::V6_4 => "6_4",
            Self::V6_5 => "6_5",
            Self::V6_6 => "6_6",
            Self::V6_7 => "6_7",
        }
    }
}

impl crate::ShaderStage {
    pub const fn to_hlsl_str(self) -> &'static str {
        match self {
            Self::Vertex => "vs",
            Self::Fragment => "ps",
            Self::Compute => "cs",
            Self::Task | Self::Mesh => unreachable!(),
        }
    }
}

impl crate::ImageDimension {
    const fn to_hlsl_str(self) -> &'static str {
        match self {
            Self::D1 => "1D",
            Self::D2 => "2D",
            Self::D3 => "3D",
            Self::Cube => "Cube",
        }
    }
}

#[derive(Clone, Copy, Debug, Hash, Eq, Ord, PartialEq, PartialOrd)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
pub struct SamplerIndexBufferKey {
    pub group: u32,
}

#[derive(Clone, Debug, Hash, PartialEq, Eq)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
#[cfg_attr(feature = "deserialize", serde(default))]
pub struct SamplerHeapBindTargets {
    pub standard_samplers: BindTarget,
    pub comparison_samplers: BindTarget,
}

impl Default for SamplerHeapBindTargets {
    fn default() -> Self {
        Self {
            standard_samplers: BindTarget {
                space: 0,
                register: 0,
                binding_array_size: None,
                dynamic_storage_buffer_offsets_index: None,
                restrict_indexing: false,
            },
            comparison_samplers: BindTarget {
                space: 1,
                register: 0,
                binding_array_size: None,
                dynamic_storage_buffer_offsets_index: None,
                restrict_indexing: false,
            },
        }
    }
}

#[cfg(any(feature = "serialize", feature = "deserialize"))]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
struct SamplerIndexBufferBindingSerialization {
    group: u32,
    bind_target: BindTarget,
}

#[cfg(feature = "deserialize")]
fn deserialize_sampler_index_buffer_bindings<'de, D>(
    deserializer: D,
) -> Result<SamplerIndexBufferBindingMap, D::Error>
where
    D: serde::Deserializer<'de>,
{
    use serde::Deserialize;

    let vec = Vec::<SamplerIndexBufferBindingSerialization>::deserialize(deserializer)?;
    let mut map = SamplerIndexBufferBindingMap::default();
    for item in vec {
        map.insert(
            SamplerIndexBufferKey { group: item.group },
            item.bind_target,
        );
    }
    Ok(map)
}

// We use a BTreeMap here so that we can hash it.
pub type SamplerIndexBufferBindingMap =
    alloc::collections::BTreeMap<SamplerIndexBufferKey, BindTarget>;

#[cfg(any(feature = "serialize", feature = "deserialize"))]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
struct DynamicStorageBufferOffsetTargetSerialization {
    index: u32,
    bind_target: OffsetsBindTarget,
}

#[cfg(feature = "deserialize")]
fn deserialize_storage_buffer_offsets<'de, D>(
    deserializer: D,
) -> Result<DynamicStorageBufferOffsetsTargets, D::Error>
where
    D: serde::Deserializer<'de>,
{
    use serde::Deserialize;

    let vec = Vec::<DynamicStorageBufferOffsetTargetSerialization>::deserialize(deserializer)?;
    let mut map = DynamicStorageBufferOffsetsTargets::default();
    for item in vec {
        map.insert(item.index, item.bind_target);
    }
    Ok(map)
}

pub type DynamicStorageBufferOffsetsTargets = alloc::collections::BTreeMap<u32, OffsetsBindTarget>;

/// Shorthand result used internally by the backend
type BackendResult = Result<(), Error>;

#[derive(Clone, Debug, PartialEq, thiserror::Error)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
pub enum EntryPointError {
    #[error("mapping of {0:?} is missing")]
    MissingBinding(crate::ResourceBinding),
}

/// Configuration used in the [`Writer`].
#[derive(Clone, Debug, Hash, PartialEq, Eq)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
#[cfg_attr(feature = "deserialize", serde(default))]
pub struct Options {
    /// The hlsl shader model to be used
    pub shader_model: ShaderModel,
    /// Map of resources association to binding locations.
    #[cfg_attr(
        feature = "deserialize",
        serde(deserialize_with = "deserialize_binding_map")
    )]
    pub binding_map: BindingMap,
    /// Don't panic on missing bindings, instead generate any HLSL.
    pub fake_missing_bindings: bool,
    /// Add special constants to `SV_VertexIndex` and `SV_InstanceIndex`,
    /// to make them work like in Vulkan/Metal, with help of the host.
    pub special_constants_binding: Option<BindTarget>,
    /// Bind target of the push constant buffer
    pub push_constants_target: Option<BindTarget>,
    /// Bind target of the sampler heap and comparison sampler heap.
    pub sampler_heap_target: SamplerHeapBindTargets,
    /// Mapping of each bind group's sampler index buffer to a bind target.
    #[cfg_attr(
        feature = "deserialize",
        serde(deserialize_with = "deserialize_sampler_index_buffer_bindings")
    )]
    pub sampler_buffer_binding_map: SamplerIndexBufferBindingMap,
    /// Bind target for dynamic storage buffer offsets
    #[cfg_attr(
        feature = "deserialize",
        serde(deserialize_with = "deserialize_storage_buffer_offsets")
    )]
    pub dynamic_storage_buffer_offsets_targets: DynamicStorageBufferOffsetsTargets,
    /// Should workgroup variables be zero initialized (by polyfilling)?
    pub zero_initialize_workgroup_memory: bool,
    /// Should we restrict indexing of vectors, matrices and arrays?
    pub restrict_indexing: bool,
    /// If set, loops will have code injected into them, forcing the compiler
    /// to think the number of iterations is bounded.
    pub force_loop_bounding: bool,
}

impl Default for Options {
    fn default() -> Self {
        Options {
            shader_model: ShaderModel::V5_1,
            binding_map: BindingMap::default(),
            fake_missing_bindings: true,
            special_constants_binding: None,
            sampler_heap_target: SamplerHeapBindTargets::default(),
            sampler_buffer_binding_map: alloc::collections::BTreeMap::default(),
            push_constants_target: None,
            dynamic_storage_buffer_offsets_targets: alloc::collections::BTreeMap::new(),
            zero_initialize_workgroup_memory: true,
            restrict_indexing: true,
            force_loop_bounding: true,
        }
    }
}

impl Options {
    fn resolve_resource_binding(
        &self,
        res_binding: &crate::ResourceBinding,
    ) -> Result<BindTarget, EntryPointError> {
        match self.binding_map.get(res_binding) {
            Some(target) => Ok(*target),
            None if self.fake_missing_bindings => Ok(BindTarget {
                space: res_binding.group as u8,
                register: res_binding.binding,
                binding_array_size: None,
                dynamic_storage_buffer_offsets_index: None,
                restrict_indexing: false,
            }),
            None => Err(EntryPointError::MissingBinding(*res_binding)),
        }
    }
}

/// Reflection info for entry point names.
#[derive(Default)]
pub struct ReflectionInfo {
    /// Mapping of the entry point names.
    ///
    /// Each item in the array corresponds to an entry point index. The real entry point name may be different if one of the
    /// reserved words are used.
    ///
    /// Note: Some entry points may fail translation because of missing bindings.
    pub entry_point_names: Vec<Result<String, EntryPointError>>,
}

/// A subset of options that are meant to be changed per pipeline.
#[derive(Debug, Default, Clone)]
#[cfg_attr(feature = "serialize", derive(serde::Serialize))]
#[cfg_attr(feature = "deserialize", derive(serde::Deserialize))]
#[cfg_attr(feature = "deserialize", serde(default))]
pub struct PipelineOptions {
    /// The entry point to write.
    ///
    /// Entry points are identified by a shader stage specification,
    /// and a name.
    ///
    /// If `None`, all entry points will be written. If `Some` and the entry
    /// point is not found, an error will be thrown while writing.
    pub entry_point: Option<(ir::ShaderStage, String)>,
}

#[derive(Error, Debug)]
pub enum Error {
    #[error(transparent)]
    IoError(#[from] FmtError),
    #[error("A scalar with an unsupported width was requested: {0:?}")]
    UnsupportedScalar(crate::Scalar),
    #[error("{0}")]
    Unimplemented(String), // TODO: Error used only during development
    #[error("{0}")]
    Custom(String),
    #[error("overrides should not be present at this stage")]
    Override,
    #[error(transparent)]
    ResolveArraySizeError(#[from] proc::ResolveArraySizeError),
    #[error("entry point with stage {0:?} and name '{1}' not found")]
    EntryPointNotFound(ir::ShaderStage, String),
}

#[derive(PartialEq, Eq, Hash)]
enum WrappedType {
    ZeroValue(help::WrappedZeroValue),
    ArrayLength(help::WrappedArrayLength),
    ImageSample(help::WrappedImageSample),
    ImageQuery(help::WrappedImageQuery),
    ImageLoadScalar(crate::Scalar),
    Constructor(help::WrappedConstructor),
    StructMatrixAccess(help::WrappedStructMatrixAccess),
    MatCx2(help::WrappedMatCx2),
    Math(help::WrappedMath),
    UnaryOp(help::WrappedUnaryOp),
    BinaryOp(help::WrappedBinaryOp),
    Cast(help::WrappedCast),
}

#[derive(Default)]
struct Wrapped {
    types: crate::FastHashSet<WrappedType>,
    /// If true, the sampler heaps have been written out.
    sampler_heaps: bool,
    // Mapping from SamplerIndexBufferKey to the name the namer returned.
    sampler_index_buffers: crate::FastHashMap<SamplerIndexBufferKey, String>,
}

impl Wrapped {
    fn insert(&mut self, r#type: WrappedType) -> bool {
        self.types.insert(r#type)
    }

    fn clear(&mut self) {
        self.types.clear();
    }
}

/// A fragment entry point to be considered when generating HLSL for the output interface of vertex
/// entry points.
///
/// This is provided as an optional parameter to [`Writer::write`].
///
/// If this is provided, vertex outputs will be removed if they are not inputs of this fragment
/// entry point. This is necessary for generating correct HLSL when some of the vertex shader
/// outputs are not consumed by the fragment shader.
pub struct FragmentEntryPoint<'a> {
    module: &'a crate::Module,
    func: &'a crate::Function,
}

impl<'a> FragmentEntryPoint<'a> {
    /// Returns `None` if the entry point with the provided name can't be found or isn't a fragment
    /// entry point.
    pub fn new(module: &'a crate::Module, ep_name: &'a str) -> Option<Self> {
        module
            .entry_points
            .iter()
            .find(|ep| ep.name == ep_name)
            .filter(|ep| ep.stage == crate::ShaderStage::Fragment)
            .map(|ep| Self {
                module,
                func: &ep.function,
            })
    }
}

pub struct Writer<'a, W> {
    out: W,
    names: crate::FastHashMap<proc::NameKey, String>,
    namer: proc::Namer,
    /// HLSL backend options
    options: &'a Options,
    /// Per-stage backend options
    pipeline_options: &'a PipelineOptions,
    /// Information about entry point arguments and result types.
    entry_point_io: crate::FastHashMap<usize, writer::EntryPointInterface>,
    /// Set of expressions that have associated temporary variables
    named_expressions: crate::NamedExpressions,
    wrapped: Wrapped,
    written_committed_intersection: bool,
    written_candidate_intersection: bool,
    continue_ctx: back::continue_forward::ContinueCtx,

    /// A reference to some part of a global variable, lowered to a series of
    /// byte offset calculations.
    ///
    /// See the [`storage`] module for background on why we need this.
    ///
    /// Each [`SubAccess`] in the vector is a lowering of some [`Access`] or
    /// [`AccessIndex`] expression to the level of byte strides and offsets. See
    /// [`SubAccess`] for details.
    ///
    /// This field is a member of [`Writer`] solely to allow re-use of
    /// the `Vec`'s dynamic allocation. The value is no longer needed
    /// once HLSL for the access has been generated.
    ///
    /// [`Storage`]: crate::AddressSpace::Storage
    /// [`SubAccess`]: storage::SubAccess
    /// [`Access`]: crate::Expression::Access
    /// [`AccessIndex`]: crate::Expression::AccessIndex
    temp_access_chain: Vec<storage::SubAccess>,
    need_bake_expressions: back::NeedBakeExpressions,
}