gfx-rs/wgpu · error

not implemented

Error message

not implemented

What it means

`map_vertex_format` in the Metal backend's `conv.rs` panics with `unimplemented!()` for 64-bit float vertex formats (Float64, Float64x2, Float64x3, Float64x4). Metal Shading Language / MTLVertexFormat has no direct double-precision vertex attribute support in the set this backend maps, so the conversion is intentionally unimplemented.

Source

Thrown at wgpu-hal/src/metal/conv.rs:245

        Vf::Snorm16x4 => MTL::Short4Normalized,
        Vf::Uint16x4 => MTL::UShort4,
        Vf::Sint16x4 => MTL::Short4,
        Vf::Float16x4 => MTL::Half4,
        Vf::Uint32 => MTL::UInt,
        Vf::Sint32 => MTL::Int,
        Vf::Float32 => MTL::Float,
        Vf::Uint32x2 => MTL::UInt2,
        Vf::Sint32x2 => MTL::Int2,
        Vf::Float32x2 => MTL::Float2,
        Vf::Uint32x3 => MTL::UInt3,
        Vf::Sint32x3 => MTL::Int3,
        Vf::Float32x3 => MTL::Float3,
        Vf::Uint32x4 => MTL::UInt4,
        Vf::Sint32x4 => MTL::Int4,
        Vf::Float32x4 => MTL::Float4,
        Vf::Unorm10_10_10_2 => MTL::UInt1010102Normalized,
        Vf::Unorm8x4Bgra => MTL::UChar4Normalized_BGRA,
        Vf::Float64 | Vf::Float64x2 | Vf::Float64x3 | Vf::Float64x4 => unimplemented!(),
    }
}

pub fn map_index_format(format: wgt::IndexFormat) -> (u64, MTLIndexType) {
    match format {
        wgt::IndexFormat::Uint16 => (2, MTLIndexType::UInt16),
        wgt::IndexFormat::Uint32 => (4, MTLIndexType::UInt32),
    }
}

pub fn map_step_mode(mode: wgt::VertexStepMode) -> MTLVertexStepFunction {
    match mode {
        wgt::VertexStepMode::Vertex => MTLVertexStepFunction::PerVertex,
        wgt::VertexStepMode::Instance => MTLVertexStepFunction::PerInstance,
    }
}

pub fn map_stencil_op(op: wgt::StencilOperation) -> MTLStencilOperation {

View on GitHub (pinned to 3e11ff59bf)

Solutions

1. Change vertex formats to f32 variants (Float32, Float32x2, Float32x3, Float32x4)
2. Convert f64 data to f32 on the CPU when uploading the vertex buffer
3. Encode large-precision values as u32/f32 pairs with a custom transform in the vertex shader

Example fix

// before
attributes: &[VertexAttribute { format: VertexFormat::Float64x3, offset: 0, shader_location: 0 }]
// after
attributes: &[VertexAttribute { format: VertexFormat::Float32x3, offset: 0, shader_location: 0 }]

fn has_f64_vertex_format(layouts: &[VertexBufferLayout]) -> bool {
    layouts.iter().flat_map(|l| &l.attributes).any(|a| matches!(a.format,
        VertexFormat::Float64 | VertexFormat::Float64x2 | VertexFormat::Float64x3 | VertexFormat::Float64x4))
}

fn is_f64_format(f: VertexFormat) -> bool {
    matches!(f, VertexFormat::Float64 | VertexFormat::Float64x2 | VertexFormat::Float64x3 | VertexFormat::Float64x4)
}

When it happens

Trigger: Creating a vertex buffer layout (or acceleration structure descriptor, via `map_acceleration_structure_descriptor`) whose `VertexFormat` is one of the four f64 formats, on the Metal backend.

Common situations: Porting Vulkan/DX12 desktop code that uses double-precision vertex data (common in CAD/scientific apps) to Metal; WGSL shaders/vertex layouts authored with f64 attributes.

Related errors

• not implemented
• VERTEX_ATTRIBUTE_64BIT feature must be enabled to use Double
• not implemented
• Ray tracing pipelines are unsupported on Metal
• {:?} is not enabled for this backend

AI-assisted analysis of gfx-rs/wgpu@3e11ff59bf (2026-09-03). Data as JSON: /api/errors/029dbfc4755b2163. Report an issue: GitHub.