gfx-rs/wgpu · error

not implemented

Error message

not implemented

What it means

describe_vertex_format in the GLES backend has no mapping for 64-bit float vertex formats (Float64, Float64x2, Float64x3, Float64x4) and panics with `unimplemented!()`. OpenGL ES does not support 64-bit (double) vertex attributes, unlike desktop GL/GL_ARB_vertex_attrib_64bit.

Source

Thrown at wgpu-hal/src/gles/conv.rs:228

        Vf::Snorm16x4 => (4, glow::SHORT, Vak::Float),
        Vf::Uint16x4 => (4, glow::UNSIGNED_SHORT, Vak::Integer),
        Vf::Sint16x4 => (4, glow::SHORT, Vak::Integer),
        Vf::Float16x4 => (4, glow::HALF_FLOAT, Vak::Float),
        Vf::Uint32 => (1, glow::UNSIGNED_INT, Vak::Integer),
        Vf::Sint32 => (1, glow::INT, Vak::Integer),
        Vf::Float32 => (1, glow::FLOAT, Vak::Float),
        Vf::Uint32x2 => (2, glow::UNSIGNED_INT, Vak::Integer),
        Vf::Sint32x2 => (2, glow::INT, Vak::Integer),
        Vf::Float32x2 => (2, glow::FLOAT, Vak::Float),
        Vf::Uint32x3 => (3, glow::UNSIGNED_INT, Vak::Integer),
        Vf::Sint32x3 => (3, glow::INT, Vak::Integer),
        Vf::Float32x3 => (3, glow::FLOAT, Vak::Float),
        Vf::Uint32x4 => (4, glow::UNSIGNED_INT, Vak::Integer),
        Vf::Sint32x4 => (4, glow::INT, Vak::Integer),
        Vf::Float32x4 => (4, glow::FLOAT, Vak::Float),
        Vf::Unorm10_10_10_2 => (4, glow::UNSIGNED_INT_2_10_10_10_REV, Vak::Float),
        Vf::Unorm8x4Bgra => (glow::BGRA as i32, glow::UNSIGNED_BYTE, Vak::Float),
        Vf::Float64 | Vf::Float64x2 | Vf::Float64x3 | Vf::Float64x4 => unimplemented!(),
    };

    super::VertexFormatDesc {
        element_count,
        element_format,
        attrib_kind,
    }
}

pub fn map_filter_modes(
    min: wgt::FilterMode,
    mag: wgt::FilterMode,
    mip: wgt::MipmapFilterMode,
) -> (u32, u32) {
    use wgt::FilterMode as Fm;
    use wgt::MipmapFilterMode as Mfm;

    let mag_filter = match mag {

View on GitHub (pinned to 3e11ff59bf)

Solutions

1. Change the vertex format to f32 variants (float32, float32x2, ...) and convert doubles to floats on the CPU (or encode with a fixed-point/split-half encoding)
2. If double precision is required, pack doubles into two u32 components (float64 emulation via mathf64x2-style unpacking in the shader)
3. Validate vertex layouts at startup and reject float64 formats when the backend is GLES

Example fix

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

fn assert_no_f64_vertex_layout(layout: &wgt::VertexBufferLayout) -> Result<(), String> {
    for a in &layout.attributes {
        if matches!(a.format, wgt::VertexFormat::Float64 | wgt::VertexFormat::Float64x2 | wgt::VertexFormat::Float64x3 | wgt::VertexFormat::Float64x4) {
            return Err(format!("float64 vertex format not supported on GLES: {:?}", a.format));
        }
    }
    Ok(())
}

fn is_f64_vertex_format(f: wgt::VertexFormat) -> bool {
    matches!(f, wgt::VertexFormat::Float64 | wgt::VertexFormat::Float64x2 | wgt::VertexFormat::Float64x3 | wgt::VertexFormat::Float64x4)
}

match device.create_render_pipeline(&desc) {
    Ok(p) => p,
    Err(e) => { log::error!("pipeline creation failed (check float64 attributes on GLES): {e}"); fallback_pipeline() }
}

When it happens

Trigger: Creating a render pipeline whose vertex buffer layout uses vertex_format: "float64", "float64x2", "float64x3" or "float64x4" while running on the GLES backend (e.g. WebGL2), which calls conv::describe_vertex_format during create_render_pipeline.

Common situations: Porting desktop (Vulkan/D3D) code using double-precision vertex data to a WebGL/GLES target; accidentally leaving debug f64 attributes in shaders/vertex layouts; targeting Android/iOS where GLES is the backend.

Related errors

• not implemented
• not implemented
• VERTEX_ATTRIBUTE_64BIT feature must be enabled to use Double
• not implemented
• metal does not support vertex ray hit return

AI-assisted analysis of gfx-rs/wgpu@3e11ff59bf (2026-09-03). Data as JSON: /api/errors/346f222995363968. Report an issue: GitHub.