meshoptimizer-rs 0.2.7

Rust port of meshoptimizer 1.3 with byte-identical output. Covers mesh simplification, meshlets, cluster LOD, vertex/index compression analysis, remeshing and optional parallel batch APIs and audited SIMD codecs.
Documentation

meshoptimizer-rs

CI crates.io docs.rs

A pure-Rust port of meshoptimizer 1.3. Its output is byte-identical to the C++ library: index order, error values and encoded bytes match. It needs no C++ toolchain and supports no_std with alloc.

[dependencies]
meshoptimizer-rs = "0.2.5"

The minimum Rust version is 1.88. The library name is meshoptimizer_rs.

Usage

Every operation takes a &mut Workspace and returns Result<_, meshoptimizer_rs::Error>. Most have an allocating form, a caller-buffer _into form and, where it applies, an _in_place form.

Simplify and optimize a mesh

Simplify to a quarter of the triangles, reorder triangles for the vertex cache and overdraw, then reorder vertices for fetch.

use meshoptimizer_rs::{
    optimize_overdraw, optimize_vertex_cache, optimize_vertex_fetch, simplify, Error, FetchedMesh,
    Positions, SimplifyOptions, SimplifySettings, VertexStream, Workspace,
};

fn make_lod(positions: &[[f32; 3]], indices: &[u32]) -> Result<FetchedMesh, Error> {
    let mut workspace = Workspace::default();
    let view = Positions::from_packed(positions);
    let target_index_count = indices.len() / 4;
    let settings = SimplifySettings { target_index_count, target_error: 0.01, options: SimplifyOptions::EMPTY };
    let lod = simplify(indices, view, settings, &mut workspace)?;
    let lod = optimize_vertex_cache(&lod.indices, positions.len(), &mut workspace)?;
    let lod = optimize_overdraw(&lod, view, 1.05, &mut workspace)?;

    let bytes: Vec<u8> = positions.iter().flatten().flat_map(|v| v.to_le_bytes()).collect();
    optimize_vertex_fetch(&lod, VertexStream::packed(&bytes, 12)?, &mut workspace)
}

To account for attributes such as normals, call simplify_with_attributes with an Attributes view and one weight per component, for example Attributes::from_interleaved(&normals, vertex_count, 3, 3, 0)? and &[0.5; 3]. Interleaved positions use Positions::from_interleaved; stride and offset are in f32 elements.

For repeated fetch-remap calls, reuse FetchIndices::new(indices, positions.len()) with optimize_vertex_fetch_remap_with_indices or its _into form to avoid repeated index validation.

Meshlets and bounds

use meshoptimizer_rs::{
    build_meshlets, compute_meshlet_bounds, Bounds, Error,
    MeshletSettings, Positions, Workspace,
};

fn meshlet_bounds(positions: &[[f32; 3]], indices: &[u32]) -> Result<Vec<Bounds>, Error> {
    let mut workspace = Workspace::default();
    let view = Positions::from_packed(positions);
    let meshlets = build_meshlets(indices, view, MeshletSettings::default(), 0.25, &mut workspace)?;

    let mut bounds = Vec::new();
    for m in &meshlets.meshlets {
        let (v, t) = (m.vertex_offset as usize, m.triangle_offset as usize);
        let vertices = &meshlets.vertices[v..v + m.vertex_count as usize];
        let triangles = &meshlets.triangles[t..t + m.triangle_count as usize * 3];
        bounds.push(compute_meshlet_bounds(vertices, triangles, view, &mut workspace)?);
    }
    Ok(bounds)
}

build_meshlets_scan, build_meshlets_flex and build_meshlets_spatial cover the other upstream builders. compute_cluster_bounds computes bounds for index-list clusters. Reuse BoundsPositions for repeated bounds queries over the same positions.

Cluster LOD DAG

With the clusterlod feature, build_dag_into builds the cluster DAG and its per-level node forest into buffers you own. Allocate each DagBuffers slice, for example with vec![Default::default(); capacity]. The returned DagCounts gives the length of each filled prefix, such as indices[..counts.indices].

#[cfg(feature = "clusterlod")]
use meshoptimizer_rs::{clusterlod::{self, DagBuffers, DagCounts, Mesh}, Workspace};

#[cfg(feature = "clusterlod")]
fn build_dag(
    indices: &[u32],
    positions: &mut [[f32; 3]],
    buffers: DagBuffers<'_>,
) -> Result<DagCounts, meshoptimizer_rs::Error> {
    let mesh = Mesh { indices, positions, attributes: None, vertex_lock: None,
                      attribute_weights: &[], attribute_protect_mask: 0 };
    let config = clusterlod::default_config(128)?;
    clusterlod::build_dag_into(config, mesh, 8, buffers, &mut Workspace::default())
}

Only the returned prefixes are output. If a call fails, including with BufferTooSmall, discard the output, restore the original positions (border dilation may have moved them) and retry with larger buffers. Caller buffers don't count toward Workspace limits.

With parallel as well, parallel::build_dags_batch_into(&mut inputs, Limits::default()) builds several meshes at once. Each DagBuildInput holds the config, mesh, node_width and destination arguments above. It returns one Result<DagCounts, Error> per input, in input order, and each item has its own limits.

Vertex and index compression

The codec module implements the codecs used by EXT_meshopt_compression, plus index sequence and meshlet codecs and the Oct, Quat, Exp and Color filters. VertexEncoding and IndexEncoding set the format version and level per call. decode_buffer_view decodes a glTF bufferView using the mode and filter names from its JSON.

use meshoptimizer_rs::codec::{
    decode_buffer_view, decode_vertex_buffer, encode_index_buffer, encode_vertex_buffer, Filter,
    IndexEncoding, Mode, VertexEncoding,
};
use meshoptimizer_rs::{Error, Workspace};

fn compress(vertices: &[u8], stride: usize, indices: &[u32]) -> Result<(Vec<u8>, Vec<u8>), Error> {
    let mut workspace = Workspace::default();
    let count = vertices.len() / stride;
    let encoding = VertexEncoding::DEFAULT;
    let packed_vertices = encode_vertex_buffer(vertices, count, stride, encoding, &mut workspace)?;
    assert_eq!(decode_vertex_buffer(count, stride, &packed_vertices, &mut workspace)?, vertices);

    let packed_indices = encode_index_buffer(indices, IndexEncoding::DEFAULT, &mut workspace)?;
    Ok((packed_vertices, packed_indices))
}

fn decode_view(mode: &str, filter: &str, n: usize, stride: usize, data: &[u8]) -> Result<Vec<u8>, Error> {
    let (mode, filter) = (Mode::parse(mode)?, Filter::parse(filter)?);
    decode_buffer_view(mode, filter, n, stride, data, &mut Workspace::default())
}

Workspace and limits

Keep one Workspace per thread and reuse it: it keeps scratch memory between calls. Its Limits cap the bytes the crate allocates and the work units a call may spend; the defaults are 1 GiB and 2^34. Set them with Workspace::new(Limits { max_bytes: 64 << 20, max_work: 1 << 30 }). A call over a limit returns Error::LimitExceeded, and Workspace::usage reports the last call's bytes and work. Call workspace.set_retention(Retention::Lean) for lower peak memory in long simplify chains at some time cost.

Features

Feature Default Enables
std yes std::error::Error for Error.
simd yes SIMD vertex, filter and meshlet decoding for SSE2/SSSE3/SSE4.1, NEON and wasm simd128.
parallel no Rayon batch APIs in the parallel module; enables std.
clusterlod no The clusterlod module, a port of upstream demo/clusterlod.h.
experimental no Upstream experimental functions, including normal generation and remeshing.

For no_std, set default-features = false; an allocator is required. Without simd the crate contains no unsafe code. Cargo feature unification applies: if another dependency enables simd, it is on for your build too.

Correctness and performance

Default output is byte-identical to meshoptimizer 1.3 at commit 4c203430ca565cb59a468a91922c76c208169536, whose src tree equals tag v1.3. Fixtures, differential sweeps, fuzzing and WASM runs check this on recorded inputs; they don't prove it for every input. SIMD decoding matches the scalar bytes. See docs/VERIFICATION.md and UPSTREAM.md. Timings against C++ 1.3 and the meshopt crate are in docs/PERFORMANCE.md, with links to the parity records. Development phases and planned work are in docs/ROADMAP.md.

Known exceptions

Intervals and method are in docs/PERFORMANCE.md.

  • Overdraw optimization is about 1.47–1.60× C++ 1.3 (scalar) on the largest real meshes measured.
  • Allocating vertex decode, vertex-v1-streaming-s4: 1.529–2.118× in 0.2.0; 1.000× and 1.058× when re-run for 0.2.2. The caller-buffer form, decode_vertex_buffer_into, is about 1.04×.
  • Allocating index decode maxima: index-2-v0-streaming-s4 1.508–3.625×, index-2-v1-streaming-s4 2.189–2.465×.
  • Allocating EXT view maxima: view-1-streaming-s4 1.343–1.636×; varied-view-3-streaming-s32 1.619–2.590× (unresolved); the filtered-view maximum is near 1.64×.
  • partition_clusters with Cargo defaults: mean 1.450×, maximum 1.778×.
  • Older records keep a sparse allocating vertex-fetch maximum failure and inconclusive overdraw and sloppy-simplification maxima.
  • A malformed meshlet tail is accepted by scalar upstream and by this crate, but rejected by upstream SSE.
  • AVX2, AVX-512 and 32-bit ARM SIMD are not implemented; index and sequence decoding is scalar.

Licence and acknowledgements

MIT; see LICENSE. meshoptimizer is written by Arseny Kapoulkine. This independent port keeps upstream's licence notice and credits its algorithms; it is not affiliated with or endorsed by the upstream project. UPSTREAM.md records provenance and intentional API differences.