Improve MeshletMesh::from_mesh performance (#13904)

This change reworks `find_connected_meshlets` to scale more linearly
with the mesh size, which significantly reduces the cost of building
meshlet representations. As a small extra complexity reduction, it moves
`simplify_scale` call out of the loop so that it's called once (it only
depends on the vertex data => is safe to cache).

The new implementation of connectivity analysis builds edge=>meshlet
list data structure, which allows us to only iterate through
`tuple_combinations` of a (usually) small list. There is still some
redundancy as if two meshlets share two edges, they will be represented
in the meshlet lists twice, but it's overall much faster.

Since the hash traversal is non-deterministic, to keep this part of the
algorithm deterministic for reproducible results we sort the output
adjacency lists.

Overall this reduces the time to process bunny mesh from ~4.2s to ~1.7s
when using release; in unoptimized builds the delta is even more
significant.

This was tested by using https://github.com/bevyengine/bevy/pull/13431
and:

a) comparing the result of `find_connected_meshlets` using old and new
code; they are equal in all steps of the clustering process
b) comparing the rendered result of the old code vs new code *after*
making the rest of the algorithm deterministic: right now the loop that
iterates through the result of `group_meshlets()` call executes in
different order between program runs. This is orthogonal to this change
and can be fixed separately.

Note: a future change can shrink the processing time further from ~1.7s
to ~0.4s with a small diff but that requires an update to meshopt crate
which is pending in https://github.com/gwihlidal/meshopt-rs/pull/42.
This change is independent.
This commit is contained in:
Arseny Kapoulkine 2024-06-18 01:29:17 -07:00 committed by GitHub
parent 91cd84fea7
commit 001cc147c6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -3,7 +3,7 @@ use bevy_render::{
mesh::{Indices, Mesh},
render_resource::PrimitiveTopology,
};
use bevy_utils::{HashMap, HashSet};
use bevy_utils::HashMap;
use itertools::Itertools;
use meshopt::{
build_meshlets, compute_cluster_bounds, compute_meshlet_bounds,
@ -11,6 +11,7 @@ use meshopt::{
simplify, simplify_scale, Meshlets, SimplifyOptions, VertexDataAdapter,
};
use metis::Graph;
use smallvec::SmallVec;
use std::{borrow::Cow, ops::Range};
impl MeshletMesh {
@ -54,18 +55,15 @@ impl MeshletMesh {
.iter()
.map(|m| m.triangle_count as u64)
.sum();
let mesh_scale = simplify_scale(&vertices);
// Build further LODs
let mut simplification_queue = 0..meshlets.len();
let mut lod_level = 1;
while simplification_queue.len() > 1 {
// For each meshlet build a set of triangle edges
let triangle_edges_per_meshlet =
collect_triangle_edges_per_meshlet(simplification_queue.clone(), &meshlets);
// For each meshlet build a list of connected meshlets (meshlets that share a triangle edge)
let connected_meshlets_per_meshlet =
find_connected_meshlets(simplification_queue.clone(), &triangle_edges_per_meshlet);
find_connected_meshlets(simplification_queue.clone(), &meshlets);
// Group meshlets into roughly groups of 4, grouping meshlets with a high number of shared edges
// http://glaros.dtc.umn.edu/gkhome/fetch/sw/metis/manual.pdf
@ -78,9 +76,13 @@ impl MeshletMesh {
for group_meshlets in groups.values().filter(|group| group.len() > 1) {
// Simplify the group to ~50% triangle count
let Some((simplified_group_indices, mut group_error)) =
simplify_meshlet_groups(group_meshlets, &meshlets, &vertices, lod_level)
else {
let Some((simplified_group_indices, mut group_error)) = simplify_meshlet_groups(
group_meshlets,
&meshlets,
&vertices,
lod_level,
mesh_scale,
) else {
continue;
};
@ -194,53 +196,70 @@ fn compute_meshlets(indices: &[u32], vertices: &VertexDataAdapter) -> Meshlets {
meshlets
}
fn collect_triangle_edges_per_meshlet(
simplification_queue: Range<usize>,
meshlets: &Meshlets,
) -> HashMap<usize, HashSet<(u32, u32)>> {
let mut triangle_edges_per_meshlet = HashMap::new();
for meshlet_id in simplification_queue {
let meshlet = meshlets.get(meshlet_id);
let meshlet_triangle_edges = triangle_edges_per_meshlet
.entry(meshlet_id)
.or_insert(HashSet::new());
for i in meshlet.triangles.chunks(3) {
let v0 = meshlet.vertices[i[0] as usize];
let v1 = meshlet.vertices[i[1] as usize];
let v2 = meshlet.vertices[i[2] as usize];
meshlet_triangle_edges.insert((v0.min(v1), v0.max(v1)));
meshlet_triangle_edges.insert((v0.min(v2), v0.max(v2)));
meshlet_triangle_edges.insert((v1.min(v2), v1.max(v2)));
}
}
triangle_edges_per_meshlet
}
fn find_connected_meshlets(
simplification_queue: Range<usize>,
triangle_edges_per_meshlet: &HashMap<usize, HashSet<(u32, u32)>>,
meshlets: &Meshlets,
) -> HashMap<usize, Vec<(usize, usize)>> {
let mut connected_meshlets_per_meshlet = HashMap::new();
for meshlet_id in simplification_queue.clone() {
connected_meshlets_per_meshlet.insert(meshlet_id, Vec::new());
}
// For each edge, gather all meshlets that use it
let mut edges_to_meshlets = HashMap::new();
for (meshlet_id1, meshlet_id2) in simplification_queue.tuple_combinations() {
let shared_edge_count = triangle_edges_per_meshlet[&meshlet_id1]
.intersection(&triangle_edges_per_meshlet[&meshlet_id2])
.count();
if shared_edge_count != 0 {
connected_meshlets_per_meshlet
.get_mut(&meshlet_id1)
.unwrap()
.push((meshlet_id2, shared_edge_count));
connected_meshlets_per_meshlet
.get_mut(&meshlet_id2)
.unwrap()
.push((meshlet_id1, shared_edge_count));
for meshlet_id in simplification_queue.clone() {
let meshlet = meshlets.get(meshlet_id);
for i in meshlet.triangles.chunks(3) {
for k in 0..3 {
let v0 = meshlet.vertices[i[k] as usize];
let v1 = meshlet.vertices[i[(k + 1) % 3] as usize];
let edge = (v0.min(v1), v0.max(v1));
let vec = edges_to_meshlets
.entry(edge)
.or_insert_with(SmallVec::<[usize; 2]>::new);
// Meshlets are added in order, so we can just check the last element to deduplicate,
// in the case of two triangles sharing the same edge within a single meshlet
if vec.last() != Some(&meshlet_id) {
vec.push(meshlet_id);
}
}
}
}
connected_meshlets_per_meshlet
// For each meshlet pair, count how many edges they share
let mut shared_edge_count = HashMap::new();
for (_, meshlet_ids) in edges_to_meshlets {
for (meshlet_id1, meshlet_id2) in meshlet_ids.into_iter().tuple_combinations() {
let count = shared_edge_count
.entry((meshlet_id1.min(meshlet_id2), meshlet_id1.max(meshlet_id2)))
.or_insert(0);
*count += 1;
}
}
// For each meshlet, gather all meshlets that share at least one edge along with shared edge count
let mut connected_meshlets = HashMap::new();
for meshlet_id in simplification_queue.clone() {
connected_meshlets.insert(meshlet_id, Vec::new());
}
for ((meshlet_id1, meshlet_id2), shared_count) in shared_edge_count {
// We record id1->id2 and id2->id1 as adjacency is symmetrical
connected_meshlets
.get_mut(&meshlet_id1)
.unwrap()
.push((meshlet_id2, shared_count));
connected_meshlets
.get_mut(&meshlet_id2)
.unwrap()
.push((meshlet_id1, shared_count));
}
// The order of meshlets depends on hash traversal order; to produce deterministic results, sort them
for (_, connected_meshlets) in connected_meshlets.iter_mut() {
connected_meshlets.sort_unstable();
}
connected_meshlets
}
fn group_meshlets(
@ -284,6 +303,7 @@ fn simplify_meshlet_groups(
meshlets: &Meshlets,
vertices: &VertexDataAdapter<'_>,
lod_level: u32,
mesh_scale: f32,
) -> Option<(Vec<u32>, f32)> {
// Build a new index buffer into the mesh vertex data by combining all meshlet data in the group
let mut group_indices = Vec::new();
@ -316,7 +336,7 @@ fn simplify_meshlet_groups(
}
// Convert error to object-space and convert from diameter to radius
error *= simplify_scale(vertices) * 0.5;
error *= mesh_scale * 0.5;
Some((simplified_group_indices, error))
}