Make meshlet processing deterministic (#13913)

This is a followup to https://github.com/bevyengine/bevy/pull/13904
based on the discussion there, and switches two HashMaps that used
meshlet ids as keys to Vec.

In addition to a small further performance boost for `from_mesh` (1.66s
=> 1.60s), this makes processing deterministic modulo threading issues
wrt CRT rand described in the linked PR. This is valuable for debugging,
as you can visually or programmatically inspect the meshlet distribution
before/after making changes that should not change the output, whereas
previously every asset rebuild would change the meshlet structure.

Tested with https://github.com/bevyengine/bevy/pull/13431; after this
change, the visual output of meshlets is consistent between asset
rebuilds, and the MD5 of the output GLB file does not change either,
which was not the case before.

crates/bevy_pbr/src/meshlet/from_mesh.rs

@@ -74,10 +74,10 @@ impl MeshletMesh {
 
             let next_lod_start = meshlets.len();
 
-            for group_meshlets in groups.values().filter(|group| group.len() > 1) {
+            for group_meshlets in groups.into_iter().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,
+                    &group_meshlets,
                     &meshlets,
                     &vertices,
                     lod_level,
@@ -101,7 +101,7 @@ impl MeshletMesh {
 
                 // For each meshlet in the group set their parent LOD bounding sphere to that of the simplified group
                 for meshlet_id in group_meshlets {
-                    bounding_spheres[*meshlet_id].parent_lod = group_bounding_sphere;
+                    bounding_spheres[meshlet_id].parent_lod = group_bounding_sphere;
                 }
 
                 // Build new meshlets using the simplified group
@@ -199,7 +199,7 @@ fn compute_meshlets(indices: &[u32], vertices: &VertexDataAdapter) -> Meshlets {
 fn find_connected_meshlets(
     simplification_queue: Range<usize>,
     meshlets: &Meshlets,
-) -> HashMap<usize, Vec<(usize, usize)>> {
+) -> Vec<Vec<(usize, usize)>> {
     // For each edge, gather all meshlets that use it
     let mut edges_to_meshlets = HashMap::new();
 
@@ -236,27 +236,19 @@ fn find_connected_meshlets(
     }
 
     // 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());
-    }
+    let mut connected_meshlets = vec![Vec::new(); simplification_queue.len()];
 
     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()
+        connected_meshlets[meshlet_id1 - simplification_queue.start]
             .push((meshlet_id2, shared_count));
-        connected_meshlets
-            .get_mut(&meshlet_id2)
-            .unwrap()
+        connected_meshlets[meshlet_id2 - simplification_queue.start]
             .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();
+    for list in connected_meshlets.iter_mut() {
+        list.sort_unstable();
     }
 
     connected_meshlets
@@ -264,36 +256,34 @@ fn find_connected_meshlets(
 
 fn group_meshlets(
     simplification_queue: Range<usize>,
-    connected_meshlets_per_meshlet: &HashMap<usize, Vec<(usize, usize)>>,
-) -> HashMap<i32, Vec<usize>> {
+    connected_meshlets_per_meshlet: &[Vec<(usize, usize)>],
+) -> Vec<Vec<usize>> {
     let mut xadj = Vec::with_capacity(simplification_queue.len() + 1);
     let mut adjncy = Vec::new();
     let mut adjwgt = Vec::new();
     for meshlet_id in simplification_queue.clone() {
         xadj.push(adjncy.len() as i32);
         for (connected_meshlet_id, shared_edge_count) in
-            connected_meshlets_per_meshlet[&meshlet_id].iter().copied()
+            connected_meshlets_per_meshlet[meshlet_id - simplification_queue.start].iter()
         {
             adjncy.push((connected_meshlet_id - simplification_queue.start) as i32);
-            adjwgt.push(shared_edge_count as i32);
+            adjwgt.push(*shared_edge_count as i32);
         }
     }
     xadj.push(adjncy.len() as i32);
 
     let mut group_per_meshlet = vec![0; simplification_queue.len()];
-    let partition_count = (simplification_queue.len().div_ceil(4)) as i32;
-    Graph::new(1, partition_count, &xadj, &adjncy)
+    let partition_count = simplification_queue.len().div_ceil(4);
+    Graph::new(1, partition_count as i32, &xadj, &adjncy)
         .unwrap()
         .set_adjwgt(&adjwgt)
         .part_kway(&mut group_per_meshlet)
         .unwrap();
 
-    let mut groups = HashMap::new();
+    let mut groups = vec![Vec::new(); partition_count];
+
     for (i, meshlet_group) in group_per_meshlet.into_iter().enumerate() {
-        groups
-            .entry(meshlet_group)
-            .or_insert(Vec::new())
-            .push(i + simplification_queue.start);
+        groups[meshlet_group as usize].push(i + simplification_queue.start);
     }
     groups
 }