Parallel Frustum Culling (#4489)

# Objective

Working with a large number of entities with `Aabbs`, rendered with an instanced shader, I found the bottleneck became the frustum culling system. The goal of this PR is to significantly improve culling performance without any major changes. We should consider constructing a BVH for more substantial improvements.

## Solution

- Convert the inner entity query to a parallel iterator with `par_for_each_mut` using a batch size of 1,024. 
- This outperforms single threaded culling when there are more than 1,000 entities. 
  - Below this they are approximately equal, with <= 10 microseconds of multithreading overhead.
  - Above this, the multithreaded version is significantly faster, scaling linearly with core count.
- In my million-entity-workload, this PR improves my framerate by 200% - 300%.

## log-log of `check_visibility` time vs. entities for single/multithreaded
![image](https://user-images.githubusercontent.com/2632925/163709007-7eab4437-e9f9-4c06-bac0-250073885110.png)

---

## Changelog

Frustum culling is now run with a parallel query. When culling more than a thousand entities, this is faster than the previous method, scaling proportionally with the number of available cores.

crates/bevy_render/Cargo.toml

@@ -56,6 +56,7 @@ once_cell = "1.4.1" # TODO: replace once_cell with std equivalent if/when this l
 downcast-rs = "1.2.0"
 thiserror = "1.0"
 futures-lite = "1.4.0"
+crossbeam-channel = "0.5.0"
 anyhow = "1.0"
 hex = "0.4.2"
 hexasphere = "7.0.0"

crates/bevy_render/src/view/visibility/mod.rs

@@ -149,9 +149,7 @@ pub fn update_frusta<T: Component + CameraProjection + Send + Sync + 'static>(
 
 pub fn check_visibility(
     mut view_query: Query<(&mut VisibleEntities, &Frustum, Option<&RenderLayers>), With<Camera>>,
-    mut visible_entity_query: ParamSet<(
+    mut visible_entity_query: Query<(
-        Query<&mut ComputedVisibility>,
-        Query<(
         Entity,
         &Visibility,
         &mut ComputedVisibility,
@@ -160,18 +158,14 @@ pub fn check_visibility(
         Option<&NoFrustumCulling>,
         Option<&GlobalTransform>,
     )>,
-    )>,
 ) {
-    // Reset the computed visibility to false
-    for mut computed_visibility in visible_entity_query.p0().iter_mut() {
-        computed_visibility.is_visible = false;
-    }
-
     for (mut visible_entities, frustum, maybe_view_mask) in view_query.iter_mut() {
-        visible_entities.entities.clear();
         let view_mask = maybe_view_mask.copied().unwrap_or_default();
+        let (visible_entity_sender, visible_entity_receiver) = crossbeam_channel::unbounded();
 
-        for (
+        visible_entity_query.par_for_each_mut(
+            1024,
+            |(
                 entity,
                 visibility,
                 mut computed_visibility,
@@ -179,15 +173,16 @@ pub fn check_visibility(
                 maybe_aabb,
                 maybe_no_frustum_culling,
                 maybe_transform,
-        ) in visible_entity_query.p1().iter_mut()
+            )| {
-        {
+                // Reset visibility
-            if !visibility.is_visible {
+                computed_visibility.is_visible = false;
-                continue;
-            }
 
+                if !visibility.is_visible {
+                    return;
+                }
                 let entity_mask = maybe_entity_mask.copied().unwrap_or_default();
                 if !view_mask.intersects(&entity_mask) {
-                continue;
+                    return;
                 }
 
                 // If we have an aabb and transform, do frustum culling
@@ -201,19 +196,18 @@ pub fn check_visibility(
                     };
                     // Do quick sphere-based frustum culling
                     if !frustum.intersects_sphere(&model_sphere, false) {
-                    continue;
+                        return;
                     }
                     // If we have an aabb, do aabb-based frustum culling
                     if !frustum.intersects_obb(model_aabb, &model, false) {
-                    continue;
+                        return;
                     }
                 }
 
                 computed_visibility.is_visible = true;
-            visible_entities.entities.push(entity);
+                visible_entity_sender.send(entity).ok();
-        }
+            },
-
+        );
-        // TODO: check for big changes in visible entities len() vs capacity() (ex: 2x) and resize
+        visible_entities.entities = visible_entity_receiver.try_iter().collect();
-        // to prevent holding unneeded memory
     }
 }