Use multidraw for shadows when GPU culling is in use.

This patch makes shadows use multidraw when the camera they'll be drawn
to has the `GpuCulling` component. This results in a significant
reduction in drawcalls; Bistro Exterior drops to 3 drawcalls for each
shadow cascade.

Note that PR #16670 will remove the `GpuCulling` component, making
shadows automatically use multidraw. Beware of that when testing this
patch; before #16670 lands, you'll need to manually add `GpuCulling` to
your camera in order to see any performance benefits.

crates/bevy_pbr/src/render/light.rs

@@ -15,6 +15,7 @@ use bevy_render::{
     batching::gpu_preprocessing::{GpuPreprocessingMode, GpuPreprocessingSupport},
     camera::SortedCameras,
     mesh::allocator::MeshAllocator,
+    view::GpuCulling,
 };
 use bevy_render::{
     diagnostic::RecordDiagnostics,
@@ -689,6 +690,7 @@ pub fn prepare_lights(
             &ExtractedView,
             &ExtractedClusterConfig,
             Option<&RenderLayers>,
+            Has<GpuCulling>,
         ),
         With<Camera3d>,
     >,
@@ -1097,14 +1099,20 @@ pub fn prepare_lights(
     let mut live_views = EntityHashSet::with_capacity_and_hasher(views_count, EntityHash);
 
     // set up light data for each view
-    for (entity, extracted_view, clusters, maybe_layers) in sorted_cameras
+    for (entity, extracted_view, clusters, maybe_layers, has_gpu_culling) in sorted_cameras
         .0
         .iter()
         .filter_map(|sorted_camera| views.get(sorted_camera.entity).ok())
     {
         live_views.insert(entity);
         let mut view_lights = Vec::new();
 
+        let gpu_preprocessing_mode = gpu_preprocessing_support.min(if has_gpu_culling {
+            GpuPreprocessingMode::Culling
+        } else {
+            GpuPreprocessingMode::PreprocessingOnly
+        });
+
         let is_orthographic = extracted_view.clip_from_view.w_axis.w == 1.0;
         let cluster_factors_zw = calculate_cluster_factors(
             clusters.near,
@@ -1232,15 +1240,15 @@ pub fn prepare_lights(
                     },
                 ));
 
+                if matches!(gpu_preprocessing_mode, GpuPreprocessingMode::Culling) {
+                    commands.entity(view_light_entity).insert(GpuCulling);
+                }
+
                 view_lights.push(view_light_entity);
 
                 if first {
                     // Subsequent views with the same light entity will reuse the same shadow map
-                    // TODO: Implement GPU culling for shadow passes.
-                    shadow_render_phases.insert_or_clear(
-                        view_light_entity,
-                        gpu_preprocessing_support.min(GpuPreprocessingMode::PreprocessingOnly),
-                    );
+                    shadow_render_phases.insert_or_clear(view_light_entity, gpu_preprocessing_mode);
                     live_shadow_mapping_lights.insert(view_light_entity);
                 }
             }
@@ -1324,14 +1332,15 @@ pub fn prepare_lights(
                 LightEntity::Spot { light_entity },
             ));
 
+            if matches!(gpu_preprocessing_mode, GpuPreprocessingMode::Culling) {
+                commands.entity(view_light_entity).insert(GpuCulling);
+            }
+
             view_lights.push(view_light_entity);
 
             if first {
                 // Subsequent views with the same light entity will reuse the same shadow map
-                shadow_render_phases.insert_or_clear(
-                    view_light_entity,
-                    gpu_preprocessing_support.min(GpuPreprocessingMode::PreprocessingOnly),
-                );
+                shadow_render_phases.insert_or_clear(view_light_entity, gpu_preprocessing_mode);
                 live_shadow_mapping_lights.insert(view_light_entity);
             }
         }
@@ -1457,15 +1466,17 @@ pub fn prepare_lights(
                         cascade_index,
                     },
                 ));
+
+                if matches!(gpu_preprocessing_mode, GpuPreprocessingMode::Culling) {
+                    commands.entity(view_light_entity).insert(GpuCulling);
+                }
+
                 view_lights.push(view_light_entity);
 
                 // Subsequent views with the same light entity will **NOT** reuse the same shadow map
                 // (Because the cascades are unique to each view)
                 // TODO: Implement GPU culling for shadow passes.
-                shadow_render_phases.insert_or_clear(
-                    view_light_entity,
-                    gpu_preprocessing_support.min(GpuPreprocessingMode::PreprocessingOnly),
-                );
+                shadow_render_phases.insert_or_clear(view_light_entity, gpu_preprocessing_mode);
                 live_shadow_mapping_lights.insert(view_light_entity);
             }
         }