bevy/crates/bevy_pbr/src/cluster/assign.rs
Patrick Walton b7bcd313ca
Cluster light probes using conservative spherical bounds. (#13746)
This commit allows the Bevy renderer to use the clustering
infrastructure for light probes (reflection probes and irradiance
volumes) on platforms where at least 3 storage buffers are available. On
such platforms (the vast majority), we stop performing brute-force
searches of light probes for each fragment and instead only search the
light probes with bounding spheres that intersect the current cluster.
This should dramatically improve scalability of irradiance volumes and
reflection probes.

The primary platform that doesn't support 3 storage buffers is WebGL 2,
and we continue using a brute-force search of light probes on that
platform, as the UBO that stores per-cluster indices is too small to fit
the light probe counts. Note, however, that that platform also doesn't
support bindless textures (indeed, it would be very odd for a platform
to support bindless textures but not SSBOs), so we only support one of
each type of light probe per drawcall there in the first place.
Consequently, this isn't a performance problem, as the search will only
have one light probe to consider. (In fact, clustering would probably
end up being a performance loss.)

Known potential improvements include:

1. We currently cull based on a conservative bounding sphere test and
not based on the oriented bounding box (OBB) of the light probe. This is
improvable, but in the interests of simplicity, I opted to keep the
bounding sphere test for now. The OBB improvement can be a follow-up.

2. This patch doesn't change the fact that each fragment only takes a
single light probe into account. Typical light probe implementations
detect the case in which multiple light probes cover the current
fragment and perform some sort of weighted blend between them. As the
light probe fetch function presently returns only a single light probe,
implementing that feature would require more code restructuring, so I
left it out for now. It can be added as a follow-up.

3. Light probe implementations typically have a falloff range. Although
this is a wanted feature in Bevy, this particular commit also doesn't
implement that feature, as it's out of scope.

4. This commit doesn't raise the maximum number of light probes past its
current value of 8 for each type. This should be addressed later, but
would possibly require more bindings on platforms with storage buffers,
which would increase this patch's complexity. Even without raising the
limit, this patch should constitute a significant performance
improvement for scenes that get anywhere close to this limit. In the
interest of keeping this patch small, I opted to leave raising the limit
to a follow-up.

## Changelog

### Changed

* Light probes (reflection probes and irradiance volumes) are now
clustered on most platforms, improving performance when many light
probes are present.

---------

Co-authored-by: Benjamin Brienen <Benjamin.Brienen@outlook.com>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
2024-12-05 13:07:10 +00:00

1155 lines
50 KiB
Rust

//! Assigning objects to clusters.
use bevy_ecs::{
entity::Entity,
query::{Has, With},
system::{Commands, Local, Query, Res, ResMut},
};
use bevy_math::{
ops::{self, sin_cos},
Mat4, UVec3, Vec2, Vec3, Vec3A, Vec3Swizzles as _, Vec4, Vec4Swizzles as _,
};
use bevy_render::{
camera::Camera,
primitives::{Aabb, Frustum, HalfSpace, Sphere},
render_resource::BufferBindingType,
renderer::RenderDevice,
view::{RenderLayers, ViewVisibility},
};
use bevy_transform::components::GlobalTransform;
use bevy_utils::{prelude::default, tracing::warn};
use crate::{
prelude::EnvironmentMapLight, ClusterConfig, ClusterFarZMode, Clusters, ExtractedPointLight,
GlobalVisibleClusterableObjects, LightProbe, PointLight, SpotLight, ViewClusterBindings,
VisibleClusterableObjects, VolumetricLight, CLUSTERED_FORWARD_STORAGE_BUFFER_COUNT,
MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS,
};
use super::ClusterableObjectOrderData;
const NDC_MIN: Vec2 = Vec2::NEG_ONE;
const NDC_MAX: Vec2 = Vec2::ONE;
const VEC2_HALF: Vec2 = Vec2::splat(0.5);
const VEC2_HALF_NEGATIVE_Y: Vec2 = Vec2::new(0.5, -0.5);
/// Data required for assigning objects to clusters.
#[derive(Clone, Debug)]
pub(crate) struct ClusterableObjectAssignmentData {
entity: Entity,
transform: GlobalTransform,
range: f32,
object_type: ClusterableObjectType,
render_layers: RenderLayers,
}
impl ClusterableObjectAssignmentData {
pub fn sphere(&self) -> Sphere {
Sphere {
center: self.transform.translation_vec3a(),
radius: self.range,
}
}
}
/// Data needed to assign objects to clusters that's specific to the type of
/// clusterable object.
#[derive(Clone, Copy, Debug)]
pub(crate) enum ClusterableObjectType {
/// Data needed to assign point lights to clusters.
PointLight {
/// Whether shadows are enabled for this point light.
///
/// This is used for sorting the light list.
shadows_enabled: bool,
/// Whether this light interacts with volumetrics.
///
/// This is used for sorting the light list.
volumetric: bool,
},
/// Data needed to assign spot lights to clusters.
SpotLight {
/// Whether shadows are enabled for this spot light.
///
/// This is used for sorting the light list.
shadows_enabled: bool,
/// Whether this light interacts with volumetrics.
///
/// This is used for sorting the light list.
volumetric: bool,
/// The outer angle of the light cone in radians.
outer_angle: f32,
},
/// Marks that the clusterable object is a reflection probe.
ReflectionProbe,
/// Marks that the clusterable object is an irradiance volume.
IrradianceVolume,
}
impl ClusterableObjectType {
/// Returns a tuple that can be sorted to obtain the order in which indices
/// to clusterable objects must be stored in the cluster offsets and counts
/// list.
///
/// Generally, we sort first by type, then, for lights, by whether shadows
/// are enabled (enabled before disabled), and then whether volumetrics are
/// enabled (enabled before disabled).
pub(crate) fn ordering(&self) -> (u8, bool, bool) {
match *self {
ClusterableObjectType::PointLight {
shadows_enabled,
volumetric,
} => (0, !shadows_enabled, !volumetric),
ClusterableObjectType::SpotLight {
shadows_enabled,
volumetric,
..
} => (1, !shadows_enabled, !volumetric),
ClusterableObjectType::ReflectionProbe => (2, false, false),
ClusterableObjectType::IrradianceVolume => (3, false, false),
}
}
/// Creates the [`ClusterableObjectType`] data for a point or spot light.
pub(crate) fn from_point_or_spot_light(
point_light: &ExtractedPointLight,
) -> ClusterableObjectType {
match point_light.spot_light_angles {
Some((_, outer_angle)) => ClusterableObjectType::SpotLight {
outer_angle,
shadows_enabled: point_light.shadows_enabled,
volumetric: point_light.volumetric,
},
None => ClusterableObjectType::PointLight {
shadows_enabled: point_light.shadows_enabled,
volumetric: point_light.volumetric,
},
}
}
}
// NOTE: Run this before update_point_light_frusta!
#[allow(clippy::too_many_arguments)]
pub(crate) fn assign_objects_to_clusters(
mut commands: Commands,
mut global_clusterable_objects: ResMut<GlobalVisibleClusterableObjects>,
mut views: Query<(
Entity,
&GlobalTransform,
&Camera,
&Frustum,
&ClusterConfig,
&mut Clusters,
Option<&RenderLayers>,
Option<&mut VisibleClusterableObjects>,
)>,
point_lights_query: Query<(
Entity,
&GlobalTransform,
&PointLight,
Option<&RenderLayers>,
Option<&VolumetricLight>,
&ViewVisibility,
)>,
spot_lights_query: Query<(
Entity,
&GlobalTransform,
&SpotLight,
Option<&RenderLayers>,
Option<&VolumetricLight>,
&ViewVisibility,
)>,
light_probes_query: Query<
(Entity, &GlobalTransform, Has<EnvironmentMapLight>),
With<LightProbe>,
>,
mut clusterable_objects: Local<Vec<ClusterableObjectAssignmentData>>,
mut cluster_aabb_spheres: Local<Vec<Option<Sphere>>>,
mut max_clusterable_objects_warning_emitted: Local<bool>,
render_device: Option<Res<RenderDevice>>,
) {
let Some(render_device) = render_device else {
return;
};
global_clusterable_objects.entities.clear();
clusterable_objects.clear();
// collect just the relevant query data into a persisted vec to avoid reallocating each frame
clusterable_objects.extend(
point_lights_query
.iter()
.filter(|(.., visibility)| visibility.get())
.map(
|(entity, transform, point_light, maybe_layers, volumetric, _visibility)| {
ClusterableObjectAssignmentData {
entity,
transform: GlobalTransform::from_translation(transform.translation()),
range: point_light.range,
object_type: ClusterableObjectType::PointLight {
shadows_enabled: point_light.shadows_enabled,
volumetric: volumetric.is_some(),
},
render_layers: maybe_layers.unwrap_or_default().clone(),
}
},
),
);
clusterable_objects.extend(
spot_lights_query
.iter()
.filter(|(.., visibility)| visibility.get())
.map(
|(entity, transform, spot_light, maybe_layers, volumetric, _visibility)| {
ClusterableObjectAssignmentData {
entity,
transform: *transform,
range: spot_light.range,
object_type: ClusterableObjectType::SpotLight {
outer_angle: spot_light.outer_angle,
shadows_enabled: spot_light.shadows_enabled,
volumetric: volumetric.is_some(),
},
render_layers: maybe_layers.unwrap_or_default().clone(),
}
},
),
);
let clustered_forward_buffer_binding_type =
render_device.get_supported_read_only_binding_type(CLUSTERED_FORWARD_STORAGE_BUFFER_COUNT);
let supports_storage_buffers = matches!(
clustered_forward_buffer_binding_type,
BufferBindingType::Storage { .. }
);
// Gather up light probes, but only if we're clustering them.
//
// UBOs aren't large enough to hold indices for light probes, so we can't
// cluster light probes on such platforms (mainly WebGL 2). Besides, those
// platforms typically lack bindless textures, so multiple light probes
// wouldn't be supported anyhow.
if supports_storage_buffers {
clusterable_objects.extend(light_probes_query.iter().map(
|(entity, transform, is_reflection_probe)| ClusterableObjectAssignmentData {
entity,
transform: *transform,
range: transform.radius_vec3a(Vec3A::ONE),
object_type: if is_reflection_probe {
ClusterableObjectType::ReflectionProbe
} else {
ClusterableObjectType::IrradianceVolume
},
render_layers: RenderLayers::default(),
},
));
}
if clusterable_objects.len() > MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS
&& !supports_storage_buffers
{
clusterable_objects.sort_by(|clusterable_object_1, clusterable_object_2| {
crate::clusterable_object_order(
ClusterableObjectOrderData {
entity: &clusterable_object_1.entity,
object_type: &clusterable_object_1.object_type,
},
ClusterableObjectOrderData {
entity: &clusterable_object_2.entity,
object_type: &clusterable_object_2.object_type,
},
)
});
// check each clusterable object against each view's frustum, keep only
// those that affect at least one of our views
let frusta: Vec<_> = views
.iter()
.map(|(_, _, _, frustum, _, _, _, _)| *frustum)
.collect();
let mut clusterable_objects_in_view_count = 0;
clusterable_objects.retain(|clusterable_object| {
// take one extra clusterable object to check if we should emit the warning
if clusterable_objects_in_view_count == MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS + 1 {
false
} else {
let clusterable_object_sphere = clusterable_object.sphere();
let clusterable_object_in_view = frusta
.iter()
.any(|frustum| frustum.intersects_sphere(&clusterable_object_sphere, true));
if clusterable_object_in_view {
clusterable_objects_in_view_count += 1;
}
clusterable_object_in_view
}
});
if clusterable_objects.len() > MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS
&& !*max_clusterable_objects_warning_emitted
{
warn!(
"MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS ({}) exceeded",
MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS
);
*max_clusterable_objects_warning_emitted = true;
}
clusterable_objects.truncate(MAX_UNIFORM_BUFFER_CLUSTERABLE_OBJECTS);
}
for (
view_entity,
camera_transform,
camera,
frustum,
config,
clusters,
maybe_layers,
mut visible_clusterable_objects,
) in &mut views
{
let view_layers = maybe_layers.unwrap_or_default();
let clusters = clusters.into_inner();
if matches!(config, ClusterConfig::None) {
if visible_clusterable_objects.is_some() {
commands
.entity(view_entity)
.remove::<VisibleClusterableObjects>();
}
clusters.clear();
continue;
}
let screen_size = match camera.physical_viewport_size() {
Some(screen_size) if screen_size.x != 0 && screen_size.y != 0 => screen_size,
_ => {
clusters.clear();
continue;
}
};
let mut requested_cluster_dimensions = config.dimensions_for_screen_size(screen_size);
let world_from_view = camera_transform.compute_matrix();
let view_from_world_scale = camera_transform.compute_transform().scale.recip();
let view_from_world_scale_max = view_from_world_scale.abs().max_element();
let view_from_world = world_from_view.inverse();
let is_orthographic = camera.clip_from_view().w_axis.w == 1.0;
let far_z = match config.far_z_mode() {
ClusterFarZMode::MaxClusterableObjectRange => {
let view_from_world_row_2 = view_from_world.row(2);
clusterable_objects
.iter()
.map(|object| {
-view_from_world_row_2.dot(object.transform.translation().extend(1.0))
+ object.range * view_from_world_scale.z
})
.reduce(f32::max)
.unwrap_or(0.0)
}
ClusterFarZMode::Constant(far) => far,
};
let first_slice_depth = match (is_orthographic, requested_cluster_dimensions.z) {
(true, _) => {
// NOTE: Based on glam's Mat4::orthographic_rh(), as used to calculate the orthographic projection
// matrix, we can calculate the projection's view-space near plane as follows:
// component 3,2 = r * near and 2,2 = r where r = 1.0 / (near - far)
// There is a caveat here that when calculating the projection matrix, near and far were swapped to give
// reversed z, consistent with the perspective projection. So,
// 3,2 = r * far and 2,2 = r where r = 1.0 / (far - near)
// rearranging r = 1.0 / (far - near), r * (far - near) = 1.0, r * far - 1.0 = r * near, near = (r * far - 1.0) / r
// = (3,2 - 1.0) / 2,2
(camera.clip_from_view().w_axis.z - 1.0) / camera.clip_from_view().z_axis.z
}
(false, 1) => config.first_slice_depth().max(far_z),
_ => config.first_slice_depth(),
};
let first_slice_depth = first_slice_depth * view_from_world_scale.z;
// NOTE: Ensure the far_z is at least as far as the first_depth_slice to avoid clustering problems.
let far_z = far_z.max(first_slice_depth);
let cluster_factors = crate::calculate_cluster_factors(
first_slice_depth,
far_z,
requested_cluster_dimensions.z as f32,
is_orthographic,
);
if config.dynamic_resizing() {
let mut cluster_index_estimate = 0.0;
for clusterable_object in &clusterable_objects {
let clusterable_object_sphere = clusterable_object.sphere();
// Check if the clusterable object is within the view frustum
if !frustum.intersects_sphere(&clusterable_object_sphere, true) {
continue;
}
// calculate a conservative aabb estimate of number of clusters affected by this light
// this overestimates index counts by at most 50% (and typically much less) when the whole light range is in view
// it can overestimate more significantly when light ranges are only partially in view
let (clusterable_object_aabb_min, clusterable_object_aabb_max) =
cluster_space_clusterable_object_aabb(
view_from_world,
view_from_world_scale,
camera.clip_from_view(),
&clusterable_object_sphere,
);
// since we won't adjust z slices we can calculate exact number of slices required in z dimension
let z_cluster_min = view_z_to_z_slice(
cluster_factors,
requested_cluster_dimensions.z,
clusterable_object_aabb_min.z,
is_orthographic,
);
let z_cluster_max = view_z_to_z_slice(
cluster_factors,
requested_cluster_dimensions.z,
clusterable_object_aabb_max.z,
is_orthographic,
);
let z_count =
z_cluster_min.max(z_cluster_max) - z_cluster_min.min(z_cluster_max) + 1;
// calculate x/y count using floats to avoid overestimating counts due to large initial tile sizes
let xy_min = clusterable_object_aabb_min.xy();
let xy_max = clusterable_object_aabb_max.xy();
// multiply by 0.5 to move from [-1,1] to [-0.5, 0.5], max extent of 1 in each dimension
let xy_count = (xy_max - xy_min)
* 0.5
* Vec2::new(
requested_cluster_dimensions.x as f32,
requested_cluster_dimensions.y as f32,
);
// add up to 2 to each axis to account for overlap
let x_overlap = if xy_min.x <= -1.0 { 0.0 } else { 1.0 }
+ if xy_max.x >= 1.0 { 0.0 } else { 1.0 };
let y_overlap = if xy_min.y <= -1.0 { 0.0 } else { 1.0 }
+ if xy_max.y >= 1.0 { 0.0 } else { 1.0 };
cluster_index_estimate +=
(xy_count.x + x_overlap) * (xy_count.y + y_overlap) * z_count as f32;
}
if cluster_index_estimate > ViewClusterBindings::MAX_INDICES as f32 {
// scale x and y cluster count to be able to fit all our indices
// we take the ratio of the actual indices over the index estimate.
// this is not guaranteed to be small enough due to overlapped tiles, but
// the conservative estimate is more than sufficient to cover the
// difference
let index_ratio = ViewClusterBindings::MAX_INDICES as f32 / cluster_index_estimate;
let xy_ratio = index_ratio.sqrt();
requested_cluster_dimensions.x =
((requested_cluster_dimensions.x as f32 * xy_ratio).floor() as u32).max(1);
requested_cluster_dimensions.y =
((requested_cluster_dimensions.y as f32 * xy_ratio).floor() as u32).max(1);
}
}
clusters.update(screen_size, requested_cluster_dimensions);
clusters.near = first_slice_depth;
clusters.far = far_z;
// NOTE: Maximum 4096 clusters due to uniform buffer size constraints
debug_assert!(
clusters.dimensions.x * clusters.dimensions.y * clusters.dimensions.z <= 4096
);
let view_from_clip = camera.clip_from_view().inverse();
for clusterable_objects in &mut clusters.clusterable_objects {
clusterable_objects.entities.clear();
clusterable_objects.counts = default();
}
let cluster_count =
(clusters.dimensions.x * clusters.dimensions.y * clusters.dimensions.z) as usize;
clusters
.clusterable_objects
.resize_with(cluster_count, VisibleClusterableObjects::default);
// initialize empty cluster bounding spheres
cluster_aabb_spheres.clear();
cluster_aabb_spheres.extend(core::iter::repeat(None).take(cluster_count));
// Calculate the x/y/z cluster frustum planes in view space
let mut x_planes = Vec::with_capacity(clusters.dimensions.x as usize + 1);
let mut y_planes = Vec::with_capacity(clusters.dimensions.y as usize + 1);
let mut z_planes = Vec::with_capacity(clusters.dimensions.z as usize + 1);
if is_orthographic {
let x_slices = clusters.dimensions.x as f32;
for x in 0..=clusters.dimensions.x {
let x_proportion = x as f32 / x_slices;
let x_pos = x_proportion * 2.0 - 1.0;
let view_x = clip_to_view(view_from_clip, Vec4::new(x_pos, 0.0, 1.0, 1.0)).x;
let normal = Vec3::X;
let d = view_x * normal.x;
x_planes.push(HalfSpace::new(normal.extend(d)));
}
let y_slices = clusters.dimensions.y as f32;
for y in 0..=clusters.dimensions.y {
let y_proportion = 1.0 - y as f32 / y_slices;
let y_pos = y_proportion * 2.0 - 1.0;
let view_y = clip_to_view(view_from_clip, Vec4::new(0.0, y_pos, 1.0, 1.0)).y;
let normal = Vec3::Y;
let d = view_y * normal.y;
y_planes.push(HalfSpace::new(normal.extend(d)));
}
} else {
let x_slices = clusters.dimensions.x as f32;
for x in 0..=clusters.dimensions.x {
let x_proportion = x as f32 / x_slices;
let x_pos = x_proportion * 2.0 - 1.0;
let nb = clip_to_view(view_from_clip, Vec4::new(x_pos, -1.0, 1.0, 1.0)).xyz();
let nt = clip_to_view(view_from_clip, Vec4::new(x_pos, 1.0, 1.0, 1.0)).xyz();
let normal = nb.cross(nt);
let d = nb.dot(normal);
x_planes.push(HalfSpace::new(normal.extend(d)));
}
let y_slices = clusters.dimensions.y as f32;
for y in 0..=clusters.dimensions.y {
let y_proportion = 1.0 - y as f32 / y_slices;
let y_pos = y_proportion * 2.0 - 1.0;
let nl = clip_to_view(view_from_clip, Vec4::new(-1.0, y_pos, 1.0, 1.0)).xyz();
let nr = clip_to_view(view_from_clip, Vec4::new(1.0, y_pos, 1.0, 1.0)).xyz();
let normal = nr.cross(nl);
let d = nr.dot(normal);
y_planes.push(HalfSpace::new(normal.extend(d)));
}
}
let z_slices = clusters.dimensions.z;
for z in 0..=z_slices {
let view_z = z_slice_to_view_z(first_slice_depth, far_z, z_slices, z, is_orthographic);
let normal = -Vec3::Z;
let d = view_z * normal.z;
z_planes.push(HalfSpace::new(normal.extend(d)));
}
let mut update_from_object_intersections = |visible_clusterable_objects: &mut Vec<
Entity,
>| {
for clusterable_object in &clusterable_objects {
// check if the clusterable light layers overlap the view layers
if !view_layers.intersects(&clusterable_object.render_layers) {
continue;
}
let clusterable_object_sphere = clusterable_object.sphere();
// Check if the clusterable object is within the view frustum
if !frustum.intersects_sphere(&clusterable_object_sphere, true) {
continue;
}
// NOTE: The clusterable object intersects the frustum so it
// must be visible and part of the global set
global_clusterable_objects
.entities
.insert(clusterable_object.entity);
visible_clusterable_objects.push(clusterable_object.entity);
// note: caching seems to be slower than calling twice for this aabb calculation
let (
clusterable_object_aabb_xy_ndc_z_view_min,
clusterable_object_aabb_xy_ndc_z_view_max,
) = cluster_space_clusterable_object_aabb(
view_from_world,
view_from_world_scale,
camera.clip_from_view(),
&clusterable_object_sphere,
);
let min_cluster = ndc_position_to_cluster(
clusters.dimensions,
cluster_factors,
is_orthographic,
clusterable_object_aabb_xy_ndc_z_view_min,
clusterable_object_aabb_xy_ndc_z_view_min.z,
);
let max_cluster = ndc_position_to_cluster(
clusters.dimensions,
cluster_factors,
is_orthographic,
clusterable_object_aabb_xy_ndc_z_view_max,
clusterable_object_aabb_xy_ndc_z_view_max.z,
);
let (min_cluster, max_cluster) =
(min_cluster.min(max_cluster), min_cluster.max(max_cluster));
// What follows is the Iterative Sphere Refinement algorithm from Just Cause 3
// Persson et al, Practical Clustered Shading
// http://newq.net/dl/pub/s2015_practical.pdf
// NOTE: A sphere under perspective projection is no longer a sphere. It gets
// stretched and warped, which prevents simpler algorithms from being correct
// as they often assume that the widest part of the sphere under projection is the
// center point on the axis of interest plus the radius, and that is not true!
let view_clusterable_object_sphere = Sphere {
center: Vec3A::from_vec4(
view_from_world * clusterable_object_sphere.center.extend(1.0),
),
radius: clusterable_object_sphere.radius * view_from_world_scale_max,
};
let spot_light_dir_sin_cos = match clusterable_object.object_type {
ClusterableObjectType::SpotLight { outer_angle, .. } => {
let (angle_sin, angle_cos) = sin_cos(outer_angle);
Some((
(view_from_world * clusterable_object.transform.back().extend(0.0))
.truncate()
.normalize(),
angle_sin,
angle_cos,
))
}
ClusterableObjectType::PointLight { .. }
| ClusterableObjectType::ReflectionProbe
| ClusterableObjectType::IrradianceVolume => None,
};
let clusterable_object_center_clip =
camera.clip_from_view() * view_clusterable_object_sphere.center.extend(1.0);
let object_center_ndc =
clusterable_object_center_clip.xyz() / clusterable_object_center_clip.w;
let cluster_coordinates = ndc_position_to_cluster(
clusters.dimensions,
cluster_factors,
is_orthographic,
object_center_ndc,
view_clusterable_object_sphere.center.z,
);
let z_center = if object_center_ndc.z <= 1.0 {
Some(cluster_coordinates.z)
} else {
None
};
let y_center = if object_center_ndc.y > 1.0 {
None
} else if object_center_ndc.y < -1.0 {
Some(clusters.dimensions.y + 1)
} else {
Some(cluster_coordinates.y)
};
for z in min_cluster.z..=max_cluster.z {
let mut z_object = view_clusterable_object_sphere.clone();
if z_center.is_none() || z != z_center.unwrap() {
// The z plane closer to the clusterable object has the
// larger radius circle where the light sphere
// intersects the z plane.
let z_plane = if z_center.is_some() && z < z_center.unwrap() {
z_planes[(z + 1) as usize]
} else {
z_planes[z as usize]
};
// Project the sphere to this z plane and use its radius as the radius of a
// new, refined sphere.
if let Some(projected) = project_to_plane_z(z_object, z_plane) {
z_object = projected;
} else {
continue;
}
}
for y in min_cluster.y..=max_cluster.y {
let mut y_object = z_object.clone();
if y_center.is_none() || y != y_center.unwrap() {
// The y plane closer to the clusterable object has
// the larger radius circle where the light sphere
// intersects the y plane.
let y_plane = if y_center.is_some() && y < y_center.unwrap() {
y_planes[(y + 1) as usize]
} else {
y_planes[y as usize]
};
// Project the refined sphere to this y plane and use its radius as the
// radius of a new, even more refined sphere.
if let Some(projected) =
project_to_plane_y(y_object, y_plane, is_orthographic)
{
y_object = projected;
} else {
continue;
}
}
// Loop from the left to find the first affected cluster
let mut min_x = min_cluster.x;
loop {
if min_x >= max_cluster.x
|| -get_distance_x(
x_planes[(min_x + 1) as usize],
y_object.center,
is_orthographic,
) + y_object.radius
> 0.0
{
break;
}
min_x += 1;
}
// Loop from the right to find the last affected cluster
let mut max_x = max_cluster.x;
loop {
if max_x <= min_x
|| get_distance_x(
x_planes[max_x as usize],
y_object.center,
is_orthographic,
) + y_object.radius
> 0.0
{
break;
}
max_x -= 1;
}
let mut cluster_index = ((y * clusters.dimensions.x + min_x)
* clusters.dimensions.z
+ z) as usize;
match clusterable_object.object_type {
ClusterableObjectType::SpotLight { .. } => {
let (view_light_direction, angle_sin, angle_cos) =
spot_light_dir_sin_cos.unwrap();
for x in min_x..=max_x {
// further culling for spot lights
// get or initialize cluster bounding sphere
let cluster_aabb_sphere =
&mut cluster_aabb_spheres[cluster_index];
let cluster_aabb_sphere =
if let Some(sphere) = cluster_aabb_sphere {
&*sphere
} else {
let aabb = compute_aabb_for_cluster(
first_slice_depth,
far_z,
clusters.tile_size.as_vec2(),
screen_size.as_vec2(),
view_from_clip,
is_orthographic,
clusters.dimensions,
UVec3::new(x, y, z),
);
let sphere = Sphere {
center: aabb.center,
radius: aabb.half_extents.length(),
};
*cluster_aabb_sphere = Some(sphere);
cluster_aabb_sphere.as_ref().unwrap()
};
// test -- based on https://bartwronski.com/2017/04/13/cull-that-cone/
let spot_light_offset = Vec3::from(
view_clusterable_object_sphere.center
- cluster_aabb_sphere.center,
);
let spot_light_dist_sq = spot_light_offset.length_squared();
let v1_len = spot_light_offset.dot(view_light_direction);
let distance_closest_point = (angle_cos
* (spot_light_dist_sq - v1_len * v1_len).sqrt())
- v1_len * angle_sin;
let angle_cull =
distance_closest_point > cluster_aabb_sphere.radius;
let front_cull = v1_len
> cluster_aabb_sphere.radius
+ clusterable_object.range * view_from_world_scale_max;
let back_cull = v1_len < -cluster_aabb_sphere.radius;
if !angle_cull && !front_cull && !back_cull {
// this cluster is affected by the spot light
clusters.clusterable_objects[cluster_index]
.entities
.push(clusterable_object.entity);
clusters.clusterable_objects[cluster_index]
.counts
.spot_lights += 1;
}
cluster_index += clusters.dimensions.z as usize;
}
}
ClusterableObjectType::PointLight { .. } => {
for _ in min_x..=max_x {
// all clusters within range are affected by point lights
clusters.clusterable_objects[cluster_index]
.entities
.push(clusterable_object.entity);
clusters.clusterable_objects[cluster_index]
.counts
.point_lights += 1;
cluster_index += clusters.dimensions.z as usize;
}
}
ClusterableObjectType::ReflectionProbe => {
// Reflection probes currently affect all
// clusters in their bounding sphere.
//
// TODO: Cull more aggressively based on the
// probe's OBB.
for _ in min_x..=max_x {
clusters.clusterable_objects[cluster_index]
.entities
.push(clusterable_object.entity);
clusters.clusterable_objects[cluster_index]
.counts
.reflection_probes += 1;
cluster_index += clusters.dimensions.z as usize;
}
}
ClusterableObjectType::IrradianceVolume => {
// Irradiance volumes currently affect all
// clusters in their bounding sphere.
//
// TODO: Cull more aggressively based on the
// probe's OBB.
for _ in min_x..=max_x {
clusters.clusterable_objects[cluster_index]
.entities
.push(clusterable_object.entity);
clusters.clusterable_objects[cluster_index]
.counts
.irradiance_volumes += 1;
cluster_index += clusters.dimensions.z as usize;
}
}
}
}
}
}
};
// reuse existing visible clusterable objects Vec, if it exists
if let Some(visible_clusterable_objects) = visible_clusterable_objects.as_mut() {
visible_clusterable_objects.entities.clear();
update_from_object_intersections(&mut visible_clusterable_objects.entities);
} else {
let mut entities = Vec::new();
update_from_object_intersections(&mut entities);
commands
.entity(view_entity)
.insert(VisibleClusterableObjects {
entities,
..Default::default()
});
}
}
}
#[allow(clippy::too_many_arguments)]
fn compute_aabb_for_cluster(
z_near: f32,
z_far: f32,
tile_size: Vec2,
screen_size: Vec2,
view_from_clip: Mat4,
is_orthographic: bool,
cluster_dimensions: UVec3,
ijk: UVec3,
) -> Aabb {
let ijk = ijk.as_vec3();
// Calculate the minimum and maximum points in screen space
let p_min = ijk.xy() * tile_size;
let p_max = p_min + tile_size;
let cluster_min;
let cluster_max;
if is_orthographic {
// Use linear depth slicing for orthographic
// Convert to view space at the cluster near and far planes
// NOTE: 1.0 is the near plane due to using reverse z projections
let mut p_min = screen_to_view(screen_size, view_from_clip, p_min, 0.0).xyz();
let mut p_max = screen_to_view(screen_size, view_from_clip, p_max, 0.0).xyz();
// calculate cluster depth using z_near and z_far
p_min.z = -z_near + (z_near - z_far) * ijk.z / cluster_dimensions.z as f32;
p_max.z = -z_near + (z_near - z_far) * (ijk.z + 1.0) / cluster_dimensions.z as f32;
cluster_min = p_min.min(p_max);
cluster_max = p_min.max(p_max);
} else {
// Convert to view space at the near plane
// NOTE: 1.0 is the near plane due to using reverse z projections
let p_min = screen_to_view(screen_size, view_from_clip, p_min, 1.0);
let p_max = screen_to_view(screen_size, view_from_clip, p_max, 1.0);
let z_far_over_z_near = -z_far / -z_near;
let cluster_near = if ijk.z == 0.0 {
0.0
} else {
-z_near
* ops::powf(
z_far_over_z_near,
(ijk.z - 1.0) / (cluster_dimensions.z - 1) as f32,
)
};
// NOTE: This could be simplified to:
// cluster_far = cluster_near * z_far_over_z_near;
let cluster_far = if cluster_dimensions.z == 1 {
-z_far
} else {
-z_near * ops::powf(z_far_over_z_near, ijk.z / (cluster_dimensions.z - 1) as f32)
};
// Calculate the four intersection points of the min and max points with the cluster near and far planes
let p_min_near = line_intersection_to_z_plane(Vec3::ZERO, p_min.xyz(), cluster_near);
let p_min_far = line_intersection_to_z_plane(Vec3::ZERO, p_min.xyz(), cluster_far);
let p_max_near = line_intersection_to_z_plane(Vec3::ZERO, p_max.xyz(), cluster_near);
let p_max_far = line_intersection_to_z_plane(Vec3::ZERO, p_max.xyz(), cluster_far);
cluster_min = p_min_near.min(p_min_far).min(p_max_near.min(p_max_far));
cluster_max = p_min_near.max(p_min_far).max(p_max_near.max(p_max_far));
}
Aabb::from_min_max(cluster_min, cluster_max)
}
// NOTE: Keep in sync as the inverse of view_z_to_z_slice above
fn z_slice_to_view_z(
near: f32,
far: f32,
z_slices: u32,
z_slice: u32,
is_orthographic: bool,
) -> f32 {
if is_orthographic {
return -near - (far - near) * z_slice as f32 / z_slices as f32;
}
// Perspective
if z_slice == 0 {
0.0
} else {
-near * ops::powf(far / near, (z_slice - 1) as f32 / (z_slices - 1) as f32)
}
}
fn ndc_position_to_cluster(
cluster_dimensions: UVec3,
cluster_factors: Vec2,
is_orthographic: bool,
ndc_p: Vec3,
view_z: f32,
) -> UVec3 {
let cluster_dimensions_f32 = cluster_dimensions.as_vec3();
let frag_coord = (ndc_p.xy() * VEC2_HALF_NEGATIVE_Y + VEC2_HALF).clamp(Vec2::ZERO, Vec2::ONE);
let xy = (frag_coord * cluster_dimensions_f32.xy()).floor();
let z_slice = view_z_to_z_slice(
cluster_factors,
cluster_dimensions.z,
view_z,
is_orthographic,
);
xy.as_uvec2()
.extend(z_slice)
.clamp(UVec3::ZERO, cluster_dimensions - UVec3::ONE)
}
/// Calculate bounds for the clusterable object using a view space aabb.
///
/// Returns a `(Vec3, Vec3)` containing minimum and maximum with
/// `X` and `Y` in normalized device coordinates with range `[-1, 1]`
/// `Z` in view space, with range `[-inf, -f32::MIN_POSITIVE]`
fn cluster_space_clusterable_object_aabb(
view_from_world: Mat4,
view_from_world_scale: Vec3,
clip_from_view: Mat4,
clusterable_object_sphere: &Sphere,
) -> (Vec3, Vec3) {
let clusterable_object_aabb_view = Aabb {
center: Vec3A::from_vec4(view_from_world * clusterable_object_sphere.center.extend(1.0)),
half_extents: Vec3A::from(clusterable_object_sphere.radius * view_from_world_scale.abs()),
};
let (mut clusterable_object_aabb_view_min, mut clusterable_object_aabb_view_max) = (
clusterable_object_aabb_view.min(),
clusterable_object_aabb_view.max(),
);
// Constrain view z to be negative - i.e. in front of the camera
// When view z is >= 0.0 and we're using a perspective projection, bad things happen.
// At view z == 0.0, ndc x,y are mathematically undefined. At view z > 0.0, i.e. behind the camera,
// the perspective projection flips the directions of the axes. This breaks assumptions about
// use of min/max operations as something that was to the left in view space is now returning a
// coordinate that for view z in front of the camera would be on the right, but at view z behind the
// camera is on the left. So, we just constrain view z to be < 0.0 and necessarily in front of the camera.
clusterable_object_aabb_view_min.z = clusterable_object_aabb_view_min.z.min(-f32::MIN_POSITIVE);
clusterable_object_aabb_view_max.z = clusterable_object_aabb_view_max.z.min(-f32::MIN_POSITIVE);
// Is there a cheaper way to do this? The problem is that because of perspective
// the point at max z but min xy may be less xy in screenspace, and similar. As
// such, projecting the min and max xy at both the closer and further z and taking
// the min and max of those projected points addresses this.
let (
clusterable_object_aabb_view_xymin_near,
clusterable_object_aabb_view_xymin_far,
clusterable_object_aabb_view_xymax_near,
clusterable_object_aabb_view_xymax_far,
) = (
clusterable_object_aabb_view_min,
clusterable_object_aabb_view_min
.xy()
.extend(clusterable_object_aabb_view_max.z),
clusterable_object_aabb_view_max
.xy()
.extend(clusterable_object_aabb_view_min.z),
clusterable_object_aabb_view_max,
);
let (
clusterable_object_aabb_clip_xymin_near,
clusterable_object_aabb_clip_xymin_far,
clusterable_object_aabb_clip_xymax_near,
clusterable_object_aabb_clip_xymax_far,
) = (
clip_from_view * clusterable_object_aabb_view_xymin_near.extend(1.0),
clip_from_view * clusterable_object_aabb_view_xymin_far.extend(1.0),
clip_from_view * clusterable_object_aabb_view_xymax_near.extend(1.0),
clip_from_view * clusterable_object_aabb_view_xymax_far.extend(1.0),
);
let (
clusterable_object_aabb_ndc_xymin_near,
clusterable_object_aabb_ndc_xymin_far,
clusterable_object_aabb_ndc_xymax_near,
clusterable_object_aabb_ndc_xymax_far,
) = (
clusterable_object_aabb_clip_xymin_near.xyz() / clusterable_object_aabb_clip_xymin_near.w,
clusterable_object_aabb_clip_xymin_far.xyz() / clusterable_object_aabb_clip_xymin_far.w,
clusterable_object_aabb_clip_xymax_near.xyz() / clusterable_object_aabb_clip_xymax_near.w,
clusterable_object_aabb_clip_xymax_far.xyz() / clusterable_object_aabb_clip_xymax_far.w,
);
let (clusterable_object_aabb_ndc_min, clusterable_object_aabb_ndc_max) = (
clusterable_object_aabb_ndc_xymin_near
.min(clusterable_object_aabb_ndc_xymin_far)
.min(clusterable_object_aabb_ndc_xymax_near)
.min(clusterable_object_aabb_ndc_xymax_far),
clusterable_object_aabb_ndc_xymin_near
.max(clusterable_object_aabb_ndc_xymin_far)
.max(clusterable_object_aabb_ndc_xymax_near)
.max(clusterable_object_aabb_ndc_xymax_far),
);
// clamp to ndc coords without depth
let (aabb_min_ndc, aabb_max_ndc) = (
clusterable_object_aabb_ndc_min.xy().clamp(NDC_MIN, NDC_MAX),
clusterable_object_aabb_ndc_max.xy().clamp(NDC_MIN, NDC_MAX),
);
// pack unadjusted z depth into the vecs
(
aabb_min_ndc.extend(clusterable_object_aabb_view_min.z),
aabb_max_ndc.extend(clusterable_object_aabb_view_max.z),
)
}
// Calculate the intersection of a ray from the eye through the view space position to a z plane
fn line_intersection_to_z_plane(origin: Vec3, p: Vec3, z: f32) -> Vec3 {
let v = p - origin;
let t = (z - Vec3::Z.dot(origin)) / Vec3::Z.dot(v);
origin + t * v
}
// NOTE: Keep in sync with bevy_pbr/src/render/pbr.wgsl
fn view_z_to_z_slice(
cluster_factors: Vec2,
z_slices: u32,
view_z: f32,
is_orthographic: bool,
) -> u32 {
let z_slice = if is_orthographic {
// NOTE: view_z is correct in the orthographic case
((view_z - cluster_factors.x) * cluster_factors.y).floor() as u32
} else {
// NOTE: had to use -view_z to make it positive else log(negative) is nan
(ops::ln(-view_z) * cluster_factors.x - cluster_factors.y + 1.0) as u32
};
// NOTE: We use min as we may limit the far z plane used for clustering to be closer than
// the furthest thing being drawn. This means that we need to limit to the maximum cluster.
z_slice.min(z_slices - 1)
}
fn clip_to_view(view_from_clip: Mat4, clip: Vec4) -> Vec4 {
let view = view_from_clip * clip;
view / view.w
}
fn screen_to_view(screen_size: Vec2, view_from_clip: Mat4, screen: Vec2, ndc_z: f32) -> Vec4 {
let tex_coord = screen / screen_size;
let clip = Vec4::new(
tex_coord.x * 2.0 - 1.0,
(1.0 - tex_coord.y) * 2.0 - 1.0,
ndc_z,
1.0,
);
clip_to_view(view_from_clip, clip)
}
// NOTE: This exploits the fact that a x-plane normal has only x and z components
fn get_distance_x(plane: HalfSpace, point: Vec3A, is_orthographic: bool) -> f32 {
if is_orthographic {
point.x - plane.d()
} else {
// Distance from a point to a plane:
// signed distance to plane = (nx * px + ny * py + nz * pz + d) / n.length()
// NOTE: For a x-plane, ny and d are 0 and we have a unit normal
// = nx * px + nz * pz
plane.normal_d().xz().dot(point.xz())
}
}
// NOTE: This exploits the fact that a z-plane normal has only a z component
fn project_to_plane_z(z_object: Sphere, z_plane: HalfSpace) -> Option<Sphere> {
// p = sphere center
// n = plane normal
// d = n.p if p is in the plane
// NOTE: For a z-plane, nx and ny are both 0
// d = px * nx + py * ny + pz * nz
// = pz * nz
// => pz = d / nz
let z = z_plane.d() / z_plane.normal_d().z;
let distance_to_plane = z - z_object.center.z;
if distance_to_plane.abs() > z_object.radius {
return None;
}
Some(Sphere {
center: Vec3A::from(z_object.center.xy().extend(z)),
// hypotenuse length = radius
// pythagoras = (distance to plane)^2 + b^2 = radius^2
radius: (z_object.radius * z_object.radius - distance_to_plane * distance_to_plane).sqrt(),
})
}
// NOTE: This exploits the fact that a y-plane normal has only y and z components
fn project_to_plane_y(
y_object: Sphere,
y_plane: HalfSpace,
is_orthographic: bool,
) -> Option<Sphere> {
let distance_to_plane = if is_orthographic {
y_plane.d() - y_object.center.y
} else {
-y_object.center.yz().dot(y_plane.normal_d().yz())
};
if distance_to_plane.abs() > y_object.radius {
return None;
}
Some(Sphere {
center: y_object.center + distance_to_plane * y_plane.normal(),
radius: (y_object.radius * y_object.radius - distance_to_plane * distance_to_plane).sqrt(),
})
}