Crate gizmo_renderer 

GPU mesh rendering for the Gizmo engine (wgpu pipelines, materials, instancing).

Frustum culling (CPU-side, before instancing)

The renderer does not iterate entities; your render loop should build the instance list. Before pushing each InstanceRaw, skip meshes outside the camera frustum using the same view * projection matrix you upload in SceneUniforms:

use gizmo_renderer::{Frustum, visible_in_frustum};
let frustum = Frustum::from_matrix(&view_proj);
if !visible_in_frustum(&frustum, &model_matrix, &mesh.bounds) {
    continue;
}

Mesh carries a local-space Aabb (bounds); visible_in_frustum transforms it by the instance model matrix and tests against the six planes. This pairs with batched draw(vertex_range, instance_start..instance_end) so culled instances are never written to the instance buffer. The demo, gizmo-studio, and gizmo::systems::default_render_pass pipelines already apply this pattern.

Implementation: frustum_cull re-exports Frustum and helpers from gizmo-math.

Re-exports

pub use frustum_cull::visible_in_frustum;

pub use web_profile::PostProcessLevel;

pub use web_profile::ShadowQuality;

pub use web_profile::WebProfile;

pub use animation::AnimationClip;

pub use animation::Keyframe;

pub use animation::SkeletonHierarchy;

pub use animation::SkeletonJoint;

pub use animation::Track;

pub use animation_state_machine::ActiveBlend;

pub use animation_state_machine::AnimationState;

pub use animation_state_machine::AnimationStateMachine;

pub use animation_state_machine::AnimationTransition;

pub use animation_system::animation_state_machine_update_system;

pub use animation_system::animation_update_system;

pub use animation_system::decompose_mat4;

pub use asset::decode_obj_vertices_for_async;

pub use asset::decode_rgba_image_file;

pub use asset::AssetManager;

pub use asset::GltfNodeData;

pub use async_assets::AsyncAssetLoader;

pub use async_assets::CompletedAsyncLoads;

pub use async_assets::GltfImportCompletion;

pub use async_assets::GltfImportError;

pub use async_assets::ObjLoadCompletion;

pub use async_assets::TextureReloadCompletion;

pub use components::Camera;

pub use components::Camera2D;

pub use components::DirectionalLight;

pub use components::LodGroup;

pub use components::LodLevel;

pub use components::Material;

pub use components::Mesh;

pub use components::MeshRenderer;

pub use components::PointLight;

pub use components::SpotLight;

pub use components::Sprite;

pub use csm::cascade_split_distances;

pub use csm::directional_cascade_view_projs;

pub use csm::CASCADE_COUNT;

pub use csm::SHADOW_MAP_RES;

pub use debug_renderer::GizmoRendererSystem;

pub use debug_renderer::Gizmos;

pub use decal::DecalState;

pub use deferred::DeferredState;

pub use gi::LightProbe;

pub use gi::ProbeGrid;

pub use gi::SHCoeffs;

pub use gpu_cull::DrawIndirectArgs;

pub use gpu_cull::GpuCullState;

pub use gpu_cull::MeshBoundsRaw;

pub use gpu_types::InstanceRaw;

pub use gpu_types::LightData;

pub use gpu_types::PostProcessUniforms;

pub use gpu_types::SceneUniforms;

pub use gpu_types::ShadowVsUniform;

pub use gpu_types::Vertex;

pub use hot_reload::AssetWatcher;

pub use pipeline::SceneState;

pub use post_process::PostProcessState;

pub use renderer::RenderContext;

pub use renderer::Renderer;

pub use ssao::SsaoParams;

pub use ssao::SsaoState;

pub use ssgi::SsgiState;

pub use taa::TaaState;

pub use fxaa::FxaaState;

Modules

animation

animation_state_machine

animation_system

asset

asset_loading

async_assets

Background-thread decoding for textures, OBJ, and GLTF import (disk + parse). GPU upload and AssetManager updates must run on the main thread — call AsyncAssetLoader::drain_completed each frame and then upload via AssetManager.

components

csm

Cascaded shadow maps (CSM) for directional lights: split the camera depth range into several cascades, each rendered to a layer of a depth2d_array with its own light orthographic projection (tighter texel density near the camera).

debug_renderer

decal

deferred

frustum_cull

CPU-side frustum culling before filling the instance buffer.

fxaa

FXAA (Fast Approximate Anti-Aliasing) — Post-Processing Pass

gi

Spherical Harmonics (SH) Probe — Global Illumination altyapısı

gpu_cull

gpu_fluid

gpu_particles

gpu_physics

gpu_types

hot_reload

Dosya izleme (AssetWatcher) ile hot-reload.

pipeline

post_process

renderer

ssao

ssgi

ssr

taa

volumetric

web_profile

Structs

Frustum

Six-plane view frustum extracted from a combined Projection × View matrix.