Struct RenderStats 

pub struct RenderStats {
    pub draw_calls: u32,
    pub objects: u32,
    pub skinned_visible: u32,
    pub skinned_pool_free: u32,
    pub gpu_frame_us: u32,
    pub vram_bytes: u64,
    pub transient_pool_bytes: u64,
    pub pass_times_us: [PassTiming; 32],
    pub auto_exposure_ev: Option<f32>,
    pub max_edr: Option<f32>,
}

Per-frame render-backend statistics.

Fields

draw_calls: u32

CPU-issued geometry draw calls this frame: the shadow, main, and composite + text passes. The optional screen-space-effect passes (SSR, SSAO, TAA, bloom) issue a fixed handful of fullscreen draws and are not counted here – gpu_frame_us still covers their GPU time.

objects: u32

Renderable objects in the scene this frame: static draw objects, every instanced-cluster instance, and skinned meshes.

skinned_visible: u32

Visible skinned objects this frame: the authored skinned meshes plus any live runtime-spawned instances, excluding the hidden pre-reserved instance-pool slots. Unlike objects (which counts the whole pre-reserved pool and so stays flat across skinned spawn/despawn), this tracks the live count, so a spawn bumps it and a despawn drops it.

skinned_pool_free: u32

Free slots remaining in the pre-reserved skinned instance pool across all templates. Drains by one when a skinned instance spawns and refills when one despawns, so a probe can watch the free-list recycle directly.

gpu_frame_us: u32

GPU execution time of the most recently completed frame, in microseconds. Reported one or more frames late: the GPU timestamps are only known once that frame’s command buffer completion handler fires.

vram_bytes: u64

Bytes of GPU memory currently allocated by the render device. On unified-memory hardware (Apple Silicon) this is the device’s share of system memory rather than dedicated VRAM.

transient_pool_bytes: u64

Bytes the render graph’s transient pool holds: the aliased footprint of the slots backing the graph-owned transients. Part of vram_bytes, which is a device-wide total; carried separately so the shared memory ledger can attribute it rather than leaving it in the unaccounted remainder.

pass_times_us: [PassTiming; 32]

Per-pass GPU microseconds for the most recently completed frame. Filled by the active backend only when its GPU supports timestamp sampling; otherwise every slot stays at the default ("", 0). Slot order is backend-defined and stable for the process lifetime.

auto_exposure_ev: Option<f32>

Current adapted exposure value (EV) from the auto-exposure EMA. None when the world did not opt in to auto-exposure (or the active backend has not yet wired the readout). The StatHud overlay reads this to render the on-screen EV chip; downstream consumers can map it to an exposure multiplier via 2^ev.

max_edr: Option<f32>

Active display’s reported maximum extended-range colour-component multiplier when the renderer is on the HDR path. Some(2.0) on a typical HDR400 panel, Some(8.0+) on HDR1000-class panels; None on SDR (because the world disabled HDR, the platform fell back, or the active backend has not wired the readout). The StatHud overlay reads this to render the on-screen EDR chip; the value is also the linear scaling factor between SDR reference white and the panel’s peak brightness, so a colour-grading consumer can interpret it directly.

Trait Implementations

impl Clone for RenderStats

fn clone(&self) -> RenderStats

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for RenderStats

impl Debug for RenderStats

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for RenderStats

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for RenderStats

fn eq(&self, other: &RenderStats) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for RenderStats