Struct RenderOptions 

pub struct RenderOptions {
    pub backend: BackendPreference,
    pub gpu: GpuRendererSettings,
    pub clear_sky: u32,
    pub gpu_mip_scan_dist: f32,
    pub gpu_chunk_upload_budget: u32,
    pub gpu_clip_upload_budget: u32,
    pub cpu_max_grid_vsid: u32,
    pub cpu_render_threads: usize,
}

Construction-time options for SceneRenderer::new.

Fields

backend: BackendPreference

Which backend to build (QE.7b — replaces want_gpu: bool; migrate want_gpu: true → BackendPreference::PreferGpu, false → BackendPreference::Cpu).

gpu: GpuRendererSettings

Settings forwarded to roxlap_gpu::GpuRenderer when the GPU backend is selected.

clear_sky: u32

Packed 0x00RRGGBB (alpha ignored) the empty/clear frame fills with until a scene render lands. Also the CPU sky-miss colour default if a frame supplies none.

gpu_mip_scan_dist: f32

GPU scene-grid LOD scan distance in world units (GPU.11.1): rays farther than this from the camera sample coarser chunk mips. QE.2a — construction-time here (was a per-frame FrameParams field); the ROXLAP_GPU_MIP_SCAN_DIST env var still overrides it as a user-side escape hatch. Ignored by the CPU backend (its mip ladder comes from OpticastSettings::mip_scan_dist).

gpu_chunk_upload_budget: u32

GPU per-frame dirty-chunk install budget (QE.2c — was env-only ROXLAP_GPU_CHUNK_BUDGET, which still overrides). Each chunk install issues several queue.write_buffer calls; a big batch can exhaust the device staging pool (seen as egui-wgpu panics on Mesa/NVK while flying fast). Small budget = bounded per-frame writes, more terrain pop-in when moving quickly. 0 = unbounded.

gpu_clip_upload_budget: u32

GPU frames-per-staging-flush while registering a flipbook clip (QE.2c — was env-only ROXLAP_GPU_CLIP_BUDGET, which still overrides). 0 = unbounded.

cpu_max_grid_vsid: u32

👎Deprecated since 0.22.0:

ignored — the DDA renderer replaced the strip-parallel opticast and needs no scratch-pool pre-sizing

Unused. Sized the strip-parallel opticast’s per-frame scratch pool; the per-pixel DDA renderer that replaced it needs no pre-sizing, so the value is ignored.

cpu_render_threads: usize

👎Deprecated since 0.22.0:

ignored — the DDA renderer parallelises over the rayon pool; bound it with RAYON_NUM_THREADS

Unused. Set the strip-parallel opticast’s thread count; the DDA renderer parallelises internally over the rayon pool (bound it with RAYON_NUM_THREADS if needed), so the value is ignored.

Trait Implementations

impl Default for RenderOptions

fn default() -> Self

Returns the “default value” for a type.