Struct TileConfig 

pub struct TileConfig {
    pub core: usize,
    pub halo: usize,
}

Tiling parameters for tiled_field_inversion.

Fields

core: usize

Output core size per axis — the region each patch contributes to the result.

halo: usize

Context margin (voxels) included on every side of each patch. Larger halos reduce tile-boundary artifacts at the cost of more work per patch.

Trait Implementations

impl Clone for TileConfig

fn clone(&self) -> TileConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for TileConfig

impl Debug for TileConfig

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

Formats the value using the given formatter.

impl Default for TileConfig

fn default() -> Self

128³ cores with an 8-voxel halo → 144³ patches (~0.4 GB of f32 activations for a typical net, comfortably within a conservative ~2 GB WASM budget). Empirically the halo barely affects accuracy here — the tiling error is dominated by global low-frequency drift (a patch can’t see distant susceptibility sources), not tile-boundary seams — so this favours the large core (few patches, ~4× less overlap compute than 64³/32) over a big halo. On real 3 T data this matched whole-volume xQSM at r≈0.94 in ~1/4 the time.