Struct DcSvdConfig 

pub struct DcSvdConfig {
    pub crossover_size: usize,
    pub compute_u: bool,
    pub compute_vt: bool,
    pub use_divide_conquer: bool,
    pub bidiagonalization: bool,
    pub deflation_tol: f64,
    pub n_threshold: usize,
}

Divide-and-conquer SVD configuration.

Fields

crossover_size: usize

Switch to QR iteration below this matrix size (default: 25).

compute_u: bool

Whether to compute U (left singular vectors).

compute_vt: bool

Whether to compute V^T (right singular vectors transposed).

use_divide_conquer: bool

Whether the divide-and-conquer algorithm is active (true for n >= n_threshold).

bidiagonalization: bool

Whether to use Householder bidiagonalization before D&C (true for n >= 256).

deflation_tol: f64

Deflation tolerance: n as f64 * eps where eps = machine epsilon.

n_threshold: usize

Minimum matrix size for the D&C path (default: 1024).

Implementations

impl DcSvdConfig

pub fn for_gpu(n: usize) -> Self

Creates a GPU-tuned configuration for a matrix of dimension n.

Enables divide-and-conquer for n >= 1024, bidiagonalization for n >= 256, and sets the deflation tolerance to n x epsilon where epsilon = 2.22e-16 (f64 epsilon).

Trait Implementations

impl Clone for DcSvdConfig

fn clone(&self) -> DcSvdConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DcSvdConfig

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

Formats the value using the given formatter.

impl Default for DcSvdConfig

fn default() -> Self

Returns the “default value” for a type.