Struct EllMatrix 

pub struct EllMatrix<T: Scalar> { /* private fields */ }

ELLPACK sparse matrix format.

Efficient for:

• GPU computation
• Vectorized operations
• Matrices with uniform row lengths

Storage

Each row stores exactly width entries. The width is typically the maximum number of non-zeros in any row. Rows with fewer non-zeros are padded with zeros and a special “invalid” column index (usually ncols or usize::MAX).

Example

use oxiblas_sparse::EllMatrix;

// Create a sparse matrix:
// [1 2 0 0]
// [0 3 4 0]
// [5 0 0 6]
let width = 2; // max 2 non-zeros per row
let data = vec![
    vec![1.0, 2.0],  // row 0
    vec![3.0, 4.0],  // row 1
    vec![5.0, 6.0],  // row 2
];
let indices = vec![
    vec![0, 1],  // row 0
    vec![1, 2],  // row 1
    vec![0, 3],  // row 2
];

let ell = EllMatrix::new(3, 4, width, data, indices).unwrap();
assert_eq!(ell.width(), 2);

Implementations

impl<T: Scalar + Clone> EllMatrix<T>

pub fn new( nrows: usize, ncols: usize, width: usize, data: Vec<Vec<T>>, indices: Vec<Vec<usize>>, ) -> Result<Self, EllError>

Creates a new ELL matrix from raw components.

Arguments

• nrows - Number of rows
• ncols - Number of columns
• width - Maximum non-zeros per row
• data - Data array (nrows × width)
• indices - Column indices array (nrows × width)

Errors

Returns an error if the input is invalid.

pub unsafe fn new_unchecked( nrows: usize, ncols: usize, width: usize, data: Vec<Vec<T>>, indices: Vec<Vec<usize>>, ) -> Self

Creates an ELL matrix without validation (unsafe but faster).

Safety

The caller must ensure:

• data.len() == nrows and each row has length width
• indices.len() == nrows and each row has length width
• All valid column indices are < ncols

pub fn zeros(nrows: usize, ncols: usize) -> Self

Creates an empty ELL matrix with given dimensions.

pub fn eye(n: usize) -> Self

where T: Field,

Creates an identity matrix in ELL format.

pub fn nrows(&self) -> usize

Returns the number of rows.

pub fn ncols(&self) -> usize

Returns the number of columns.

pub fn shape(&self) -> (usize, usize)

Returns the shape (nrows, ncols).

pub fn width(&self) -> usize

Returns the width (max non-zeros per row).

pub fn nnz(&self) -> usize

where T: Field,

Returns the number of non-zero elements.

Note: This counts actual non-zeros, not stored values.

pub fn nstored(&self) -> usize

Returns the total stored values (including padding).

pub fn efficiency(&self) -> f64

where T: Field,

Returns the storage efficiency (nnz / nstored).

pub fn data(&self) -> &[Vec<T>]

Returns a reference to the data array.

pub fn data_mut(&mut self) -> &mut [Vec<T>]

Returns a mutable reference to the data array.

pub fn indices(&self) -> &[Vec<usize>]

Returns a reference to the indices array.

pub fn get(&self, row: usize, col: usize) -> Option<&T>

Gets the value at (row, col), returning None if not present.

pub fn get_or_zero(&self, row: usize, col: usize) -> T

where T: Field,

Gets the value at (row, col), returning zero if not present.

pub fn row_iter(&self, row: usize) -> impl Iterator<Item = (usize, &T)>

Returns an iterator over non-zeros in a row as (col, value).

pub fn iter(&self) -> impl Iterator<Item = (usize, usize, &T)> + '_

Returns an iterator over all non-zeros as (row, col, value).

pub fn matvec(&self, x: &[T], y: &mut [T])

where T: Field,

Matrix-vector product: y = A * x.

pub fn mul_vec(&self, x: &[T]) -> Vec<T>

where T: Field,

Matrix-vector product returning a new vector: y = A * x.

pub fn to_csr(&self) -> CsrMatrix<T>

where T: Field,

Converts to CSR format.

pub fn to_dense(&self) -> Mat<T>

where T: Field + Zeroable,

Converts to dense matrix.

pub fn from_dense(dense: &MatRef<'_, T>, max_width: Option<usize>) -> Self

where T: Field,

Creates an ELL matrix from a dense matrix.

Arguments

• dense - Source dense matrix
• max_width - Maximum width (if None, uses actual max non-zeros per row)

pub fn from_csr( csr: &CsrMatrix<T>, max_width: Option<usize>, ) -> Result<Self, EllError>

where T: Field,

Creates an ELL matrix from CSR format.

Arguments

• csr - Source CSR matrix
• max_width - Maximum width (if None, uses actual max non-zeros per row)

pub fn scale(&mut self, alpha: T)

Scales all values by a scalar.

pub fn scaled(&self, alpha: T) -> Self

Returns a scaled copy of this matrix.

pub fn transpose(&self) -> Self

where T: Field,

Returns the transpose of this matrix.

Note: This is less efficient than CSR/CSC transpose.

Trait Implementations

impl<T: Clone + Scalar> Clone for EllMatrix<T>

fn clone(&self) -> EllMatrix<T>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T: Debug + Scalar> Debug for EllMatrix<T>

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

Formats the value using the given formatter.