Struct SparseMatrixGpu 

pub struct SparseMatrixGpu {
    pub n: usize,
    pub nnz: usize,
    pub row_ptr: Vec<usize>,
    pub col_idx: Vec<usize>,
    pub values: Vec<f64>,
}

Compressed Sparse Row (CSR) matrix stored in GPU-friendly layout.

Rows have entries in col_idx[row_ptr[i]..row_ptr[i+1]] with corresponding values in values[row_ptr[i]..row_ptr[i+1]].

Fields

n: usize

Number of rows (and columns — assumed square).

nnz: usize

Number of non-zero entries.

row_ptr: Vec<usize>

Row pointer array (length n+1).

col_idx: Vec<usize>

Column index array (length nnz).

values: Vec<f64>

Value array (length nnz).

Implementations

impl SparseMatrixGpu

pub fn new(n: usize) -> Self

Create a new empty n×n sparse matrix.

pub fn from_csr( n: usize, row_ptr: Vec<usize>, col_idx: Vec<usize>, values: Vec<f64>, ) -> Self

Construct from pre-built CSR arrays.

Panics

Panics in debug mode if row_ptr.len() != n + 1.

pub fn add_entry(&mut self, row: usize, col: usize, val: f64)

Append a single entry (row, col, val) in COO style.

This invalidates the CSR structure until finalize is called.

pub fn finalize(&mut self)

Convert count-per-row in row_ptr to proper CSR offsets.

Must be called after a series of add_entry calls and before any SpMV or solver operations.

pub fn nnz(&self) -> usize

Return the number of non-zero entries.

pub fn rows(&self) -> usize

Return the number of rows.

pub fn diagonal(&self) -> Vec<f64>

Extract the main diagonal as a dense vector.

Missing diagonal entries are filled with zero.

pub fn is_symmetric(&self) -> bool

Check whether the matrix is structurally symmetric (A[i,j] ↔ A[j,i]).

pub fn frobenius_norm(&self) -> f64

Frobenius norm: √(∑ aᵢⱼ²).

Trait Implementations

impl Clone for SparseMatrixGpu

fn clone(&self) -> SparseMatrixGpu

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseMatrixGpu

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

Formats the value using the given formatter.