Struct SimdAdjacency 

pub struct SimdAdjacency;

SIMD-accelerated adjacency matrix-vector product operations.

Provides optimized dense and sparse matrix-vector products that are central to many graph algorithms (PageRank, spectral methods, centrality).

Implementations

impl SimdAdjacency

pub fn dense_matvec( matrix: &Array2<f64>, vector: &Array1<f64>, ) -> Result<Array1<f64>>

Dense adjacency matrix-vector product using SIMD.

Computes y = A * x where A is the adjacency matrix. Each row uses SIMD dot product for acceleration.

Arguments

• matrix - Dense adjacency matrix (rows x cols)
• vector - Input vector (length cols)

Returns

Result vector of length rows.

pub fn dense_matvec_scaled( alpha: f64, matrix: &Array2<f64>, vector: &Array1<f64>, ) -> Result<Array1<f64>>

Dense matrix-vector product with alpha scaling: y = alpha * A * x.

Arguments

• alpha - Scaling factor
• matrix - Dense matrix (rows x cols)
• vector - Input vector (length cols)

Returns

Result vector of length rows.

pub fn sparse_csr_matvec( row_ptr: &[usize], col_idx: &[usize], values: &[f64], x: &[f64], n_rows: usize, ) -> Result<Vec<f64>>

CSR-style sparse matrix-vector product using SIMD.

Computes y = A * x where A is stored in compressed sparse row format. For each row, gathers x values at column indices and performs SIMD dot product.

Arguments

• row_ptr - Row pointer array (length n+1)
• col_idx - Column index array
• values - Non-zero values array
• x - Input vector
• n_rows - Number of rows

Returns

Result vector of length n_rows.

pub fn weighted_matvec( matrix: &Array2<f64>, vector: &Array1<f64>, row_weights: &Array1<f64>, ) -> Result<Array1<f64>>

Compute weighted adjacency matrix-vector product.

Like dense_matvec, but scales each row by a per-node weight (e.g., inverse degree for normalization).

Arguments

• matrix - Adjacency matrix (n x n)
• vector - Input vector (length n)
• row_weights - Per-row weight vector (length n)

Returns

Result vector where y[i] = row_weights[i] * sum_j(A[i,j] * x[j]).

pub fn adjacency_to_transition(adjacency: &Array2<f64>) -> Result<Array2<f64>>

Compute the transition matrix from an adjacency matrix using SIMD.

Each row of the adjacency matrix is divided by the row sum (out-degree), producing a row-stochastic matrix.

Arguments

• adjacency - Dense adjacency matrix (n x n)

Returns

Row-stochastic transition matrix.