Module cache_ops 

Cache-aware matrix operations for high-performance computing.

This module provides cache-optimized implementations of common matrix operations, including tiled matrix multiplication and cache-oblivious transpose. The CacheAwareConfig struct exposes cache topology information and derives optimal blocking parameters so that working sets fit in the appropriate cache level.

Examples

use scirs2_core::cache_ops::{CacheAwareConfig, tiled_matmul};
use ndarray::Array2;

let config = CacheAwareConfig::detect();
let a = Array2::<f64>::eye(4);
let b = Array2::<f64>::eye(4);
let c = tiled_matmul(&a, &b);
assert_eq!(c, a);

Structs

CacheAwareConfig

Cache topology description used to derive blocking parameters.

Functions

cache_oblivious_transpose

In-place cache-oblivious transpose of a square Array2<f64>.

prefetch_matmul

Matrix multiplication with software prefetch hints for pipelined execution.

tiled_matmul

Cache-efficient tiled matrix multiplication C = A × B.