public static operator@(notnull first as NDArray<T>, notnull second as NDArray<T>) as NDArray<T>
Matrix multiplication (2D x 2D). Float16, Float and Double products use packed CPU kernels; sufficiently large products share persistent workers. Independent result rows are computed in parallel without changing each dot product's accumulation order.
a = NDArray<Int>.full([2, 3], 1) b = NDArray<Int>.full([3, 4], 2) c = a @ b