Function gemm 

pub fn gemm<L: LoopOrder>(
    alpha: bool,
    a: MatrixTileSlice<'_>,
    b: MatrixTileSlice<'_>,
    beta: bool,
    c: MatrixTileSliceMut<'_>,
)

Performs tile-level GEMM with a specified loop ordering.

This is the sequential (non-parallel) version. For large matrices, use gemm_concurrent instead.

Loop Ordering

The choice of loop order affects cache locality and performance. Benchmarking suggests RIC is optimal for most cases, but this depends on matrix dimensions.