Blocked Cholesky decomposition factorizes a Hermitian positive-definite matrix into an upper triangular factor \(R\), in the form \(A = R^H R\). The implementation processes the input matrix as compile-time sized square blocks to reduce local storage pressure and improve reuse in HLS.
This API supports complex float block Cholesky use cases, including the \(256 \times 256\) matrix size with a \(32 \times 32\) block size.