The DFT library element is able to split the workload into multiple kernels connected in a chain with the use of the TP_CASC_LEN template parameter. The required computation, as well as input data buffer and twiddle values, will be split and distributed between the number of kernels determined by TP_CASC_LEN. Using multiple kernels, support for TP_POINT_SIZE is extended beyond what can be stored on a single kernel. The data must be split in a “card-dealing” manner across the kernels in cascade. This is necessary as the generated twiddles for the DFT are also split this way.