TDM FIR can be configured to operate on multiple cascaded AI Engine Tiles using TP_CASC_LEN template parameter.
When used (TP_CASC_LEN > 1), an array of TP_CASC_LEN kernels will be created and connected through cascade interface. Each kernel will operate on a fraction of the TP_FIR_LEN requested (TP_FIR_LEN / TP_CASC_LEN).
FIR taps will be split in a balanced way that ensures maximum efficiency of the cascaded chain.
For example, a 16 tap FIR split over 2 cascaded kernels will result in each operating on 8 taps, while cascade of 3 will result in kernels operating on 6 taps on first kernel and 5 taps on remaining kernels.