The Finder subgraph performs a fine-grained search of the MUSIC spectrum regions tagged by the Scanner subgraph to identify a negative-to-positive gradient change in the spectrum as highlighted in the diagram above. When a gradient change is found, the tag is set to the index of the bin corresponding to the local minimum. To meet the 1 $\mu s$ throughput target, this search is partitioned over a \(16\) tile pipeline where each tile performs the fine-grained search on two of the 8-bin regions. The diagram below shows the AI Engine physical array view for the Finder subgraph.