In this paper we present a genetic scheduling algorithm to support the synthesis of structured data paths with the aim of producing designs with a predictable layout structure and conserving on-chip wiring resources. The data path is organized as architectural blocks (A-block) with local functional unit (FU), memory elements and internal interconnections. The A-blocks are interconnected by a few global buses. Our scheduling algorithm delivers the schedule of operations, the A-block in which each operation is scheduled and also the schedule of transfers over the buses, including transfers required to define variables in the basic block which remain live after its execution, all satisfying specified architectural constraints. The make up of the FUs in each A-block in terms of specific implementations of operators from a module database is also provided.

Keywords: Scheduling, Allocation, High-Level Synthesis (HLS), Genetic Algorithm (GA).