A general constraint-centric scheduling framework for spatial architectures
Proceedings of the 34th ACM SIGPLAN conference on Programming language design and implementation
| Hi-index | 0.03 |
Synthesizing a digital system from a functional description is a complex process requiring the solution of various different problems. TASS (Tabu Search Synthesis System) is a functional synthesis system made up of interdependent modules based on new and more efficient algorithms. First, a control and data flow graph model for system representation is developed and presented. This model generates a single graph representing both the data and control flows of a VHDL behavioral description. A new representation of conditional branches and a mutual exclusion testing procedure offering optimized resource sharing and critical path reduction possibilities have been developed. This graph model is an environment used for various synthesis needs starting from high-level transformations to FSM synthesis for controller implementation. A new mathematical formulation of the scheduling problem is developed using a new approach based on penalty weights. This approach avoids the inflexibility of the ILP formulations developed in related works where the functional unit performing each type of operation is fixed prior to scheduling. The Tabu Search technique, which has been effective in finding optimal solutions for many types of large and difficult combinatorial optimization problems, has been adapted for this purpose. This technique, which performs an intelligent and fast solution space exploration, combined with an effective mathematical formulation makes the scheduling algorithm presented here very powerful