Efficient encoding for exact symbolic automata-based scheduling

Abstract

1. ABSTRACT This paper presents an efficient encoding and automaton construction which improves performance of automata-based scheduling techniques. The encoding preserves knowledge of what operations occurred previously but excludes when they occurred, allowing greater sharing among scheduling traces. The technique inherits all of the features of BDD-based control dominated scheduling including systematic speculation. Without conventional pruning, all schedules for several large samples are quickly constructed. KeywordsHigh-level synthesis, scheduling, BDD, automata INTRODUCTIONThe scheduling problem occurs across diverse areas of application from networking to manufacturing to high-level synthesis of digital systems (HLS). Scheduling which assigns operations to time-slots in a synchronous system subject to data and control-flow dependencies as well as resource constraints is a key component of many HLS systems. Consequently, solving this problem efficiently is a direct way to enhance the abilities of such systems.Most solutions to the scheduling problem fall into two categories: i) heuristics and ii) integer linear programming (ILP

A model for scheduling protocol-constrained components and environments

“…Furthermore, control structures were not directly addressed in this short paper. Our previous work in [4] demonstrates a possible way of adding control. Finally, our use of symbolic reachability to determine valid schedules will be refined by related work in symbolic traversal techniques for verification.…”

A symbolic approach for the combined solution of scheduling and allocation

“…Although we use CDFGs as the input specification for our tool, we therefore decided to extend the model introduced by [6], that is at the same time formal (based on concurrent automata), efficient (we can use symbolic representation techniques [3] with enhancements derived from concurrent specification models), control-oriented (condition evaluation and speculative execution were specific features of [6]), and flexible (we can represent I/O constraints by restrictions on the automata state space).…”

“…In this paper we also address this issue, by combining the scheduling and allocation steps together, while keeping an implicit representation of the complete solution space (as [6] did for scheduling alone). The designer must still explore the design space by defining the acceptable maximum numbers of functional units and registers.…”

“…As [6] we represent implicitly the full solution space by means of the state space of a product of automata, and we represent resource (allocation) constraints by reducing the concurrency of the automata.…”

“…Once the set of valid schedules is computed symbolically, the extra information we encode allows us to find a schedule with best allocation cost, whereas in [6] a schedule is only checked to fit within the given resource bounds.…”