Critical-Path-Aware High-Level Synthesis for Fast Timing Closure

Abstract

Rapid advancement of process technology enables designers to integrate various functions onto a single chip and to realize diverse requirements of customers, but productivity of system designers has improved too slowly to make optimal design in time-to-market. Since designing at higher levels of abstraction reduces the number of design instances to be considered to acquire an optimal design, it improves quality of system as well as reduces design time and cost. High-level synthesis, which maps behavioral description models to register-transfer models, can improve design productivity drastically, and thus, it has been one of the important issues in electronic system level design. Centralized controllers commonly used in high-level synthesis often require long wires and cause high load capacitance, and that is why critical paths typically occur on paths from controllers to data registers instead of paths from data registers to data registers. However, conventional high-level synthesis has focused on delays within a datapath, making it difficult to solve the timing closure problem during physical synthesis. This thesis presents hardware architecture with a distributed controller, which makes the timing closure problem much easier. A novel critical-path-aware high-level synthesis flow is also presented for synthesizing such hardware through datapath partitioning, register binding, and controller optimization. We explore the design space related to the number of partitions, which is an important design parameter for target architecture. According to our experiments, the proposed approach reduces the critical path delay excluding FUs by 29.3% and that including FUs by 10.0%, with 2.2% area overhead on average compared to centralized controller architecture. We also propose two approaches, clock gating and register constrained flow, to alleviate high peak current problem which is caused by the proposed approach. These approaches suppress the peak current overhead to keep it less than 3.6%.