Performance driven resynthesis by exploiting retiming-induced state register equivalence
P. Kalla, M.J. Ciesielski
Abstract
P. Kalla, M.J. Ciesielski
Abstract
This paper presents a retiming and resynthesis technique for cycle-time minimization of sequential circuits with feedback (finite state machines). Operating on the delay critical paths of the circuit, we perform a set of controlled local retimings of registers across fanout stems and logic gates, followed by local node simplifications. We guide the retiming of registers across fanout stems to induce equivalence relations among them, which are exploited for subsequent logic simplification. Our technique is able to analyze correlation of logic across register boundaries during simplification. We strive to minimize the increase in number of registers without sacrificing the cycle-time performance. The results demonstrate a favourable performance/area trade-off when compared with optimally retimed circuits.
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This paper presents a retiming and resynthesis technique for cycle-time minimization of sequential circuits with feedback (finite state machines). Operating on the delay critical paths of the circuit, we perform a set of controlled local retimings of registers across fanout stems and logic gates, followed by local node simplifications. We guide the retiming of registers across fanout stems to induce equivalence relations among them, which are exploited for subsequent logic simplification. Our technique is able to analyze correlation of logic across register boundaries during simplification. We strive to minimize the increase in number of registers without sacrificing the cycle-time performance. The results demonstrate a favourable performance/area trade-off when compared with optimally retimed circuits.
Key concepts: Retiming, Sequential logic, Computer science, Logic synthesis, Equivalence (formal languages), Shift register, Minification, Logic gate