Serial decomposition of finite state machines for FPGA-based implementation
Grzegorz Borowik
Abstract
Grzegorz Borowik
Abstract
This paper presents a new method of the finite state machine logic synthesis intended for the modern FPGAs with embedded memory blocks. Although the functional decomposition is recognized as the most efficient method of digital circuits synthesis for implementation with FPGAs, none of the known state encoding algorithms is effective. This is caused by the fact that traditional methods comprise two steps: internal states encoding and, then, mapping of the encoded state transition table into target architecture. In this paper a new method of FSM state encoding is presented. It is an inherent part of the serial decomposition process and therefore no separate encoding step is required. It is shown that such state encoding guarantees the best solution. The paper presents examples from standard benchmark set, which confirm that the proposed method allows for a reduction of utilization of logic cells and embedded memory blocks.
A significance statement is not available in the OpenAlex record.
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 new method of the finite state machine logic synthesis intended for the modern FPGAs with embedded memory blocks. Although the functional decomposition is recognized as the most efficient method of digital circuits synthesis for implementation with FPGAs, none of the known state encoding algorithms is effective. This is caused by the fact that traditional methods comprise two steps: internal states encoding and, then, mapping of the encoded state transition table into target architecture. In this paper a new method of FSM state encoding is presented. It is an inherent part of the serial decomposition process and therefore no separate encoding step is required. It is shown that such state encoding guarantees the best solution. The paper presents examples from standard benchmark set, which confirm that the proposed method allows for a reduction of utilization of logic cells and embedded memory blocks.
Key concepts: Computer science, Encoding (memory), Field-programmable gate array, Finite-state machine, Benchmark (surveying), State (computer science), Decomposition, Parallel computing