Minimalizacja poboru mocy wspólnego modelu automatów skończonych

W artykule przedstawiono nowy algorytm kodowania stanów wewnętrznych automatu skończonego o obniżonym poborze mocy. Zastosowano w nim wspólny model automatu klas ADE co pozwoliło to na zmniejszenie ilości przerzutników przechowujących kod stanu. Badania symulacyjne przeprowadzone z wykorzystaniem standardowych układów testowych potwierdziły skuteczność kodowania z wykorzystaniem proponowanego algorytmu w porównaniu z algorytmami JEDI oraz NOVA, jak i zawartymi we wcześniejszych pracach autorów.

---

In this paper there is addressed the problem of power minimisation of the finite state machine (FSM). Power reduction is of great importance in design of digital systems as it can improve the speed and extend the time between recharging the batteries in mobile systems. In the common model of the FSM of class ADE (Section 2) the set A of internal states consists of three subsets: AA, AD, and AE. AA is the set of internal states of the FSM of class A, AD is the set of internal states of the FSM of class D (the output vector is identical to the next state code), and AE is the set of internal states of the FSM of class E (the input vector is identical to the next state code) [12]. The common model of the FSM of class ADE requires an additional register used for storing the input and output vector values. These registers are present in modern programmable logic devices. In Section 3 there is proposed a new algorithm of the FSM state assignment that makes use of the common model. The assigned code consists of three parts: G - input vector, Z - output vector and E - state code. G and Z are stored in the input and output registers, respectively. With this algorithm it is possible to assign codes that are shorter than those assigned with use of classical methods, and thus less power is dissipated in registers storing the current state code during every transition. The experimental results (Section 4, Tables 1 and 2) show the significant reduction (of 13 to 51%) in power dissipation compared to classic (JEDI, NOVA, column-based) and recent (sequential and iterating) algorithms.