Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
1406 JOURNAL OF COMPUTERS, VOL. 8, NO. 6, JUNE 2013<br />
State Assignment for F<strong>in</strong>ite State Mach<strong>in</strong>e<br />
Synthesis<br />
Meng Yang<br />
State Key Lab of ASIC and Systems, Fudan University, Shanghai, Ch<strong>in</strong>a<br />
Email: mengyang@fudan.edu.cn<br />
Abstract—This paper proposes simulated anneal<strong>in</strong>g based<br />
algorithm for the synthesis of a f<strong>in</strong>ite state mach<strong>in</strong>e to<br />
determ<strong>in</strong>e the optimal state assignment with less area and<br />
power dissipation. The algorithm has two anneal<strong>in</strong>g stages.<br />
In the first rough anneal<strong>in</strong>g stage it tries to search <strong>in</strong> global<br />
scope by the proposed rough search method. In the second<br />
focus<strong>in</strong>g anneal<strong>in</strong>g stage it tries to search <strong>in</strong> local scope by<br />
us<strong>in</strong>g proposed focus<strong>in</strong>g search methods <strong>in</strong>tend<strong>in</strong>g chang<strong>in</strong>g<br />
solution slightly. In both stages, the experience of past<br />
solution is utilised by comb<strong>in</strong>g the best solution <strong>in</strong> the past<br />
and the current solution. The experiments performed on a<br />
large suite of benchmarks have established the fact that the<br />
proposed method outperforms the published GA-based<br />
algorithms. The results have shown the effectiveness of the<br />
proposed method <strong>in</strong> achiev<strong>in</strong>g optimal state assignment for<br />
f<strong>in</strong>ite state mach<strong>in</strong>e.<br />
Index Terms—state assignment, f<strong>in</strong>ite state mach<strong>in</strong>e,<br />
optimisation algorithm, simulated anneal<strong>in</strong>g algorithm<br />
I. INTRODUCTION<br />
As the mobile applications are emerg<strong>in</strong>g, the power<br />
consumption of the circuits has become a major concern.<br />
Numerous researches have been <strong>in</strong>vestigated concern<strong>in</strong>g<br />
the power issues [1-3]. F<strong>in</strong>ite state mach<strong>in</strong>e (FSM) is<br />
mathematical model of the sequential circuits with<br />
discrete <strong>in</strong>puts, <strong>in</strong>ternal states and discrete outputs. The<br />
problem of f<strong>in</strong>d<strong>in</strong>g an optimal state assignment is NPhard<br />
[4]. As a result, the synthesis of an FSM plays an<br />
important role.<br />
The genetic algorithm (GA) technique [5] has been<br />
successfully applied to a variety of computationally<br />
complex problems s<strong>in</strong>ce it has a large search space. Many<br />
<strong>in</strong>vestigations have shown that GA can f<strong>in</strong>d good state<br />
assignments. Alma<strong>in</strong>i, et al [6] have demonstrated that<br />
the GA method produced significantly simpler solutions.<br />
In [7] multi objective GA (MOGA) has been used to<br />
optimise area and power simultaneously. Xia and<br />
Alma<strong>in</strong>i [8] have used GAs to optimise both area and<br />
power with good tradeoffs. Pradhan, et al [3] report on<br />
the application of power gat<strong>in</strong>g <strong>in</strong> the higher level of<br />
system design <strong>in</strong> the form of f<strong>in</strong>ite state mach<strong>in</strong>e (FSM)<br />
synthesis. In [9] Chattopadhyay has used GAs to obta<strong>in</strong><br />
power optimised two- and multilevel FSM realisations.<br />
Chattopadhyay [10] considers D flip-flops to store the<br />
state bits and <strong>in</strong>vestigates the avenue of GAs to achieve<br />
area reduction under flip-flop and output polarity<br />
selection.<br />
Other than GA based methods, a number of heuristic<br />
algorithms have been proposed, which are based on<br />
different cost functions estimat<strong>in</strong>g the effect of state<br />
assignment on logic m<strong>in</strong>imisation. It has shown a new<br />
comprehensive method <strong>in</strong> [11] consist<strong>in</strong>g of an efficient<br />
state m<strong>in</strong>imisation and state assignment technique. Goren,<br />
et al [12] present a heuristic for state reduction of<br />
<strong>in</strong>completely specified f<strong>in</strong>ite state mach<strong>in</strong>es. The<br />
proposed heuristic is based on a branch-and-bound search<br />
technique and identification of sets of compatible states<br />
of a given <strong>in</strong>completely specified f<strong>in</strong>ite state mach<strong>in</strong>e<br />
specification. In [13], the usage of a stochastic search<br />
technique <strong>in</strong>spired by simulated anneal<strong>in</strong>g is explored to<br />
solve the state assignment problem.<br />
Generally speak<strong>in</strong>g, it is relatively easy to f<strong>in</strong>d state<br />
assignments to m<strong>in</strong>imise the area only or the power<br />
dissipation only. However, it is known that m<strong>in</strong>imisation<br />
of either the power or logic complexity could be at the<br />
expense of the other and <strong>in</strong> most cases it is hard to f<strong>in</strong>d a<br />
solution that is optimum <strong>in</strong> both doma<strong>in</strong>s. For large<br />
circuits, there are millions or possibly billions of<br />
assignments [14] and hence to f<strong>in</strong>d the state assignment<br />
for the m<strong>in</strong>imisation of power consumption and area at<br />
the same time is computationally difficult. Besides, GA<br />
selects the next generation via a rank<strong>in</strong>g system, which is<br />
not always necessary but takes significant runtime. In this<br />
paper, <strong>in</strong> order to reduce the computational time but at the<br />
same time reta<strong>in</strong> the quality if the solution, simulated<br />
anneal<strong>in</strong>g based algorithm is proposed to solve FSM<br />
problem with low-power and small-area requirements.<br />
The rema<strong>in</strong>der of the paper is organised as follows.<br />
Section II gives term<strong>in</strong>ology of state assignment of the<br />
FSM. The two anneal<strong>in</strong>g stage simulated anneal<strong>in</strong>g<br />
approach is given <strong>in</strong> Section III. Section IV discusses the<br />
comparison results <strong>in</strong> details with respect to other<br />
approaches. Conclusions are then given <strong>in</strong> Section V.<br />
II. TERMINOLOGY<br />
An FSM can be characterised by a 5-tuple (I, O, M, X,<br />
Y) where I and O are the sets of primary <strong>in</strong>puts and<br />
primary outputs, M is the <strong>in</strong>ternal states, X and Y are the<br />
output and the next state functions, respectively. An FSM<br />
with M states requires a m<strong>in</strong>imum of S state variables for<br />
the assignment, where S = ⎡log<br />
2 M ⎤ and ⎡g ⎤ is the<br />
smallest <strong>in</strong>teger equal to or greater than g. The number of<br />
logically unique assignments for an FSM N is given as<br />
follows.<br />
© 2013 ACADEMY PUBLISHER<br />
doi:10.4304/jcp.8.6.1406-1410