A PLA-Based FSM Design Technique

We propose a system for optiniul design of PL.4- based FSMs. Previous approaches using symbolic mzn- imization use greater than minirniirn code len,gth re- sulting in higher area counts. We have implemented algorithms using minimum code length uiid utlempt io maximize the size/frequency of the cominoii cubes in the encoded machine. Our algoriilrirzs iiicliide impor- tant constraints and these are testrd for areu, delay and test-vector size. The resulis show iinprot~emeni in most of the cases of FSM benchmarks. Introduction The automatic synthesis of PLA-based FShI iii- volves functional design, logic design, topological cle- sign and physical design (l) which maps tlie functioiial description into a logic representation in terms of logic variables. State assignment profoundly alrects the area,delay and testability of the comhinationa~l corn- ponent of the FSM. Several heuristics ha.ve been SUE;- gested for state assignment. Symbolic minimization is a technique that yields a minimal encoding-independent suiii-of-products rep- resentation of a symbolic function. Symbolic miiii- mization based approaches like l<ISS(2),NOVA((3) and DIET(? were found to be effective in mininiiziiig the cover o the PLA. The complexity of the combinational conipoiient of the FSM depends heavily on the statc assignment. The PLA area is proportional to tlie product of t,he number of the rows times the nuinber of columns. While the number of rows is the cartlinalit,y of the cover of the PLA, the number of hits used to represent the states is related to the number of PLA columiis. Hence, even though the syinbolic minimizat8ion basc:tl approaches were effective in miniiiiizing tlie nunilm of rows of the PLA, they gave higher area cout1t.s because these approaches use patter than imininrii ti1 codelength.