Computation of Sequential Flexibility in Netlists by Windowing - The Unknown Component Problem: Theory and Applications - page 177

Fig. 11.2 Topology
of the result of latch splitting

The known solution for and in this case the outputs v are the outputs of

the latches. Note that because the outputs u of

X is called A

are from latches, the configuration
in Fig. 11.2 can have no combinational loops. Generally,

F

S is a Mealy machine, so

the input ,asshown.

The method of latch splitting will be illustrated with an example in more detail
in Sect. 13.3; there we will set up some extra correspondences between signals that
will force a simulation relation between state signals. The use of simulation relations
is the topic of Chap. 13.
Another way to select a window is to focus on a particular section of a netlist
for optimization. Around this section we can construct a surrounding window
which will serve as the specification for the section. How to compute a reasonable
surrounding window is the next topic.

i needs to be an input of X

11.3 Efficient Window Selection

Windowing is a method of limiting the scope of logic synthesis to work only on
a small portion of a Boolean network. This method is indispensable for scalability
when working with large Boolean networks arising in industrial applications.

11.3.1 Efficient Window Selection for Combinational Networks

The material in this section is adapted from [96], where windowing is used to
compute don't cares in large combinational networks.

Definition 11.3.1. Two non-overlapping subsets of nodes, the leaf set and the root
set, are in a leaf/root relation if every path from the PIs to any node in the root set
passes through some node in the leaf set.