I think the work list algorithm for dataflow analysis presented in Fig. 6.5 works only for live variable analysis but not many other dataflow analyses.
The problem is in the initialization of the mapping, which are all initialized to bottom.
In the general case, one needs to distinguish between extremal program points (entry/exit) and the other program points. Only the non-extremal program points must be initialized with bottom; the extremal points need to be initialized with some other value iota which should be an additional parameter (along with the extremal points).
For instance, for an available expression analysis, bottom is the set of all expressions in the program, whereas iota is the empty set. See also Table 2.8 and Figure 2.6 in "Principles of Program Analysis".
I think the work list algorithm for dataflow analysis presented in Fig. 6.5 works only for live variable analysis but not many other dataflow analyses.
The problem is in the initialization of the mapping, which are all initialized to bottom.
In the general case, one needs to distinguish between extremal program points (entry/exit) and the other program points. Only the non-extremal program points must be initialized with bottom; the extremal points need to be initialized with some other value
iotawhich should be an additional parameter (along with the extremal points).For instance, for an available expression analysis, bottom is the set of all expressions in the program, whereas
iotais the empty set. See also Table 2.8 and Figure 2.6 in "Principles of Program Analysis".