Questions for dataflow:

1.  Describe what is meant by dead code elimination.  Give an algorithm
    acting on a program represented as a collection of basic blocks
    which marks all dead code in some fashion.  Give two uses of such
    information.  Comment on how label variables would influence
    your answer.  What would your algorithm make of:
    let start() be
    $( g()
       return
    l: h()
       goto l
    $)

2.  Explain what is meant by the call graph of a (bcpl) program.
    To what extent does it indicate the actual flow of control
    at run time?  How do function valued variables affect
    your definition.  Give an algorithm to report on procedures
    which are demonstrably never called.  Can the algorithm
    be modified to remark on procedures which are unlikely to
    be called?

3.  Define path expressions.  Explain how certain path expressions
    can be used to indicate:
    1. data flow anomalies (indicate which data flow you regard as
       anomalous).
    2. whether x (a variable) is live at a particular point
    3. whether x (a variable) is available at a particular point.

4.  Define the notion of a variable being live at a particular point
    in a flowgraph.  Give an algorithm for deriving the set of
    live variables at each program point.  Explain how such
    information may help register allocation.

5.  Explain the notion of an expression e being available at a
    particular point in a flowgraph.  By considering a suitable
    example program (with some nodes having multiple predecessors)
    how a program may be optimised to take account of such
    information.  Give also a 'near miss' where an expression
    is not available on every route to a given program point.
    Give an algorithm for determining the set of avaiable
    expressions at every program point.