Abstract:

Data dependence analysis determines the ordering relationships between
instructions in a unit of code that must be satisfied for the code to
execute correctly. It is a fundamental program analysis technique
used by optimizing compilers to identify constraints on data flow,
code motion, and instruction reordering. Dependence graphs are often
constructed as the result of data dependence analysis.

In this talk we describe a new approach to performing data dependence
analysis for Java in the presence of Java's "non-traditional" language
features such as exception, synchronization, and memory consistency.
We introduce new classes of edges in a dependence graph to model code
motion constraints arising from these language features. We also present
a linear-time algorithm for constructing this augmented dependence graph
for an extended basic block.