On the Importance of Program Representations in Static Analysis

Static analysis is a set of automatic or semi-automatic techniques to gather the semantic knowledge of a program without executing it. An important kind of semantic knowledge is the control flow and data flow information, which characterizes the instruction execution paths and the data generation/consumption paths, respectively. The obtained information is encoded in various graphical data structures or in the program itself directly for subsequent use. These data structures or instrumented programs are called program representations. In this survey, we first study four data structures that can help us dissect the code structure and dig deep structural properties. Then, we discuss the popular representations used in contemporary compiler, the single static assignment (SSA) form and its variants, which enable efficient program analysis algorithms by exploiting the sparsity of the data flow problem. We believe our surveyed knowledge can shed light to our future work.