Abstract:

Existing pointer analysis algorithms usually adopt a lower-level intermediate representation which can not sufficiently represent the syntactical structure and semanteme of programs, which makes them difficult apply to source-to-source program transformation or analysis. To solve this problem， a flow-sensitive and contextsensitive pointer analysis algorithm based on system dependence graph is presented. An improved representation of alias information is proposed to describe the subscript details of elements in arrays， especially in pointer arrays. Then an iterative alias analysis for each procedure is performed on the system dependence graph of the program. An improved Banning algorithm or an inlining method is used to analyze the alias information respectively at each calling node according to a rule whether it’s an iterative call. Test results show that the precision of this approach is higher than those of Wilson’s and Emami’s approaches. The pointer analyzing results can be directly applied to source-to-source program transformation and analysis.