Partitioned Memory Models for Program Analysis

Wei Wang, Clark W. Barrett, Thomas Wies

Abstract: Scalability is a key challenge in static analysis. For imperative languages like C, the approach taken for modeling memory can play a significant role in scalability. In this paper, we explore a family of memory models called partitioned memory models which divide memory up based on the results of a points-to analysis. We review Steensgaard’s original and field-sensitive points-to analyses as well as Data Structure Analysis (DSA), and introduce a new cell-based points-to analysis which more precisely handles heap data structures and type-unsafe operations like pointer arithmetic and pointer casting. We give experimental results on benchmarks from the software verification competition using the program verification framework in Cascade. We show that a partitioned memory model using our cell-based points-to analysis outperforms models using other analyses.