Chapter 1.Introduction 4 We introduce BEAN to improve the precision of points-to analysis with small overhead increases. -BEAN is a new object-sensitivity approach for points-to analysis.Object- sensitivity [53,55,75]is usually considered as the most precise context- sensitivity for points-to analysis for Java [34].BEAN further improves its precision by automatically identifying and eliminating the redundant context elements in distinguishing contexts.Unlike traditional object- sensitivity,which obtains better precision by increasing the limit of con- text length k with usually significantly worse efficiency,BEAN is able to achieve better precision with still the same k-limiting at only small increases in analysis cost. - We thoroughly evaluate the effectiveness of BEAN by comparing it with traditional object-sensitivity [53,75]and the state-of-art hybrid context- sensitivity [34]for points-to analysis.Our evaluation shows that BEAN can always achieve better precision with small increases in analysis cost for real-world Java programs in practice. We introduce MAHJONG to significantly improve the efficiency of points- to analysis while maintaining nearly the same precision for type-dependent clients such as call graph construction. MAHJONG is a new heap abstraction that models the heap by iden- tifying and merging type-consistent objects,which are distinguished by the mainstream allocation-site-based points-to analysis.However, the allocation-site-based heap abstraction often over-partitions the heap without improving the precision much for an important class of type- dependent clients such as call graph construction,devirtualization andChapter 1. Introduction 4 • We introduce Bean to improve the precision of points-to analysis with small overhead increases. – Bean is a new object-sensitivity approach for points-to analysis. Object￾sensitivity [53, 55, 75] is usually considered as the most precise context￾sensitivity for points-to analysis for Java [34]. Bean further improves its precision by automatically identifying and eliminating the redundant context elements in distinguishing contexts. Unlike traditional object￾sensitivity, which obtains better precision by increasing the limit of con￾text length k with usually significantly worse efficiency, Bean is able to achieve better precision with still the same k-limiting at only small increases in analysis cost. – We thoroughly evaluate the effectiveness of Bean by comparing it with traditional object-sensitivity [53, 75] and the state-of-art hybrid context￾sensitivity [34] for points-to analysis. Our evaluation shows that Bean can always achieve better precision with small increases in analysis cost for real-world Java programs in practice. • We introduce Mahjong to significantly improve the efficiency of points￾to analysis while maintaining nearly the same precision for type-dependent clients such as call graph construction. – Mahjong is a new heap abstraction that models the heap by iden￾tifying and merging type-consistent objects, which are distinguished by the mainstream allocation-site-based points-to analysis. However, the allocation-site-based heap abstraction often over-partitions the heap without improving the precision much for an important class of type￾dependent clients such as call graph construction, devirtualization and