Section: New Results

Symbolic Analyses of pointers

Participants : Laure Gonnord, Maroua Maalej, Fernando Pereira [(UFMG, Brasil)] , Leonardo Barbosa [(UFMG, Brasil)] , Vitor Paisante [(UFMG, Brasil)] , Pedro Ramos [(UFMG, Brasil)] .

Alias analysis is one of the most fundamental techniques that compilers use to optimize languages with pointers. However, in spite of all the attention that this topic has received, the current state-of-the-art approaches inside compilers still face challenges regarding precision and speed. In particular, pointer arithmetic, a key feature in C and C++, is yet to be handled satisfactorily.

A first work presents a new range-based alias analysis algorithm to solve this problem. The key insight of our approach is to combine alias analysis with symbolic range analysis. This combination lets us disambiguate fields within arrays and structs, effectively achieving more precision than traditional algorithms. To validate our technique, we have implemented it on top of the LLVM compiler. Tests on a vast suite of benchmarks show that we can disambiguate several kinds of C idioms that current state-of-the-art analyses cannot deal with. In particular, we can disambiguate 1.35x more queries than the alias analysis currently available in LLVM. Furthermore, our analysis is very fast: we can go over one million assembly instructions in 10 seconds.

A second work starts from an obvious, yet unexplored, observation: if a pointer is strictly less than another, they cannot alias. Motivated by this remark, we use the abstract interpretation framework to build strict less-than relations between pointers. To this end, we construct a program representation that bestows the Static Single Information (SSI) property onto our dataflow analysis. SSI gives us an efficient sparse algorithm, which, once seen as a form of abstract interpretation, is correct by construction. We have implemented our static analysis in LLVM. It runs in time linear on the number of program variables, and, depending on the benchmark, it can be as much as six times more precise than the pointer disambiguation techniques already in place in that compiler.

This work has been published in the International Symposium of Code Generation and Optmization [31] and at CGO'17 [29].