NeonGoby: Effective Dynamic Detection of Alias Analysis Errors

We are developing NeonGoby, a system for effectively detecting errors in alias analysis implementations, improving their correctness and hopefully widening their adoption. NeonGoby has effectively found 29 bugs in two very popular alias analyses implemented in LLVM, with very low false positive rate and reasonable performance overhead. 4 bugs we found have already been fixed by developers.

Read More…

Download NeonGoby

We released it open-source at GitHub, along with our error detection results and proposed patches.

Columbia University Department of Computer Science