Fencing off unwanted behavior: Improving and evaluating the Fency static analysis tool

Pieter van den Ham

Date: Fri, December 02, 2022
Time: 15:00
Room: Van der Poelzaal (Building 36)
Note: This is a MSc thesis defense

Computer architectures with weak memory models, such as ARMv8 and ARMv7, allow memory accesses to be reordered in many situations. Fency is a static analysis tool that inserts memory fences to ensure that a program exhibits the same behavior when run on a weaker memory model. We expand Fency with a new dependency tracking analysis, integrate it with LLVM’s alias analysis infrastructure, and improve its usability. (thesis link)

Previous: Liam Clark |
Next: Jesper Cockx |