Jubi Taneja presents "Improving Compiler Construction Using Formal Methods"

Jubi Taneja explains how formal-methods-based techniques can help compiler developers create better static analyses and automatic peephole optimizers.

Talk by Jubi Taneja in the Berkeley Programming Systems Seminar on June 4, 2020 Talk Abstract: As programming languages are becoming more high-level to improve programmers' efficiency, the processors, in contrast, are moving towards specialization. The role of compilers is becoming crucial in bridging the ever-increasing gap between these two. We want compilers that are correct and generate high-quality code. Many researchers and developers invested efforts in solving these problems, but without making direct use of formal methods. In this talk, Jubi will describe how formal-methods-based techniques can help compiler developers create better static analyses, generate an automatic peephole optimizer for a compiler, and thus be used to improve overall compiler construction. Speaker Bio: Jubi is a PhD student at the University of Utah. Her research broadly focuses on compiler optimizations, correctness, and static analysis. She received the Best Paper Award and Best Student Presentation Award for her work on testing LLVM's static analyses at the CGO 2020. She started her learning of compilers at IIT Bombay as an undergraduate research fellow. She earned her B.E. from Punjabi University (Patiala, India). Jubi's homepage: cs.utah.edu/~jubi/ Seminar schedule: ps.berkeley.edu/seminar