The highlight has been the creation of an open-source modular and reusable compiler infrastructure for machine learning and artificial computations called MLIR (Multi-level Intermediate Representation), where the faculty member was a co-founder and the key contributor to its polyhedral compiler aspects. MLIR provides a new intermediate representation to progressively lower dataflow compute graphs through loop nests to high-performance target-specific code. It is meant to serve as the infrastructure to drive compilation of the next generation of programming languages/models to a wide variety of general-purpose as well as specialized hardware. Within its first year, the project has had significant adoption in the compiler stacks of several ML/AI programming models.
Chris Lattner, Mehdi Amini, Uday Bondhugula, Albert Cohen, Andy Davis, Jacques A. Pienaar, River Riddle, Tatiana Shpeisman, Nicolas Vasilache, Oleksandr Zinenko: “MLIR: Scaling Compiler Infrastructure for Domain Specific Computation. IEEE/ACM International Symposium on Code Generation and Optimization.” CGO 2021: 2-14.
Chris Lattner, Mehdi Amini, Uday Bondhugula, Albert Cohen, Andy Davis, Jacques Pienaar, River Riddle, Tatiana Shpeisman, Nicolas Vasilache, Oleksandr Zinenko, “MLIR: A Compiler Infrastructure for the End of Moore’s Law.” arXiv:2002.11054 https://arxiv.org/abs/2002.11054.