About

We develop programming languages and compiler systems that make high-level applications portable across different data representations and different machines.

In practice, that means building compilers for sparse and structured computation — tensor algebra, array programming, relational operations, and more — together with the languages and libraries that sit on top of them and the meta-compilers that generate compilers for new languages and targets. The unifying idea is to decouple a computation from how its data is stored and where it runs, so that one program can move across formats and hardware without being rewritten.

The group is not only about any single application area. Sparse computation is a major thread, but the broader project is the separation of computation from representation and target — a principle that reaches from CPUs and GPUs to accelerators and clusters.

The group

The Stanford Compilers Lab was established in 2021 and is directed by Fredrik Kjolstad, assistant professor of computer science. Currently, the group comprises 10 PhD students, several of whom are jointly advised by faculty in programming languages, systems, computer architecture, and graphics. The group’s research has appeared at PLDI, OOPSLA, ASPLOS, ISCA, CGO, and SC, and its software is released as open source.