The Department of Computing Science at Umeå University (Sweden) invites applications for postdoctoral and PhD positions to conduct research on the compilation of linear algebra expressions.
Expressions containing vectors and matrices are ubiquitous in science and engineering. Many programming languages make it possible to input such expressions with almost no effort. However, the efficient evaluation (compilation) of linear algebra expressions is a problem still largely unsolved, even by the most advanced languages (e.g., Julia, Matlab).
The state-of-the-art compiler for linear algebra expressions is Linnea (https://linnea.cs.umu.se/). Linnea's most significant limitation is that all operands' sizes need to be known at compile time.
Recently, we received a grant from the Swedish Research Council to investigate how to effectively evaluate linear algebra expressions when the matrix sizes are unknown at compile-time. The project aims to address the problem using e-graphs. We hypothesize that an e-graph would be an effective tool for discovering and encoding a large set of alternative ways of evaluating a particular expression with unspecified matrix sizes. Once the matrix sizes become available (at run time), an extraction algorithm would then identify the optimal evaluation scheme.
If you are interested, these are the links to apply.
PhD student: https://umu.varbi.com/what:job/jobID:780040/
Postdoc: https://umu.varbi.com/what:job/jobID:782568/The successful candidates will be part of the High-Performance and Automatic Computing group (HPAC), working with Paolo Bientinesi and Lars Karlsson. HPAC’s webpage: https://hpac.cs.umu.se/
