The section for Programming Languages and the Theory of Computation (PLTC) at the Department of Computer Science invites applicants for a PhD fellowship in the implementation and optimisation of data-parallel functional programming languages. More information: https://employment.ku.dk/phd/?show=157471 Start date is (expected to be) January 2023 or as soon as possible thereafter. Note that the application deadline is quite soon (20th of October). Modern massively parallel processors such as GPUs are highly efficient, but notoriously difficult and awkward to program for humans, particulary for irregular problems or those with nested structure. At DIKU, we are a small team that researches programming and compilation techniques for making high-performance parallel programming more accessible to non-expert programmers. We conduct applied research with a focus on actual real-world performance on mainstream hardware, and we use the functional array language Futhark and its implementation as our main experimental vessel. The precise research problems to be investigated by the PhD candidate depends on the interests of the candidate, but we are particularly interested in problems such as: * How to efficiently map fully or partially irregular nested parallelism to hardware, while still taking advantage of locality. * The construction of profiling tools that can precisely measure the performance of heavily compiler-transformed code and connect it to the original program, allowing the programmer to understand the high-level performance of their program. * Optimising data layout based on how it is accessed by the program; possibly also taking distributed execution or NUMA behaviour into account. For more information, feel free to contact Troels Henriksen (athas@diku.dk).