Research

The overarching aim of our research group is the development of methods, algorithms, and software for scientific computing and uncertainty quantification.

Scientific computing on GPU-accelerated architectures has become an important direction in our research group. We pursue this research in close collaboration with international research teams developing algorithms and open-source software libraries for such systems, in particular, the Kokkos and Trilinos libraries.

Open-source software repositories

We develop our software in open-source repositories:

uliegecsm/kokkos-executionCustomization of std::execution for Kokkos
uliegecsm/reprospectA framework for reproducible prospecting of CUDA applications.
uliegecsm/graph-dispatchingThis repository is a playground for graph-based dispatching of asynchronous workloads using Kokkos and the P2300 formalism.
uliegecsm/kokkos-utilsThis repository contains utilities related to Kokkos.

Tools and utilities

We also develop a collection of smaller tools that support everyday work with CI, containers and GPU toolchains:

uliegecsm/amd-comgr-helpersCollection of helper scripts related to AMD Code Object Manager.
uliegecsm/hwloc-xml-parserPython tool to parse xml output of hwloc.
uliegecsm/system-helpersThis repository contains useful standalone helper scripts for Linux systems.