Research · Jülich Supercomputing Centre · 2026
Adaptive Trees on a GPU
Adaptive octrees for the Fast Multipole Method, rebuilt on the GPU every time step.
High-Performance Computing · GPU Computing
Nils
Rottstegge
M.Sc. Computer Science, ETH Zurich
Studying at
M.Sc. Computer Science, major in Data Management Systems
Worked at
Jülich Supercomputing Centre, Guest Student Programme 2026
Computed on
Skills: CUDA, C/C++, CUDA-aware MPI, NCCL, OpenMP, OpenACC, HIP / hipify, CUDA Graphs, CUB, cuBLAS, cuSPARSE, Nsight Compute, Nsight Systems, Python, Julia, Slurm, CMake, Roofline analysis, CUDA Fortran, cuFFT, cuRAND, cuTENSOR, Git, Catch2 / CTest, Compute Sanitizer, cuda-gdb, Java, SQL / PostgreSQL.
Projects
Selected work
Adaptive trees, sparse matrix formats, multi-GPU solvers and PDEs on GPUs.
Hover an entry to see where it lands. Columns are cut into blocks of 8; a diagonal with enough entries in a block goes to the GPU, the rest to the CPU.
Bachelor thesis · ETH Zurich, SPCL · 2025
HBDIA: Hybrid Blocked-DIA
A sparse matrix format that splits the work between the GPU and the CPU of a GH200.
Project · ALPS supercomputer
Multi-GPU HPCG
Ported a single-GPU HPCG to many GPUs, then optimised it until it beat the original on one GPU.
Course project · ETH Zurich
Porous Convection on GPUs
Thermal convection solvers in Julia, distributed over GPUs with MPI.
About
I study Computer Science at ETH Zurich (M.Sc., Data Management Systems) and work on high-performance and GPU computing: sparse linear algebra, multi-GPU solvers and adaptive trees. Before computer science I earned a B.Eng. in Audiovisual Media, and I have worked as a live-TV video engineer since 2015.
Contact
Questions, opportunities or a good sparse matrix? Write me.