2026-06-11 –, Plenary
We implemented an N-body astrophysical simulation code and offloaded its most computationally intensive kernel to Tenstorrent RISC-V–based accelerators using the TT-Metalium programming interface. Performance was assessed on the Wormhole n300 card in terms of execution time and energy consumption, and compared with both an optimized CPU implementation and a CUDA version. The TT-Metalium implementation achieves a speedup of 2× over the CPU baseline, although its performance still slightly lags behind the CUDA implementation. Finally, we investigated strategies for scaling the application across multiple Tenstorrent accelerators, evaluating configurations with up to four devices.
Elisabetta Boella received her M.Sc. degree in Energy and Nuclear Engineering in 2009 from Politecnico di Torino (Turin, Italy) and her Ph.D. in Computational Plasma Physics in 2014 from the same institution. She currently works as HPC product specialist at E4 Computer Engineering (Scandiano, Italy), where she leads the company effort in several European projects, including MaX, SPACE and EoCoE. Her research interests include numerical modelling, parallel programming, and co-design practices. She has a long-time experience in the development and optimisation of parallel codes using the Message Passing Interface protocol. She is one of the main developers of the massively parallel plasma code ECsim. She also has extensive experience in Graphical Processing Unit (GPU) programming and off-loading of legacy codes to GPU.