I have seen the development on OpenMP target offload at GitHub - exasmr/openmc: OpenMC Monte Carlo Code, are there plans to merge these two efforts? What are blockers in pursuing such a merge?
I read that, wondering is the question mark on slide 31 has changed at all since May? This work looks great, and would love to see progress and maybe even help with perf on NVIDIA systems. Know that there are projects hoping to use this.
2 Likes
From a technical perspective it hasn’t changed since May (still on a separate branch), but the development team at Argonne was awarded a new grant to push this forward in collaboration with GE, so there will be some efforts starting up shortly to get the GPU functionality in the main branch. @mattb please reach out to us if you’d like to chat further!
1 Like
Just in case it is useful this text is in the nuclear energy section of the website.
Transforming Reactor Design: Fast-Tracking Industrial Adoption of GPU-Powered OpenMC
Lead Lab: Argonne National Laboratory
The DOE has invested heavily in the development of the OpenMC Monte Carlo particle transport code via the recent Exascale Computing Project. Global Nuclear Fuels-Americas (GNF-A), the cost-share partner for this project, is seeking to apply OpenMC for reactor engineering design and analysis work in order to exploit OpenMC’s GPU-powered performance on modern GPU HPC resources. The significant speedups offered by OpenMC on GPU vs. CPU are expected to both 1) compress the costs and design timelines of existing engineering workflows for LWRs, and 2) empower advanced reactor design efforts. As proven by ExaSMR milestone results, OpenMC on GPU has already demonstrated a high technology readiness level, leaving relatively few barriers to commercial adoption by GNF-A. This proposal aims to address those few remaining barriers so as to bridge the final gap between DOE research and industrial adoption.
1 Like