As part of our ongoing work to transfer knowledge gained during our runtime, we held our latest webinar this week on our experience with VerifiCarlo, an open-source framework designed to verify and optimize numeric aaccuracy in complex programs. Beyond the information offered in our best practice guide for measuring energy to solution, we hope this webinar will further support you in maximizing the energy eficiency of your codes. Join Pablo de Oliveira Castro, Université Paris-Saclay UVSQ, and our own Roman Iakymchuk, Umeå University on behalf of CEEC, to learn more about VerifiCarlo below.
Built on the LLVM infrastructure, VerifiCarlo provides various floating-point backends to simulate the effects of numerical errors and lower precision. By leveraging alternative floating-point models, such as Stochastic Rounding, Verificarlo pinpoints numerical bugs in simulation codes. A probabilistic definition of the number of significant digits allows one to estimate computational accuracy accurately.
Through its variable precision backend, Verificarlo enables one to explore the trade-offs between precision and performance by simulating lower precisions in software. It identifies specific code regions that benefit from reduced floating-point formats without sacrificing numerical correctness. This approach has been successfully applied in high-performance computing (HPC) domains such as neuroimaging pipelines, DFT quantum mechanical modeling, structural simulations, and now CFD.
In this webinar, we introduced Verificarlo, showcased its backends for numerical bug detection and mixed-precision analysis, and presented a success story highlighting the road from analysis of codes with Verificarlo to reliable mixed-precision codes.
Further Reading
- R. Iakymchuk et al., ‘Best Practice Guide — Harvesting energy consumption on European HPC systems: Sharing Experience from the CEEC project’, Zenodo, Aug. 2024. doi: 10.5281/zenodo.13306639.
- Y. Chen, P. de O. Castro, P. Bientinesi, N. Jansson, and R. Iakymchuk, ‘Enabling Mixed-Precision in Computational Fluids Dynamics Codes’, Mar. 03, 2025, arXiv: arXiv:2503.02134. doi: 10.48550/arXiv.2503.02134.
- Y. Chen, P. de O. Castro, P. Bientinesi, and R. Iakymchuk, ‘Enabling mixed-precision with the help of tools: A Nekbone case study’, May 17, 2024, arXiv: arXiv:2405.11065. doi: 10.48550/arXiv.2405.11065.