Mitigating computational bottlenecks using graphics processors

I have become fascinated by the intersection of computer hardware and the fidelity of our numerical analyses. For example, graphics processors (GPUs), real-time computer chips used for gaming and machine learning, could be leveraged to accelerate our computational models, enabling faster and higher-fidelity simulations. In using GPU-based computing, the governing equations of motion would remain the same, but how they are solved would be fundamentally different.

I presented the challenges of GPU acceleration at the 3rd Joint UniValle/MECHS Workshop, held virtually on July 15-16. The topic area, “computational bottlenecks” is a major research thrust for real-time hybrid simulation with complex and larger (more degrees-of-freedom) models.

More information on the workshop can be found here: https://mechs.designsafe-ci.org/workshops/