At AIAA SciTech 2020, the CFD 2030 Integration Committee hosted a Forum 360 discussion on the topic of High Performance Computing’s Impact on Aerospace Prediction. The blue ribbon panel included representatives from NASA, the DoE, the DoD, and the aircraft business.
HPC is a central component of the CFD 2030 Vision Study. Based on a predicted 30 exaFLOPS capability, the Study (published in 2014) cited several impediments to effective use of HPC platforms for achieving the Study’s vision.
- Current state of the art computing systems consume too much power.
- The entire CFD software stack doesn’t scale to the level of future HPC systems nor is it sufficiently fault-tolerant.
- Programming environments are insufficient to allow CFD experts to effectively implement algorithms that map to future computing environments.
- The 800 pound gorilla in the room is the simple fact that current CFD algorithms aren’t sufficiently scalable and new techniques must be developed.
- And as bad as things are for the CFD solver itself, the situation for pre- and post-processing software is even worse.
The Forum 360 panel addressed these issues and more in a two-hour discussion and presentation.
All of the panelists presentations and a video recording of the entire event are now available on cfd2030.com, the website of the CFD 2030 Integration Committee.
Take a look and listen and let me know what you think in the comments below.