Most scientific applications require considerable computer resources, but use applications and software libraries that have been developed during the last decades. Consequently, these programs are not suitable for modern and future HPC systems that contain millions of cores, many-core architectures, or GPGPUs. This seminar focuses on applications and solutions that have high demands on HPC systems. Scientific problems and their algorithmic formulation and efficient implementation on high performance computers will be addressed. The expected developments in the coming years will also be considered.
The seminar will cover a representative cross-section of scientific applications in physics and quantum chemistry, attempting to identify common problems and solutions. The selected applications range from on-line analysis using “capabilty computing” on hundreds of thousands of processor cores to data-driven “high throughput computing”.
This seminar is generously funded by the Wilhelm und Else Heraeus-Stiftung.