One of the main focus points of ISC’12 will be HPC programming tools and application performance measurements — seen from different directions and tackled in various conference sessions and panels.

The panel “Programming Models in the Years to come,” will highlight the newest developments for programming tools intended for solving large-scale simulation problems for upcoming HPC systems. Four world-class panelists will contribute to this panel, offering their insights into future programming models.

 

The session “Application Performance: Lessons Learnt from Petascale Computing” will reflect the programming and development efforts that were critical to entering the petascale era and to significantly improve application performances for future HPC-Systems. Four speakers will present: 

• Performance engineering for multi-/many cores

• Automatic tuning frameworks to address the many degrees of freedom in code optimization

• The need for new application structures and libraries: From LAPACK to MAGMA

• Application performance matters: Design principles of the K-Computer

It is a well-known fact that the LINPACK algorithm, the basis for the TOP500 list, needs significant improvements to reduce the LINPACK runtime on HPC systems — especially at the petascale and beyond. In the meantime, other programs such as Green500 and, Graph500 are coming up to measure supercomputer performance in different ways. An entire session, “New Developments for Ranking Supercomputers,” is dedicated to these topics with:

• Reduced LINPACK to keep the run time manageable for future TOP500 lists

• Graph500 -philosophy, challenges, and the state of the list

• Tracking the effects of technology and architecture on energy through the TOP500, Green500 and Graph500

And, finally, it is evident that the lessons learnt from petascale computing have deep impacts on new HPC applications. One of the many application sessions at ISC’12 is “Novel HPC Simulations in Research & Industry,” where three top-level application areas will demonstrate their HPC efforts in performance improvements for:

• Supernovae and cosmology

• Aircaft design

• Challenges in material science