1. Domain Applications: Broader Community Perspective Mixture of apprehension and excitement about programming for emerging architectures

2. MPI+OpenMP? Charm++? MPI+CUDA? MPI+OpenACC? X10?

3. Domain Applications: Broader Community Perspective Mixture of apprehension and excitement about programming for emerging architectures Early adopters are the tip of the iceberg – how does this trickle down to the typical computational scientist? – What will be the analog(s) to MPI’s standardization of the communicating sequential process programming model? – “the MPI standard has been the most significant advancement in practical parallel programming in over a decade, and it is the foundation of the vast majority of modern parallel programs … MPI is evil … MPI’s success further stifled adoption of advanced parallel programming techniques” – Dunning, Harrison, and Nichols (2006).

4. Domain Applications on Leadership-Class Platforms There is a healthy mix of: – Evolutionary porting* of today’s codes and algorithms to new architectures, to stress limits of performance and scalability – Cross-platform comparison as proxies of future exascale systems, with various node (complexity/heterogeneity of cores, memory bandwidth/complexity) and network (topology and performance) characteristics (and programming models) – Revolutionary development of hybrid/hierarchical algorithms tailored to hybrid/hierarchical platforms (e.g. scale-bridging, process engineering examples) *Often including significant rewrites, e.g. Fortran+PETSc  C+MPI+X={OpenMP,CUDA} for GTC-P.

5. Domain Applications – Common Themes (1/2) Importance of international collaborations – E.g., E8 exascale climate and fusion projects to tackle shared societal (climate, energy) and technical challenges Continuing adoption of multiphysics* coupling in various domains *And multigrid, multiscale, multiprecision, etc. Need for performance measurement, analysis/visualization, and optimization tools appropriate to complexity (hierarchical compute & memory) and scale (millions to billions of threads, variability) of emerging systems

6. Domain Applications – Common Themes (2/2) Increased computational power is needed to: – Improve physical fidelity, e.g. empirical force fields for biomolecular & materials MD – Increase resolution, e.g. Earth system and industrial modeling – Increase time scales, e.g. protein folding, PIC convergence – Increase ensemble size, e.g. UQ and parameter/design space exploration – Tackle non-traditional problems, e.g. computational systems biology, integration of modeling & experiment (“Computational Steering 2.0”) Need to distinguish between “voracious” (more of same – just bigger & faster) vs. “transformational”.(achievement of major new levels of scientific understanding)